Construction and Clinical Validation of a Predictive Model for Postoperative Adjuvant Therapy in Hepatocellular Carcinoma Based on Whole-Slide Digital Pathological Images and Deep Learning

NCT07417800 | Recruiting DMC

• 1 other identifier: 2025-1125
• Study Type: observational
• Target: 11,000
• Locations: 1 country
• Sites: 1

Brief Summary

Hepatocellular carcinoma (HCC) is a high-mortality global malignancy with a heavy disease burden in China. Although curative surgical resection improves survival for early-stage HCC patients, the 5-year postoperative recurrence rate remains as high as 50%-70%. Postoperative adjuvant TACE and systemic TKIs are standard treatments for high-risk HCC, yet both therapies have prominent drawbacks, including limited response rates, unavoidable toxicities, and inconsistent clinical benefits. Current treatment decisions rely on conventional clinical and pathological features without precise biomarkers, leading to inadequate individualized therapy and wasted medical resources. Tumor immune microenvironment and multimodal imaging-pathological features critically determine HCC treatment sensitivity. Artificial intelligence and deep learning based on preoperative radiomics and postoperative H\&E whole-slide imaging (WSI) can capture hidden tumor biological characteristics and predict therapeutic responses. However, no validated multimodal AI model is available for predicting postoperative TACE and TKI treatment outcomes in HCC, lacking large-scale multicenter prospective evidence. This study aims to construct and validate a multimodal deep learning model integrating preoperative contrast-enhanced CT/MRI, postoperative WSI, pathological reports, and clinical data, to precisely identify HCC patients sensitive to postoperative adjuvant TACE or TKI therapy and optimize individualized treatment strategies. This is a hybrid retrospective-training and prospective observational multicenter study with no clinical intervention. A total of 10,000 retrospective HCC surgical patients will be enrolled to develop an AI classification model for predicting responses to four postoperative treatment strategies: surgery alone, surgery plus TACE, surgery plus TACE combined with systemic therapy, and surgery plus exclusive systemic therapy. Subsequently, 1,000 eligible postoperative HCC patients will be prospectively and consecutively enrolled from 10-15 centers. The AI model will generate adjuvant therapy predictions without interfering with real clinical decisions. Patients will be divided into prediction-consistent and prediction-inconsistent cohorts based on the match between model predictions and actual treatments. Long-term follow-up will be performed to compare prognostic outcomes and validate the model's real-world performance and stability. Key inclusion criteria: histopathologically confirmed HCC; aged 18-75 years; received R0 curative resection; available qualified H\&E-stained FFPE slides for digital scanning; complete clinical, pathological and follow-up data; high-quality preoperative contrast-enhanced CT/MRI images eligible for AI analysis. Key exclusion criteria: prior preoperative anti-tumor therapy with unavailable baseline data; concurrent other primary malignancies; non-R0 resection; unqualified pathological slides or imaging data; severe missing clinical or follow-up information.

Conditions

Hepatocellular Carcinoma (HCC); Artificial Intelligent; Adjuvant Chemoradiotherapy; TACE; Lenvatinib; Liver Surgery

Keywords

Hepatocellular Carcinoma (HCC); Artificial Intelligent; TACE; lenvatinib; Adjuvant Chemoradiotherapy

Outcome Measures

Primary Outcomes (2)

• recurrence free survival

  Recurrence-Free Survival (RFS) refers to the length of time from the completion of curative hepatectomy for hepatocellular carcinoma (such as hepatectomy or liver transplantation) until the first documented recurrence of the tumor or the patient's death from any cause, whichever occurs first.

  Up to 3 years after curative hepatectomy

• recurrence rate

  rate of recurrence after the surgery

  up to 3 years after the surgery

Secondary Outcomes (1)

• overall survival

  Up to 5 years after curative hepatectomy

Study Arms (3)

Retrospective Cohort for Deep Learning Model Construction

This cohort is a large-scale retrospective observational cohort enrolled primarily for the construction, feature screening, and preliminary internal verification of the deep learning predictive model. A total of approximately 10,000 postoperative hepatocellular carcinoma patients with contrast-enhanced computed tomography (CT) / magnetic resonance imaging ,clinial data, pathological, treatment, and follow-up data will be included. All enrolled subjects received standard surgical resection for HCC and completed standardized postoperative follow-up in participating centers. No trial-related intervention is imposed on patients. Core clinical endpoints include postoperative tumor recurrence time, recurrence pattern, overall survival, and disease-free survival. All real-world data of this cohort will be used to train, optimize, and calibrate the AI model to identify high-risk recurrence populations and generate individualized postoperative adjuvant therapy prediction schemes.

Prospective Consistent Adjuvant Therapy Cohort (AI Prediction-Matched Actual Treatment)

This is a prospective observational cohort consisting of postoperative HCC patients whose clinically implemented adjuvant therapy regimens are completely consistent with the individualized adjuvant therapy schemes predicted by the validated deep learning model. All subjects undergo routine curative resection and receive standardized postoperative management in strict accordance with clinical guidelines. The AI model only provides predictive treatment recommendations without forcing or intervening clinical decision-making, and the final treatment plan is independently determined by attending physicians. This cohort mainly verifies the clinical accuracy and practical value of the AI model. Long-term follow-up will be performed to record tumor recurrence, metastasis, survival status and adverse reactions, aiming to confirm that AI-matched adjuvant therapy can effectively reduce postoperative recurrence and improve long-term prognosis of HCC patients.

