Age estimation from MR images via 3D convolutional neural network and densely connect

Abstract

The estimation of brain age from magnetic resonance (MR) images is useful for computer-aided diagnosis (CAD) in neurodegenerative diseases. Some deep learning methods has been proposed for age estimation from MR images recently. These methods release the burden of pre-processing dramatically, and they outperform the methods with hand-crafted features as well. However, the existing models of brain age estimation simply stack several convolution layers together, whose fitting ability is still limited. In this paper, we propose a deep learning framework based on 3D convolution neural network and dense connections to predict brain ages from MR images. The densely connect block in the proposed framework has a stronger fitting ability. Besides, combined with the domain knowledge of brain age estimation, the high-frequency structures of brain MR images are extracted and then are sent into the deep network. The proposed method is evaluated on a public brain MRI dataset. With the comparisons with existing methods, the experimental results demonstrated that our method achieved the state-of-the-art performances with the accuracy of 4.28 years on mean absolute error (MAE).