Capsule endoscopy: Emerging technologies and future clinical applications

Abstract

Capsule endoscopy has evolved from a diagnostic tool for small bowel evaluation into a complex platform with expanding diagnostic and prognostic potential. This mini review aims to summarize recent advances in capsule endoscopy and explore potential clinical applications and future directions of this technology. The introduction of pan-enteric capsule endoscopy has enabled comprehensive evaluation of the entire gastrointestinal tract in a single, minimally invasive examination, supporting treat-to-target approaches and improving disease phenotyping in inflammatory bowel disease. Artificial intelligence (AI) represents one of the most important recent developments in capsule endoscopy. Deep learning algorithms have demonstrated high accuracy in lesion detection while reducing reading times. More recently, AI has been applied to automated inflammatory activity scoring, bowel cleansing assessment, and computer-assisted reporting, facilitating standardized interpretation and improving report quality. Magnetically controlled capsule endoscopy has further expanded the role of capsule technology by enabling active navigation within the upper gastrointestinal tract. Recent innovations include detachable-string systems for esophageal examination, remote 5G-assisted capsule navigation, and integration with AI-assisted image analysis, opening new perspectives for noninvasive evaluation from mouth to anus. Experimental robotic capsule prototypes have explored the feasibility of interventional functions, including tissue sampling and targeted drug delivery. These technologies remain at the preclinical stage, and their potential clinical applicability requires further investigation.

Key Words: Capsule endoscopy; Gastrointestinal bleeding; Crohn’s disease; Pan-enteric capsule endoscopy; Magnetically controlled capsule endoscopy; Artificial intelligence; Robotic endoscopy

Core Tip: Capsule endoscopy is rapidly evolving beyond conventional small bowel imaging. Pan-enteric systems enable comprehensive gastrointestinal assessment, while artificial intelligence improves lesion detection, standardizes examination quality, and reduces reading time. Magnetically controlled capsules extend minimally invasive examination to the upper gastrointestinal tract and provide remote navigation. Emerging robotic platforms may further expand these capabilities by integrating active navigation, tissue sampling, and targeted drug delivery. This review summarizes these recent advances and highlights the transition of capsule endoscopy from a passive diagnostic device dedicated to small bowel visualization to an intelligent, minimally invasive platform with expanding applications throughout the gastrointestinal tract.

INTRODUCTION

Since its introduction in 2000, capsule endoscopy has enabled minimally invasive visualization of the small bowel, a segment that was previously difficult to assess using conventional endoscopy. Today, capsule endoscopy is an integral part of the diagnostic workup for small bowel disorders across a broad range of clinical settings, including gastrointestinal bleeding, inflammatory bowel disease, and small bowel neoplasia[1]. Moreover, its ease of use and the high diagnostic yield have progressively broadened the application of capsule endoscopy beyond the small bowel, extending its use to the lower gastrointestinal tract, particularly the colon, as well as to the upper gastrointestinal tract, including the esophagus and stomach. This mini review aims to provide an overview of the emerging indications for capsule endoscopy and to explore its emerging clinical applications.

LITERATURE REVIEW

Relevant literature was identified through searches of PubMed/MEDLINE and Scopus from inception to June 28, 2026. The search included combinations of terms “capsule endoscopy”, “pan-enteric capsule endoscopy”, “artificial intelligence”, “deep learning”, “magnetically controlled capsule endoscopy”, “magnet-assisted capsule endoscopy”, “robotic capsule endoscopy”, and “therapeutic capsule endoscopy”. Only articles published in English and available in full text were considered eligible for inclusion. Priority was given to international guidelines, systematic reviews and meta-analyses, randomized controlled trials, and prospective multicenter studies, with additional relevant studies selected according to their clinical relevance and contribution to emerging technological developments.

