Simple brain atrophy quantification method using MR images

In this paper, age estimation models introduced with automatic preprocessing of the T-1 weighted images, dimension reduction via principal component analysis, training of a multiple regression model, and then estimating the age of the subjects from the test samples. The regression model is automatically trained from a diverse set of 80 adult subjects (age 60-92 years) exhibiting significant variation to discover anatomical structure related to age and deformation. The methods proved to be a reliable one for age estimation in healthy subjects, yielding a correlation of r = 0.780 between the estimated and real age in the test samples and a mean absolute error of 2.155 years for PCAR method, and r = 0.834 and a mean absolute error of 2.092years for the PCA-ML method. To test the potential of these proposed age estimation models in the clinical situation, very mild to moderate Alzheimer's disease (AD) subject's age has been estimated.