The Application of Machine Learning in the Diagnosis of Lung Cancer

Abstract

 Lung cancer remains the main cause of cancer-related incidence and mortality globally, with more than 2,220,000 newly diagnosed cases annually and a five-year survival rate of less than 25%. The complexity of diagnosis and treatment is exacerbated by Intra-tumoral Heterogeneity (ITH), which drives therapy resistance. Recent advances in the field of Machine Learning (ML) and Deep Learning (DL) offer promising solutions by enabling the analysis of high-dimensional medical data beyond human capability. This review explores the applications of ML in lung cancer diagnosis, focusing on Deep Neural Network (DNN), Convolutional Neural Network (CNN), and Vision Transformer (ViT)-based models across radiomics, histopathology, and gene expression analysis. Innovative techniques such as semi-supervised learning, data augmentation, and optimization algorithms have enhanced model performance, achieving high accuracy in classifying lung cancer subtypes and predicting genetic mutations. Federated learning emerges as a privacy-preserving approach for collaborative training across institutions, addressing critical data security concerns. However, significant challenges remain, including limited model interpretability, generalizability across diverse populations, and integration into clinical workflows. Future research should prioritize interpretable Artificial Intelligence (AI) frameworks and privacy-preserving technologies to enable earlier diagnosis and tailored therapies for lung cancer patients.