Material Detail

This video was recorded at Joint IAPR International Workshops on Structural and Syntactic Pattern Recognition (SSPR) and Statistical Techniques in Pattern Recognition (SPR), Cesme 2010. We present the possibility of detecting dementia in an early stage using dissimilarity-based classification. The approach is based on the nonrigid registration of MR brain scans, which lends itself very well for a dissimilarity-based treatment. As discussed, there is no need to go trough the difficult process of building an atlas or the like, as the registration can be easily performed in a pairwise fashion and a dissimilarity measure can be computed directly in this way. The experiments show that a classification accuracy of 88% can attained for prediction of dementia within 5 years [with a specificity of 82% and a sensitivity of 93%], demonstrating the potential of the method to aid in early diagnosis of dementia.

Disciplines:

• Science and Technology  / Computer Science