Point-of-care ultrasound in the acute abdomen and gastrointestinal emergencies: An evidence-based guide and the case for standardisation

Abstract

Point-of-care ultrasound (POCUS) is embedded in emergency assessment of acute abdominal pain, yet its evidence is scattered across emergency, surgical, anaesthetic and radiology journals. This structured narrative review synthesises four applications: (1) Small-bowel obstruction; (2) Gastric aspiration-risk assessment; (3) Non-traumatic free intraperitoneal fluid; and (4) Biliary disease including cholangitis. Sources came from a purposive search of PubMed, Scopus and Web of Science through August 2026, favouring the highest available evidence per application; it was English-language only, unregistered, and without review-level risk-of-bias appraisal or de novo pooling. We report thresholds, pooled sensitivities and specificities, likelihood ratios and, where retrievable, heterogeneity and risk-of-bias data. POCUS for small-bowel obstruction and cholecystitis is primarily a rule-in tool: A negative scan does not reliably exclude disease, and neither defines aetiology, ischaemia or perforation, for which computed tomography remains necessary. For non-traumatic free fluid the strongest rule-in evidence concerns suspected ectopic pregnancy; gastric POCUS stratifies aspiration risk against surrogate endpoints. We contribute an integrated, symptom-based, pretest-probability-driven algorithm with a competency precondition, a non-diagnostic exit and a defined action for every result; it is not prospectively validated. Dispersion across operators, thresholds and reference standards exceeds the margins bedside decisions turn on; we argue standardisation, not further accuracy estimation, is rate-limiting.

Key Words: Point-of-care ultrasound; Acute abdomen; Small bowel obstruction; Gastric ultrasound; Cholecystitis; Emergency medicine; Diagnostic accuracy; Standardization

Core Tip: This review integrates emergency point-of-care ultrasound for small-bowel obstruction, gastric aspiration risk, non-traumatic free fluid and biliary disease into one symptom-based, pretest-probability-driven algorithm, giving numeric thresholds, likelihood ratios and an explicit action for every result, including non-diagnostic scans. The algorithm has not been prospectively validated. Assembling the evidence also shows that dispersion across operators, thresholds and reference standards, together with absent reliability data for several core measurements, makes standardisation rather than further accuracy estimation the most likely rate-limiting step, a proposition this review frames as testable.

INTRODUCTION

Acute abdominal pain is the most common reason for an adult emergency department (ED) visit, and no standardised diagnostic pathway exists for it; in nationally representative United States data the proportion of adult ED visits with abdominal pain as the primary reason for visit rose between 2007 and 2019, reaching 11.6% at ages 18-45 years, 9.0% at ages 46-64 years and 6.5% at ages 65 years and older[1]. Its differential ranges from benign and self-limiting conditions to surgical emergencies that threaten life within hours (15%-20% of emergency patients with acute abdominal pain require interventional or surgical treatment, and mortality of 2%-12% rises with every hour of delay to specific treatment[2]) and the clinician must separate the two quickly, often before laboratory results return and sometimes without immediate access to computed tomography (CT). Point-of-care ultrasound (POCUS) has become part of that first assessment.

POCUS in emergency and critical-care medicine has developed considerably over the past twenty years. What began as focused trauma and cardiac imaging now extends to the biliary tract, the bowel, the stomach, and the dependent peritoneal recesses[3-5], supported by consensus statements and scoping reviews mapping its scope in acute care[6-8]. The evidence supporting each of these applications, however, has accumulated in separate literatures: (1) Small-bowel obstruction (SBO) has been studied largely by emergency physicians (EPs)[9-13]; (2) Gastric assessment almost entirely by anaesthetists[14-17]; (3) Free-fluid detection by trauma surgeons[18,19]; and (4) Biliary imaging by radiologists and surgeons[20-22]. Two syntheses do take a multi-application approach and are the closest antecedents of the present work: (1) A practical scan-protocol review of POCUS in gastrointestinal disease[23]; and (2) A systemic sonographic approach to the acute abdomen in the ED[24]. Both establish that several applications can be organised into one bedside examination, and the algorithm proposed here builds on that premise rather than displacing it. What neither provides, and what this review adds, is three-fold: (1) The presenting symptom cluster is used to set pretest probability and thereby to order the applications by expected yield; (2) Every result is assigned an explicit action, so that a positive, a negative and a non-diagnostic study each have a defined next step; and (3) The non-diagnostic examination is represented as a third exit rather than being collapsed into a negative one. Absent those three elements, a clinician at the bedside of an undifferentiated patient still has no single reference specifying which scan to perform first, what threshold defines a positive result, and what to do with the answer.

Two problems follow from this fragmentation, and this review addresses both. The first is practical: No integrated pathway exists. The review’s principal contribution is a symptom-based, pretest-probability-driven scanning algorithm that links the presenting syndrome to a prioritised sequence of POCUS applications and specifies an explicit action for each result (Figure 1). The second problem becomes apparent only when the evidence for the four applications is assembled in one place, and it is the more consequential. Viewed together, the pooled estimates do not describe a single stable examination. Sensitivity for SBO is 92.4% at study level but 83.0% in individual-patient-data (IPD) analysis[9,10]; attending physicians reach 87.7% sensitivity against 73.0% for residents[10]; specificity for cholecystitis is 85% across all operators and 92% when the analysis is restricted to EPs (Figure 2)[25]; and the antral cross-sectional area (CSA) thresholds used in gastric assessment differ by more than twofold across populations and body positions[26-29]. Such dispersion is not the behaviour of a fixed test. It is the behaviour of undefined combinations of operator training, scanning protocol, measurement convention and reference standard.

Figure 1 Integrated, symptom-based, pretest-probability-driven point-of-care ultrasound scanning algorithm for the acute abdomen. The pathway opens with two entry conditions: A mandatory instability and red-flag check (red) – shock, peritonitis, suspected ruptured ectopic pregnancy, ruptured abdominal aortic aneurysm, bowel ischaemia or perforation divert the patient to a resuscitation-first track with expedited surgical consultation, and bedside scanning never delays it – and an operator-competency precondition, below which a negative scan is treated as uninformative. For the stable patient, each presenting symptom cluster sets the pretest probability and selects an entry scan. Every scan node has three exits: Adequate and positive (green, rule-in actions), adequate and negative (orange, rule-out caution and redirect), and non-diagnostic (grey), which returns the clinician to the pretest probability and mandates confirmatory imaging, consultation or a serial scan. A rule-in biliary scan routes to surgical consultation with Tokyo Guidelines 2018 severity grading, which determines the destination, and the cholangitis pathway is shown separately. The aortic branch distinguishes detection of an aneurysmal aorta from detection of rupture. A parallel gastric branch (purple) stratifies aspiration risk at any point. The algorithm has not been prospectively validated. POCUS: Point-of-care ultrasound; AAA: Abdominal aortic aneurysm; RUQ: Right upper quadrant; TG18: Tokyo Guidelines 2018; CT: Computed tomography; CSA: Cross-sectional area.

Figure 2 Pooled diagnostic accuracy, heterogeneity and threshold dispersion in emergency point-of-care ultrasound. A: Circles denote pooled sensitivity and squares pooled specificity, with 95%CIs where the source reports them. The four core applications appear above the divider and the key differentials of the acute abdomen below it. Where more than one meta-analysis exists for an application, each source is plotted separately rather than combined, so that between-source dispersion is visible. The right-hand columns give the number of contributing studies and patients and the heterogeneity statistic reported by each source; values of I² at or above 60% are shown in orange, and NR indicates that the value could not be recovered from the retrievable source text; B: Published antral cross-sectional area thresholds for gastric point-of-care ultrasound plotted on a common scale. Gastric assessment is shown separately because it stratifies aspiration risk against a surrogate endpoint rather than functioning as a diagnostic accuracy test. Each threshold is specific to the body position, population and target volume in which it was derived, and the thresholds are not interchangeable. SBO: Small-bowel obstruction; IPD: Individual-patient-data; QUADAS-2: Quality Assessment of Diagnostic Accuracy Studies 2; EP: Emergency physician; TG18: Tokyo Guidelines 2018; AAA: Abdominal aortic aneurysm; RLD: Right lateral decubitus; POCUS: Point-of-care ultrasound; N/A: Not applicable.

This observation reframes what limits these applications, and it is the position this review argues: The binding constraint is more likely to be the absence of a standardised procedure than any further increment in diagnostic accuracy. The Focused Assessment with Sonography for Trauma (FAST) examination is the instructive precedent. FAST achieved universal adoption despite unexceptional accuracy, pooled sensitivity for thoracoabdominal injury in blunt trauma is 0.74 (95%CI: 0.65-0.81)[18], because a consensus process first fixed the binary question, the acoustic windows and the minimum image set. Only once the procedure was defined could curricula, case-number requirements, credentialing and quality assurance be built upon it. No equivalent framework exists for the four applications reviewed here.

This review therefore has a deliberately narrow scope and two linked aims. It synthesises the diagnostic performance of four high-yield emergency POCUS applications (SBO, gastric assessment for aspiration risk, free intraperitoneal fluid beyond trauma, and biliary disease) reporting explicit thresholds, pooled sensitivities and specificities, likelihood ratios and the heterogeneity and risk-of-bias data behind them. It then integrates the four into a single algorithm for bedside use, which has not been prospectively validated, and sets out what a consensus process would need to fix before these applications can be standardised as FAST was. The four applications were selected because each is high-yield, each falls within routine emergency practice, and each currently sits in a different literature.

METHODOLOGY

Four emergency POCUS applications form the focus of this structured narrative review: (1) SBO; (2) Gastric content and aspiration risk; (3) Free intraperitoneal fluid beyond trauma; and (4) Biliary disease including acute cholecystitis. We searched three bibliographic databases for English-language articles: (1) PubMed/MEDLINE; (2) Scopus; and (3) The Web of Science Core Collection (editions SCI-EXPANDED, SSCI, AHCI, CPCI-S, CPCI-SSH, BKCI-S, BKCI-SSH and ESCI). Records identified through citation chaining of included articles and through publisher alerts were also screened.

In each database a POCUS concept block (“point-of-care ultrasound“, “point of care ultrasound“, “POCUS“, “bedside ultrasound“, “emergency ultrasound“, “point-of-care ultrasonography“ or “point-of-care sonography“) was combined with an application-specific block using the Boolean operator AND. The application blocks were: (1) For SBO, “small bowel obstruction“, “small-bowel obstruction“, “bowel obstruction“ or “intestinal obstruction“; (2) For gastric assessment, “gastric ultrasound“, “gastric content“, “gastric antrum“, “antral cross-sectional“, “aspiration risk“ or “gastric volume“; (3) For free fluid, “free fluid“, “free intraperitoneal fluid“, “peritoneal free fluid“, “ascites“, “hemoperitoneum“ or “intraperitoneal fluid“; and (4) For biliary disease, “acute cholecystitis“, “cholecystitis“, “biliary colic“, “gallbladder“, “cholelithiasis“, “choledocholithiasis“ or “biliary tract“. Searches used the Topic field (TS =) in Web of Science, the TITLE-ABS-KEY field in Scopus, and title and abstract fields in PubMed. Analogous term sets covered the principal differentials of the acute abdomen [appendicitis, renal colic and hydronephrosis, abdominal aortic aneurysm (AAA)], inter-observer reliability, artificial intelligence (AI), special populations, low-resource settings, and training, competency and quality assurance. The last search was run on July 3, 2026 and was updated on August 8, 2026 and again on August 17, 2026 during preparation of this revision. The complete database-specific search strings as executed, with field tags, run dates, records retrieved per query, the number remaining after deduplication, and the numbers screened, assessed in full text and included, are reported in Supplementary Table 1; the corresponding source-selection flow is shown in Figure 3. Because selection was purposive rather than exhaustive, these counts document what was retrieved and how it was reduced, and are not offered as evidence of complete capture of the literature.

Figure 3 Source identification and selection for this structured narrative review.

Two reviewers (Danış F and Kudu E) independently screened titles and abstracts against the review scope and examined the full text of potentially relevant records; the final set was agreed by consensus, with disagreement resolved by discussion. We included English-language guidelines, systematic reviews, meta-analyses, and diagnostic-accuracy or clinically influential primary studies addressing the four target applications and their principal differentials. For the core diagnostic-accuracy synthesis we prioritised studies from emergency and acute-care settings. Studies from perioperative, obstetric, radiology or other settings were retained when they provided threshold, reliability, special-population or implementation data not available from emergency cohorts. We excluded non-English reports and articles reporting neither a diagnostic threshold, nor an accuracy estimate, nor a directly relevant clinical principle. Source selection was structured but purposive: Rather than aiming at exhaustive capture, we preferentially selected the highest available level of evidence for each application, prioritising current society guidelines first, then systematic reviews and meta-analyses, and finally clinically influential or recent primary diagnostic-accuracy studies to fill gaps.