PAN-ENTERIC CAPSULE ENDOSCOPY

The introduction of pan-enteric capsule endoscopy (PCE) has represented a major technological advancement, evolving from colon capsule endoscopy and enabling the assessment of both the small bowel and colon within a single procedure. This approach may reduce the need for multiple diagnostic investigations while providing a more comprehensive evaluation of disease extent in Crohn’s disease. Moreover, in patients with gastrointestinal bleeding and a negative upper gastrointestinal endoscopy, PCE offers the potential to assess both the small bowel and colon in a single examination.

Regarding its use in inflammatory bowel disease, early clinical evidence came from the prospective multicenter study by Leighton et al[2], which compared PCE with conventional colonoscopy in 66 patients with active Crohn’s disease. The per-patient diagnostic yield was significantly higher with capsule endoscopy than with colonoscopy [83.3% vs 69.7%; absolute difference, 13.6%; 95% confidence interval (CI): 2.6%-24.7%], while the per-segment analysis showed a diagnostic yield of 40.6% with capsule endoscopy compared with 32.7% with colonoscopy. Notably, lesions were detected exclusively by capsule endoscopy in 12 patients, highlighting the added value of concomitant small bowel assessment.

These findings were subsequently supported by the meta-analysis by Tamilarasan et al[3], which included 14 prospective studies involving a total of 797 patients, seven of which specifically evaluated patients with Crohn’s disease. PCE showed a diagnostic yield comparable to that of magnetic resonance, colonoscopy, and enteroscopy, although diagnostic performance was numerically higher with capsule endoscopy (pooled odds ratio: 1.25; 95%CI: 0.85-1.86). From a clinical perspective, PCE showed a numerically higher diagnostic yield than magnetic resonance and colonoscopy, although the pooled difference did not reach statistical significance. The meta-analysis also showed that newer generation capsule systems achieved better diagnostic performance than earlier devices, supporting the role of PCE in both the diagnosis of Crohn’s disease and the assessment of disease extent.

Subsequently, the role of PCE expanded beyond diagnosis to include disease phenotyping and prognostic assessment. The prospective multicenter BOMIRO study[4], which included 282 adult and pediatric patients with Crohn’s disease undergoing PCE, demonstrated that PCE can significantly influence the clinical characterization and management of the disease. In particular, PCE identified previously unrecognized proximal small bowel involvement in a substantial proportion of patients, leading to reclassification of disease phenotype according to the Montreal classification in 32% of cases. Furthermore, PCE findings resulted in a change in therapeutic strategy in 79% of patients, predominantly through treatment intensification. During follow-up, the extent of small bowel involvement emerged as an independent predictor of adverse clinical outcomes, indicating that PCE may provide not only diagnostic but also prognostic information, with potential implications for disease monitoring and treat-to-target strategies.

More recently, the first meta-analysis specifically addressing capsule endoscopy for therapeutic monitoring in Crohn’s disease was published[5]. The meta-analysis included nine studies comprising 321 patients with small bowel or ileocolonic involvement and evaluated the use of capsule endoscopy, including both small bowel capsule endoscopy and PCE, to guide treatment optimization. Active disease was present at baseline in 76.3% of patients, and capsule endoscopy findings led to treatment modification in 93% of patients with endoscopically active disease. After a mean follow-up of approximately 33 weeks, mucosal healing was achieved in 27% of patients (95%CI: 18%-38%), while mucosal response was observed in 43%; conversely, 34% showed no endoscopic response. Capsule retention was extremely rare, with only one case reported across the included studies. Overall, these findings support the evolving role of capsule endoscopy beyond diagnosis, as an objective tool for monitoring therapeutic response within treat-to-target strategies.

The introduction of PCE has also opened new perspectives in the management of obscure gastrointestinal bleeding. Traditionally, following a negative upper gastrointestinal endoscopy, the diagnostic pathway involves colonoscopy, followed, if negative, by small bowel capsule endoscopy[1]. More recently, an alternative approach has been proposed, in which PCE is used as a first-line investigation after a negative upper gastrointestinal endoscopy, allowing comprehensive evaluation of both the small bowel and colon within a single procedure.

