Complementary diagnostic value of conventional ultrasound and contrast-enhanced magnetic resonance imaging for post-hepatectomy hepatocellular carcinoma surveillance

Abstract

BACKGROUND

Hepatocellular carcinoma (HCC) is one of the most common cancers and the second leading cause of cancer-related death globally. Recurrence rates range from 50% to 70% within 5 years after radical hepatectomy, with most recurrences occurring during the first 2 postoperative years. Hence, precise identification of recurrent lesions is essential to optimize patient outcomes. Conventional ultrasound (US) and contrast-enhanced magnetic resonance imaging (CE-MRI), each with distinct advantages and limitations, are the main imaging modalities used for surveillance. However, the combined diagnostic performance of these modalities in a systematic post-hepatectomy surveillance program remains uncharacterized.

AIM

To assess the complementary diagnostic efficacy of conventional US and CE-MRI in early recurrence surveillance following radical hepatectomy for HCC. Specifically, we evaluated the diagnostic performance of each modality alone and in combination for detecting early recurrent lesions, as well as their respective strengths and weaknesses, to identify an optimal imaging follow-up strategy.

METHODS

We performed a retrospective study of 186 patients with HCC who underwent radical hepatectomy in the Department of Hepatobiliary Surgery at Fengjie County People’s Hospital from January 2020 to December 2024. All patients underwent regular surveillance with conventional US and CE-MRI during the first 2 years after surgery. The study protocol was approved by the hospital ethics committee. Patients were categorized into recurrence and non-recurrence groups based on clinical follow-up and confirmation of recurrent lesions by pathological examination (where biopsy or surgical re-exploration was performed) or by multidisciplinary clinical-imaging consensus using Liver Imaging Reporting and Data System criteria on sequential imaging in conjunction with alpha-fetoprotein trajectory and clinical progression. Detection rate, diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were compared among US alone, CE-MRI alone, and the combined US-MRI approach. We also examined the size distribution, location patterns, and imaging features of recurrent lesions detected by each modality.

RESULTS

Early recurrence occurred in 68 (36.6%) of 186 patients within 24 months after surgery. US detected 42 of the 68 recurrent cases (detection rate, 61.8%), whereas CE-MRI detected 63 cases (detection rate 92.6%). The combined US-MRI approach achieved a detection rate of 97.1% (66/68). For lesions measuring ≥ 2 cm, US demonstrated a sensitivity of 85.7% and specificity of 93.2%, compared with a sensitivity of 97.1% and specificity of 95.8% for CE-MRI. However, for lesions measuring < 1 cm, US sensitivity was only 38.5%, compared with 84.6% for CE-MRI (P < 0.001). Compared with CE-MRI, US had significantly higher sensitivity for detecting superficial subcapsular recurrences (88.9% vs 72.2%, P = 0.032) and provided real-time assessment. CE-MRI showed superior performance in detecting deep parenchymal lesions (sensitivity 95.2% vs 52.4% for US, P < 0.001), portal vein tumor thrombus, and lesion vascularity patterns. The combined approach achieved an overall sensitivity, specificity, positive predictive value, and negative predictive value of 97.1%, 96.6%, 94.3%, and 98.3%, respectively, outperforming each modality alone (all P < 0.05). The greatest benefit of complementary diagnostic testing was observed for 1- to 2-cm lesions and lesions in difficult-to-visualize locations.

CONCLUSION

The findings of this study provide compelling preliminary evidence supporting a complementary role for a dual-modality strategy in post-hepatectomy HCC surveillance. However, the combined detection rate of 97.1% did not achieve absolute sensitivity because 2 subcentimeter hepatic dome lesions were not detected even with the combined approach. These findings warrant consideration for incorporation into clinical practice, pending confirmation in larger multicenter prospective trials.