Diagnostic estimates (pooled or study-level sensitivity, specificity, likelihood ratios and diagnostic thresholds) were extracted from the primary sources by one reviewer (Danış F) and checked against the source article by the second (Kudu E); this is verification of a single extraction rather than independent duplicate extraction. Every numeric value reported in the text and in Table 1 was checked in this way, and any disagreement was resolved by both reviewers re-reading the original article together[9-11,18,19,25-27,30-37]. Verification here means that each value was confirmed against its source, not that every desired value exists: Where a source does not report a quantity, the corresponding cell records NR, and these entries are findings about the source rather than unchecked data. Supplementary Table 2 gives the source, the location of the value within it, the population, the outcome definition and the extracted number for every estimate in Table 1 and Figure 2[9-11,18,19,25-27,30-37]. For each pooled estimate we additionally sought the number of contributing studies and patients, the reference standard, the heterogeneity statistic and any formal risk-of-bias or certainty-of-evidence appraisal; these are reported in Table 1[9-11,18,19,25-27,30-37], and where a value could not be recovered from the retrievable text this is stated rather than left blank. Every reference was cross-checked against PubMed and Crossref for bibliographic accuracy and screened for retraction or expression of concern using PubMed publication types and Crossref update records; none of the cited articles is retracted or under an expression of concern, and the five entries carrying a published erratum are cited in their corrected form.

Table 1 Diagnostic performance, heterogeneity and risk of bias of emergency point-of-care ultrasound applications in the acute abdomen.

Ref.	Population/setting	Operator	Reference standard	Diagnostic threshold	Sensitivity (95%CI)	Specificity (95%CI)	LR+/LR-	Studies/patients	Heterogeneity (%)	Risk of bias/certainty	Interpretation and limitations
SBO: Study-level meta-analysis (Gottlieb et al[9], 2018)	ED and inpatient adults with suspected SBO	Mixed (EP and others)	CT, surgery, or final diagnosis	Dilated loop > 25 mm	92.4% (89.0%-94.7%)	96.6% (88.4%-99.1%)	27.5/0.08	11/1178	NR in retrievable text	QUADAS-2 not reported in retrievable text; no GRADE	Strong rule-in; study-level pooling may overestimate sensitivity
SBO: IPD meta-analysis (Shokoohi et al[10], 2023)	Multicentre ED, adults	EP (attending and resident)	CT/surgery/final diagnosis	Dilated loop > 25 mm	83.0% (71.7%-90.4%)	93.0% (55.3%-99.3%)	11.9 (1.2-114.9)/0.2	5/433	NR in retrievable text	IPD design reduces reporting bias; no GRADE	Most conservative estimate; specificity CI so wide the parameter is effectively unestimated
SBO: Meta-analysis with QUADAS-2 (Motavaselian et al[30], 2024)	ED adults with suspected SBO	Mixed	CT/surgery/final diagnosis	Dilated loop > 25 mm	NR here (see source)	0.80 (0.77-0.83)	N/A	21/1977	62% sensitivity; 88% specificity (both P < 0.01)	QUADAS-2: High risk for patient selection and flow/timing in several studies; Egger P = 0.04	Specificity 16 points below the two estimates above; publication bias detected
SBO: Bedside-threshold analysis (Shokoohi et al[11], 2024)	ED adults	EP	CT/surgery	Loop > 2.75 cm (AUC = 0.76, 0.71-0.81); ≤ 1.7 cm 100% sensitivity; > 4 cm	N/A	Approximately 90.7% at > 4 cm	N/A	IPD secondary analysis	N/A	Unvalidated threshold-derivation study	Candidate threshold only; no measurement convention specified; age/sex shift optimum (2.95 cm at ≥ 65 years)
Acute cholecystitis: EP-POCUS meta-analysis (Wilson et al[32], 2024)	ED adults, suspected acute cholecystitis	EP	Final diagnosis	Stones + wall > 3 mm + Murphy sign + pericholecystic fluid	70.9% (62.3%-78.2%)	94.4% (88.2%-97.5%)	12.7/0.31	10 included; estimates from 7/1772	NR in retrievable text	Not reported in retrievable text	Rule-in test; negative scan does not exclude
Acute cholecystitis: All-operator meta-analysis (Huang et al[25], 2023)	Mixed settings, all specialties (EP subgroup)	All operators; EP subgroup	Final diagnosis/pathology	As above	71% (69%-72%); EP 71% (67%-74%)	85% (84%-86%); EP 92% (90%-93%)	N/A	40/8652	89.7% (87%-92%)	QUADAS-2: > 25% of studies high risk (case-control design, non-consecutive enrolment)	Operator specialty changes specificity by 7 points
Acute cholecystitis: Prospective TG18-adjudicated cohort (Nipitkul et al[33], 2026)	ED adults, high disease prevalence	EP	TG18	POCUS criteria vs TG18 (wall > 3 mm vs ≥ 4 mm scales)	86.6% (83.4%-89.4%)	17.1% (12.5-22.5)	NR/NPV 37.0%	1/783	N/A (primary study)	Prospective; single reference standard	Discordant with pooled specificity of 85%-94%; only gallstones were significantly associated with confirmed disease
Gastric: ED intubation cohort (Asokan et al[27], 2023)	ED adults undergoing urgent intubation	EP	Aspiration-relevant gastric content	Antral CSA ≥ 9.27 cm²	100%	87%	N/A	1/100	N/A	Single-centre	Risk stratification against a surrogate endpoint, not diagnosis
Gastric: Fasting IPD meta-analysis (Perlas et al[26], 2024)	Fasting adults (reference range)	Trained sonographers	Endoscopic/measured volume	95th-percentile CSA 9.9 cm² (9.4-10.4); volume 2.3 mL/kg (2.3-2.4)	Grade 0/1 → 98% probability below threshold	N/A	N/A	12/1203	NR in retrievable text	Reference-range derivation	Thresholds do not generalise across position or population
Free fluid: Ectopic pregnancy (Moore et al[31], 2007)	ED women, suspected ectopic pregnancy	EP	Operative intervention	Any free fluid in Morison's pouch	N/A	N/A	LR+ 112 (15-831)	1/242	N/A	Single-centre; estimate rests on 10 fluid-positive patients	Decisive rule-in in a high-pretest obstetric population only
Free fluid: Prehospital FAST meta-analysis (Lin et al[19], 2024)	Prehospital trauma	Prehospital providers	CT/laparotomy	Any free fluid	0.596 (0.345-0.822)	0.970 (0.953-0.983)	N/A	6/1356	94%a	GRADE: Sensitivity very low (inconsistency, imprecision, publication bias); specificity high	Trauma population, indirect evidence for non-traumatic disease; indicates the sensitivity ceiling
Free fluid: Cochrane review, blunt trauma (Stengel et al[18], 2018)	Blunt thoracoabdominal trauma, adults and children	Mixed	CT/laparotomy/clinical follow-up	Any free fluid	0.74 (0.65-0.81); children 0.63 (0.46-0.77)	0.96 (0.94-0.98)	N/A	34 studies	Reported in source	Cochrane methodology	At 28% prevalence, 73 injured adults and 118 injured children missed per 1000 scanned
Appendicitis: POCUS meta-analysis (Matthew Fields et al[35], 2017)	ED adults and children	Non-radiologist physicians; EP subgroup	Pathology/final diagnosis	Non-compressible, dilated appendix	91% (83%-96%); EP subgroup 80%	97% (91%-99%); EP subgroup 92%	N/A	21/NR	99% (99%-100%)	Not reported in retrievable text	Extreme heterogeneity; EP-only performance lower than the headline figure
Appendicitis: Paediatric meta-analysis (Miller et al[36], 2026)	Children	Mixed	Pathology/final diagnosis	Non-compressible, dilated appendix	85.6% (68.9%-94.1%)	90.2%	N/A	8/993	Sensitivity range 53%-100% across studies	Recent; see source	Between-study range wider than the pooled CI implies
Hydronephrosis/renal colic (Wong et al[37], 2018)	ED adults, suspected renal colic	EP	CT	Any hydronephrosis	70.2% (67.1%-73.2%)	N/A	N/A	NR in retrievable text	NR in retrievable text	Not reported in retrievable text	Risk stratification only; cannot exclude
AAA (Shaban et al[34], 2025)	ED adults, symptomatic	EP	CT/operative/final diagnosis	Aortic diameter ≥ 3 cm (aneurysm definition)	98.33%	99.84%	N/A	9/NR	NR in retrievable text	Not reported in retrievable text	Identifies an aneurysmal aorta; does NOT identify rupture, for which CT angiography is the reference standard

^aFor Lin et al[19], 2024 the source reports this 94% number for sensitivity from a χ² test of heterogeneity, not as an I² statistic, and reports 45% for specificity by the same test; the value is reproduced here as published and the column is therefore headed heterogeneity rather than I².

Values are reproduced from the cited sources; no estimates were recalculated. Every value in this table was checked against the cited source; NR denotes a value that the source itself does not report, and is therefore a finding about the source rather than an unverified entry. Full source-by-source extraction details, including the location of each value within its source, are given in Supplementary Table 2. AAA: Abdominal aortic aneurysm; AUC: Area under the receiver-operating-characteristic curve; CI: Confidence interval; CSA: Cross-sectional area; CT: Computed tomography; ED: Emergency department; EP: Emergency physician; GRADE: Grading of Recommendations Assessment, Development and Evaluation; IPD: Individual-patient-data; LR+: Positive likelihood ratio; LR-: Negative likelihood ratio; NPV: Negative predictive value; NR: Not report; QUADAS-2: Quality Assessment of Diagnostic Accuracy Studies 2; SBO: Small-bowel obstruction; TG18: Tokyo Guidelines 2018; POCUS: Point-of-care ultrasound; N/A: Not applicable.

The conduct and reporting of this narrative review were informed by the domains of the Scale for the Assessment of Narrative Review Articles[38]. Consistent with that design, the review was not registered, and we did not perform a formal risk-of-bias assessment of individual studies, a certainty-of-evidence [Grading of Recommendations Assessment, Development and Evaluation (GRADE)] evaluation, or any de novo quantitative synthesis; all pooled diagnostic estimates are reproduced from the cited meta-analyses rather than recalculated here, and the risk-of-bias and GRADE information in Table 1 is that reported by those source meta-analyses[9-11,18,19,25-27,30-37]. Because this is a review of published literature involving no human subjects and no patient-level data, institutional review board approval and informed consent were not required for the review itself.

GENERAL PRINCIPLES OF ABDOMINAL POCUS

POCUS is ultrasonography performed and interpreted at the bedside by the treating clinician to answer one focused clinical question. This differs from radiology’s comprehensive consultative examination in purpose, in scope and in the interpretive standard applied[3,6,39]. The distinction matters for how the accuracy data in this review should be read: The estimates that follow describe a focused question asked by a clinician who already knows the patient, not a complete diagnostic study.

All four applications share the same technical basis. A low-frequency curvilinear transducer (2-5 MHz) provides the depth needed for the gallbladder, the dependent recesses and the gastric antrum, while a high-frequency linear transducer (5-12 MHz) improves resolution of superficial bowel in thinner patients[7,24,40]. Image acquisition follows the same sequence in each: (1) Identify the anatomical landmark; (2) Sweep systematically through the region of interest in two planes; and (3) Interpret the finding against the clinical question that prompted the scan.

A single principle governs interpretation throughout this review: A POCUS finding is meaningful only when weighed against the pretest probability of disease, and does not function as a stand-alone rule-in or rule-out test. The post-test probability depends jointly on the likelihood ratio of the finding and on the probability of disease before the transducer touched the patient. A positive scan in a patient with a compatible syndrome carries very different weight from the same image in a patient with low pretest probability, and this Bayesian framing is what allows the four applications to be sequenced into a single algorithm.

Two further principles underpin the argument of this review. First, a diagnostic estimate is a property of a procedure rather than of a technology: Any pooled sensitivity reported below applies to the specific combination of operator, protocol, measurement convention and reference standard from which it was derived. Second, the between-study heterogeneity that pervades this literature is therefore not statistical noise but evidence that the procedure itself is not fixed. Operator experience, body habitus and bowel gas recur as performance modifiers across all four applications; each section quantifies them, and the section on standardisation draws them together.