Initial evidence came from a European prospective study including 128 patients presenting with melena and a negative upper gastrointestinal endoscopy, 23 had a negative EGD and 12 underwent PCE[6]. PCE identified the bleeding source in 83% of patients, enabling small bowel endoscopic treatment in approximately half of the cases and thereby avoiding colonoscopy in these patients.

These findings were subsequently supported by a prospective study including 100 patients with iron-deficiency anemia or overt gastrointestinal bleeding following a negative upper gastrointestinal endoscopy[7]. PCE identified lesions potentially responsible for bleeding in 95.7% of patients with positive findings, compared with 50.0% for colonoscopy (P < 0.01). Notably, lesions were detected exclusively by PCE in 65.2% of cases, while more than half of the patients (54%) had no clinically significant lesions, potentially avoiding the need for further invasive procedures. PCE showed a sensitivity of 95.7% and a specificity of 100%, compared with a sensitivity of 50% for colonoscopy. Based on these findings, the authors proposed a novel diagnostic algorithm, the “PCE first strategy”, using PCE after a negative upper gastrointestinal endoscopy to identify patients in whom further invasive procedures may be warranted for tissue sampling or endoscopic therapy. Overall, these findings support the potential role of PCE as an alternative diagnostic approach in obscure gastrointestinal bleeding, with subsequent invasive endoscopy guided by capsule findings.

ARTIFICIAL INTELLIGENCE AND IMAGE ANALYSIS

The introduction of artificial intelligence (AI)-assisted reading systems has substantially transformed the field of capsule endoscopy. Although capsule endoscopy represents one of the most important diagnostic modalities for small bowel evaluation, reviewing a complete examination remains a time-consuming and demanding process, requiring the assessment of tens of thousands of images and sustained attention from the endoscopist. Prolonged reading times and reader fatigue may therefore increase the risk of missing clinically relevant lesions. A recent meta-analysis showed that the application of deep-learning algorithms, predominantly based on convolutional neural networks, has enabled the development of AI-assisted systems capable of automatically identifying and selecting abnormal frames, thereby improving diagnostic accuracy while significantly reducing reading time[8].

Beyond improving reading efficiency, AI may also help reduce interobserver variability and support the training of less experienced endoscopists. Several studies included in the meta-analysis showed that the diagnostic sensitivity achieved by AI-assisted junior readers may reach or even exceed that of experienced endoscopists performing conventional reading. These findings suggest a potential role for AI not only as a diagnostic support tool, but also in standardizing capsule endoscopy interpretation and facilitating endoscopist training.

The first AI algorithms developed for capsule endoscopy focused primarily on the detection of small bowel vascular lesions, particularly angioectasias, which represent the most common cause of suspected small bowel bleeding. In this setting, the first prospective European multicenter study evaluated the ProScan system for AI-assisted capsule endoscopy reading in patients with suspected small bowel bleeding. AI-assisted reading was shown to be noninferior to conventional reading and superior for several study endpoints, increasing the diagnostic yield (73.7% vs 62.4%) while markedly reducing the mean reading time from 33.7 minutes to 3.8 minutes[9].

The application of AI was subsequently extended to the assessment of Crohn’s disease, with the aim not only of detecting inflammatory lesions but also of quantifying disease activity. In this setting, a recent European study evaluated the OMOM SmartScan system in 141 patients with Crohn’s disease undergoing small bowel capsule endoscopy[10]. The study compared conventional full video reading with AI-assisted review, with agreement in the assessment of the Lewis Score and Capsule Endoscopy Crohn’s Disease Activity Index as the main endpoints.

The results showed excellent agreement between SmartScan and conventional reading for both the Lewis Score (intraclass correlation coefficient: 0.99) and Capsule Endoscopy Crohn’s Disease Activity Index (intraclass correlation coefficient: 0.98), demonstrating the noninferiority of the AI-assisted system for the assessment of inflammatory activity. At the same time, mean reading time was reduced from 30.2 ± 10.8 minutes to 2.9 ± 1.8 minutes (P < 0.001), corresponding to an approximately 90% reduction. In this setting, AI-assisted systems may facilitate the routine use of endoscopic activity scores in capsule endoscopy while improving the efficiency and reproducibility of disease activity assessment.

