Immunotherapy Efficacy Prediction for Non-Small Cell Lung Cancer Using Multi-View Adaptive Weighted Graph
IEEE Journal of Biomedical and Health Informatics
|August 29, 2023
Summary
Predicting non-small cell lung cancer (NSCLC) immunotherapy efficacy is crucial. A novel multi-view adaptive weighted graph convolutional network (MVAW-GCN) effectively predicts treatment response using radiomic and phenotypic data.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Immunotherapy is a key treatment for non-small cell lung cancer (NSCLC), but efficacy varies significantly among patients.
- Predicting immunotherapy response before surgery is vital to optimize treatment and manage potential side effects.
- Current radiomics approaches often overlook inter-patient correlations and complex multi-type feature interactions.
Purpose of the Study:
- To develop and evaluate a novel multi-view adaptive weighted graph convolutional network (MVAW-GCN) for predicting NSCLC immunotherapy efficacy.
- To integrate radiomic features from different image types and phenotypic information within a graph convolutional network framework.
- To address limitations of single-view models by considering correlations among multiple feature types and inter-patient relationships.
Main Methods:
- Radiomic features were grouped into distinct views based on filtered image types.
- Graphs were constructed within each view using radiomic features and patient phenotypic data.
- A multi-view adaptive weighted graph convolutional network (MVAW-GCN) with an attention mechanism and separable graph convolution was proposed to model view correlations and feature interactions.
Main Results:
- The MVAW-GCN model was evaluated on a cohort of 107 NSCLC patients.
- The proposed method achieved a prediction accuracy of 77.27% and an AUC of 0.7780 for immunotherapy efficacy.
- These results demonstrate the model's effectiveness in predicting treatment response.
Conclusions:
- The MVAW-GCN model offers a promising approach for predicting NSCLC immunotherapy efficacy by leveraging multi-view radiomic and phenotypic data.
- The model's ability to capture inter-view correlations and feature interactions enhances prediction performance.
- This approach could aid in personalized treatment strategies for NSCLC patients undergoing immunotherapy.