SBO

Rationale and technique

SBO accounts for a substantial share of ED visits for abdominal pain and of surgical admissions, and CT remains the conventional confirmatory modality[41,42]. POCUS offers a radiation-free, repeatable bedside alternative that fits into the initial workup. The operator sweeps all four abdominal quadrants in a “lawnmower“ pattern, using a curvilinear transducer, or a linear one for thinner patients, to look for bowel dilatation, altered peristalsis and free fluid[24,43]. Obstruction is often segmental, so missing a quadrant can mean missing the finding. Graded compression and systematic coverage of all quadrants are therefore required components of the examination.

Diagnostic criteria and thresholds

Dilated fluid-filled small-bowel loops beyond normal calibre are the sonographic hallmark of SBO[44,45]. The conventional threshold was borrowed from CT: A loop diameter greater than 25 mm (2.5 cm, Figure 4A)[9,11]. A more recent bedside-specific secondary analysis of individual-patient data identified 2.75 cm as the most discriminative threshold, with an area under the receiver-operating-characteristic curve of 0.76 (95%CI: 0.71-0.81); the same analysis reported that a diameter of 1.7 cm or less carried 100% sensitivity, and that specificity reached approximately 90.7% at thresholds above 4 cm[11]. Optimal thresholds also shifted with age and sex, at 2.95 cm in patients aged 65 years or older and in men against 2.75 cm in younger patients and in women[11].

Figure 4 Annotated sonographic criteria on representative point-of-care ultrasound images acquired by the authors. A: Small-bowel obstruction – a fluid-filled dilated loop with valvulae conniventes (the keyboard sign), with a second dilated loop also labelled; B: Acute cholecystitis – an intraluminal gallstone with posterior acoustic shadowing, adjacent to the anechoic gallbladder lumen; C: Acute cholecystitis – the gallbladder wall, the anechoic lumen and a gallstone. Arrows indicate the labelled structures. Diagnostic thresholds (dilated small-bowel loop > 2.5 cm conventionally, with 2.75 cm proposed as a bedside-specific candidate; gallbladder wall > 3 mm in the emergency literature and ≥ 4 mm in the Tokyo Guidelines 2018) are defined, referenced and compared in the main text. GB: Gallbladder.

Three cautions attach to the 2.75 cm number, and this review presents it as a candidate threshold rather than an established one. First, it derives from a single unvalidated secondary analysis and has not been externally validated in an independent prospective cohort. Second, an area under the receiver-operating-characteristic curve of 0.76 indicates modest discrimination; it does not support the arithmetic precision that a two-decimal threshold implies. Third, and most consequential for reproducibility, no published convention specifies how the measurement is to be made. The source analysis reports maximal diameter but does not state whether the loop is measured outer-wall-to-outer-wall or inner-wall-to-inner-wall, in longitudinal or transverse plane, at the point of maximal distension or as an average across loops[11]. A 2.5-mm difference separates the conventional 2.5 cm threshold from the proposed 2.75 cm one, and outer-to-outer vs inner-to-inner measurement of a bowel loop with a 3-mm wall can differ by more than that.

Several supportive findings can sharpen the diagnosis: To-and-fro or whirling peristalsis within a dilated loop, the keyboard sign traced out by the valvulae conniventes, thickened bowel wall, and free fluid pooling between loops. Free interloop fluid in particular tends to indicate a more advanced or already complicated process[12,24].

Diagnostic accuracy

POCUS performs well for the specific question of luminal dilatation[9], although it does not identify the transition point, the cause of obstruction, bowel ischaemia or perforation, for which CT remains necessary[41,42,45]. The aetiology may also be one that no cross-sectional modality characterises definitively: Infiltrative and infectious causes such as gastrointestinal tuberculosis mimic other intra-abdominal pathology and are established only by a combination of biochemical, microbiological, radiological and endoscopic assessment[46]. A study-level systematic review and meta-analysis of 11 studies and 1178 patients reported a sensitivity of 92.4% (95%CI: 89.0%-94.7%) and specificity of 96.6% (95%CI: 88.4%-99.1%), with a positive likelihood ratio (LR+) of 27.5 and a negative likelihood ratio (LR-) of 0.08[9]. A second study-level meta-analysis of 15 studies reported a pooled sensitivity of 92% (95%CI: 89%-95%) with specificity of 93% (95%CI: 85%-97%), and found specificity significantly lower when the examination was performed in the ED and when CT was the sole reference standard[13]. The most conservative estimate comes from an IPD meta-analysis of five prospective studies and 433 patients, which reported a pooled sensitivity of 83.0% (95%CI: 71.7%-90.4%), specificity of 93.0% (95%CI: 55.3%-99.3%), LR+ of 11.9 (95%CI: 1.2-114.9) and LR- of 0.2[10].

Those estimates should not be read as three measurements of one quantity. The specificity interval in the IPD analysis runs from 55.3% to 99.3%, which is wide enough that the parameter is effectively unestimated, and the LR+ interval (1.2-114.9) spans the range from clinically useless to decisive[10]. A third meta-analysis, of 21 studies and 1977 patients, pooled specificity at 0.80 (95%CI: 0.77-0.83), roughly 13-17 percentage points below the two numbers above, with I² of 62% for sensitivity and 88% for specificity (both P < 0.01), a Quality Assessment of Diagnostic Accuracy Studies 2 appraisal showing high risk of bias for patient selection and for flow and timing in several studies, and an Egger test indicating significant publication bias (P = 0.04)[30]. Earlier reviews extend the evidence to paediatric populations and to the question of using POCUS to rule out obstruction[44,47-49]. A systematic review addressing implementation in resource-limited settings reported sensitivity ranging from 85.0% to 100% and specificity from 54.0% to 100% across studies[49].

The divergence between study-level and IPD pooling should not be read as a simple correction of bias in study-level meta-analysis. The lower estimate in the IPD analysis may reflect differences in study inclusion, harmonised participant-level definitions, patient spectrum and analytic approach[9,10]. IPD analysis avoids the between-study reporting bias that inflates study-level estimates and therefore yields a lower, more conservative sensitivity[9,10]. But the specificity spread from 0.80 through 0.93 to 0.966 across four meta-analyses of the same application is not explained by pooling method alone, and reflects genuine differences in the populations, operators, thresholds and reference standards contributing to each, a conclusion one of those sources reaches itself, reporting that specificity varied significantly by study setting, by operators from different continents, and by reference standard[13]. The clinical conclusion is nonetheless stable across all of them: POCUS is a strong rule-in test for SBO, while its sensitivity, most conservatively estimated at 83.0%, is not high enough for a negative scan to exclude the diagnosis. Where pretest probability is high, a negative bedside examination does not rule out obstruction and confirmatory CT should still be obtained[9,10,13].

Performance modifiers

Operator skill largely determines accuracy. In the IPD meta-analysis, attending physicians reached a sensitivity of 87.7% and specificity of 91.4%, against 73.0% and 88.2% for residents, a 14.7%-percentage-point sensitivity gap for the same nominal examination[10]. Performance also shifted with body habitus in the same analysis: Sensitivity and specificity were 88.6% and 84.0% in patients with a body mass index below 30, against 72.0% and 89.5% at body mass index of 30 or above[10]. An early comparison of emergency-medicine against radiology residents for SBO[50] established the pattern that training background alters performance for the same examination. Prospective ED studies show that EPs across a range of training levels can perform the examination with clinically useful accuracy, and that POCUS shortens time to diagnosis while changing management and disposition[12,51-53]. Experience is associated with better clinician-performed accuracy[54], and guideline-concordant management of adhesive SBO tracks with better outcomes[55]. Adhesive SBO remains a more difficult target; here POCUS works best as an early triage test that speeds surgical consultation and CT in equivocal cases rather than as a basis for operative management on its own[41,42,45].

GASTRIC POCUS

Framing: Aspiration-risk stratification, not disease diagnosis

Gastric POCUS differs from the other three applications in this review: It is not a diagnostic test, because there is no disease to detect. It assesses the nature and volume of gastric contents in order to stratify aspiration risk before procedural sedation, emergency airway management or anaesthesia[14,15]. The test characteristics reported below describe how well antral measurements classify a stomach as empty rather than at-risk, not how well they confirm or exclude a condition[14,56,57]. Because it stratifies risk rather than diagnosing disease, gastric POCUS must never delay definitive airway management or any other time-sensitive resuscitative intervention; it informs how and when the airway is secured, not whether urgent management proceeds.

One limitation of this evidence base deserves to be stated plainly, because it constrains every number that follows. Every cited study links antral measurement to gastric volume, whether aspirated, endoscopically measured or model-predicted. No study links antral measurement to aspiration itself, or to any other patient-centred outcome[16,17,26-28]. Gastric volume is a surrogate endpoint, and its relationship to clinical aspiration, an uncommon event requiring large samples to study, remains unquantified. The recommendations in this section therefore rest on a mechanistic assumption that a fuller stomach carries greater risk, not on demonstrated outcome benefit.

Technique and qualitative grading

To image the gastric antrum, a curvilinear transducer (or a linear probe in slimmer patients) is swept through a sagittal or parasagittal epigastric plane, using the left lobe of the liver and the aorta or superior mesenteric vessels as anchor points. The patient is scanned supine first, then in the right lateral decubitus (RLD) position[15,16]. Perlas and colleagues developed a three-point qualitative grading system on this basis: (1) Grade 0, an antrum empty in both positions; (2) Grade 1, clear fluid visible only in RLD, indicating a small volume; and (3) Grade 2, clear fluid present in both positions, indicating a larger volume[16,17]. The qualitative grade correlates with measured volume and allows rapid bedside categorisation; a grade of 0 or 1 indicates a high probability that gastric volume lies below the threshold associated with elevated aspiration risk[16,26].

Quantitative thresholds

The quantitative route uses antral CSA as the input to models predicting gastric volume (Figure 5)[15,17]. An individual-participant-data meta-analysis of 12 studies and 1203 fasting subjects established the reference range: A 95^th-percentile antral CSA of 9.9 cm² (95%CI: 9.4-10.4) and gastric volume of 2.3 mL/kg (95%CI: 2.3-2.4); when the qualitative grade is 0 or 1, the probability that antral area falls below this 95^th percentile is 98%[26]. In the ED, among 100 patients undergoing urgent endotracheal intubation, a calculated CSA cut-off of 9.27 cm² or greater identified the at-risk stomach with 100% sensitivity and 87% specificity[27]. In 73 labouring women scanned supine, an antral area of 3.81 cm² (381 mm²) identified an empty, grade-0 stomach with 81% sensitivity and 76% specificity[28].

Figure 5 Gastric point-of-care ultrasound for aspiration-risk stratification. A: Representative antral image acquired by the authors, showing the gastric antrum with the anterior liver as the acoustic window; antral cross-sectional area is obtained from the anteroposterior and craniocaudal diameters as described in the text; B: The Perlas qualitative grading system (grades 0-2) with the corresponding aspiration-risk category, and the emergency-department and fasting-reference antral cross-sectional area cut-offs. Each threshold applies only to the body position and population in which it was derived, and the full set of published thresholds and their dispersion is shown in Figure 2B. All cited studies relate antral measurement to gastric volume; none relates it to aspiration or to any other patient-centred outcome. RLD: Right lateral decubitus.

Placed side by side, these values are not interchangeable. Non-interchangeable here means that a threshold derived in one body position, population and target volume cannot be substituted for another: Each was fitted to a different reference condition, so applying one outside its derivation setting changes what the cut-off actually classifies. The cited literature contains at least nine such thresholds, reported here in a common unit: (1) 9.9 cm² as a fasting 95^th percentile in RLD[26]; (2) 9.27 cm² for predicting aspiration risk at ED intubation[27]; (3) 3.81 cm² (381 mm²) supine for grade 0 in labour[28]; (4) 6.08 cm² (608 mm²) semi-recumbent and 9.60 cm² (960 mm²) in RLD in pregnancy[29]; (5) Separate values of 6.7 cm² and 10.0 cm² for two subgroups within a single dyspepsia study[58]; and (6) 6.92 cm² supine and 10.65 cm² in RLD in older adults[59]. The published thresholds therefore span 3.81 cm² to 10.65 cm², a 2.8-fold range (Figure 2B).