Other: AI adjuvant therapy

Prospective Inconsistent Adjuvant Therapy Cohort (AI Prediction-Mismatched Actual Treatment)

This is a prospective observational cohort composed of postoperative HCC patients whose actual clinical adjuvant therapy regimens are inconsistent with the optimal adjuvant therapy schemes predicted by the deep learning AI model. All enrolled patients meet the surgical resection indications for HCC and receive conventional postoperative clinical management, with all treatment decisions made by clinicians based on traditional clinical experience, guidelines and individual patient conditions, free from any mandatory intervention of the AI model. Through long-term real-world follow-up of tumor recurrence, disease-free survival and overall survival of patients in this cohort, the study aims to quantitatively compare the prognostic differences between AI-predicted optimal treatment schemes and conventional empirical treatment schemes, further validate the clinical guiding significance and superiority of the AI predictive model for HCC postoperative adjuvant therapy.

Other: AI adjuvant therapy

Interventions

AI adjuvant therapy OTHER

This is a purely observational study involving no clinical intervention. The multimodal AI model analyzes patients' preoperative imaging, postoperative digital pathological slides, and clinical indicators to predict HCC postoperative recurrence risk and optimal adjuvant therapy regimens. All AI outputs are used only for research recording and outcome comparison. No model predictions will affect physicians' real clinical decisions, treatment plans, or patient management throughout the study.

Prospective Consistent Adjuvant Therapy Cohort (AI Prediction-Matched Actual Treatment); Prospective Inconsistent Adjuvant Therapy Cohort (AI Prediction-Mismatched Actual Treatment)

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Healthy Volunteers: No
• Age Groups: Adult (18-64), Older Adult (65+)
• Sampling Method: Probability Sample

Study Population

The study population consists of patients histopathologically diagnosed with hepatocellular carcinoma who underwent curative R0 surgical resection. Eligible participants are aged 18-75 years with complete and high-quality preoperative and postoperative contrast-enhanced CT/MRI images, qualified postoperative H\&E-stained FFPE pathological slides, and complete clinicopathological and follow-up data. Patients with prior non-collaborative-center antitumor treatment, concurrent other malignancies, positive surgical margins, or severely unqualified imaging and pathological specimens are excluded.

You may qualify if:

• Histopathologically confirmed hepatocellular carcinoma;
• Aged more than 18 years;
• Underwent radical resection of primary liver cancer (R0 resection);
• Availability of postoperative H\&E-stained paraffin embedded tissue sections suitable for digital whole-slide imaging;
• Had complete and accessible clinicopathological data and follow-up data;
• Has complete and evaluable preoperative and postoperative contrast-enhanced CT or MRI imaging with standardized scanning parameters and no severe artifacts, meeting the quality requirements for radiomic and artificial intelligence analysis.

You may not qualify if:

• Significant missing clinical or follow-up data;
• Concurrent primary malignancy in other organs;
• Positive surgical margin (R1 or R2 resection);
• Tissue sections of poor quality (e.g., severe fading, folding, damage) unsuitable for digital scanning or analysis;

Sponsors & Collaborators

• Second Affiliated Hospital, School of Medicine, Zhejiang University: lead

Study Sites (1)

the Second Affiliated Hospital Zhejiang University School of Medicine

Hangzhou, Zhejiang, 310009, China

RECRUITING

Related Publications (8)

• Guo W, Li S, Qian Y, Li L, Wang F, Tong Y, Li Q, Zhu Z, Gao WQ, Liu Y. KDM6A promotes hepatocellular carcinoma progression and dictates lenvatinib efficacy by upregulating FGFR4 expression. Clin Transl Med. 2023 Oct;13(10):e1452. doi: 10.1002/ctm2.1452.

  PMID: 37846441 BACKGROUND

• Vayrynen JP, Lau MC, Haruki K, Vayrynen SA, Dias Costa A, Borowsky J, Zhao M, Fujiyoshi K, Arima K, Twombly TS, Kishikawa J, Gu S, Aminmozaffari S, Shi S, Baba Y, Akimoto N, Ugai T, Da Silva A, Song M, Wu K, Chan AT, Nishihara R, Fuchs CS, Meyerhardt JA, Giannakis M, Ogino S, Nowak JA. Prognostic Significance of Immune Cell Populations Identified by Machine Learning in Colorectal Cancer Using Routine Hematoxylin and Eosin-Stained Sections. Clin Cancer Res. 2020 Aug 15;26(16):4326-4338. doi: 10.1158/1078-0432.CCR-20-0071. Epub 2020 May 21.