Key Words: Hepatocellular carcinoma; Radical hepatectomy; Early recurrence; Conventional ultrasound; Contrast-enhanced magnetic resonance imaging

Core Tip: Postoperative recurrence remains a major challenge in the management of hepatocellular carcinoma. This study evaluated the complementary diagnostic value of conventional ultrasound and contrast-enhanced magnetic resonance imaging for detecting early recurrence after radical hepatectomy. Contrast-enhanced magnetic resonance imaging demonstrated superior sensitivity for detecting small and deep lesions and vascular invasion, whereas ultrasound showed advantages in detecting superficial lesions and facilitating frequent surveillance. The combined use of ultrasound and contrast-enhanced magnetic resonance imaging significantly improved diagnostic accuracy and detection rates. An integrated surveillance strategy using both modalities may enhance early detection and facilitate timely therapeutic intervention.

INTRODUCTION

Hepatocellular carcinoma (HCC) is one of the most commonly diagnosed malignant tumors worldwide, being the sixth most commonly diagnosed cancer and the third leading cause of cancer-related death globally[1]. Despite advances in surgical techniques and perioperative management, the prognosis of HCC remains poor, with high recurrence rates being a particular barrier to long-term survival. Radical hepatectomy remains the most effective curative strategy for HCC at present; however, 50%-70% of patients develop recurrence within 5 years after surgery, and the majority of recurrences occur during the first 2 postoperative years[2]. Early detection and timely intervention for recurrent lesions are key factors influencing patient prognosis and long-term survival outcomes in patients with HCC[3].

Imaging surveillance of patients following hepatectomy for HCC is recommended in clinical practice guidelines to detect recurrence at the earliest possible stage[4]. Several imaging modalities are available for this purpose, each with its own advantages and limitations. Commonly employed as a first-line screening method, conventional ultrasound (US) offers several benefits, including low invasiveness, real-time imaging capability, no X-ray exposure, repeatability, cost-effectiveness, and widespread availability[5]. Despite its widespread acceptance, US has limitations such as operator dependency, restricted acoustic windows, limited sensitivity for small lesions (< 2 cm), and obscuration of deep hepatic segments and lesions by bowel gas or obesity[6].

Contrast-enhanced magnetic resonance imaging (CE-MRI) has recently become a highly accurate imaging modality for both the diagnosis of HCC and surveillance. Compared with other imaging techniques, CE-MRI offers excellent soft tissue contrast resolution, multiphase imaging capability, and high sensitivity for small lesions and vascular invasion, as well as greater diagnostic confidence through dynamic contrast-enhancement patterns that facilitate differentiation between malignant and benign lesions[7]. The use of hepatobiliary-specific contrast agents allows more confident detection and characterization of HCC lesions, especially in a cirrhotic liver background[8]. Nonetheless, CE-MRI has disadvantages, including high cost, prolonged examination time, and limited availability. In addition, it cannot be performed in patients with renal insufficiency or metallic implants, and it cannot provide real-time assessment for clinical use[9].

Because of the unique features and limitations of each imaging modality, there is growing interest in understanding the complementary roles of US and CE-MRI in postoperative HCC surveillance. Thus far, studies have focused on the standalone comparison of these modalities as diagnostic tools, with limited assessment of their complementary value within a systematic surveillance program[10]. Prior work by Jeon et al[10] and Kim et al[6] evaluated each modality separately but did not investigate their combined utility in a systematic manner. Further insights into the specific contexts in which each modality achieves superior performance, as well as guidance on optimal combined approaches for early recurrence detection, could lead to substantial improvements[11,12].

Consequently, this retrospective study was conducted to explore the complementary diagnostic value of conventional US and CE-MRI in early recurrence monitoring following radical hepatectomy for HCC. Through a focused assessment of modality-specific advantages in conjunction with their detection rates and diagnostic performance characteristics, we endeavoured to derive evidence-based recommendations for optimal imaging follow-up strategies in the postoperative management of HCC.

