Machine Learning Model Based on Prognostic Nutritional Index for Predicting Long-Term Outcomes in Patients With HCC
Nan Zhang1, Ke Lin1, Bin Qiao1
1Division of Interventional Ultrasound, Department of Medical Ultrasonics, Institute of Diagnostic and Interventional Ultrasound, The First Affiliated Hospital of Sun Yat-sen University, Guangzhou, Guangdong, China.
Cancer Medicine
|October 23, 2024
Summary
A new machine learning model using the prognostic nutritional index (PNI) accurately predicts long-term survival in hepatocellular carcinoma (HCC) patients after ablation. This PNI-based Aorsf model offers improved strategies for HCC surveillance, prevention, and treatment.
Area of Science:
- Hepatocellular Carcinoma Research
- Machine Learning in Oncology
- Prognostic Factor Analysis
Background:
- Hepatocellular carcinoma (HCC) survival prediction after local ablation remains challenging.
- Identifying reliable prognostic markers is crucial for optimizing patient management.
- The prognostic nutritional index (PNI) is a potential indicator of nutritional status and inflammation, impacting cancer outcomes.
Purpose of the Study:
- To develop and validate machine learning (ML) models for predicting long-term survival in HCC patients post-local ablation.
- To identify the optimal ML model based on the prognostic nutritional index (PNI).
- To compare the performance of the developed ML model against existing prognostic tools.
Main Methods:
- Analysis of data from 848 primary HCC patients undergoing local ablation (2009-2019).
- Construction and evaluation of multiple ML models using concordance index (C-index), C/D AUC, and Brier scores.
- Interpretation of the optimal model using Partial Dependence Plots (PDP) and SHapley Additive exPlanations (SHAP).
Main Results:
- Alkaline phosphatase, preoperation alpha-fetoprotein, PNI, tumor number, and size were key prognostic factors.
- The Aorsf model demonstrated superior performance in both training and validation cohorts (e.g., validation C-index: 0.793).
- The Aorsf model showed significant predictive accuracy for 1- to 9-year survival, outperforming other models in clinical utility.
Conclusions:
- The PNI-based Aorsf model is effective for predicting long-term survival in HCC patients after ablation.
- This model enhances HCC research by informing surveillance, prevention, and treatment strategies.
- The Aorsf model offers a valuable tool for personalized HCC patient management.
