MULTIMODEL DEEP LEARNING MODEL INTEGRATING CNN AND TRANSFORMER FOR PREDICTING FOR PREDICTING CH

Abstract

Chemotherapy-induced peripheral neuropathy (CIPN) is a major side effect of cancer treatment that can significantly affect patient quality of life and lead to treatment interruptions. This project proposes a Transformer-based multimodal deep learning framework to predict the risk of CIPN at an early stage. The system integrates multiple healthcare data sources, including clinical records, imaging data, wearable sensor information, and genomic features, to improve prediction accuracy. Advanced preprocessing and feature extraction techniques are used to handle heterogeneous data efficiently. The Transformer model captures complex relationships among different modalities and provides better performance compared to conventional machine learning and deep learning approaches. Experimental results show high accuracy, sensitivity, specificity, and AUC, demonstrating the effectiveness of the proposed system. Explainable AI techniques such as SHAP and Grad-CAM are incorporated to improve transparency and help identify key risk factors influencing prediction. This approach supports early intervention, personalized treatment planning, and continuous patient monitoring, making it a promising solution for precision oncology and improved cancer care.