In addition to lesion detection, AI is also being applied to the standardization of qualitative parameters in capsule endoscopy. Adequate bowel cleansing is a major determinant of the diagnostic accuracy of capsule endoscopy; however, its assessment still relies largely on the subjective judgment of the reader[11]. To overcome this limitation, a deep-learning algorithm was developed to automatically assess bowel cleanliness based on the proportion of visible mucosa. In the first study[12], using images obtained from three centers, the software generated a continuous cleansing score ranging from 1 to 5 and showed excellent correlation with expert assessment. Moreover, better bowel preparation quality was associated with increased detection of small lesions, supporting the clinical relevance of the automated score.

In a later study, the same group validated an algorithm using 400000 images obtained from small bowel capsule endoscopy examinations. The algorithm achieved an accuracy of 93% for image classification and substantial agreement with expert assessment (Cohen’s κ = 0.672). A cutoff value of 2.95 identified adequate bowel preparation with an area under the receiver operating characteristic curve of 0.913, a sensitivity of 81%, and a specificity of 84%, demonstrating the reproducibility of the automated assessment in an independent population[13].

The role of AI in capsule endoscopy is expanding from lesion detection to the automated, objective, and reproducible assessment of examination quality. AI may therefore serve not only to facilitate image review but also as a valuable tool to support the interpretation and reporting of capsule endoscopy examinations.

MAGNETICALLY CONTROLLED AND ROBOTIC CAPSULE ENDOSCOPY

One of the main limitations of conventional capsule endoscopy is the inability to actively control capsule movement within the gastrointestinal tract. While passive propulsion is generally adequate for small bowel examination, the large gastric cavity and its complex anatomy may result in incomplete visualization of the gastric mucosa. To overcome this limitation, magnetically controlled capsule endoscopy (MCCE) systems have been developed, in which an external magnetic field enables real-time navigation of the capsule, allowing a more comprehensive inspection of the gastric mucosa without the need for conventional endoscopy.

When compared with conventional gastroscopy, MCCE demonstrated a sensitivity of 90.4%, a specificity of 94.7%, and an overall diagnostic accuracy of 93.4% for the detection of gastric lesions, with almost perfect agreement with conventional gastroscopy (κ = 0.884). Furthermore, the gastric examination completion rate exceeded 98%, with no relevant adverse events reported, supporting the feasibility and safety of the technique[14].

The potential of minimally invasive magnetically controlled capsule examination has also been explored in the screening of upper gastrointestinal malignancies. For the detection of upper gastrointestinal neoplasms, MCCE showed a sensitivity of 100% and a specificity of 99.9%, while sensitivity and specificity for focal gastric lesions were 95.1% and 96.6%, respectively. The procedure was also well accepted by patients, who showed a significant preference for MCCE over conventional gastroscopy. Although these findings demonstrate promising diagnostic performance, current evidence does not support MCCE as a replacement for conventional gastroscopy or establish its effectiveness in population-based screening programs. Nevertheless, its minimally invasive nature and high patient acceptability warrant further evaluation of its potential role in upper gastrointestinal cancer screening[15].

Another emerging application of MCCE is its use in gastrointestinal emergencies. In a prospective multicenter study[16], MCCE was evaluated as an initial diagnostic investigation in approximately 100 hemodynamically stable patients with acute upper gastrointestinal bleeding. The overall diagnostic yield was 95.8%, with 92 lesions identified in 96 patients. In more than half of the cases (52.5%), MCCE avoided a negative gastroscopy during hospitalization, while patients with high-risk lesions or requiring endoscopic therapy were selectively referred for the most appropriate conventional endoscopic procedure according to lesion location.

