Deep Learning Performance on Medical Image, Data and Signals

Abstract

In this study, the recent medical studies with deep learning between 2009-2019 have been researched for observing the performance of deep learning on medical images, data and signal. Recent studies attained from Web of Science have been evaluated and selected according to the citation numbers. Studies have been listed as a table, according to the publication year, deep network structure, database used training and testing, evaluation metric and results. The studies have also been classified into the organs and the types of important diagnosis. The results have shown that the deep learning network structures, applied on fundus images, have attained nearly %99 percent accuracy. Although most of the studies between the range, made by Radiology and Nuclear Medicine Imaging, the accuracy of the results are 80-90% range. The current studies especially focus on automatic detection or classification of the tumor as benign or malign. Studies are mostly on medical CT, ultrasound, radiography and MRI images. This results show that computer aided medical diagnosis systems will be used in a very near future with fully performance.