Comparative analysis of efficient deep learning models for breast cancer identification using relevant genes

Abstract

Cancer remains a formidable global health challenge, with early detection critical in improving patient outcomes. In this context, applying deep learning techniques to relevant gene analysis has emerged as a promising avenue for enhancing cancer detection and diagnosis. This study presents an investigation into utilizing dimensionality reduction methods and deep learning techniques to analyze gene sequences with the primary aim of detecting breast cancer. Various dimensionality reduction techniques reduce the amplitude by selecting a subset of relevant characteristics or variables from a larger collection of available features. The selected relevant features are then transformed into meaningful representations obtained using SDAE (Stacked Denoising Autoencoder) which are used to familiarize the data with different noise and outliers. Various expert learning architectures have been studied to evaluate the effectiveness of compact functions resulting from SDAE transforms. In addition, we apply discrete classification algorithms such as SVM, ANN, and SVM-RBF to distinguish between cancer and normal samples. The primary objective is cancer detection, encompassing the identification of breast cancer in its early stages, the recurrence of cancer, and the pathological response based on the genetic data. The study contributes to the growing body of knowledge in bioinformatics and medical research. It holds the potential to translate its findings into practical clinical applications, thereby advancing our ability to combat this devastating disease. This work represents a significant step towards realizing more effective and precise cancer diagnostics, offering hope for improved patient care and outcomes in the fight against cancer.