A novel lightweight CNN for chest X-ray-based lung disease identification on heterogeneous embedded system
Theodora Sanida, Minas Dasygenis
Abstract
Abstract The global spread of epidemic lung diseases, including COVID-19, underscores the need for efficient diagnostic methods. Addressing this, we developed and tested a computer-aided, lightweight Convolutional Neural Network (CNN) for rapid and accurate identification of lung diseases from 29,131 aggregated Chest X-ray (CXR) images representing seven disease categories. Employing the five-fold cross-validation method to ensure the robustness of our results, our CNN model, optimized for heterogeneous embedded devices, demonstrated superior diagnostic performance. It achieved a 98.56% accuracy, outperforming established networks like ResNet50, NASNetMobile, Xception, MobileNetV2, DenseNet121, and ViT-B/16 across precision, recall, F1-score, and AUC metrics. Notably, our model requires significantly less computational power and only 55 minutes of average training time per fold, making it highly suitable for resource-constrained environments. This study contributes to developing efficient, lightweight networks in medical image analysis, underscoring their potential to enhance point-of-care diagnostic processes.