BrainSegNet: a convolutional neural network architecture for automated segmentation of human brain structures

Abstract

Automated segmentation of cortical and noncortical human brain structures has been hitherto approached using nonrigid registration followed by label fusion. We propose an alternative approach for this using a convolutional neural network (CNN) which classifies a voxel into one of many structures. Four different kinds of two-dimensional and three-dimensional intensity patches are extracted for each voxel, providing local and global (context) information to the CNN. The proposed approach is evaluated on five different publicly available datasets which differ in the number of labels per volume. The obtained mean Dice coefficient varied according to the number of labels, for example, it is [Formula: see text] and [Formula: see text] for datasets with the least (32) and the most (134) number of labels, respectively. These figures are marginally better or on par with those obtained with the current state-of-the-art methods on nearly all datasets, at a reduced computational time. The consistently good performance of the proposed method across datasets and no requirement for registration make it attractive for many applications where reduced computational time is necessary.

Spatially Localized Atlas Network Tiles Enables 3D Whole Brain Segmentation from Limited Data

“…To address the challenges of training a network on a small number of manually traced brains, patch-based DCNN methods have been proposed. de Brbisson et al, [2] proposed to learn 2D and 3D patches as well as spatial information, which was extended to include 2.5D patches by BrainSegNet [10]. Recently, DeepNAT [13] was proposed to perform hierarchical multi-task learning on 3D patches.…”

Accuracy of deep learning to differentiate the histopathological grading of meningiomas on MR images: A preliminary study

“…These maps are then analyzed to create increasingly more complex and abstract representation of the items represented in the images. A broad number of applications of DCNNs in neuroradiology are being studied, and the most thoroughly investigated are: 1) automatic brain segmentation; 2) automatic detection of Alzheimer's disease‐associated lesions from functional MRI; 3) automatic detection of stroke‐related lesions from CT images; 4) prediction of genetic mutation; and 5) grading of gliomas …”

“…These maps are then analyzed to create increasingly more complex and abstract representation of the items represented in the images. A broad number of applications of DCNNs in neuroradiology are being studied, and the most thoroughly investigated are: 1) automatic brain segmentation 11,12 ; 2) automatic detection of Alzheimer's disease-associated lesions from functional MRI 13 ; 3) automatic detection of stroke-related lesions from CT images 14 ; 4) prediction of genetic mutation 15 ; and 5) grading of gliomas. 16,17 Recently, the scope to predict the grading of spontaneously occurring meningiomas from routine MRI sequences by means of DCNNs has been explored in dogs; an 82% accuracy in discrimination between benign, atypical, and malignant lesions was achieved on a small-sized dataset (56 patients).…”

Development of accurate human head models for personalized electromagnetic dosimetry using deep learning

“…Recently, it becomes interesting to investigate how deep learning can contribute to the personalized electromagnetic dosimetry (Rashed et al, 2019b). For anatomical brain structure segmentation, several methods have been presented (Chen et al, 2018a,b;de Brebisson and Montana, 2015;Mehta et al, 2017;Milletari et al, 2017;Moeskops et al, 2016;Wachinger et al, 2018;Zhang et al, 2015). Additionally, methods for the segmentation of small regions within the brain has also been proposed (Choi and Jin, 2016;Dolz et al, 2018;Kushibar et al, 2018;Roy et al, 2019a,b).…”