ABSTRACT
A deep neural network (DNN) is a multineural layer architecture model whose layers learn to represent the data at multiple abstract levels. It makes DNN a powerful tool to address the existing problems in traditional architectures of machine learning. To learn large databases semantically at a high level with deep features learning emerging its applications in various medical and nonmedical fields. The current review, in its first phase, focuses on how the deep supervised architecture works and, in a second phase, examines the present state of deep supervised algorithm's applications in the detection, diagnosis, and classification of neurosurgery, diabetes, and heart diseases and the efficiency of applying algorithms. For these analyses, publications from the Scopus and PubMed libraries are selected after a rigorous screening process. In most of these studies, deep neural networks outperformed the machine learning algorithms. The detail overview present in this article would be of interest to scientific groups working in the area of automated detection, diagnosis, and classification of neurosurgery, diabetes mellitus, and heart disease in humans using DNN-based decision support systems.
