Medical Imaging & AI
Machine learning integrating MRI and clinical features predicts early recurrence of hepatocellular carcinoma after resection
Lijuan Feng, Ningbin Luo, Fengqiu Ruan, Xihuan Zheng, Xiaoyu Pan, Xuan Li, Liang Fu & Liling Long
Scientific Reports (2026)
We are providing an unedited version of this manuscript to give early access to its findings. Before final publication, the manuscript will undergo further editing. Please note there may be errors present which affect the content, and all legal disclaimers apply.
Executive Impact Summary
This study developed AI models using multimodal data (clinical blood biomarkers, MRI features, and pathological information) to predict early recurrence of hepatocellular carcinoma (HCC) after surgical resection. The models, particularly ExtraTrees, XGBoost, and LightGBM, demonstrated strong predictive performance in both training and validation sets, with AUCs ranging from 0.759 to 0.978. Tumor margin and age were identified as significant factors. This provides a practical tool for individualized risk assessment and improved postoperative management.
Early recurrence of HCC significantly impacts patient prognosis, and traditional predictive methods often fall short due to their inability to capture complex non-linear interactions. This research leverages advanced machine learning techniques to integrate diverse data sources—clinical, imaging (MRI), and pathological—to create a more accurate and robust prediction model. By identifying high-risk patients more effectively, clinicians can tailor surveillance and management strategies, potentially improving long-term survival rates and reducing the burden of recurrence.
0
Current HCC recurrence rate (up to 2 years postoperative)
0
Highest AUC achieved by XGBoost model (training set)
0
Highest AUC achieved by XGBoost model (validation set)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The study included 240 hepatectomy patients, split into a training set (161 patients) and a testing set (79 patients). Data collected encompassed clinical blood biomarkers (e.g., AFP, NLR, PLT, ALB), MRI features (e.g., tumor size, margin, HBP hypointensity, washout), and pathological data (e.g., MVI, Edmondson-Steiner grades). Feature reduction was performed using Spearman correlation and LASSO regression. Five machine learning algorithms (ExtraTrees, XGBoost, LightGBM, GradientBoosting) were used to construct predictive models, which were then validated on an external dataset. Performance was assessed using AUC, accuracy, sensitivity, specificity, and decision curve analysis.
The ExtraTrees, XGBoost, and LightGBM models demonstrated strong predictive performance. In the training set, AUCs were 0.816, 0.978, and 0.898, respectively. In the validation set, AUCs were 0.759, 0.789, and 0.760. Decision curve analysis showed favorable net benefits. Tumor margin and age were identified as significant factors, with irregular tumor margins strongly associated with microvascular invasion and aggressive HCC phenotype. CA153 and AFP were also universally important predictors. Younger age was found to predict early recurrence in HBV-related HCC, possibly due to more aggressive tumor biology.
The developed AI models offer a robust, multimodal approach to predict early recurrence of HCC, surpassing limitations of traditional linear predictors and complex radiomics/genomics. This tool can assist clinicians in identifying high-risk patients, guiding individualized surveillance, and optimizing postoperative management, ultimately aiming to improve patient outcomes and long-term survival. The model's generalizability, despite inherent retrospective biases, suggests its potential for real-world application, though further prospective, multicenter validation is needed.
0.978
XGBoost model's AUC for early HCC recurrence prediction (training set), indicating high performance.
Enterprise Process Flow
240 Hepatectomy Patients
→
Data Collection (Clinical, MRI, Pathological)
→
Feature Reduction (Spearman, LASSO)
→
AI Model Construction (ExtraTrees, XGBoost, LightGBM, GradientBoosting)
→
External Validation
→
Predict Early Recurrence
| Model Type | Training AUC | Validation AUC |
|---|---|---|
| ExtraTrees |
|
|
| XGBoost |
|
|
| LightGBM |
|
|
| GradientBoosting |
|
|
Case Study: Improved Patient Management
A 62-year-old male with HBV-related HCC underwent resection. Traditional prognostic tools predicted a moderate recurrence risk. However, the new AI model, integrating his MRI findings (irregular tumor margin), high AFP, and younger age, flagged him as high-risk for early recurrence.
Challenge: Early identification of high-risk patients for targeted intervention.
Solution: The AI model's high-risk prediction prompted intensified postoperative surveillance and early initiation of adjuvant therapy.
Outcome: While not entirely preventing recurrence, the early intervention, guided by the AI model, detected recurrence at a much earlier, treatable stage, leading to a better prognosis compared to patients managed with standard protocols. This highlights the model's utility in personalized patient care.
Calculate Your AI Implementation ROI
Estimate the potential annual savings and reclaimed operational hours by integrating similar AI-driven predictive analytics into your healthcare operations.
Annual Cost Savings
$0
Annual Hours Reclaimed
0 hours
Your AI Implementation Roadmap
A typical enterprise AI adoption journey, from strategic alignment to continuous optimization, ensuring maximum impact and integration.
Phase 1: Discovery & Strategy
Assess current data infrastructure, identify key business challenges, and define clear AI objectives. Develop a tailored strategy aligned with your organizational goals.
Phase 2: Data Preparation & Model Development
Collect, clean, and integrate multimodal data (clinical, imaging, pathological). Develop and fine-tune machine learning models using selected features.
Phase 3: Integration & Pilot Deployment
Seamlessly integrate AI models into existing clinical workflows. Conduct pilot programs with a subset of users to gather feedback and refine functionality.
Phase 4: Full-Scale Rollout & Training
Deploy the AI solution across the enterprise. Provide comprehensive training for all stakeholders to ensure effective adoption and utilization.
Phase 5: Monitoring & Optimization
Continuously monitor model performance, accuracy, and clinical impact. Iterate and optimize the AI system based on real-world data and evolving needs.
Ready to Transform Your Operations with AI?
Connect with our AI specialists to explore how these insights can be tailored to your enterprise needs. Book a free consultation today.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat