Age-related Classification and Prediction Based on MRI: A Sparse Representation Method
Longfei Su, Lubin Wang, Hui Shen, Dewen Hu · Procedia Environmental Sciences · 2011
Through analysis of structural magnetic resonance imaging (MRI) images, classification and prediction of the age of adults (19-79) were implemented to make analysis of the age-related changes of the grey matter (GM) concentration. Due to the distributed nature of the aging spatial pattern in human brain, a multivariate voxel selection method based on sparse representation was introduced to identify the most discriminative brain regions. It can effectively pick out the isolating voxels as well as the clustered voxels which contribute enormously to the classification and age prediction. By using this multivariate voxel selection method, the binary classification can get a higher accuracy compared to univariate voxel selection method. Age prediction of all the subjects via sparse representation (SR) was carried out in our study. Four different models were used to fit the predicted age of all subjects in maturity index (MI) space. Difference trend of the brain development between the senior and the junior was observed. That the development or decline of GM of the senior over 60 accelerates was found in our study.