The integration of AI has further expanded the potential applications of MCCE. In 2024, the first deep-learning algorithm designed to simultaneously analyze gastric and small bowel images acquired by MCCE was developed[17], shortening the reading time. Remote control represents a promising development in MCCE, enabled by high-speed telecommunication networks. In a prospective study[18], the feasibility of the first remotely controlled MCCE system was evaluated using a hub-and-spoke model, in which capsule navigation was performed by an endoscopist located at a referral center (hub), while the patient underwent the examination at a peripheral facility (spoke). Remote MCCE was successfully completed in all patients (100%), with no adverse events or connection interruptions. The quality of gastric examination, procedure time (approximately 23 minutes), and diagnostic yield were comparable between conventional and remote MCCE.

5G technology may enable MCCE to be performed in geographically remote areas while maintaining quality standards comparable to those of conventional MCCE. In the future, tele-endoscopy could facilitate wider implementation of MCCE by providing peripheral centers with access to highly specialized expertise and potentially reducing disparities in access to advanced endoscopic procedures.

To overcome one of the main limitations of conventional capsule endoscopy, namely the rapid esophageal transit time, detachable-string MCCE was developed[19]. In this system, a thin detachable string allows active control of capsule movement during esophageal examination, enabling repeated inspection of the esophageal mucosa before the string is detached and the capsule proceeds to the stomach and small bowel. Initially developed for the evaluation of esophageal disorders, detachable-string MCCE was later investigated as a potential screening tool for esophageal neoplasia[20]. The technique showed a sensitivity of 95.4% and a specificity of 89.0% for the detection of esophageal mucosal lesions. For early esophageal squamous cell carcinoma, sensitivity was 66.7% and specificity was 99.2%. The same platform has also been evaluated for the screening of gastroesophageal varices in patients with cirrhosis, achieving a sensitivity of approximately 97.5%[21].

Economic considerations may represent another relevant factor in the future implementation of MCCE. More recently, Tai et al[22] specifically evaluated the economic implications of magnet-assisted capsule endoscopy (MACE). The authors compared the costs of conventional oral gastroscopy, transnasal endoscopy, and MACE in patients with dyspepsia. Although MACE offers the advantages of a minimally invasive and well-tolerated examination, its estimated cost per procedure was higher than that of conventional oral gastroscopy (€407.10 vs €184.10), mainly because of the cost of the single-use capsule.

However, the cost of an individual procedure does not necessarily reflect the economic impact of the overall diagnostic pathway. Previous evidence from Marmo et al[23] demonstrated that capsule endoscopy was a cost-effective strategy in patients with obscure gastrointestinal bleeding, primarily as a result of its higher diagnostic yield and the consequent reduction in additional diagnostic investigations. Whether similar economic benefits could be achieved with MCCE remains to be established. This may be particularly relevant in selected clinical settings in which a single capsule examination can evaluate both the upper gastrointestinal tract and the small bowel. Future cost-effectiveness studies should therefore assess MACE within the entire diagnostic pathway rather than considering procedural costs alone.

EXPERIMENTAL PROTOTYPES: FROM DIAGNOSIS TO THERAPY

The evolution of capsule endoscopy is progressively moving beyond its role as a purely diagnostic modality toward the development of robotic platforms capable of performing interventional procedures within the gastrointestinal tract. Among the most promising applications are targeted tissue biopsy and localized drug delivery, two functions that could substantially expand the role of capsule endoscopy.

In this context, a magnetically controlled soft capsule equipped with a fine-needle biopsy system has recently been developed[24]. Capsule navigation is achieved through an external magnetic field, while a flexible mechanism enables controlled tissue sampling while minimizing the risk of injury to the gastrointestinal wall. Experimental studies have demonstrated the feasibility of tissue biopsy in ex vivo models, representing one of the first examples of a capsule platform combining active navigation with diagnostic intervention.