MATERIALS AND METHODS

Study design and patient selection

This retrospective study included all patients with HCC who underwent radical hepatectomy at the Department of Hepatobiliary Surgery of our hospital from January 2020 to December 2024. Approval of the study protocol was granted by the Ethics Committee of People’s Hospital of Fengjie, and the requirement for informed consent was waived because of the retrospective nature of the analysis.

The inclusion criteria were non-HCC, local resection or non-R0 excision, age < 18 years or > 75 years at diagnosis, Child-Pugh class C liver function, and incomplete clinical information.

The exclusion criteria were positive surgical margins, evidence of extrahepatic metastases at the time of surgery, prior malignancy, simultaneous transplantation, incomplete surveillance data for HCC recurrence after surgery, contraindications to gadolinium-based MRI, severe renal failure with a glomerular filtration rate < 30 mL/minute/1.73 m², and death from non-HCC causes during follow-up.

Imaging surveillance protocol

US examinations were conducted on high-end systems (GE Logiq E9 or Philips EPIQ 7) with convex transducers (2-5 MHz) by experienced sonographers (> 10 years’ experience). All imaging examinations were performed on 1.5T or 3.0T MR scanners (Siemens or GE) using standard hepatic acquisition protocols consisting of T1-weighted imaging, T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast-enhanced sequences with gadolinium-based agents. Surveillance was performed as follows: US every 3 months during the first year and every 6 months during the second year, and CE-MRI every 3 months during the first year and every 6 months thereafter.

Recurrence confirmation

Of the 68 recurrence events, 31 (45.6%) were histopathologically verified by biopsy or surgical re-exploration. The remaining 37 recurrences (54.4%) were confirmed by multidisciplinary clinical-imaging consensus using Liver Imaging Reporting and Data System criteria on sequential imaging in conjunction with alpha-fetoprotein (AFP) trajectory and clinical progression. Regarding blinding, US and CE-MRI findings were reported separately by radiologists blinded to the other modality’s results; however, access to clinical history was permitted, which represents a limitation of the retrospective study design that we explicitly acknowledge.

Statistical analysis

Statistical analysis was conducted using SPSS 26.0 and MedCalc 20.0. Continuous variables were reported as mean ± SD or median (interquartile range). Categorical data were presented as n (%). Continuous variables were analyzed using the independent t-test or Mann-Whitney U test, while categorical variables were analyzed using either the χ² test or Fisher’s exact test. Diagnostic performance parameters were calculated with 95% confidence intervals. Paired diagnostic performances were compared using McNemar’s test. Diagnostic performance was evaluated using receiver operating characteristic curve analysis and compared based on the area under the curve (AUC) using DeLong’s test. P < 0.05 was considered statistically significant.

RESULTS

Patient characteristics

In total, 186 patients were included. Of these patients, early recurrence occurred in 68 (36.6%) within a median of 11 months (range, 4-23 months). The patients’ baseline characteristics are shown in Table 1. The median time to first recurrence detection was 9.5 months (interquartile range, 6.0-14.0 months). The only significant difference between the recurrence and non-recurrence groups was the preoperative AFP level (P = 0.018) (Table 1).

Table 1 Baseline characteristics of the study population.

Characteristic	Total (n = 186)	Recurrence (n = 68)	Non-recurrence (n = 118)
Age, years	56.8 ± 11.2	57.3 ± 10.8	56.5 ± 11.4
Male sex	158 (84.9)	59 (86.8)	99 (83.9)
HBV infection	162 (87.1)	60 (88.2)	102 (86.4)
Child-Pugh A	168 (90.3)	60 (88.2)	108 (91.5)
Tumor size, cm	5.2 ± 2.8	5.8 ± 3.1	4.9 ± 2.6
AFP, ng/mL	185.5 (28.7-756.3)	298.4 (56.2-1024.5)	142.8 (22.3-528.6)
Liver cirrhosis	148 (79.6)	58 (85.3)	90 (76.3)
Solitary tumor	142 (76.3)	48 (70.6)	94 (79.7)
Microvascular invasion	52 (28.0)	28 (41.2)	24 (20.3)
ALT, U/L	42.6 ± 28.3	45.8 ± 31.2	40.7 ± 26.5
Albumin, g/L	38.5 ± 4.8	37.8 ± 5.1	38.9 ± 4.6

Data are presented as mean ± SD, n (%), or median (interquartile range).

