Lung cancer is among the major causes of cancer-related deaths in the world, and the problem requires proper and timely diagnostic measures. The proposed study is HyMoIF-Net, a new radiogenomic deep learning framework that combines computer tomography (CT) imaging characteristics with gene expression data to classify lung cancer. TCIA provides CT images and corresponding genomic profiles are obtained in TCGA-LUAD and TCGA-LUSC. The convolutional neural network is used to extract spatial features of CT scans whereas a fully connected network is used to process gene expression features. The fusion mechanism that is presented is a cross-attention fusion mechanism that is effective in integrating multimodal representations. Empirical testing on cohorts of patients proves that the given model can attain a better classification outcome in contrast to unimodal methods. The findings demonstrate the possibility of radiogenomic fusion usage to improve the quality of the diagnosis and assist with individual treatment planning. The outcome obtained supports that the hybrid approach introduced enhances diagnostic performance of conventional systems with an accuracy rate up to 96.3% and provides a rich resource for lung cancer early detection, staging and personalized treatment planning in future.
Rosy et al. (Fri,) studied this question.