Review on Image Processing Techniques of Early Detection of Neurodegenerative Disorders

Parkinson's and Alzheimer's are disorders of the central nervous system and are severe chronic disorders that progress in time, and today they are endangering the lives of millions of individuals. These disorders end up killing the brain slowly, causing memory loss and mental and physical disability. The most difficult aspect of the treatment of such conditions is that they are always realized too late when their levels have gone to the extreme. Early diagnosis, therefore, is a worthy thing to do, as this would produce certain important effects on the modes of treatment and patient reaction. Doctors used observations and clinical tests in the past in the identification of these conditions. However, these ways are problematic and may lead to missing early symptoms. The past years made new opportunities in detecting disease- related alterations in the brain MRIs and PET scans, achieved with the help of digital image processing (DIP), in particular with the method of deep learning. In this survey, we discuss six recent articles in which DIP is used in the detection and identification of the early precursors of neurodegenerative conditions. The methods that are implemented are also different with each of the papers we will review—the convolutional neural networks (CNNs) and hybrid networks consisting of CNNs. These approaches have been quite promising in terms of accuracy and automation. They, however, also have limitations, such as large requirements in hardware, the danger of overfitting, and generalization challenges across diseases. This survey shows the level at which the current study is placed and the direction it can run based on the process, findings, strengths, and weaknesses that these researchers conducted. Our conclusion is a recommendation for future studies.