AFP: Alpha-fetoprotein; ALT: Alanine aminotransferase; HBV: Hepatitis B virus.

Overall diagnostic performance

Among the 68 confirmed recurrences (comprising 76 unique lesions in total; the 68 patients with recurrence had a combined total of 76 lesions, which explains the discrepancy in stratified lesion-level analyses), US identified 42 (61.8%) cases, CE-MRI identified 63 (92.6%) cases, and the combined modality detected 66 (97.1%) cases. The detailed diagnostic performance is shown in Table 2 and Figure 1. CE-MRI had a much higher sensitivity than US (92.6% vs 61.8%, P < 0.001) but similar specificity (95.8% vs 93.2%, P = 0.412). The diagnostic performance of each individual modality is shown in Table 2, and the combined modality approach achieved the best overall results (sensitivity 97.1%, specificity 96.6%, positive predictive value 94.3%, negative predictive value 98.3%, and accuracy 96.8%) (Table 2).

Figure 1 Receiver operating characteristic curves comparing diagnostic performance. Receiver operating characteristic (ROC) curves demonstrating the diagnostic performance of conventional ultrasound (US), contrast-enhanced magnetic resonance imaging (CE-MRI), and the combined approach for early recurrence detection after hepatocellular carcinoma resection. The combined approach [blue line, area under the curve (AUC) = 0.969] showed significantly superior performance compared with CE-MRI alone (purple line, AUC = 0.942, P = 0.035) and US alone (orange line, AUC = 0.775, P < 0.001). CE-MRI also demonstrated significantly better performance than US (P < 0.001).

Table 2 Overall diagnostic performance of different imaging modalities.

Parameter	US alone	CE-MRI alone	Combined
Sensitivity, %	61.8 (49.2-73.3)	92.6 (83.7-97.6)	97.1 (89.8-99.6)
Specificity, %	93.2 (87.1-97.0)	95.8 (90.4-98.6)	96.6 (91.6-99.1)
PPV, %	84.0 (70.9-92.8)	92.6 (83.7-97.6)	94.3 (86.0-98.4)
NPV, %	81.5 (73.4-87.9)	95.8 (90.4-98.6)	98.3 (93.8-99.8)
Accuracy, %	82.3 (76.0-87.5)	94.6 (90.3-97.4)	96.8 (93.1-98.8)
AUC	0.775 (0.708-0.834)	0.942 (0.897-0.971)	0.969 (0.932-0.989)
Detection rate, %	61.8 (42/68)	92.6 (63/68)	97.1 (66/68)
Positive LR	9.09 (5.12-16.13)	22.05 (11.36-42.78)	28.56 (13.12-62.19)
Negative LR	0.41 (0.30-0.56)	0.08 (0.03-0.18)	0.03 (0.01-0.12)
Youden index	0.550 (0.417-0.683)	0.884 (0.793-0.952)	0.937 (0.872-0.978)
F1 score	0.716 (0.612-0.808)	0.926 (0.867-0.964)	0.957 (0.911-0.983)

Data are presented as point estimate (95% confidence interval), with the exception of the detection rate.

AUC: Area under the curve; CE-MRI: Contrast-enhanced magnetic resonance imaging; LR: Likelihood ratio; NPV: Negative predictive value; PPV: Positive predictive value; US: Ultrasound.

