Semi-supervised Learning using Denoising Autoencoders for Brain Lesion Detection and Segmentation

The work presented explores the use of denoising autoencoders (DAE) for brain lesion detection, segmentation and false positive reduction. Stacked denoising autoencoders (SDAE) were pre-trained using a large number of unlabeled patient volumes and fine tuned with patches drawn from a limited number of patients (n=20, 40, 65). The results show negligible loss in performance even when SDAE was fine tuned using 20 patients. Low grade glioma (LGG) segmentation was achieved using a transfer learning approach wherein a network pre-trained with High Grade Glioma (HGG) data was fine tuned using LGG image patches. The weakly supervised SDAE (for HGG) and transfer learning based LGG network were also shown to generalize well and provide good segmentation on unseen BraTS 2013 & BraTS 2015 test data. An unique contribution includes a single layer DAE, referred to as novelty detector(ND). ND was trained to accurately reconstruct non-lesion patches using a mean squared error loss function. The reconstruction error maps of test data were used to identify regions containing lesions. The error maps were shown to assign unique error distributions to various constituents of the glioma, enabling localization. The ND learns the non-lesion brain accurately as it was also shown to provide good segmentation performance on ischemic brain lesions in images from a different database.