A further step toward therapeutic capsule endoscopy has recently been described with the development of the Magnetically Actuated Robotic Capsule Endoscope[25]. This device integrates magnetic navigation, real-time visualization, and a retractable microneedle system designed to deliver drugs directly to the gastrointestinal mucosa. As a proof of concept, the authors used epinephrine, demonstrating controlled and targeted drug delivery at the intended site. However, this technology remains at the preclinical stage and its role in clinical practice is not well established.

Overall, these advances point toward a new generation of robotic capsule platforms in which magnetic navigation, real-time localization, AI, and interventional capabilities converge within a single device. In the future, further technological development may enable capsule platforms to combine diagnostic imaging with functions such as tissue sampling, targeted drug delivery, and real-time AI-assisted analysis, although these applications remain largely investigational and require further preclinical and clinical validation (Table 1).

Table 1 Representative studies on emerging technologies and clinical applications of capsule endoscopy.

Ref.	Study design	Capsule/technology	Clinical setting
Leighton et al[2], 2017	Prospective multicenter feasibility study	PCE	Diagnostic yield in active Crohn’s disease
Tamilarasan et al[3], 2022	Systematic review and meta-analysis	PCE	Diagnosis and assessment of Crohn’s disease extent
Oliva et al[4], 2025 (BOMIRO)	Prospective multicenter study	PCE	Phenotypic characterization and prognostic assessment in Crohn’s disease
Marmo et al[5], 2026	Systematic review and meta-analysis	PCE/SBCE	Treatment monitoring in Crohn’s disease
Mussetto et al[6], 2021	Prospective feasibility study	PCE	Diagnostic yield in melena after negative upper GI endoscopy
Rosa et al[7], 2024	Prospective study	PCE	GI bleeding/iron-deficiency anemia after negative upper GI endoscopy
Spada et al[9], 2024	Prospective multicenter study	AI-assisted SBCE	AI-assisted lesion detection in suspected small-bowel bleeding
Mesquita et al[10], 2026	Prospective study	AI-assisted SBCE	AI-assisted assessment of Crohn’s disease activity
Nam et al[12], 2021	Multicenter algorithm development study	AI-assisted SBCE	Automated assessment of small-bowel cleansing
Liao et al[14], 2016	Prospective comparative study	MCCE	Detection of gastric lesions
Li et al[15], 2021	Prospective feasibility study	MCCE	Upper GI cancer screening
Yu et al[16], 2023	Prospective multicenter study	MCCE	Acute upper GI bleeding
Xie et al[17], 2024	AI development/validation study	MCCE + AI	AI-assisted detection of gastric and small-bowel lesions
Zhang et al[18], 2023	Prospective study	Remote MCCE	Remote upper GI examination using 5G technology
Chen et al[19], 2019	Prospective feasibility study	DS-MCCE	Complete visualization of the esophagus and stomach
Dai et al[20], 2026	Prospective diagnostic study	DS-MCCE	Detection of esophageal lesions and early esophageal squamous cell carcinoma
Son et al[24], 2020	Preclinical/ex vivo study	Robotic capsule	Preclinical tissue sampling
Chen et al[25], 2026	Preclinical proof-of-concept study	Robotic capsule	Preclinical targeted drug delivery

PCE: Pan-enteric capsule endoscopy; SBCE: Small-bowel capsule endoscopy; GI: Gastrointestinal; AI: Artificial intelligence; MCCE: Magnetically controlled capsule endoscopy; DS-MCCE: Detachable-string magnetically controlled capsule endoscopy.

DISCUSSION

The evidence summarized in this review illustrates the progressive expansion of capsule endoscopy beyond its established role in small-bowel evaluation, driven by technological advances that have broadened its diagnostic applications across the gastrointestinal tract and introduced new opportunities for AI-assisted interpretation and magnetically controlled navigation, while interventional applications remain largely experimental.

