Age Estimation from MR Images via 3D Convolutional Neural Network and Densely Connect

摘要

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 handcrafted 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).