Enhancing Health Care with Optimized Computational Models for Disease Diagnosis Using Gene Expression Dataset

Advancements in microarray technology allow scientists to study thousands of genes, offering valuable insights into cellular functions and aiding in the diagnosis and prediction of diseases like cancer, diabetes, and heart disease. However, the complexity of gene data, characterized by numerous variables, noise, and limited samples, poses challenges in identifying crucial genes efficiently. To streamline this process and facilitate cost-effective testing, a specialized Rider-Chicken Optimization (RCO) algorithm was employed to construct a cancer diagnosis model. The selection of genes for this model is based on their relevance to understanding cancer. Subsequently, a Recurrent Neural Network (RNN) was trained using the Rider-Chicken Optimization (RCO) for diagnosis, achieving an accuracy rate of 95.1%, a maximum sensitivity of 95.03%, and a maximum specificity of 95.26%. Expanding its scope, the research explores diabetes prediction through an innovative Invasive Weed Bird Swarm Optimization Algorithm (IWBSOA). This unique algorithm integrates strategies from Enhanced Invasive Weed Optimization (IWO) and Bird Swarm Algorithm (BSA) and, when coupled with deep learning classifiers like Recurrent Neural Networks (RNN) and Support Vector Machines (SVM), yields a hybrid model with an impressive accuracy rate of 96.2%, a sensitivity of 97.1%, and a specificity of 94.4%. In the domain of heart disease detection, an effective model is introduced, utilizing the Political Deer Hunting Optimization (PDHO) algorithm-driven Deep Q-Network. PDHO is a fusion of the Political Optimizer (PO) and Deer Hunting Optimization (DHO) algorithms. The Deep Q-network, rooted in PDHO, achieves a maximum accuracy of 93.4%, a sensitivity of 96.2%, and a specificity of 89.2%. These results emphasize the capability of sophisticated algorithms to improve diagnostic accuracy across a range of medical scenarios.