Application of Deep Learning for Classification of CT Images in Order to Predict EGFR and KRAS Mutations in Oncology

This study explores the application of advanced deep learning models to classify lung cancer tumors based on CT images into three groups: mutant EGFR, mutant KRAS, and wildtype. Utilizing a comprehensive radiogenomic dataset from 211 patients, we developed models using YOLOv10 and convolutional neural networks like ResNet and DenseNet to analyze and predict genetic mutations directly from CT scans. The results demonstrate high precision and recall in mutation detection, particularly for EGFR, with the potential to significantly enhance non-invasive diagnostic methods. This approach promises to improve clinical decision-making and treatment planning by providing rapid, accurate assessments of mutation status, thereby facilitating personalized therapy strategies in oncology.