Computer Aided Diagnosis of Alzheimers Disease from MRI Images by Merging Segmentation Based Fractal Texture Analysis Features and Gray Level Co Occurrence Matrix features

Abstract

Alzheimer is found to be the fourth largest cause of death in United States and it is gradually increasing every year. It is also found to be moderately growing in Asian countries and by 2050, half of the Alzheimer patients will be from this region. There are no proven medicines that could cure Alzheimer completely, but early detection of Alzheimer could help the doctors provide the right treatment to the patient, which would increase the quality of living and also increase the life span of them. Computer detection system could be used for early detection of the disease and to assist the doctors in detecting Alzheimer s. In past years many CAD system were developed; yet most of them possess a lower accuracy rate. Hence an efficient CAD system which could efficiently predict the presence of AD has to be developed. newlineThe proposed system goes through several stages namely, image acquisition, image preprocessing, image segmentation, feature extraction and classification. In the initial step the images are acquired from publicly available datasets such as ADNI and OASIS. To enhance the image, the intensity values of the pixel are increased using Wiener filter. The next step is to segment the image to extract the Region of Interest (ROI). Edge based Active contour method is implemented and later GLCM and SFTA features are extracted. Finally, the classification is performed using Logical Regression, Support Vector Machine, K- Nearest Neighbor and bagged tree classifier. Brain area was calculated using morphological operators and compared with original image for better prediction. newlineThe proposed research work aims to develop an efficient CAD system which could assist doctors in identifying the presence of Alzheimer among patients. The system was evaluated at various stages and found to be performing well and finally compared with existing systems among the performance metrics namely accuracy, sensitivity and specificity as the values 91%, 89.7%, 92.74% respectively newline