Alzheimer's Detection By Using Neural Networks

Alzheimer’s disease is disorder in which person slowly starts to lose his/her memory, thinking ability and finally he/she is unable to do the simplistic day to day task. It is irreversible damage to the brain. It affects mostly the older people but can rarely be seen in people between 30 to 60 years. It was discovered by Dr. Alois Alzheimer in 1906. While examining he started noticing abnormalities in brain tissues of a dead woman. These abnormalities are now known as main features of the disease. The other significant finding was damage to interconnect between nerve cells. By studying these abnormalities we can detect the onset of the disease. The people affected by the disease can look normal in day to day life but steadily start having difficulties in day to day activities like reading, recognizing object, reasoning, driving and others. Eventually the brain damage is so much that its tissue shrunk. The disease is fatal and causes number of deaths every year. It is really necessary to diagnose the disease in early stages to take better care of the patient. Mild cognitive impairment (MCI) can be considered as the early stage of the Alzheimer. But not every person with MCI develops Alzheimer. Thus proper classification of the brain scans is necessary to segregate the affected person from non-affected one. There is pattern in brain damage as disease progresses. Hippocampus is one of the regions in the brain which first gets affected. This region is responsible for episodic and spatial memory and also provides interconnects between brain and body. Present day medicines only help to keep the symptoms in check. But they are not able to stop the progression of the disease. The disease is not curable and there is no treatment in current time to reverse the damage done to the brain. Computer based system for Alzheimer’s disease focused on taking number of data samples and splitting them in training and testing samples. It uses morphological operations to separate normal brain scans from disease affected brain scans. Morphometry analysis is also used to study brain MRI images to classify them in normal or abnormal categories. It helps in computation of the structural differences throughout the brain tissues. Another form of Morphometry is voxel based Morphometry which allows to compare differences in local concentration of brain tissues using voxel wise comparison of multiple brain images. Study has been conducted to measure volumetric atrophy of the gray matter in the areas of the neocortex of the Alzheimer’s patients and non-affected patients. In this method subject images are spatially normalized and then tissue classes are segmented using priori probability maps. Next step is to use denoising and remove distortions. The data is then tested using general linear model for residual variability and experimental results. This results in statistical parametric map which are thrsholded for required result. There are also experiments using machine learning techniques for classifying brain images into normal or specific neurodegenerative disease. In these experiments various forms of artificial neural network are used for classification. Some studies also show use of support vector machine with linear as well as non-linear kernels. Feature extraction can be done in different ways like using morphological operation, based on region of interest, or using grey level co-occurrence matrix on segmented image. Computer techniques are widely used in medical diagnosis mostly in cancer related cases. The research is done in detecting brain, liver and breast cancer detection using various image processing techniques using human organ photos or images. MRI scans are commonly used in neuroscience to study the brain tissues and any damage to them.