  PMID: 32439699 BACKGROUND

• Vanguri RS, Luo J, Aukerman AT, Egger JV, Fong CJ, Horvat N, Pagano A, Araujo-Filho JAB, Geneslaw L, Rizvi H, Sosa R, Boehm KM, Yang SR, Bodd FM, Ventura K, Hollmann TJ, Ginsberg MS, Gao J; MSK MIND Consortium; Hellmann MD, Sauter JL, Shah SP. Multimodal integration of radiology, pathology and genomics for prediction of response to PD-(L)1 blockade in patients with non-small cell lung cancer. Nat Cancer. 2022 Oct;3(10):1151-1164. doi: 10.1038/s43018-022-00416-8. Epub 2022 Aug 29.

  PMID: 36038778 BACKGROUND

• Jia G, He P, Dai T, Goh D, Wang J, Sun M, Wee F, Li F, Lim JCT, Hao S, Liu Y, Lim TKH, Ngo NT, Tao Q, Wang W, Umar A, Nashan B, Zhang Y, Ding C, Yeong J, Liu L, Sun C. Spatial immune scoring system predicts hepatocellular carcinoma recurrence. Nature. 2025 Apr;640(8060):1031-1041. doi: 10.1038/s41586-025-08668-x. Epub 2025 Mar 12.

  PMID: 40074893 BACKGROUND

• Zeng Q, Klein C, Caruso S, Maille P, Laleh NG, Sommacale D, Laurent A, Amaddeo G, Gentien D, Rapinat A, Regnault H, Charpy C, Nguyen CT, Tournigand C, Brustia R, Pawlotsky JM, Kather JN, Maiuri MC, Lomenie N, Calderaro J. Artificial intelligence predicts immune and inflammatory gene signatures directly from hepatocellular carcinoma histology. J Hepatol. 2022 Jul;77(1):116-127. doi: 10.1016/j.jhep.2022.01.018. Epub 2022 Feb 7.

  PMID: 35143898 BACKGROUND

• da Fonseca LG, Reig M, Bruix J. Tyrosine Kinase Inhibitors and Hepatocellular Carcinoma. Clin Liver Dis. 2020 Nov;24(4):719-737. doi: 10.1016/j.cld.2020.07.012. Epub 2020 Sep 28.

  PMID: 33012455 BACKGROUND

• Nevola R, Ruocco R, Criscuolo L, Villani A, Alfano M, Beccia D, Imbriani S, Claar E, Cozzolino D, Sasso FC, Marrone A, Adinolfi LE, Rinaldi L. Predictors of early and late hepatocellular carcinoma recurrence. World J Gastroenterol. 2023 Feb 28;29(8):1243-1260. doi: 10.3748/wjg.v29.i8.1243.

  PMID: 36925456 BACKGROUND

• Hwang SY, Danpanichkul P, Agopian V, Mehta N, Parikh ND, Abou-Alfa GK, Singal AG, Yang JD. Hepatocellular carcinoma: updates on epidemiology, surveillance, diagnosis and treatment. Clin Mol Hepatol. 2025 Feb;31(Suppl):S228-S254. doi: 10.3350/cmh.2024.0824. Epub 2024 Dec 26.

  PMID: 39722614 BACKGROUND

Biospecimen

Retention: SAMPLES WITHOUT DNA

This study retrospectively and prospectively collects residual postoperative tumor specimens from patients with hepatocellular carcinoma (HCC) undergoing curative surgical resection. All biospecimens are routine formalin-fixed paraffin-embedded (FFPE) H\&E-stained pathological slides generated during standard postoperative clinical diagnosis. No additional biopsy, invasive sampling, or extra tissue collection is performed for this research, thus bringing no additional physical burden or safety risk to patients. All slides are prepared in accordance with standardized pathological protocols to ensure qualified tissue structure and staining quality. Qualified pathological slides are digitally scanned to obtain high-resolution whole-slide images (WSI).

MeSH Terms

Conditions

Carcinoma, Hepatocellular

Condition Hierarchy (Ancestors)

Adenocarcinoma; Carcinoma; Neoplasms, Glandular and Epithelial; Neoplasms by Histologic Type; Neoplasms; Liver Neoplasms; Digestive System Neoplasms; Neoplasms by Site; Digestive System Diseases; Liver Diseases

Study Officials

• ding yuan, doctor

  Second Affiliated Hospital, School of Medicine, Zhejiang University

  STUDY CHAIR

Central Study Contacts

ding yuan, doctor

CONTACT

+86 18858101960; dingyuan@zju.edu.cn

wang weilin, doctor

CONTACT

+86 13606642087; wam@zju.edu.cn

Study Design

• Study Type: observational
• Observational Model: COHORT
• Time Perspective: CROSS SECTIONAL
• Sponsor Type: OTHER
• Responsible Party: SPONSOR

Study Record Dates

• First Submitted: February 2, 2026
• First Posted: February 18, 2026
• Study Start: November 1, 2025
• Primary Completion (Estimated): January 1, 2029
• Study Completion (Estimated): December 1, 2029
• Last Updated: June 10, 2026

Record last verified: 2025-10

Data Sharing

• IPD Sharing: Will not share

Locations

CN (1)