Integrating Deep Learning Models and Swarm Intelligence Algorithms for Lung Disease Detection

Lung diseases have a substantial worldwide impact, contributing significantly to morbidity and mortality. Lung diseases can vary widely in severity and impact, ranging from mild and manageable conditions to severe, life-threatening illnesses. Lung diseases constitute almost 16 % of the most common causes of death globally. So early detection is the need of the hour. A comprehensive dataset of X-Rays and CT-Scans is formed which constitutes a Chest X-ray NIH dataset of 14 diseases along with tuberculosis, COVID-19, and CT-Scans of cancer, COVID-19 and normal. The proposed methodology integrates the power of pre-trained models for feature extraction, followed by the application of swarm intelligence algorithms, namely Monarch Butterfly Optimization, Cuckoo Search Optimization and Elephant Herding Optimization for optimization purposes. MobileNetv2 has been taken as the base model for fine-tuning. The best model with the best results is taken in the end. Through Monarch Butterfly, the best model for CT-Scans has an accuracy of 97.20% and through Cuckoo Search, the best model for X-Rays has an accuracy of 96.15%. By addressing the critical importance of timely diagnosis, the system helps the society to mitigate the global burden of lung diseases and improve overall public health outcomes.