Leveraging Radiomics and Genetic Algorithms to Improve Lung Infection Diagnosis in X-Ray Images Using Machine Learning

Radiomics, an emerging discipline in medical imaging, focuses on extracting detailed quantitative features from images to unveil subtle patterns imperceptible to the naked eye. This study specifically employs radiomics and machine learning techniques to discern cases of viral pneumonia and COVID-19. By harnessing intricate radiomic features derived from medical images, the objective is to train a machine learning model capable of accurately distinguishing between patients with viral pneumonia, COVID-19, and those unaffected. To optimize the performance of machine learning models, the paper incorporates genetic algorithms for hyperparameter optimization. A comparative analysis is conducted among the genetic algorithm-based TPOT (Tree-based Pipeline Optimization Tool) settings, namely TPOT-Default, TPOT-Light, and TPOT-Sparse, to select the most effective hyperparameters. Custom modifications are introduced to the TPOT model to align it with the specific requirements of the current model, resulting in a noteworthy achievement. The proposed model attains a remarkable 94% AUC (Area Under the Curve) when employing the Random Forest algorithm. Furthermore, the study systematically evaluates the execution time taken by each TPOT model. The model's performance is comprehensively assessed through key metrics, including accuracy, precision, sensitivity, specificity, and F1-score. The techniques suggested in this article could aid radiologists in identifying anomalies in chest X-ray (CXR) images, offering a more accurate and efficient interpretation to improve medical decision-making.