Performance analyzed by lesion size

The lesions comprised 13 small (< 1 cm), 28 medium (1-2 cm), and 35 large (> 2 cm) lesions. Table 3 shows the size-stratified performance. For large lesions, there was no statistically significant difference in performance between the 2 modalities (P = 0.065). For medium lesions, CE-MRI had significantly higher sensitivity (89.3% vs 60.7%, P < 0.001). The difference was most evident for small lesions, where US had a sensitivity of only 38.5% compared with 84.6% for CE-MRI (P < 0.001) (Table 3).

Table 3 Diagnostic performance stratified by lesion size.

Lesion size, cm	n	US sensitivity, %	CE-MRI sensitivity, %	P value
< 1	13	38.5 (13.9-68.4)	87.5 (61.7-98.4)	< 0.001
1-2	28	60.7 (40.6-78.5)	89.3 (71.8-97.7)	< 0.001
> 2	35	85.7 (69.7-95.2)	97.1 (85.1-99.9)	0.065
2-3	15	80.0 (51.9-95.7)	93.3 (68.1-99.8)	0.091
3-5	12	91.7 (61.5-99.8)	100.0 (73.5-100.0)	0.478
> 5	8	87.5 (47.3-99.7)	100.0 (63.1-100.0)	0.302
Overall	76	61.8 (49.2-73.3)	92.6 (83.7-97.6)	< 0.001

Data are presented as sensitivity (95% confidence interval).

CE-MRI: Contrast-enhanced magnetic resonance imaging; US: Ultrasound.

Performance by lesion location

Of the 68 recurrences, 27 (39.7%) were superficial subcapsular lesions and 41 (60.3%) were deep parenchymal lesions. Table 4 shows the location-stratified performance. US performed better than CE-MRI for superficial lesions (sensitivity 88.9% vs 72.2%, P = 0.032), whereas the opposite was true for deep lesions, again significantly so (sensitivity 95.2% vs 52.4%, P < 0.001). Using the combined approach, optimal detection was achieved for both superficial (96.3%) and deep (97.6%) lesions (Table 4).

Table 4 Diagnostic performance stratified by lesion location.

Lesion location	n	US sensitivity (%)	CE-MRI sensitivity (%)	P value
Superficial subcapsular	27	88.9 (70.8-97.6)	72.2 (52.8-87.3)	0.032
Deep parenchymal	41	52.4 (36.4-68.0)	95.2 (83.8-99.4)	< 0.001
Near surgical margin	18	72.2 (46.5-90.3)	88.9 (65.3-98.6)	0.045
Dome region (VII/VIII)	14	42.9 (17.7-71.1)	92.9 (66.1-99.8)	0.003
Caudate lobe (Seg I)	8	37.5 (8.5-75.5)	87.5 (47.3-99.7)	0.021
Right posterior section	12	50.0 (21.1-78.9)	91.7 (61.5-99.8)	0.008

Data are presented as sensitivity (95% confidence interval).

Superficial subcapsular: Lesions < 2 cm from the liver capsule; deep parenchymal: Lesions ≥ 2 cm from the liver capsule.

CE-MRI: Contrast-enhanced magnetic resonance imaging; Seg: Segment; US: Ultrasound.

Detection of vascular invasion

Of the 68 recurrences, 16 (23.5%) patients developed portal vein tumor thrombus, whereas 5 (7.4%) patients demonstrated hepatic vein invasion. Table 5 shows the detection of vascular invasion. CE-MRI performed significantly better than US for detecting both portal vein tumor thrombus (sensitivity 93.8% vs 62.5%, P = 0.012) and hepatic vein invasion (100% vs 40.0%, P = 0.025). For both types of vascular invasion, the combined method achieved a 100% detection rate (Table 5).

Table 5 Detection of vascular invasion.

