Precise Lung Disease Classification Through Bagging Ensemble Method

Lung disease pose a significant global health challenge, with their prevalence and impact on public well-being necessitating advanced diagnostic methods for timely and accurate identification. As a vital organ for respiratory function, lungs are susceptible to an infinite number of conditions, ranging from infectious diseases like pneumonia and tuberculosis to chronic ailments such as chronic obstructive pulmonary disease (COPD) and lung cancer. The study proposes an innovative approach for accurate lung disease classification utilizing a Bagging ensemble technique (Bootstrap Aggregating). Three pre-trained models, namely Modified VGG-16, VGG-19, and MobileNet V2, are individually trained on a common computed tomography (CT) image dataset. The fine-tuned models predict class probabilities and the bagging ensemble technique amalgamates their predictions by calculating the average probability score. The input images are correctly classified into five distinct groups by the ensemble model such as Normal, Covid-19, Viral Pneumonia, Bacterial Pneumonia, and Tuberculosis with an impressive accuracy of 97.99%, aiming to alleviate individual model biases, enhance overall predictive accuracy and provide a stable architecture for the precise classification of various lung diseases.