An approach for robust and explainable lung disease detection and localization

Pulmonary diseases, including lung and respiratory diseases, are a significant global health issue, causing nearly 4 million deaths in 2019. Chronic obstructive pulmonary disease is the third leading cause of death worldwide, with lung cancer being the leading cause. In 2021, the EU’s respiratory disorders death rate was 65.5 deaths per 100,000 inhabitants. This study applies Deep Learning models to classify RX thoracic images of various lung diseases, achieving high accuracy values (97/98%). The study compares different architectures, including MobileNet and DenseNet, highlighting that while these models yield optimal quantitative results, they often produce less informative heatmaps. In contrast, the Standard_CNN model, despite slightly lower metrics, provides more localized heatmaps, which are crucial for clinical diagnosis. The research emphasizes the importance of visual explainability in automated diagnosis, utilizing CAM algorithms to provide visual insights into model decision-making, which clinicians evaluated. Additionally, the paper introduces index measures as tools to quantify qualitative outcomes, enhancing the robustness and trustworthiness of the models.