Vascular invasion type	n	US detection (%)	CE-MRI detection (%)	P value
PVTT	16	62.5 (35.4-84.8)	93.8 (69.8-99.8)	0.012
Hepatic vein invasion	5	40.0 (5.3-85.3)	100.0 (47.8-100.0)	0.025
Bile duct invasion	7	42.9 (9.9-81.6)	85.7 (42.1-99.6)	0.035
Inferior vena cava invasion	3	33.3 (0.8-90.6)	100.0 (29.2-100.0)	0.046
Microvascular invasion (imaging)	22	45.5 (24.4-67.8)	86.4 (65.1-97.1)	0.002
Multiple vascular invasion	9	44.4 (13.7-78.8)	88.9 (51.8-99.7)	0.023

Data are presented as detection rate (95% confidence interval).

CE-MRI: Contrast-enhanced magnetic resonance imaging; PVTT: Portal vein tumor thrombus; US: Ultrasound.

Comparison of imaging characteristics

The imaging characteristics and comparative advantages of each modality are summarized in Table 6. The advantages of US included excellent real-time assessment, superior visualization of superficial lesions, and immediate availability, whereas CE-MRI offered superior soft tissue contrast, multiplanar capability, and comprehensive vascular assessment. The combined approach utilized the strengths of each modality to achieve the best diagnostic results. Receiver operating characteristic curve analysis showed that among the 4 approaches, the combined approach had better diagnostic performance (AUC = 0.969) than CE-MRI alone (AUC = 0.942; P = 0.035) and US alone (AUC = 0.775; P < 0.001), with CE-MRI also demonstrating better diagnostic performance than US (P < 0.001) (Figure 1). Overall, 66 cases were detected by both methods. Recurrence was first detected by CE-MRI in 51 (77.3%) cases, by US in 11 (16.7%) cases, and simultaneously by both imaging modalities in the remaining 4 (6.1%) cases. The median first-detection advantage for CE-MRI was 3.2 months, supporting the superior performance of this modality (Figure 2).

Figure 2 Time distribution of first recurrence detection by imaging modality. Bar chart showing which imaging modality first detected recurrence among the 66 cases detected by the combined approach. Contrast-enhanced magnetic resonance imaging (CE-MRI) was the first modality to detect recurrence in 51 cases (77.3%, blue bar), ultrasound (US) was first in 11 cases (16.7%, orange bar), and both modalities detected recurrence simultaneously in 4 cases (6.1%, light blue bar). The median time advantage of CE-MRI for first detection was 3.2 months (range, 1.5-6.0 months) compared with subsequent detection by US. IQR: Interquartile range.

Table 6 Comparative imaging characteristics and advantages.

Characteristic	Conventional US	CE-MRI
Spatial resolution	Excellent for superficial lesions	Excellent for deep lesions
Soft tissue contrast	Moderate	Superior
Small lesion detection	Limited (< 1 cm)	Excellent (< 1 cm)
Vascular assessment	Limited (Doppler)	Comprehensive
Real-time capability	Yes	No
Cost-effectiveness	High	Moderate
Availability	Widespread	Limited in some areas
Operator dependency	High	Low

CE-MRI: Contrast-enhanced magnetic resonance imaging; US: Ultrasound.

DISCUSSION

The objective of this study was to systematically assess the complementary diagnostic value of conventional US and CE-MRI in monitoring for early HCC recurrence. Although both modalities have unique advantages, the overall performance of CE-MRI was superior, and combined use remarkably enhanced early recurrence detection. This is one of only a handful of studies to comprehensively evaluate the complementary roles, as opposed to the competitive roles, of both modalities in post-hepatectomy surveillance.

The 36.6% recurrence rate within 24 months is consistent with previous reports[13,14] and is in line with the prevailing view that intrahepatic metastasis underlies early recurrence after partial liver resection[2,15-17]. The detection rates of US (61.8%) and CE-MRI (92.6%) reveal the performance gap between the individual modalities, whereas the combined approach achieved a detection rate of 97.1%[15]. The only 2 cases missed by the combined approach were subcentimeter lesions at the liver dome with marked respiratory motion artifacts.