Emergency-department relevance and modifiers

Sedation and intubation in the ED rarely occur in a patient with a truly empty stomach, and gastric POCUS provides a concrete tool when the fasting history is unclear or unavailable. Point-of-care studies show that a substantial proportion of ED and peri-procedural patients have a full stomach despite nominal fasting intervals, and that bedside antral assessment predicts aspiration-relevant content in both adults and children[56,57,60]. Anaesthesiology and regional-anaesthesia societies have incorporated gastric POCUS into consensus recommendations for aspiration-risk assessment in medically complex and non-elective patients[14,61]. The technique has been described across groups relevant to the ED’s mixed population, including pregnancy and the postpartum period, patients with comorbidities such as diabetes, and scenarios involving thick or mixed gastric content[29,58,62].

Reliability data for the measurement are thinner than the precision of the published thresholds implies. The one available study reported intra-rater and inter-rater correlation coefficients above 0.8 with maximum inter-rater differences within 13%, but it studied 22 healthy volunteers, three raters and clear fluid only, outside the emergency setting[63]. Two recent studies bear on reliability, and together they locate the problem precisely. In 122 adults scanned before ED procedural sedation, three blinded reviewers classifying antral contents as low or high risk achieved a kappa of 0.79; but that number describes the qualitative binary judgement, not measurement of antral CSA[64]. When CSA measurement itself was tested formally in 28 term-pregnant patients scanned by anaesthesiologists of two expertise levels, intra-rater reliability was good to excellent while inter-rater reliability was only poor to moderate, and the authors framed the finding as the difficulty of transferring gastric ultrasound skills to novice operators[65]. The qualitative grading therefore transfers between observers considerably better than the quantitative measurement on which most published thresholds depend, and neither study addresses whether either predicts aspiration. Body habitus, prior gastric surgery and difficulty quantifying solid or mixed content remain practical limitations.

FREE FLUID BEYOND TRAUMA

Scope and technique

Detection of free intraperitoneal fluid is the POCUS skill best known from trauma, but it is equally useful in the non-traumatic acute abdomen. Ectopic pregnancy with haemodynamically significant complications is the setting in which it matters most, although ascites, spontaneous haemoperitoneum and ruptured viscera also fall within its scope[31,66,67]. The technique mirrors the FAST examination: Using a curvilinear transducer, the sonographer scans the dependent regions, the hepatorenal recess (Morison’s pouch), the splenorenal recess and the pelvis (rectouterine pouch of Douglas; Figure 6). In early pregnancy, transvaginal or transabdominal pelvic ultrasonography is added to characterise the uterus and adnexa[31,68,69].

Figure 6 Point-of-care ultrasound of free intraperitoneal fluid beyond trauma, on images acquired by the authors. A: Sagittal pelvic view showing the urinary bladder, the uterus and anechoic free fluid in the dependent pelvis; B: Sagittal pelvic view showing free fluid in the rectouterine pouch (pouch of Douglas) posterior to the uterus. Arrows indicate the labelled structures. In a high-pretest-probability obstetric population, even a small anechoic collection in a dependent recess is clinically meaningful; the significance of free fluid in other populations is interpreted in its own clinical context.

Diagnostic value and thresholds

Outside trauma, free-fluid POCUS is useful mainly because it is specific rather than sensitive: (1) A positive result carries weight; and (2) A negative one usually carries little. The evidence is population-specific and must not be generalised across all causes of non-traumatic intraperitoneal fluid. In suspected ectopic pregnancy, a landmark ED study of 242 patients found that free fluid in Morison’s pouch predicted the need for operative intervention with an LR+ of 112 (95%CI: 15-831)[31]. The width of that interval reflects the fact that the estimate rests on the ten patients in whom free fluid was identified, and the point estimate should be read accordingly. In this high-pretest-probability obstetric population, even a small volume of anechoic fluid in a dependent recess is clinically important and supports expedited surgical and gynaecological involvement. The finding should not be extended unchanged to non-obstetric haemoperitoneum from other ruptured viscera, or to ascites, where volume, character and cause carry different implications[31,66]. In the obstetric setting, clinical series of ectopic pregnancy presenting with free intraperitoneal fluid, together with sonographic studies of tubal rupture, support the pattern that fluid volume and distribution correlate with the need for intervention[70,71].

Sensitivity is a different matter, and the honest statement of the evidence is that no meta-analysis of free-fluid POCUS sensitivity in a non-traumatic abdominal population exists. The closest quantitative estimate comes from a meta-analysis of six studies and 1356 patients of prehospital free-fluid ultrasound in trauma, which reported a pooled sensitivity of 0.596 (95%CI: 0.345-0.822) against a pooled specificity of 0.970 (95%CI: 0.953-0.983)[18]. That source graded the certainty of its sensitivity estimate as very low under GRADE, downgrading for serious inconsistency (heterogeneity 94% for sensitivity by χ² test, 45% for specificity), serious imprecision and strongly suspected publication bias, while grading specificity as high[19]. It is a trauma population and therefore indirect evidence for non-traumatic disease; it is cited here as an indication of the sensitivity ceiling of free-fluid detection, not as a population-specific number. The same asymmetry appears in the Cochrane review of POCUS in blunt trauma, which translates a pooled sensitivity of 0.74 into absolute terms: At a median prevalence of 28%, POCUS would miss 73 injured patients per 1000 scanned[19].

AI-assisted interpretation of free-fluid POCUS has shown promising accuracy and may eventually widen access to the examination. A meta-analysis of seven retrospective studies and 2332 patients (predominantly, though not exclusively, trauma FAST, and including one non-trauma ascites cohort) reported pooled sensitivity of 91.4% (95%CI: 81.6%-96.3%) and specificity of 96.8% (95%CI: 86.5%-99.3%) in its broader abdominal free-fluid analysis[72]. All included studies were retrospective and image-based rather than prospective and patient-based, and this evidence is discussed further in the section on AI below.

The implication of these data is asymmetric. In the appropriate clinical context a positive scan is decisive and should accelerate definitive care. A negative scan must always be weighed against pretest probability and should not override strong clinical suspicion[19,31,73]. The point is sharpened by data from patients in whom a false negative is most dangerous: Among patients with non-compressible torso haemorrhage, FAST was positive in only 51% of those with abdominal or pelvic bleeding, a false-negative rate of 49%, rising to 61% for pelvic sources[74].

Applications and modifiers

Bedside detection of free fluid does more than raise suspicion for ectopic pregnancy: It supports the diagnosis of symptomatic ascites, guides paracentesis and contributes to the workup of spontaneous haemoperitoneum[67,73]. During first-trimester pelvic assessment, POCUS may identify an intrauterine pregnancy, which substantially lowers the probability of ectopic pregnancy, or may reveal adnexal findings and free fluid that raise it. A structured emergency approach improves diagnostic accuracy for the causes of acute pelvic pain[66-68], and POCUS performed for vaginal bleeding or abdominal pain in early pregnancy has been evaluated directly in the ED[75]. Performance depends on the acoustic window and the volume of fluid present; an appropriate transducer and the transvaginal window, used where indicated, improve detection[40,69].

How reliably clinicians agree on this finding is less well established than its accuracy. In trauma FAST, inter-observer agreement between EPs and radiologists has been reported at kappa 0.525 in one study and 0.865 in another, for nominally the same binary task[76,77], a spread of 0.34 that is itself evidence of an unstandardised examination. Agreement is worst precisely where it matters for the non-traumatic abdomen: In a study of 81 paediatric FAST examinations, distinguishing trace or small collections from moderate or large ones yielded kappa of 0.65 (95%CI: 0.63-0.66) among trained paediatric EPs and 0.48 (95%CI: 0.47-0.49) among paediatric radiologists[78]. For non-traumatic emergency populations, no inter-observer reliability data for the detection of a small free-fluid collection were identified.

BILIARY TRACT, ACUTE CHOLECYSTITIS AND CHOLANGITIS

Rationale and technique

Right upper quadrant pain is a frequent ED presentation, and among abdominal applications biliary POCUS has strong validation[32,79]. The gallbladder is scanned with a curvilinear transducer using subcostal and intercostal windows in both long and short axis. Scanning begins with the patient supine, moving to left lateral decubitus when stones need repositioning or the acoustic window is poor[79,80]. The examination assesses gallstones, wall thickness, pericholecystic fluid, a sonographic Murphy sign, and common bile duct (CBD) calibre[20,79].

Diagnostic criteria and thresholds

Several sonographic signs support the diagnosis of acute cholecystitis: Stones or sludge, a thickened gallbladder wall, pericholecystic fluid, and a positive sonographic Murphy sign, defined as maximal tenderness when the transducer presses directly over the sonographically localised gallbladder[20,79]. Representative gallstones with posterior acoustic shadowing, the gallbladder wall and the anechoic lumen are shown in Figure 4B and C.

The wall-thickness threshold is itself unsettled, and the discrepancy is instructive. Emergency-medicine POCUS literature conventionally uses a wall thicker than 3 mm, whereas the Tokyo Guidelines 2018 (TG18) list wall thickening of 4 mm or more among the imaging findings characteristic of acute cholecystitis[21]. Studies that apply a 3 mm POCUS criterion while adjudicating the final diagnosis against TG18 are therefore measuring the index test and the reference standard on different scales[33]. Both thresholds are in current use; the 3 mm number is retained in this review and in Figure 1 because it is the value used in the emergency literature from which the accuracy estimates derive, but the discordance should be recognised rather than concealed.

No single finding suffices; diagnosis depends on how features combine against pretest probability. That is the logic of TG18, which requires one local plus one systemic sign of inflammation for a suspected diagnosis and adds characteristic imaging for a definite one[21,81]. imaging is therefore not optional in that framework, and this is where POCUS enters the pathway. The 2020 World Society of Emergency Surgery guidelines name abdominal ultrasound the first-line study for suspected acute calculous cholecystitis on grounds of availability, low cost and absence of ionising radiation, while noting its limited sensitivity[22].

CBD dilatation provides supporting information, but its performance for complicated biliary disease in the ED is modest: One ED point-of-care study of 158 patients reported a sensitivity of 23.7% and specificity of 77.9%, so a normal-calibre duct does not rule out complicated disease[82]. Interpretation of CBD calibre must also account for age. Duct diameter increases by approximately 0.07 mm per year, and in the oldest asymptomatic groups mean diameter approaches 8 mm[83], while a pooled analysis of normal biliary morphometry gives a mean ultrasound-measured CBD of 3.82 mm (95%CI: 3.15-4.49) in adults against 1.32 mm (95%CI: 1.03-1.61) in children[84]. In ED patients with normal liver function tests, a CBD wider than 7 mm was found in only 1.98% (95%CI: 1.08%-3.6%), which bears on how often the measurement changes management[85].

Diagnostic accuracy

Biliary POCUS performed by EPs has high specificity for acute cholecystitis and more limited sensitivity. In a systematic review and meta-analysis of ten studies (n = 2356), the subset of emergency-physician-performed POCUS using final diagnosis as the reference standard (7 studies, n = 1772) gave a pooled sensitivity of 70.9% (95%CI: 62.3%-78.2%) and specificity of 94.4% (95%CI: 88.2%-97.5%), with an LR+ of 12.7 and LR- of 0.31[32]. A broader meta-analysis of 40 studies and 8652 patients pooling operators from every specialty reported overall sensitivity of 71% (95%CI: 69%-72%) and specificity of 85% (95%CI: 84%-86%); restricted to EPs, sensitivity was 71% (95%CI: 67%-74%) and specificity 92% (95%CI: 90%-93%)[25]. That analysis reported I² of 89.7% (95%CI: 87%-92%) and a Quality Assessment of Diagnostic Accuracy Studies 2 appraisal in which more than a quarter of studies were at high risk of bias from case-control design or non-consecutive enrolment[25]. Earlier prospective ED work showed bedside biliary ultrasound comparing favourably with radiology-performed studies and shortening time to diagnosis[79].

A 2026 prospective study of 783 ED patients adjudicated against TG18 reports markedly different numbers and should be set alongside those above rather than omitted: POCUS sensitivity was 86.6% (95%CI: 83.4-89.4) but specificity only 17.1% (95%CI: 12.5-22.5), with a negative predictive value of 37.0% in a cohort with high disease prevalence, and the authors concluded that POCUS showed poor agreement with the definitive TG18 diagnosis[33]. The contrast with a pooled specificity of 92%-94% in the meta-analyses above is stark, and it is not readily explained by chance. It reflects differences in the definition of a positive study, in the reference standard, and in disease prevalence between the contributing cohorts, which is the standardisation problem stated in the form of a single discordant result.

