AI-Based Prediction of HCC Recurrence Patterns After Resection (APAR)
Trial statusRecruiting
Trial phaseNot listed
Trial typeObservational
Biological sexAll
Age18-75
SponsorTongji Hospital
About this trial
This observational study aims to validate a deep learning model for predicting aggressive recurrence patterns in patients with early-stage liver cancer (HCC) after surgery.
The main question it aims to answer is: Can the AI model accurately identify patients at high risk of cancer recurrence within 2 years after surgery? Participants will provide clinical data and undergo standard surgery, followed by 2-year imaging surveillance. Their data will be used for both AI prediction and validation of recurrence patterns.
Eligibility criteria
This trial does not accept healthy volunteersQualifiers
Aged 18-75 years, regardless of gender.
BCLC stage 0-A, scheduled for curative liver resection.
Preoperative clinical diagnosis of hepatocellular carcinoma (HCC).
Availability of dynamic contrast-enhanced MRI within 1 month before surgery, with acceptable image quality.
Disqualifiers
Concurrent other malignancies (except cured non-melanoma skin cancer or cervical carcinoma in situ).
Postoperative pathology confirms non-HCC diagnosis.
Pregnant or lactating women.
History of organ transplantation.
Trial population
This study will enroll patients with early-stage hepatocellular carcinoma (BCLC 0-A) scheduled for curative liver resection at tertiary academic medical centers in China. Participants will be consecutively recruited from hepatobiliary surgery clinics, with preoperative MRI and postoperative pathology confirmation of HCC. The population reflects real-world clinical practice, including both surgery-only patients and those receiving neoadjuvant/adjuvant therapies per physician discretion.
Trial design
Design model
Cohort
Time perspective
Prospective
Treatments tested in this trial
Curative liver resection
Procedure/SurgeryStandard radical hepatectomy performed according to 2024 HCC guidelines. No neoadjuvant or adjuvant therapies administered. Follows institutional surgical protocols for BCLC 0-A HCC.
Real-world multimodal therapy
Procedure/SurgeryCurative resection combined with clinically indicated therapies (e.g., TACE, targeted drugs, immunotherapy) as per treating physician's decision. Treatments recorded but not protocol-mandated.
Treatment groups
353 Participants
are divided into 2 treatment groupsGroup A: Surgery-Only Validation Cohort1 intervention
Group B: Exploratory Treatment Cohort1 intervention
Trial outcomes
Primary outcomes
1
Accuracy of AI Model in Predicting Aggressive HCC Recurrence (AUC)
The area under the receiver operating characteristic curve (AUC) of the multimodal deep learning model (PRE/POST) for predicting postoperative recurrence beyond Milan criteria within 2 years after resection, validated against actual imaging/histopathology-confirmed recurrence patterns. Unit : Dimensionless (0-1)
Time frame
2 years post-surgerySecondary outcomes
1
Recurrence-Free Survival (RFS)
Time from surgery to first radiologically confirmed recurrence (any pattern) or death from any cause, analyzed by Kaplan-Meier method and compared between model-predicted high/low-risk groups. Unit : Months
Time frame
Up to 3 years2
Overall Survival (OS)
Time from surgery to death from any cause, compared between patients stratified by AI model predictions (high-risk vs. low-risk) and treatment cohorts (surgery-only vs. real-world therapy). Unit : Months
Time frame
Up to 5 yearsOther outcomes
1
Therapeutic Efficacy in Exploratory Cohort
Objective response rate (ORR) and RFS/OS benefits of neoadjuvantin model-predicted high-risk patients, assessed descriptively (non-randomized comparison). Unit : Percentage (%)
Time frame
Up to 1 yearsSponsors and contacts
Click on the lead sponsor to view all of their trials.
Source ClinicalTrials.gov