Analysis of Logistic Regression and CNN Using Parallel Computing for Chest X-ray Pneumonia Detection

Pneumonia, a severe respiratory infection, remains a leading global cause of death, particularly among vulnerable populations such as children and the elderly, underscoring the importance of early and accurate diagnosis. This study evaluates the performance of Logistic Regression (LR) and Convolutional Neural Networks (CNN) for classifying pneumonia using chest X-ray images. It also investigates the role of parallel processing techniques such as OpenMP for CPU parallelism and CUDA for GPU parallelism in accelerating model training and inference. Logistic Regression, known for its computational simplicity, achieves moderate performance gains with OpenMP. In contrast, CNNs, which are more suited for image-based tasks, demand higher computational resources but benefit significantly from CUDA, with substantial reductions in training time. Using a publicly available chest X-ray dataset, the models are assessed based on classification accuracy, training time, and computational efficiency, both with and without parallelization. The results indicate that CNNs outperform Logistic Regression in accuracy, while CUDA-enabled parallelism greatly enhances CNN training efficiency, highlighting the balance between computational demands and diagnostic performance in medical image classification.