The direction of clinical use is nonetheless consistent across these sources: Biliary POCUS rules in disease, a negative scan does not exclude acute cholecystitis, and persistent suspicion despite a normal-appearing gallbladder calls for confirmatory imaging (radiology ultrasound, CT or hepatobiliary scintigraphy) rather than reassurance[25,32,86]. The residual error is documented: A scoping review found approximately 30% of cases may be missed and 20%-36% of initial diagnoses incorrect, attributing much of this to variability in diagnostic criteria and to reliance on clinical judgement rather than standardised frameworks such as TG18[87].

Severity grading and acute cholangitis

Ruling in cholecystitis does not by itself determine management, and an algorithm that routes every positive scan to “surgical referral“ is incomplete. TG18 grades severity in three tiers, and the grade determines the destination[21,81]: (1) Grade I, no organ dysfunction with mild local inflammation; (2) Grade II, any one of four markers of advanced local disease; and (3) Grade III, dysfunction of any one of six organ systems[21]. Each grade carries its own pathway (early laparoscopic cholecystectomy, medical treatment with gallbladder drainage and interval surgery, or urgent biliary drainage with delayed cholecystectomy) subject to Charlson comorbidity index and American Society of Anesthesiologists physical status limits[81]. The defining criteria and those limits are given in Table 2[21,81,88]. Serum total bilirubin is required to apply the grade III predictive factors[21].

Table 2 Tokyo Guidelines 2018 diagnostic and severity criteria relevant to bedside sonography.

Item	Acute cholecystitis	Acute cholangitis
Diagnostic criteria	One local sign of inflammation + one systemic sign = suspected; + characteristic imaging = definite[21]	One item of systemic inflammation + one item of cholestasis or imaging = suspected; all three = definite[88]
Systemic inflammation	N/A	Fever > 38 °C, or laboratory evidence of an inflammatory response[88]
Cholestasis	N/A	Jaundice with total bilirubin ≥ 2 mg/dL, or liver function tests > 1.5 × upper limit of normal[88]
Imaging item	Characteristic imaging findings (wall thickening ≥ 4 mm among them)[21]	Biliary dilatation, or evidence of aetiology (stricture, stone, stent)[88]
Grade I (mild)	No organ dysfunction; mild local inflammatory change[21]	Neither grade II nor grade III criteria met[88]
Grade II (moderate)	Any one of: WBC > 18000/mm³; palpable tender RUQ mass; complaints > 72 hours; marked local inflammation (gangrenous or emphysematous cholecystitis, pericholecystic or hepatic abscess, biliary peritonitis)[21]	Any two of five criteria, one of which is age ≥ 75 years[88]
Grade III (severe)	Dysfunction of any one organ system: Cardiovascular (hypotension requiring dopamine ≥ 5 μg/kg per minute, or any dose of noradrenaline); neurological (decreased consciousness); respiratory (PaO2/FiO2 < 300); renal (oliguria or creatinine > 2.0 mg/dL); hepatic (PT-INR > 1.5); haematological (platelets < 100000/mm³)[21]	Organ dysfunction as for cholecystitis[88]
Management by grade	Grade I and grade II (experienced surgeons): Early laparoscopic cholecystectomy where CCI ≤ 5 and ASA-PS ≤ 2; otherwise, medical treatment and/or gallbladder drainage with interval cholecystectomy. Grade III: Early or urgent biliary drainage with delayed cholecystectomy; upfront surgery only with favourable organ-system failure, no negative predictive factors, CCI ≤ 3, ASA-PS ≤ 2, in an advanced centre[81]	Biliary drainage, urgently in severe disease[88]
Sonographic component of grading	None: Grading uses vital signs, organ-failure laboratory values, WBC, symptom duration and comorbidity indices[21]	None[88]

Criteria are reproduced from Tokyo Guidelines 2018 as cited. No element of severity grading in either condition is sonographic; ultrasound establishes the imaging item of the diagnostic criteria, after which grading is clinical and laboratory-based. ASA-PS: American Society of Anesthesiologists Physical Status; CCI: Charlson Comorbidity Index; PT-INR: Prothrombin time-international normalised ratio; RUQ: Right upper quadrant; WBC: White blood cell count; N/A: Not applicable.

One point follows directly and belongs in any bedside algorithm: No element of TG18 severity grading is sonographic. Grading is driven by vital signs, organ-failure laboratory values, white cell count, symptom duration and comorbidity indices. POCUS establishes the imaging element of the diagnosis; it neither grades severity nor selects the operative pathway. Figure 1 therefore routes a rule-in biliary scan to expedited surgical consultation together with TG18 severity assessment, rather than to an operation.

Cholangitis shares the same acoustic window, and its omission from a biliary bedside pathway is a safety gap. TG18 requires, for a suspected diagnosis, one item of systemic inflammation plus one item of either cholestasis or imaging (biliary dilatation, or evidence of the aetiology such as a stricture, stone or stent) and all three for a definite diagnosis; the numeric criteria are given in Table 2[88]. Charcot’s triad is an unreliable gate, jaundice being present in only 60%-70% of patients[88].

Sonographic performance places cholangitis firmly in the rule-in category. TG18 cites transabdominal ultrasound sensitivity of 42% (95%CI: 28%-56%) and specificity of 96% (95%CI: 94%-98%) for a dilated CBD, and sensitivity of 38% (95%CI: 27%-49%) with specificity of 100% (95%CI: 99%-100%) for bile duct stones (high specificity, insufficient sensitivity) while still recommending it as the initial study[88]. A visualised duct stone or biliary dilatation in a clinically compatible patient substantially raises the probability of cholangitis and should prompt TG18 severity assessment and timely biliary drainage, urgently in severe disease[88]; a normal-appearing duct does not exclude it. Severity grading differs from that for cholecystitis in two bedside-relevant respects: Grade II requires any two of five criteria rather than any one, and age of 75 years or above is itself one of them[88].

Limitations and special situations

Contemporary clinical reviews frame ultrasound as the initial imaging study in this pathway[89]. Acalculous cholecystitis, mostly a disease of the critically ill, lacks the gallstone anchor: Ultrasound is less reliable, CT performs better, and diagnosis depends on sustained clinical suspicion with complementary imaging[90-92]. Structured reporting of cholecystitis probability has been proposed to reduce error[20,87,93], and operator experience and body habitus affect results, with stones in the gallbladder neck or a contracted gallbladder liable to be missed[25,79].

Reliability of the individual biliary signs is uneven, and the sonographic Murphy sign is the weakest link in the chain. Agreement between trained emergency medicine residents and blinded radiologists has been reported at kappa 0.75 for gallbladder wall thickening and kappa 0.95 for the sonographic Murphy sign in a single 200-patient ED study[94]. That study is single-centre and reports no confidence interval, and it specifies no standardised transducer pressure. A multi-reader ED study of 446 patients reported inter-reader kappa of 0.79-0.96 across sonographic image features and found that the Murphy sign did not add significantly to the ultrasound model[95]. The analgesia problem is not hypothetical: In 686 patients, opioid analgesia before scanning raised the proportion of indeterminate Murphy signs from 3.0% to 7.9% and was associated with 24 cases of cholecystitis missed on radiology interpretation[96], a question raised in the emergency literature for two decades without a standardised answer[97,98]. A tenderness-based sign altered by a routine emergency intervention, and for which no transducer-pressure convention exists, is a clear candidate for consensus specification. Other work gives kappa 0.63 (95%CI: 0.42-0.83) for wall thickening against expert review of archived ED studies, with technical error in wall measurement named as an observed failure mode[99], and a study of bedside ultrasound for gallstone disease reported kappa 0.85 (95%CI: 0.76-0.95) for the simple gallstone anchor but only 0.37 (95%CI: 0.18-0.56) for the composite imaging diagnosis of acute cholecystitis[100]. Agreement is high for the anchoring finding and falls sharply for the composite judgement built on it. Consistent with this, the TG18-adjudicated prospective study found that of the individual sonographic signs, only gallstones were significantly associated with confirmed cholecystitis[33]. No study reporting an inter-observer statistic for millimetre measurement of gallbladder wall thickness in an emergency setting was identified.

AN INTEGRATED, SYMPTOM-BASED, PRETEST-PROBABILITY-DRIVEN SCANNING ALGORITHM

Each of these four applications is well described individually, but the literature does not integrate them, leaving the emergency clinician to construct a bedside strategy for the undifferentiated patient. The algorithm proposed here ties the presenting syndrome to a prioritised sequence of POCUS applications and specifies an explicit action for each result (Figure 1), on the principle set out above: A POCUS finding updates pretest probability rather than replacing it, and the action it triggers depends on whether the application rules in, rules out, or stratifies risk.

Everything in this section and the next is the authors’ proposal rather than a synthesis of evidence, and should be read as such: The diagnostic estimates above are drawn from the literature, whereas the sequencing of applications, the action assigned to each result and the standardisation specification in Table 3 are our own constructions built on those estimates. The algorithm is a conceptual, evidence-informed framework. It has not been prospectively validated as a diagnostic pathway, no cohort data demonstrate that it reduces CT use or missed surgical disease, and it is intended to support rather than replace clinical judgement. Its component estimates carry the heterogeneity documented throughout this review, and it should be read as a structure for reasoning rather than as a validated instrument.

Table 3 The authors’ proposed specification: What a consensus process would need to fix before these applications can be standardised.

Consensus item	SBO	Gastric assessment	Free fluid beyond trauma	Biliary tract
Minimum image set	Number of quadrants required for a complete study is undefined; segmental obstruction means an incomplete survey can be falsely negative. Fix: Mandated quadrants and documented images per quadrant	Supine and RLD antral views are used inconsistently; some thresholds are supine-derived and others RLD-derived. Fix: Both positions mandatory, each documented	Number and identity of dependent recesses varies outside the FAST protocol; pelvic views may be omitted in non-trauma scanning. Fix: Defined recess set for the non-traumatic indication	Long-axis and short-axis gallbladder views, wall measurement site and duct view are not standardised. Fix: Mandated views including the point of wall measurement
Measurement convention	No published convention specifies outer-to-outer vs inner-to-inner, plane, or whether maximal or averaged loop diameter is used, a difference capable of exceeding the 2.5 mm separating the 2.5 cm and 2.75 cm thresholds. Fix: Single defined convention	Antral CSA is derived by different formulae and at different antral levels. Fix: One formula, one anatomical level, one patient position per stated threshold	Volume estimation and the distinction between trace, small, moderate and large collections are undefined; agreement falls to kappa 0.48-0.65 for this distinction. Fix: Defined volume categories with anchoring images	Wall thickness is measured at unspecified sites and on discordant scales, > 3 mm in the emergency literature against ≥ 4 mm in TG18. Fix: One threshold, one measurement site, one wall (anterior)
Definition of a positive study	Whether dilatation alone suffices, or dilatation plus abnormal peristalsis, or plus interloop fluid, varies between studies. Fix: Explicit rule for the composite	A positive study means different target volumes in different sources (empty stomach, > 1.25 mL/kg, aspiration-relevant content). Fix: State the target volume with the threshold	Any anechoic fluid vs a clinically significant volume is not distinguished. Fix: Population-specific definition, since physiological pelvic fluid differs by sex and pregnancy status	Number and combination of signs required is unstated; agreement is kappa 0.85 for gallstones but 0.37 for the composite diagnosis. Fix: Defined minimum criteria set, aligned to TG18
Reporting lexicon	Terms for peristaltic patterns and the valvulae sign are used inconsistently across the literature. Fix: Controlled vocabulary with defined terms	Perlas grades 0-2 are well defined but quantitative and qualitative results are reported interchangeably. Fix: Report grade and CSA separately, never as equivalents	Descriptors of volume and location vary. Fix: Standardised recess names and volume categories	Sonographic Murphy sign has no standardised elicitation, defined transducer pressure or accounting for prior analgesia. Fix: Defined elicitation protocol, or removal of the sign from criteria
Reference standard	CT, surgical findings and final clinical diagnosis are used interchangeably across pooled studies. Fix: Single specified standard, or pre-specified stratification by standard	Aspirated volume, endoscopic volume and model-predicted volume are used as equivalents. Fix: One standard; and outcome studies linking measurement to aspiration rather than to volume	Trauma-derived standards (CT, laparotomy) are applied to non-traumatic populations. Fix: Indication-specific reference standard	Final diagnosis, pathology and TG18 give materially different specificity (17.1% vs 85%-94%). Fix: TG18 as the uniform standard, with the index test measured on the same scale
Training volume for credentialing	No application-specific minimum; the attending-resident sensitivity gap is 14.7 percentage points. Fix: Application-specific case minimum with competency assessment	No emergency-setting training standard exists; the available reliability study used three raters and healthy volunteers. Fix: Defined supervised case volume	FAST credentialing exists but is trauma-specific and does not cover non-traumatic indications. Fix: Extension of credentialing to the non-traumatic indication	Credentialing exists in most curricula but is not tied to demonstrated agreement on wall thickness or the Murphy sign. Fix: Competency demonstrated per sign, not per application
Quality assurance	No archived-image review requirement; loop measurement cannot be audited if images are not stored. Fix: Mandatory archiving of the measured loop	No QA mechanism described in any cited source. Fix: Archived antral images with periodic measurement audit	Where archiving exists it is trauma-oriented. Fix: QA review of non-traumatic studies, with attention to small collections	QA review is described in stewardship literature but is not sign-specific. Fix: Periodic audit of wall measurement and of Murphy-sign documentation
Priority reliability studies required	Inter-observer agreement for loop-diameter measurement: No published kappa or ICC was identified for this measurement, on which the entire branch turns	Inter-observer agreement for antral CSA in an emergency population; and any study linking measurement to aspiration rather than to volume	Inter-observer agreement for detecting a small collection in a non-traumatic emergency population, none identified	Inter-observer agreement for millimetre wall-thickness measurement in an emergency setting (none identified) and a second, multicentre study of the sonographic Murphy sign

Each cell states the current position in the literature followed by the specific item a consensus process would need to fix. Statements that no data were identified refer to the searches described in the METHODOLOGY section. CSA: Cross-sectional area; CT: Computed tomography; FAST: Focused Assessment with Sonography for Trauma; ICC: Intraclass correlation coefficient; QA: Quality assurance; RLD: Right lateral decubitus; TG18: Tokyo Guidelines 2018; SBO: Small-bowel obstruction.

