Joint Classification and Regression via Deep Multi-Task Multi-Channel Learning for Alzheimer's Disease Diagnosis

摘要: In the field of computer-aided Alzheimer's disease (AD) diagnosis, jointly identifying brain diseases and predicting clinical scores using magnetic resonance imaging (MRI) have attracted increasing attention since these two tasks are highly correlated. Most existing joint learning approaches require hand-crafted feature representations for MR images. Since features MRI classification/regression models may not coordinate well with each other, conventional methods lead to sub-optimal performance. Also, demographic information ( e.g. , age, gender, education) subjects also be related status, thus can help improve diagnostic However, seldom incorporate such into models. To this end, we propose a deep multi-task multi-channel (DM $^2$ L) framework simultaneous classification score regression, data subjects. Specifically, first identify discriminative anatomical landmarks from images in data-driven manner, then extract multiple image patches around detected landmarks. We convolutional neural network regression. Our DM L only automatically learn images, but explicitly process. evaluate proposed method on four large multi-center cohorts 1984 subjects, experimental results demonstrate that is superior several state-of-the-art both