Alzheimer's disease diagnosis via interested structure selection in MRIs

摘要

Diagnosing Alzheimer's disease (AD) with MRI has attracted more and more attention. This paper proposed a new feature selection method which selected interested structures in MRI brain images based on P-Value. The P-Value can help to obtain the independent principal features. P-Value for every voxel was calculated by T-test between different image classes, then the average value of P-Value for every brain tissue was calculated. The structures which have large value were selected as interested structures, then the AD diagnosis was completed by the collaborative representation based classification (CRC) using these features. The classification accuracy between normal controls (NC) and AD is 92.2%, between NC and Mild Cognitive Impairment (MCI) is 75.0%, between AD and MCI is 83.5%. Compared with existing single mode methods, the proposed classification achieved a better performance.