Entry conditions and the competency precondition

Two conditions govern entry. The first is the instability and red-flag check: A patient with shock, peritonitis, or a picture suggesting ruptured ectopic pregnancy, ruptured AAA, bowel ischaemia or perforation is managed on a resuscitation-first track, with immediate haemodynamic support, expedited surgical consultation and guideline-concordant assessment proceeding in parallel; bedside scanning is never allowed to delay them. In this setting POCUS is confirmatory and time-sparing rather than gating[34,101].

The second condition is operator competency, and it is a precondition rather than an assumption. Every accuracy estimate in this review was generated by operators who had met some training standard, and the attending-versus-resident gap of 14.7 percentage points in SBO sensitivity[10] indicates what happens when that condition is not met. A clinician who has not achieved local credentialing for an application should treat a negative scan as uninformative and proceed as though the examination had not been performed. Current emergency-medicine and multispecialty curricula define core abdominal applications, milestones and minimum experience for credentialing[102-104].

Symptom-based entry points

The point of entry depends on the presenting symptom cluster, which sets the pretest probability and determines the first scan. For biliary-type pain in the right upper quadrant or epigastrium, evaluation starts with biliary POCUS. Gallstones plus a positive sonographic Murphy sign and wall thickening rule in cholecystitis and prompt expedited surgical consultation together with TG18 severity assessment, which determines the destination[21,81]. Where fever, jaundice or a dilated duct accompany the presentation, cholangitis enters the differential and the pathway shifts toward urgent biliary drainage[88]. A negative scan does not end the evaluation where suspicion remains high, because sensitivity is only moderate[22,32].

Crampy, distension-heavy pain with vomiting or obstipation begins with SBO POCUS. Dilated fluid-filled loops beyond the bedside threshold, combined with abnormal peristalsis, rule in obstruction and prompt both surgical consultation and CT; free interloop fluid should raise concern for complicated obstruction[10-12]. For lower-abdominal or pelvic pain in a patient of reproductive potential with a positive pregnancy test, priority shifts to free-fluid and pelvic POCUS: Free fluid in Morison’s pouch is a strong and decisive finding mandating immediate gynaecological and surgical involvement, while an identified intrauterine pregnancy substantially lowers the likelihood of ectopic pregnancy[31,66,68]. A specific safety rule applies here, because it is the mechanism by which ruptured ectopic pregnancies are missed: Emergency-physician ultrasound excludes ectopic pregnancy safely by demonstrating an intrauterine pregnancy, with pooled sensitivity 99.3% (95%CI: 96.6%-100%) and negative predictive value 99.96%[105], and not by demonstrating the absence of free fluid. A scan in which no intrauterine pregnancy is seen is not a reassuring scan.

For flank pain or pain with haemodynamic instability, the aorta and kidneys are examined. Running in parallel with all entry points, any patient requiring urgent sedation or airway management is evaluated with gastric POCUS, which stratifies aspiration risk and informs airway technique and timing regardless of the diagnostic path being followed[14,27].

Design rules that keep the algorithm safe

Four rules govern the pathway. The first is the instability and red-flag check described above. The second addresses the failure mode common to every application in this set: Each rules in far better than it rules out, so a negative entry scan must not close the evaluation in a patient who began with high pretest probability; the algorithm instead redirects toward the next most likely application or to confirmatory imaging.

The third rule handles the examination that fails, which the previous version of this algorithm did not represent. Body habitus and bowel gas are the commonest reasons a study is non-diagnostic, and a non-diagnostic study is not a negative study. Figure 1 therefore carries an explicit third exit from every scan node: (1) Adequate and positive; (2) Adequate and negative; and (3) Non-diagnostic. A non-diagnostic examination returns the clinician to the pretest probability that prompted the scan and mandates an alternative route: (1) Confirmatory imaging; (2) Specialist consultation; and (3) A repeat scan after an interval. Serial scanning is an explicit option where the clinical picture is evolving and the initial study was equivocal or non-diagnostic, since the abdominal findings of obstruction and of biliary inflammation both develop over hours.

The fourth rule applies when the presenting syndrome is undifferentiated: A brief structured multi-region survey through the biliary tract, bowel and dependent recesses is preferred to a single-organ scan, because a protocolised survey yields more findings[23,24]. The algorithm also incorporates the principal differentials sharing POCUS windows, so findings pointing away from the entry diagnosis are captured rather than missed.

Secondary redirect applications: Key differentials

These applications (appendicitis, renal colic and hydronephrosis, and AAA) are not core topics of this review but secondary redirect points: When the four principal scans are negative or non-diagnostic, they direct the clinician to the next highest-yield step. In suspected appendicitis, a meta-analysis of 21 studies reported pooled sensitivity of 91% (95%CI: 83%-96%) and specificity of 97% (95%CI: 91%-99%) for POCUS performed by non-radiologist physicians overall, while the subset restricted to emergency-physician-performed scans gave sensitivity of 80% and specificity of 92%; heterogeneity was extreme, with I² of 99% (95%CI: 99%-100%)[35]. A 2026 paediatric meta-analysis of eight studies and 993 patients reported pooled sensitivity of 85.6% (95%CI: 68.9%-94.1%) and specificity of 90.2%, with sensitivity ranging from 53% to 100% between studies[36]; a multicentre paediatric study reported comparable performance with kappa of 0.87 (95%CI: 0.70-1.00) against radiology-performed ultrasound[106]. A positive scan supports the diagnosis; a negative scan does not exclude it[107,108].

Flank pain with suspected renal colic follows a different pattern: POCUS detection of hydronephrosis has a pooled sensitivity of approximately 70.2% (95%CI: 67.1%-73.2%), sufficient for risk stratification but not to exclude the diagnosis[37,109]. EPs identify and grade hydronephrosis with acceptable agreement against radiology interpretation[110].

Suspected AAA requires a safety-critical distinction. POCUS is highly accurate for identifying an aneurysmal aorta, a meta-analysis of nine studies reported pooled sensitivity of 98.33% and specificity of 99.84%[34], but those numbers describe detection of aneurysm, not of rupture. POCUS is unreliable for identifying rupture itself, for which CT angiography is the reference standard[101]. A completely visualised normal-calibre abdominal aorta substantially lowers the probability of ruptured AAA as the cause of shock, whereas an aneurysmal aorta in a shocked patient should prompt immediate vascular involvement without waiting for sonographic demonstration of rupture[34,101]. Supporting studies extend bedside aortic evaluation[111,112], as does a systematic review of thoracic aortic assessment[112]; one of the supporting studies is, however, a single-centre convenience sample dating from 2000[113].

Records were identified by searching PubMed (n = 639), Scopus (n = 972) and Web of Science (n = 676) with the concept blocks reported in Supplementary Table 1, giving 2287 records in total. Cross-database deduplication, performed by matching digital object identifiers and then normalised titles, removed 1121 duplicates and left 1166 unique records. Title and abstract screening against the review scope excluded 952 records, 214 full texts were assessed for eligibility, and 63 were excluded because they had been superseded by a later synthesis or reported no extractable numeric estimate. A purposive selection step then retained the highest available level of evidence for each application, and 151 sources are cited. Sources identified outside the database search from the reference lists of retrieved syntheses, current society guidelines, and two ongoing studies located in ClinicalTrials.gov, NCT06803628 and NCT07580456, were added at this stage. Because selection was purposive rather than exhaustive, this diagram documents what was retrieved and how it was reduced; it is not a PRISMA flow diagram and is not offered as evidence of complete capture of the literature.

STANDARDISATION AS A PRECONDITION FOR IMPLEMENTATION

The estimates assembled in this review support the algorithm above. They also motivate a second argument, which is less comfortable and in our view more important, and which we advance as an interpretation of those estimates rather than as a finding derived from them: What limits these four applications may be less their accuracy than the absence of a standardised procedure to which any accuracy number could belong.

The evidence for that claim is internal to this review. For SBO, pooled sensitivity is 92.4% at study level and 83.0% in IPD analysis, and the IPD specificity estimate is so imprecise (95%CI: 55.3%-99.3%) that it constrains the parameter very little, while a third meta-analysis pools specificity at 0.93 and a fourth at 0.80 with I² of 88%[9,10,13,30]. Attending physicians outperform residents by 14.7 percentage points of sensitivity for the same nominal examination[10]. Cholecystitis specificity is 85% across all operators and 92% among EPs in one meta-analysis with I² of 89.7%[25], while a prospective cohort adjudicated against TG18 reports 17.1%[33]. Gastric assessment offers at least nine non-interchangeable antral thresholds spanning a 2.8-fold range[26-29], with further position-specific and population-specific values in pregnancy, dyspepsia and older adults[58,59]. Pooled free-fluid sensitivity outside trauma cannot be stated at all, and the nearest surrogate is 0.596 (95%CI: 0.345-0.822) with GRADE certainty rated very low[19].

Heterogeneity of this magnitude is consistent with pooling studies that did not perform the same examination, and we advance that as the most actionable explanation rather than the only one. Variation in the clinical spectrum of enrolled patients, in the reference standard against which POCUS was adjudicated, and in study design will each contribute to the dispersion, and the available data do not permit their contributions to be partitioned. The claim made here is narrower and testable: Procedural variation is the component that is remediable by agreement rather than by further pooling, and if it were removed the residual heterogeneity should fall. That is a prediction, and it is the prediction a standardisation programme would test. For SBO there is no agreement on the number of quadrants surveyed, the plane of measurement, or whether a loop is measured outer-wall-to-outer-wall or inner-wall-to-inner-wall, a distinction capable of exceeding the 3 mm that separates the two competing thresholds. For biliary disease, the definition of a positive study varies between authors, the wall-thickness criterion differs between the emergency literature (3 mm) and TG18 (4 mm)[21,33], and the sonographic Murphy sign (which carries substantial weight in every published biliary pathway, including Figure 1) has no standardised elicitation, no defined transducer pressure and no accounting for analgesia given before the scan. For gastric assessment, position, population and target volume all vary between the studies from which thresholds derive.

Reliability data, where they exist at all, point the same way: Agreement is acceptable for binary detection tasks and falls sharply for graded or composite judgements: Kappa 0.525 to 0.865 for free-fluid detection on trauma FAST[76,77], falling to 0.48-0.65 when a small collection must be distinguished from a moderate one[78]. More striking is what is missing. Despite an explicit search, we identified no study reporting kappa, weighted kappa, an intraclass correlation coefficient or percent agreement for the measurement of small-bowel loop diameter, the single measurement on which the SBO branch of every published algorithm turns. Nor did we identify any inter-observer study of millimetre measurement of gallbladder wall thickness in an emergency setting (the nearest evidence, from radiology, found that two observers measuring prospectively in 50 patients differed significantly[114]) or any reliability datum for detecting a small free-fluid collection in a non-traumatic emergency population. Where reliability data have recently appeared they sharpen rather than resolve the problem. Qualitative antral classification transfers well between observers (kappa 0.79 in an ED cohort)[64], but formal testing of antral CSA measurement across expertise levels found inter-rater reliability only poor to moderate[65], that is, the quantitative measurement underpinning every published threshold is the part that does not transfer. And the sonographic Murphy sign turns out to be altered by an intervention as routine as opioid analgesia[96]. Accuracy without reproducibility cannot support an algorithm that routes patients toward an operating theatre, and the same problem recurs wherever a millimetre-level threshold is derived from a single emergency cohort. A prospective ED study of optic nerve sheath diameter in hypertensive crisis (112 patients) identified an optimal cut-off of 5.8 mm with sensitivity of 68%, specificity of 93.1%, LR+ of 9.86 and LR- of 0.34, a threshold that rules in but cannot rule out, reported to a precision of 0.1 mm from one centre and without an accompanying inter-observer statistic[115]. The bowel-diameter and gallbladder-wall thresholds in this review share that structure exactly.

