Abstract
Individuals age differently depending on a multitude of different factorssuch as lifestyle, medical history and genetics. Often, the globalchronological age is not indicative of the true ageing process. An organ-basedage estimation would yield a more accurate health state assessment. In thiswork, we propose a new deep learning architecture for organ-based ageestimation based on magnetic resonance images (MRI). The proposed network is a3D convolutional neural network (CNN) with increased depth and width madepossible by the hybrid utilization of inception and fire modules. We apply theproposed framework for the tasks of brain and knee age estimation. Quantitativecomparisons against concurrent MR-based regression networks illustrated thesuperior performance of the proposed work.