In the lesion size-stratified analysis, CE-MRI maintained high sensitivity (84.6%) for small (< 1 cm) lesions, whereas the sensitivity of US decreased to 38.5% (P < 0.001). US is limited by its resolution for deep or obscured small lesions, whereas CE-MRI utilizes the hepatobiliary phase of gadoxetate disodium, in which small HCC lesions appear hypointense against the hyperintense background liver parenchyma[7,8]. For lesions measuring 1-2 cm, CE-MRI achieved a sensitivity of 89.3% compared with only 60.7% for US (P < 0.001). This represents a clinically important window because lesions of this size are often amenable to curative ablation therapies.

Location-based analysis indicated superior US performance for superficial lesions (88.9% vs 72.2%), attributable to the excellent near-field resolution of US, whereas CE-MRI was clearly superior for deep lesions (95.2% vs 52.4%), reflecting its superior tissue penetration and multiplanar capabilities[18,19]. Lesions in the liver dome were detected in only 42.9% of patients by US, which significantly improved to 92.9% with CE-MRI (P = 0.003)[6,22]. The sensitivity for caudate lobe lesions was 37.5% vs 87.5% (P = 0.021), reflecting well-known anatomical challenges. Eleven (16.7%) recurrences were detected by US before CE-MRI, including superficial subcapsular lesions; however, this first-detection advantage must be interpreted in the context of the more frequent US surveillance interval used in this study and should not be taken as evidence of true modality superiority. More frequent US examinations may increase surveillance density while avoiding the cost burden of additional MRI examinations, but modality-specific diagnostic performance must be evaluated independently of surveillance frequency.

In terms of vascular invasion detection (portal vein tumor thrombus and hepatic vein invasion), which guides treatment decisions[20,21], CE-MRI had a clear advantage, with higher detection rates than US (93.8% vs 62.5%, P = 0.012; 100% vs 40.0%, P = 0.025). A major limitation of color Doppler US is its reduced sensitivity for partial or early-stage vascular invasion and its inability to reliably differentiate tumor thrombus from bland thrombus[22,23]. In addition, CE-MRI was superior in detecting microvascular invasion (86.4% vs 45.5%; P = 0.002), an independent predictor of poor recurrence-free survival that contributes to prognostic assessment[20,21]. As described in the American College of Radiology Liver Imaging Reporting and Data System, the combination of arterial phase hyperenhancement with washout on the portal venous or delayed phase allows confident noninvasive diagnosis of HCC[4,24,25].

Although CE-MRI is superior to US, US offers several practical and economic advantages, including its noninvasive nature, absence of ionizing radiation, wide availability in most hospitals, lower cost per examination, and real-time imaging capability[22,23]. CE-MRI is reported to be 5-6 times more expensive than US in the United States, with longer scheduling wait times. US also allows immediate image-guided biopsy or ablation planning and is better tolerated by older or claustrophobic patients, leading to improved surveillance adherence—a critical factor for program effectiveness.

Based on these findings, we suggest a risk-stratified surveillance approach in which standard-risk patients undergo alternating US every 3 months with CE-MRI every 6 months during the first 2 years, whereas high-risk patients (those with microvascular invasion, AFP > 200 ng/mL, multiple tumors, or tumor size > 5 cm) undergo US every 3 months combined with CE-MRI every 3-4 months during the first year, followed by every 6 months during the second year[24,25]. Any indeterminate US finding should prompt earlier CE-MRI rather than waiting for the next scheduled examination. The preoperative AFP level was significantly higher in the recurrence group than in the non-recurrence group (median, 298.4 ng/mL vs 142.8 ng/mL; P = 0.018), suggesting that monitoring AFP trajectories may help identify patients who would benefit from earlier imaging in future surveillance strategies.