FAST is the counter-example. Its pooled sensitivity of 0.74 (95%CI: 0.65-0.81) for thoracoabdominal injury[18] is lower than that of several applications reviewed here, yet it scaled into universal practice because a consensus process first fixed the question it answers, the windows it uses and the minimum image set constituting a complete study. Curricula, case-number requirements, credentialing and quality assurance could then attach to that fixed procedure, and estimates from different centres became commensurable. The four applications reviewed here have the technology without the scaffolding.

Table 3 sets out what a consensus process would need to fix for each application: (1) The minimum image set; (2) The measurement convention; (3) The definition of a positive study; (4) The reporting lexicon; (5) The reference standard; (6) The training volume; and (7) The quality-assurance mechanism. It is deliberately concrete, because the general call for standardisation with which such reviews usually close has not produced one.

A staged research agenda

The sequence matters as much as the content, because each stage is uninterpretable without the one before it. First, consensus: A multi-society process fixing, per application, the items in Table 3. The machinery for this already exists in emergency medicine. A national society clinical policy on tranexamic acid was produced by exactly this route, a research committee combining systematic literature review with the GRADE approach and structured expert consensus to answer pre-specified clinical questions on which the trial evidence was contested and heterogeneous[116]. That is the same procedural problem the four applications here present, and the same instrument would serve; what is absent is not the method but its application to these examinations. Second, reliability: Inter-observer studies of the specific measurements (loop diameter, gallbladder wall thickness, sonographic Murphy sign, small free-fluid collections) conducted in emergency populations by emergency operators, reporting kappa or intraclass correlation with confidence intervals. Third, accuracy: Prospective multicentre diagnostic-accuracy studies using the agreed protocol and a uniform reference standard, powered and pre-specified to report performance stratified by operator training level rather than pooled across it. Fourth, outcomes: Trials asking whether a standardised pathway changes time to diagnosis, CT utilisation, missed surgical disease and patient-centred endpoints. The algorithm proposed in this review is a candidate for that fourth stage, and it should be validated rather than adopted. The design is already established for emergency decision rules: Prospective consecutive enrolment, management left to the treating physician so that performance is measured against real practice, structured follow-up to capture outcomes missed at the index visit, and sensitivity reported with confidence intervals against a clinically important outcome, the approach taken in the external validation of a falls decision rule in 800 older adults, in which the rule reached 97.9% sensitivity for clinically important intracranial bleeding[117].

Two registered studies indicate that the second and third stages are beginning, while also showing how far the fourth remains. A multicentre prospective diagnostic cohort in France (NCT06803628, recruiting since March 2025, planned enrolment 667, estimated primary completion March 2027) is testing precisely the rule-out question this review identifies as the weak point of the SBO branch: The negative predictive value of emergency-physician POCUS in patients with low or moderate clinical probability, with the explicit aim of reducing unnecessary CT. A proof-of-concept study in Canada (NCT07580456, recruiting since May 2026, 30 healthy volunteers, estimated primary completion December 2027) is training and testing an artificial-intelligence model to distinguish an empty stomach from clear fluid and from solid content on gastric POCUS. We identified no registered study evaluating an integrated multi-application abdominal POCUS algorithm of the kind proposed here, none addressing the reliability of bowel-loop diameter or gallbladder-wall measurement, and no registered trial with a patient-centred or CT-utilisation endpoint for such a pathway; the searches supporting these statements are given in Supplementary Table 1. Running the stages out of order produces the present literature. Diagnostic-accuracy studies conducted before the procedure was fixed generate estimates that cannot be pooled meaningfully, and outcome studies built on unfixed procedures would inherit the same problem at greater cost.

SPECIAL POPULATIONS, AI AND LOW-RESOURCE SETTINGS

Children, older adults and pregnancy

Most diagnostic estimates in this review derive from adult ED populations, and thresholds do not transfer automatically across age or pregnancy. Where population-specific data exist they are summarised here; where they do not, the gap is stated.

In children, the strongest data concern intussusception and appendicitis rather than the four core applications. Reliability evidence for paediatric abdominal POCUS is sparser still: We identified no study reporting kappa or an intraclass correlation coefficient for bowel-loop diameter, gallbladder-wall thickness or small-volume free fluid in a paediatric emergency population, so the age-adjusted thresholds below rest on accuracy data alone. POCUS for intussusception has a pooled sensitivity of 95.1% (95%CI: 90.3%-97.2%) and specificity of 98.1% (95%CI: 95.8%-99.2%)[118], and a multicentre study found POCUS non-inferior to radiology-performed ultrasound, with sensitivity 96.6% (95%CI: 87.2%-99.1%) and specificity 98.0% (95%CI: 94.7%-99.2%)[119]; expert and novice interpretation nonetheless differ materially[120]. Paediatric gastric assessment has its own thresholds and its own models: An antral CSA cut-off of 2.40 cm² in RLD identified an empty stomach in infants with 100% sensitivity and 68.6% specificity[121], a paediatric prediction model relates RLD antral CSA to fluid volume with a mean bias of 0.01 mL/kg[122], and qualitative paediatric gastric assessment gives sensitivity of 75% (95%CI: 58%-88%) and specificity of 85% (95%CI: 73%-93%)[123]. European paediatric fasting guidance incorporates gastric ultrasound in this context[61]. Ultrasound for intra-abdominal free air in children has sensitivity of 71.4% against specificity of 100%[124], and the Cochrane trauma data show children fare worse than adults: Pooled sensitivity 0.63 (95%CI: 0.46-0.77) vs 0.78 (95%CI: 0.69-0.84), corresponding to 118 missed children per 1000 scanned[18]. No paediatric-specific dilated-loop threshold for SBO, and no paediatric-derived gallbladder wall-thickness threshold, were identified.

In older adults, two adjustments are supported. The optimal SBO bowel-diameter threshold rises to 2.95 cm at age 65 or above[11], though this age-stratified value has not itself been prospectively validated. Antral CSA cut-offs also differ, with elderly-specific values of 6.92 cm² supine and 10.65 cm² in RLD, and the authors report that applying a young-adult cut-off in older patients reduced specificity, positive predictive value and accuracy[59]. CBD calibre increases with age at approximately 0.07 mm per year[83], so a duct of 7-8 mm in an asymptomatic older patient may be physiological. Age is also embedded in the TG18 cholangitis severity grading, where 75 years or above is one of the Grade II criteria[88].

In pregnancy, gastric thresholds are position-specific: 6.08 cm² (608 mm²) semi-recumbent and 9.60 cm² (960 mm²) in RLD[29], with a 2026 review giving 6.08 cm² semi-recumbent, 7.19 cm² (719 mm²) in RLD and 9.60 cm² for a full stomach[125]. POCUS for appendicitis in pregnancy has sensitivity of 66.7% (95%CI: 47.1%-82.7%) and specificity of 96.8%[126]. For suspected ectopic pregnancy, emergency-physician ultrasound achieves a pooled sensitivity of 99.3% (95%CI: 96.6%-100%) with a negative predictive value of 99.96% when it demonstrates an intrauterine pregnancy[105], and a 2026 meta-analysis reports pooled sensitivity of 86.3% (95%CI: 75.6%-92.8%, I² = 89.8%) for identifying intrauterine pregnancy[127]. Set against this, 72% of interstitial ectopic pregnancies were initially misdiagnosed in one series[128]. No pregnancy-specific gallbladder wall-thickness threshold, and no data on the prevalence or volume of physiological free fluid in pregnancy, were identified, meaning the free-fluid branch of the algorithm is applied in pregnancy without a population-specific normal range.

AI and the operator-experience gap

Operator dependence is the dominant modifier throughout this review, and AI-assisted acquisition and interpretation is the most plausible near-term route to narrowing it. The evidence that it can do so is real but sits almost entirely outside abdominal POCUS. Outside the abdomen, AI assistance has raised non-expert interpretation accuracy in lung ultrasound from 68.1% (95%CI: 57.9%-78.2%) to 93.4% in prospective real-time testing[129], and AI-guided acquisition has enabled non-experts and untrained nurses to obtain diagnostic-quality studies in 98% or more of cases[130,131]. Most directly relevant to this review, deep-learning-guided novices achieved adequate scan quality for aortic assessment in 87.5% of cases against 91.3% for physicians (P = 0.310)[132]. In the one biliary application identified, an AI tool raised the area under the curve of both junior (0.773-0.835) and senior radiologists[133]. For gastric assessment, machine-learning models predicting gastric volume outperformed the Perlas model[134], and automated ascites detection has been demonstrated[135].

The limits deserve equal plainness. We identified no AI system for POCUS diagnosis of adult SBO, none for cholecystitis validated in an emergency setting, none for cholangitis, and no randomised trial comparing AI-assisted against unassisted abdominal POCUS for any of the four applications. The free-fluid AI evidence is retrospective and image-based[72]. A 2026 review of AI-enhanced POCUS in emergency care characterises the field as one that may lower the expertise threshold, and notes that models not trained on point-of-care images lose performance when applied to them[136]; a scoping review of AI-POCUS in low-resource settings found the evidence concentrated in pulmonology, obstetrics and emergency medicine, with abdominal applications sparse[137]. AI is a plausible answer to the operator-experience gap and is not yet an evidenced one for the abdomen. A model trained on unstandardised images also inherits the variability of its training data.

CT-limited and low-resource settings

Where CT is unavailable, POCUS carries more diagnostic weight and its limitations become more consequential, the setting in which the rule-in/rule-out asymmetry matters most. A systematic review of POCUS for SBO addressing implementation barriers in resource-limited settings reported sensitivity from 85.0% to 100% and specificity from 54.0% to 100%[49]. In a South African tertiary ED, POCUS utilisation was 5.2% of attendances and 73.2% of examinations were performed by the five credentialed providers[138], indicating that credentialing capacity rather than equipment constrained use. In rural New Zealand, FAST and AAA scans had sensitivities of 75% and 100% and specificities of 100% and 93%, and altered care significantly in 17% and 31% of cases[139]. A scoping review of POCUS training in low-resource settings found curricula dominated by lung (54%), cardiac (53%) and obstetric (43%) applications[140], abdominal protocols of the kind this review addresses are under-taught in exactly the settings where they would carry the most weight. Curricula adapted to such settings have nonetheless been implemented and evaluated[141].

Cost-effectiveness cannot yet be claimed for these applications. A systematic review of the cost-effectiveness of emergency ultrasound found the evidence concentrated in paediatric appendicitis (11 studies) and abdominal trauma (6 of 35 studies)[142]. We identified no cost-effectiveness analysis of POCUS for SBO, gastric assessment, non-traumatic free fluid or acute cholecystitis, none conducted in a low- or middle-income setting, and no study quantifying CT examinations avoided by an integrated symptom-based POCUS algorithm. The economic case for the pathway proposed here is therefore untested. The settings that depend most on POCUS because cross-sectional imaging is unavailable are also the least represented in the literature that would establish its value there, so this gap is unlikely to close without studies designed and funded for those settings specifically.

PITFALLS, RISK OF OVER-RELIANCE, AND TRAINING

Using abdominal POCUS safely requires knowing where it breaks down. Body habitus and bowel gas erode the acoustic window across all four applications and are the commonest reason a study is non-diagnostic; obesity in particular reduces sensitivity for free fluid, gallstones and dilated bowel[10,25,69]. Interpreting the literature is difficult in its own right, because reference standards vary: Pooled estimates for SBO, cholecystitis and appendicitis variously use surgical findings, CT, or final clinical diagnosis, and the resulting heterogeneity is quantified in Table 1 and Figure 2[9-11,18,19,25-27,30-37].

