Machine learning model using immune indicators to predict outcomes in early liver cancer

Area of Science:

• Hepatocellular Carcinoma Research
• Machine Learning in Oncology
• Biomarker Discovery

Background:

• Early-stage hepatocellular carcinoma (HCC) patients often have good outcomes after surgery, but some experience recurrence within five years.
• Identifying patients at high risk of mortality post-surgery is crucial for personalized treatment strategies.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting mortality risk in early-stage HCC patients.
• To compare the performance of ML models against traditional scoring systems for HCC prognosis.

Main Methods:

• A cohort of 808 early-stage HCC patients was randomly assigned to training (2:1) and validation sets.
• Prognostic models were built using random survival forests and artificial neural networks (ANNs).
• A decision-tree model assessed the impact of immune-inflammatory markers on long-term outcomes.

Main Results:

• Immune-inflammatory markers, albumin-bilirubin scores, alpha-fetoprotein, tumor size, and INR were significant prognostic factors.
• The ANN model achieved a 5-year AUC of 0.85 in the training set and 0.82 in the validation set.
• The ANN model effectively stratified patients into high-risk and low-risk groups with a significant survival difference (HR 7.98, P < 0.0001).

Conclusions:

• A non-invasive, cost-effective ML-based model can identify high-risk early-stage HCC patients.
• This model aids clinicians in predicting poor postoperative prognosis after surgical resection.
• The developed ANN model offers a valuable tool for risk stratification and management of HCC.