This risk-stratified strategy seeks to optimize the breadth of diagnostic assessment while minimizing resource use and patient burden. The subgroup of high-risk patients in our study, defined by variables including microvascular invasion (41.2% in the recurrence group vs 20.3% in the non-recurrence group, P < 0.05), would particularly benefit from more intensive MRI follow-up because these patients have a higher likelihood of developing small, multifocal, or vascular-invasive recurrences that are best detected by this technique[3]. Moreover, any equivocal finding on US should prompt expedited CE-MRI rather than waiting for the next scheduled MRI. This approach takes advantage of the complementary nature of the dual-modality strategy and minimizes delays in diagnosing potentially treatable recurrences.

In this cohort, preoperative AFP was significantly elevated in the recurrence group (median, 298.4 ng/mL vs 142.8ng/mL;P = 0.018), a finding that underscores the potential value of biomarker-imaging integration. However, detailed investigation of AFP trajectory monitoring or novel biomarkers as triggers for intensified imaging is beyond the scope of the present study and will be addressed in future investigations.

Limitations of the current study include its retrospective design, which inherently confers selection and information biases; the use of data from a single center, which may limit generalizability to institutions with different equipment, expertise, and patient populations; the lack of a cost-effectiveness analysis, which would be important for healthcare policy recommendations; and the modest sample size in subgroup analyses, particularly for vascular invasion subtypes and small lesion categories. Furthermore, our study did not assess the role of contrast-enhanced US, which has emerged as an important diagnostic tool for bridging part of the imaging gap between conventional US and CE-MRI. The lack of contrast-enhanced US in our protocol is both a limitation and an opportunity for further investigation. In addition, although inter-reader variability for US examinations was not formally evaluated in this retrospective study, the high operator dependency of US[22-25] is an important confounder that could limit the generalizability of our sensitivity estimates. Importantly, US examinations in this study were performed by experienced sonographers using high-end equipment, and the observed sensitivity of 61.8% may therefore overestimate real-world performance in community settings with less specialized personnel. Conversely, the potentially higher sensitivity observed with CE-MRI under controlled institutional conditions may complement conventional US in clinical practice, particularly in centers where US performance is limited by operator experience or equipment quality.

Future studies should involve larger multicenter prospective trials to confirm our results and assess the effect of the composite surveillance strategy on survival outcomes. Cost-effectiveness analyses of alternative surveillance protocols will provide important evidence for better utilization of healthcare resources. The incorporation of emerging techniques such as contrast-enhanced US, abbreviated MRI protocols, and artificial intelligence-based diagnostic tools represents a promising approach to further enhance the efficiency and accuracy of surveillance[26-30]. More specifically, shortened MRI protocols that reduce cost and scan time while preserving diagnostic performance for HCC detection may make MRI-based surveillance more affordable and feasible[9,19,30]. Artificial intelligence-based tools for automated lesion detection on both US and MRI have shown promising results in preliminary studies and may reduce operator dependency for US while increasing the efficiency of MRI interpretation[27,29]. Longitudinal studies examining how earlier recurrence detection translates into treatment eligibility and improved survival would provide the ultimate clinical validation of optimized surveillance strategies.

CONCLUSION

Conventional US and CE-MRI provide unique yet complementary diagnostic advantages in early HCC recurrence surveillance. Although CE-MRI offers superior overall accuracy, particularly for detecting small, deep, and vascular-invasive recurrences, US retains important advantages in terms of cost-effectiveness, accessibility, and detection of superficial lesions. The combined use of both modalities achieved a diagnostic accuracy and detection rate exceeding 97%. An optimal surveillance strategy should therefore incorporate both modalities according to individual patient risk profiles. Such an evidence-based approach may improve early detection and facilitate timely treatment. The combined use of US and CE-MRI provides compelling preliminary evidence supporting a complementary dual-modality strategy for post-hepatectomy HCC surveillance; however, larger multicenter prospective trials are needed before incorporation into routine clinical practice.