The use of brain imaging techniques for predicting early diagnosis of alzheimer's disease

Brains of dementia patients show characteristic differences according to the type of disease. Brain measurements such as cerebral cortical thickness, volumes or surface areas of some specific regions are effective in determining the type of disease. Brain images that are appropriate to the medical imaging standards can be obtained from magnetic resonance imaging devices. The headers of the images contain many different types of technical or non-technical information about patients, diseases, also imaging studies. Through brain imaging files and headers, the physical characteristics of patient brains can be extracted with the help of image processing techniques. After the numerical data is expressed as vectors in the plane during the classification, samples can be classified. In this study, magnetic resonance scans of 63 samples, having three disease types: 19 Alzheimer's disease, 19 frontotemporal dementia, and 25 vascular dementia, are used. Sliced brain image sets are processed with Freesurfer brain analyzing software tool. Different feature groups are created via generated statistical information after successful analysis of the program. Feature matrices are sent to the genetic algorithm that is used as a wrapper method for feature selection. Significant features, the brain regions that are effective in identifying the disease, are discussed. Applying different classification algorithms, and also genetic algorithm parameters, accuracy results up to %95.2 and confusion matrices are achieved.