Real Time Implementation of Medical Data Analysis Approaches

Abstract

The proposed work initially explored the advantages of using unsupervised clustering ap- proaches, k-means and k-medoids, for the early diagnosis of AD. The proposed framework applied the clustering approaches on VBM features extracted from MRI. The achieved results show that k-means obtained slightly higher accuracy than k-medoids and consumed nearly half of the k-medoids run time. The work explored also the effect of choosing certain ROIs for analysis compared to the whole-brain analysis. The whole-brain approach achieved more than 10% higher accuracy than the ROI approach. However, the run time of the ROI approach is 12.5% lower than the whole-brain approach. The proposed approach managed successfully to obtain an early AD diagnosis with an accuracy of 76%. The clustering techniques used