DeepLesionBrain: Towards a broader deep-learning generalization for multiple sclerosis lesion segmentation

Recently, segmentation methods based on Convolutional Neural Networks (CNNs) showed promising per-formance in automatic Multiple Sclerosis (MS) lesions segmentation. These techniques have even out-performed human experts in controlled evaluation conditions such as Longitudinal MS Lesion Segmenta-tion Challenge (ISBI Challenge). However, state-of-the-art approaches trained to perform well on highly-controlled datasets fail to generalize on clinical data from unseen datasets. Instead of proposing another improvement of the segmentation accuracy, we propose a novel method robust to domain shift and per-forming well on unseen datasets, called DeepLesionBrain (DLB). This generalization property results from three main contributions. First, DLB is based on a large group of compact 3D CNNs. This spatially dis-tributed strategy aims to produce a robust prediction despite the risk of generalization failure of some individual networks. Second, we propose a hierarchical specialization learning (HSL) by pre-training a generic network over the whole brain, before using its weights as initialization to locally specialized net-works. By this end, DLB learns both generic features extracted at global image level and specific features extracted at local image level. Finally, DLB includes a new image quality data augmentation to reduce dependency to training data specificity (e.g., acquisition protocol). DLB generalization was validated in cross-dataset experiments on MSSEG'16, ISBI challenge, and in-house datasets. During experiments, DLB showed higher segmentation accuracy, better segmentation consistency and greater generalization per-formance compared to state-of-the-art methods. Therefore, DLB offers a robust framework well-suited for clinical practice. (c) 2021 Published by Elsevier B.V.