An Approach to Detect Cardiomegaly, COVID-19, Pneumonia, Pneumothorax and Tuberculosis from CXR Images Using Ensembles of Deep Learning

Recent advancements in medical image analysis have harnessed the potential of deep learning, especially in diagnosing thoracic ailments using chest X-ray (CXR) images. This study introduces a novel approach to detecting a range of thoracic diseases, including COVID-19, pneumonia, pneumothorax, cardiomegaly, and tuberculosis, through ensembles of deep learning models. The development of a reliable and efficient diagnostic system is crucial given the increasing significance of early disease detection and the growing number of CXR images in medical databases. The proposed ensemble method combines features of multiple deep learning models to improve overall performance, leveraging their distinct strengths. The study includes extensive experimentation and comparative analyses on a diverse and extensive dataset to highlight the superior performance of this approach. The results demonstrate that the ensemble-based model outperforms individual models, achieving a higher accuracy of 98.37%, sensitivity of 0.9837, and precision of 0.9836, thus presenting significant potential for robust and multi-disease detection in CXR images. Our results hold promise for advancing healthcare and clinical decision-making, underlining the valuable contribution of this research. This work not only underscores the potency of deep learning in medical imaging but also highlights its pivotal role in elevating patient care.