The harm of over-trusting a negative scan

The clinical consequence of the rule-in/rule-out asymmetry is best expressed as counts rather than adjectives. In blunt trauma, at a median prevalence of 28%, POCUS would miss 73 injured adults per 1000 scanned, and 118 injured children per 1000[18]. Among patients with non-compressible torso haemorrhage, those in whom a false negative is least survivable, FAST was falsely negative in 49% of cases with abdominal or pelvic bleeding[74].

Two failure modes deserve specific attention because they recur. The first is the confidently wrong scan rather than the empty one. In a prospective study of ultrasound for complicated (including perforated) diverticulitis, sensitivity was 55% (95%CI: 32%-76%), and 8 of the 10 missed cases were not reported as negative but as simple diverticulitis[143]. The danger is not only the scan that shows nothing; it is the scan that shows something reassuring and thereby downgrades the perceived severity. The same caution applies to any bedside measurement adopted because it is easy to obtain rather than because it has been shown to discriminate. In a prospective study of 100 intubated cardiac arrest patients, the perfusion index (non-invasive, instantly displayed and inexpensive to acquire) did not differ between patients who achieved return of spontaneous circulation and those who did not at any minute of resuscitation, and showed no correlation with end-tidal carbon dioxide, which did discriminate from the fifth minute onward[144]. Ease of acquisition and discriminative value are independent properties, and the standard of proof for an adjunctive bedside measurement is the same as for a sonographic sign.

The second is the ectopic pregnancy that is missed because absence of free fluid was read as reassurance. In a published series of near-miss ruptured ectopic pregnancies, the formal radiology study was reported as showing no intrapelvic abnormality or as inconclusive, and independent emergency-physician review of the same images raised the concern that led to intervention[145]. The lesson is not that POCUS is unreliable, nor that formal imaging is; it is that no single negative sonographic study, whoever performs it, should override a concerning clinical picture. Applied to the algorithm, the rule is the one stated above: Ectopic pregnancy is excluded by demonstrating an intrauterine pregnancy, not by failing to demonstrate free fluid.

Missed diagnosis in the acute abdomen carries measurable downstream cost. Potentially missed appendicitis has been documented in 6.0% of adults and 4.4% of children in a large administrative analysis[146], missed diagnoses are well described among ED patients with acute abdominal pain[147], and approximately 30% of acute cholecystitis cases may be missed on initial assessment[87]. These numbers are not attributable to POCUS specifically, but they define the diagnostic environment in which a negative bedside scan is interpreted.

Training, competency and quality assurance

Among the variables shaping POCUS performance, operator experience is the one a department can address directly. Accuracy improves with training and case volume[10]. Structured curricula, competency assessment and image-quality review each improve performance measurably, and shared graduate-level curricula paired with web-based training improve both acquisition and interpretation[102-104,148].

Sustaining quality across a department requires more than one skilled sonographer. It also requires that the finding reach the rest of the team. A bedside diagnosis changes outcome only when it is transmitted into a coordinated response. In the time-critical presentations this pathway serves, the steps that determine outcome (recognition, resuscitation, transfusion preparation and specialty escalation) run in parallel across nursing and physician roles rather than sequentially after the diagnosis is made. An algorithm that stops at the image is incomplete. POCUS stewardship (quality-assurance image review, structured documentation and integration into the electronic record) is increasingly treated as necessary at scale, and implementation studies describe workable models for archiving, review and workflow integration[149-151]. A fixed minimum image set is enforceable only where images are archived and reviewed, which makes these mechanisms the practical route to any agreed protocol.

LIMITATIONS

This is a narrative rather than a systematic review, and its synthesis remains exposed to selection bias in which studies were included, notwithstanding the transparent search strategy described above. We did not perform a formal risk-of-bias assessment or a GRADE certainty rating of our own; where such appraisals are reported in Table 1[9-11,18,19,25-27,30-37], they are those of the source meta-analyses, and for several pooled estimates the source did not report them or the full text was not retrievable. Specifically, we could not recover a numeric heterogeneity statistic for the two SBO estimates on which the manuscript relies most[9,10], nor for the emergency-physician cholecystitis meta-analysis[32] (whose study and patient counts we were able to recover), the hydronephrosis meta-analysis[37] or the AAA meta-analysis[34]; GRADE certainty was reported in only one source in the pool[19]. Values that could not be verified against retrievable source text are identified as such in Table 1 rather than omitted[9-11,18,19,25-27,30-37].

The pooled estimates themselves draw on meta-analyses that used heterogeneous reference standards, operators of varying experience and thresholds that differed between studies; the numbers here should be treated as representative rather than definitive, and the standardisation argument above is in part a consequence of that fact. The integrated algorithm combines published evidence with explicit clinical reasoning but has not been prospectively validated as a decision instrument; no data demonstrate that it reduces CT use, shortens time to diagnosis or lowers the rate of missed surgical disease, and its economic impact is untested. Most primary studies were conducted in resource-rich EDs with confirmatory imaging readily available, which limits generalisability to settings where POCUS is the only available tool. Because the search was restricted to English-language articles, relevant evidence in other languages may have been missed. Finally, the reliability gaps documented above are conclusions about the retrievable published literature; they indicate the absence of published evidence rather than proof that no such data exist.

CONCLUSION

POCUS has matured into an evidence-backed adjunct for the acute abdomen, and its yield depends on the question that prompted it. For SBO and acute cholecystitis, it functions as a strong rule-in test while remaining unable to define aetiology, transition point, ischaemia, perforation or disease extent, for which CT is still required. Free fluid outside trauma rules in but does not rule out. Gastric assessment stratifies aspiration risk against a surrogate endpoint rather than diagnosing disease. This review integrates the four into a symptom-based, pretest-probability-driven algorithm that sequences each examination and ties its result to a concrete action, and that algorithm requires prospective validation before it is adopted (Table 4)[7-11,18,19,21,25-27,30-32,34-37,81,88].

Table 4 Evidence-versus-proposal matrix for each action node of the integrated algorithm (Figure 1): What the node rests on, and whether it is derived from evidence or proposed by the authors.

Algorithm action node	Evidence type	Evidence supporting the node	Status	Ref.
Step 1: Instability/red-flag check precedes any scan	Expert recommendation	Consensus statements on emergency POCUS scope; no comparative trial of scan-first vs resuscitate-first	Authors’ proposal	Kameda and Taniguchi[7], Oto et al[8]
Step 2: Operator competency precondition	Evidence-informed	Operator-level subgroup estimates differ by application and do not all run the same way: For appendicitis the EP subgroup is lower than the pooled estimate on both axes (sensitivity 80% vs 91%, specificity 92% vs 97%), whereas for cholecystitis the EP subgroup matches the pooled sensitivity (71% vs 71%) and exceeds the pooled specificity (92% vs 85%). Competency is therefore required as a precondition because performance is operator-dependent in an application-specific direction, not because EP performance is uniformly lower	Evidence-based (subgroup estimates), authors’ proposal (credentialling requirement)	Huang et al[25], Matthew Fields et al[35]
Step 3: Symptom cluster sets pretest probability and orders the scans	Authors’ construction	No study compares symptom-ordered POCUS sequencing against a fixed protocol	Authors’ proposal	N/A
Biliary branch: Gallstones + sonographic Murphy sign + wall > 3 mm as the rule-in trigger	Evidence-based	Pooled specificity 94.4% (88.2%-97.5%) for EP-performed POCUS; TG18 uses wall ≥ 4 mm	Evidence-based	Yokoe et al[21], Wilson et al[32]
Biliary branch: Negative scan does not exclude cholecystitis	Evidence-based	Pooled EP sensitivity 70.9% (62.3%-78.2%); LR- accordingly poor	Evidence-based	Wilson et al[32]
Biliary branch; fever + jaundice + dilated duct triggers the cholangitis pathway	Guideline-based	TG18 diagnostic criteria (systemic inflammation + cholestasis + imaging) and severity grading	Evidence-based (guideline)	Okamoto et al[81], Kiriyama et al[88]
SBO branch: Dilated loops > 2.5 cm as the rule-in trigger	Evidence-based	Pooled sensitivity 92.4% (89.0%-94.7%), specificity 96.6% (88.4%-99.1%), LR+ 27.5	Evidence-based	Gottlieb et al[9]
SBO branch: 2.75 cm as a candidate refined threshold	Authors’ construction	Proposed in a bedside-threshold analysis; no reliability data exist for bowel-loop diameter at this precision	Authors’ proposal: Explicitly unvalidated	Shokoohi et al[11]
SBO branch, CT still required for transition point, aetiology, ischaemia, perforation	Evidence-based	Consistent statement across the SBO syntheses; POCUS not evaluated for these endpoints	Evidence-based	Gottlieb et al[9], Shokoohi et al[10], Motavaselian et al[30]
Free-fluid branch, free fluid in Morison’s pouch or pelvis as the rule-in trigger	Evidence-based	Ectopic pregnancy cohort; prehospital FAST pooled specificity 0.970 (0.953-0.983)	Evidence-based	Lin et al[19], Moore et al[31]
Free-fluid branch: Ectopic pregnancy excluded only by seeing an intrauterine pregnancy	Evidence-based	Pooled FAST sensitivity 0.74 (0.65-0.81) in blunt trauma; absent free fluid carries no exclusionary value for ectopic pregnancy	Evidence-based	Stengel et al[18], Moore et al[31]
Aorta/renal branch, aortic diameter ≥ 3 cm; POCUS does not identify rupture	Evidence-based	Aneurysm detection sensitivity 98.3%, specificity 99.8%; no pooled estimate for rupture detection	Evidence-based	Shaban et al[34]
Gastric parallel branch: Perlas grade and antral CSA stratify aspiration risk	Evidence-based	IPD meta-analysis: Grade 0/1 gives 98% probability of being below the risk threshold; ED intubation cohort sensitivity 100%, specificity 87%	Evidence-based (against a surrogate endpoint)	Perlas et al[26], Asokan et al[27]
Gastric parallel branch; thresholds are position- and population-specific, not interchangeable	Evidence-based	Distinct CSA thresholds by position and population across the source studies	Evidence-based	Perlas et al[26]
Third exit; non-diagnostic study treated separately from a negative study	Authors’ construction	No source study reports a distinct action pathway for non-diagnostic examinations	Authors’ proposal, a principal original contribution	N/A
Serial scanning after an interval where the picture is evolving	Expert recommendation	No comparative evidence for serial abdominal POCUS in this setting	Authors’ proposal	N/A
Secondary redirect applications when the entry scan is negative or non-diagnostic	Evidence-based (component estimates), authors’ construction (the redirect rule)	Appendicitis EP sensitivity 80%, specificity 92%; hydronephrosis sensitivity 70.2%; AAA detection 98.3%	Mixed, estimates evidence-based, sequencing proposed	Shaban et al[34], Matthew Fields et al[35], Miller et al[36], Wong et al[37]
Governing principle, every application rules in better than it rules out	Evidence-based	Direct consequence of the sensitivity/specificity asymmetry documented in Table 1	Evidence-based	Gottlieb et al[9], Shokoohi et al[10], Stengel et al[18], Lin et al[19], Huang et al[25], Wilson et al[32]

N/A in the reference column indicates that no source study addresses the node, which is therefore an authors’ proposal rather than an evidence-derived step. It is provided so that a reader can separate the evidence synthesised in this review from the pathway the authors construct from it; the algorithm as a whole has not been prospectively validated. AAA: Abdominal aortic aneurysm; CSA: Cross-sectional area; CT: Computed tomography; ED: Emergency department; EP: Emergency physician; IPD: Individual-patient-data; LR+: Positive likelihood ratio; LR-: Negative likelihood ratio; SBO: Small-bowel obstruction; TG18: Tokyo Guidelines 2018; POCUS: Point-of-care ultrasound; N/A: Not applicable; FAST: Focused Assessment with Sonography for Trauma.

The wider conclusion is that the evidence assembled here makes a stronger case about heterogeneity than about accuracy. Pooled sensitivities and specificities for these applications are respectable; their dispersion across operators, protocols, thresholds and reference standards is larger than the differences on which bedside decisions turn, and the reproducibility data that would anchor them are in several places simply absent. On the evidence assembled here, standardisation is therefore best treated not as a closing aspiration but as the high-priority prerequisite for scalable and reliable implementation – a proposition this review argues and that the staged sequence set out above is designed to test: Consensus on what constitutes each examination, then reliability, then prospective accuracy stratified by operator level, then outcomes. FAST became universal on the strength of a fixed procedure rather than exceptional accuracy. The four applications reviewed here have the technology and the evidence; what they lack is the agreement.