PCE offers the possibility of assessing multiple gastrointestinal segments within a single examination, providing simultaneous evaluation of the small bowel and colon. This approach has shown encouraging results in patients with suspected small bowel or lower gastrointestinal bleeding after a negative upper endoscopy, as well as in the assessment of disease activity in patients with Crohn’s disease[2-7]. However, its use should be considered in appropriately selected patients, and its precise role within established diagnostic algorithms and international guidelines remains to be clearly defined. In addition, the evidence supporting a PCE-first strategy in gastrointestinal bleeding is still preliminary, being derived from a pilot study in which only 12 patients underwent PCE[6] and from a prospective study of 100 patients[7]; larger comparative studies are needed before this strategy can be incorporated into routine diagnostic algorithms.

One of the main limitations of PCE is the variable completion rate reported across studies. In the meta-analysis by Tamilarasan et al[3], capsule completion rates were reported in 13 of 14 studies and ranged from 69% to 100% across studies in inflammatory bowel disease, whereas in the study by Mussetto et al[6] in patients with gastrointestinal bleeding, complete small-bowel and colonic examinations were achieved in 100% (12/12) and 91% (11/12) of patients, respectively. PCE requires adequate bowel preparation to ensure satisfactory colonic cleansing, together with subsequent boosters to promote capsule progression through the gastrointestinal tract. The complexity and volume of these regimens may represent a barrier to patient adherence. In the prospective study by Brodersen et al[26] which evaluated adherence to a PCE preparation regimen consisting of 2 + 2 L of polyethylene glycol (PEG) and a sodium phosphate booster in 59 patients with suspected Crohn’s disease, only 17% of patients were able to ingest the full 4 L of PEG. Lower PEG intake was associated with poorer bowel cleansing; however, diagnostic yield did not differ significantly between adequately and inadequately prepared examinations.

Low-volume regimens may represent a strategy to improve the tolerability of bowel preparation while maintaining adequate cleansing and examination completion rates[27]. Therefore, rather than considering bowel preparation an intrinsic limitation of PCE, optimization of preparation protocols may help improve patient adherence and overall feasibility. Further studies in larger populations are needed to identify the most acceptable preparation regimen while maintaining adequate bowel cleansing and completion rates.

The introduction of AI-assisted reading systems represents another major advance in capsule endoscopy. As capsule endoscopy is inherently based on images, AI algorithms are particularly well suited to identifying potentially pathological findings among the thousands of frames generated during each examination. This approach has been particularly investigated for the detection of lesions with bleeding potential and inflammatory lesions in patients with Crohn’s disease[9,10]. AI-assisted reading can substantially reduce the time required for image review and may also provide a valuable tool for the training of less experienced endoscopists. Nevertheless, the implementation of AI in clinical practice requires caution. The endoscopist remains responsible for the interpretation and validation of AI-generated findings, and adequate expertise is therefore essential to distinguish true pathological lesions from false-positive findings. Moreover, AI systems may fail to identify uncommon or atypical lesions that are underrepresented or absent from their training datasets. Further external validation, standardized performance assessment, regulatory oversight, and appropriate endoscopist training will therefore be required before widespread clinical implementation. At present, AI should be regarded as a tool to support, rather than replace, the clinical judgment of the endoscopist.

The possibility of performing gastric examination by capsule endoscopy using external magnetic fields to actively control capsule movement within the gastric cavity represents another important technological advance. Available studies have reported encouraging results and suggest that MCCE is feasible[14-17]. Compared with conventional gastroscopy, MCCE may be better accepted by patients because of its minimally invasive nature, the absence of sedation, and the possibility of resuming normal activities immediately after the procedure. However, appropriate patient selection remains essential. Patients with permanent pacemakers, other implantable electromedical devices, or certain metallic implants have generally been excluded from clinical studies of MCCE because of potential interactions with the magnetic field used for capsule navigation. Consequently, clinical safety data in these populations remain limited, and the safety of MCCE in patients with such devices cannot currently be established[14,16]. A further important limitation of currently available MCCE systems is their exclusively diagnostic nature. Despite their high diagnostic performance for gastric lesions, they do not allow tissue sampling or therapeutic intervention. This limitation is clinically relevant, as conventional endoscopy remains necessary when histological confirmation or endoscopic treatment is required. In the study by Liao et al[14], 110 of 350 patients (31.4%) required biopsy during conventional gastroscopy. Similarly, in the study by Yu et al[16], subsequent conventional endoscopy was required in selected patients for tissue biopsy, endoscopic hemostasis, or further evaluation after inadequate visualization. Therefore, the inability to obtain histology or provide treatment during the same examination should be considered when selecting patients for MCCE and when defining its role relative to conventional gastroscopy[14]. Of note, nearly all clinical studies of MCCE discussed in this review, including those on gastric cancer screening, acute bleeding, AI-assisted reading, remote navigation, and detachable-string systems, were conducted in China[14-20]; their generalizability to Western healthcare settings, which differ in disease prevalence, access to endoscopy, and reimbursement models, remains to be established.

Despite these advances, several additional barriers apply across capsule-based technologies. Capsule retention remains an important safety concern, particularly in patients with established Crohn’s disease or suspected small-bowel stenosis, emphasizing the importance of appropriate patient selection and, when indicated, patency assessment or cross-sectional imaging[1]. Diagnostic performance may also be affected by inadequate bowel preparation and incomplete examination. Swallowing disorders should be considered before capsule administration because dysphagia may increase the risk of aspiration; in selected patients, endoscopic capsule placement may be required[1]. Cost and limited availability of newer technologies, particularly magnetically controlled systems, represent additional barriers to widespread adoption. Notably, in a United Kingdom cost analysis the higher per procedure cost of MACE was driven by the single-use capsule rather than by equipment costs, which were lower than those of flexible endoscopy[22].

The next step is to move from diagnostic accuracy to outcome-based evidence. A model for this transition is provided by a recent randomized controlled trial in quiescent Crohn’s disease, in which proactive treatment optimization guided by capsule endoscopy every 6 months reduced the rate of clinical flares at 24 months in high-risk patients compared with continued standard care (25% vs 70%; odds ratio: 0.14; 95%CI: 0.04-0.57)[28]. Comparable studies are needed to establish whether a PCE first strategy in suspected small bowel or lower gastrointestinal bleeding, MCCE as the initial examination in stable patients with acute upper gastrointestinal bleeding, and AI-assisted reading translate into fewer invasive procedures, shorter time to diagnosis, lower overall costs, and better outcomes. In parallel, the combination of pan-enteric or magnetically controlled examination with AI-assisted reading and remote navigation may ultimately allow a single minimally invasive examination of the gastrointestinal tract to be offered within hub-and-spoke networks[16,18]; such integrated pathways will, however, require prospective validation before they can be incorporated into clinical guidelines.

Finally, the level of clinical evidence differs substantially among the technologies discussed in this review. Conventional small bowel capsule endoscopy is supported by international guidelines; PCE and selected magnetically controlled applications are supported by prospective studies and meta-analyses, although randomized evidence remains limited; AI-assisted reading has been evaluated in prospective studies but still requires external, real-world validation across platforms; and remote navigation has so far been assessed only in feasibility studies. Robotic capsules incorporating tissue sampling or targeted drug delivery remain investigational and are currently largely confined to preclinical studies.

CONCLUSION

Capsule endoscopy is now firmly established as a key diagnostic tool for small bowel evaluation; however, its clinical applications have progressively expanded beyond the small bowel. The development of PCE has enabled comprehensive assessment of the small bowel and colon within a single procedure, while magnetically controlled systems have extended the examination to the upper gastrointestinal tract. At the same time, the integration of AI has substantially reduced reading and reporting times while maintaining high diagnostic accuracy. Nevertheless, the relatively high costs associated with capsule technologies may currently limit their widespread implementation. The widespread adoption of these technological advances, particularly in Western healthcare systems, will likely depend on reductions in device costs and greater affordability as these technologies become more widely available.