Investigating CNN Models Efficacy in Spotting Lung Conditions using X-Ray Images

Lung diseases significantly impact global health, necessitating early and accurate classification for effective treatment. However, manual interpretation of X-ray images is of- ten time-consuming and prone to human error, leading to delayed or inaccurate diagnoses. This study evaluates four deep learning models—EfficientNetB0, MobileNetV2, RegNet, and ResNet18—using the “Lung X-Ray Image Dataset” for three- class classification (Normal, Lung Opacity, and Viral Pneumonia). The primary objective is to determine the most effective model for automating lung disease classification, thereby improving diagnostic efficiency and reducing reliance on manual analysis. The models were trained for 15 epochs, with RegNet achieving the highest accuracy of 92.53%, followed by EfficientNetB0 (90.42%), ResNet18 (90.23%), and MobileNetV2 (85.25%). These findings highlight RegNet as the most suitable model for reliable lung disease classification, potentially aiding radiologists in faster and more accurate diagnosis.