Authors - Hetal M. Bhatt, Sunil Bajeja Abstract - Background: DCNNs have led to revolution of medical imaging for disease diagnosis, segmentation and prognosis. Such application has brought undeniable gains in accuracy on diverse clinical domains. Objective: This study offers a systematic review of the applicability, challenges, and proposed work directions in medical imaging in DCNNs. Methods: PubMed, IEEE Xplore, Scopus and Google Scholar were searched in a comprehensive manner, and the studies were deemed relevant, methodologically rigorous and performance metrics based. Results: Overall, DCNNs have resulted in much higher diagnostic accuracy when used for tumour classification, lung disease detection, brain lesion segmentation and cardiac abnormality identification. However, there are still a few challenges, including the issue of scarcity of data, the lack of the model interpretability and the clinical adoption.
