Abstract
Anatomical segmentation of brain scans is highly relevant for diagnostics and neuroradiology research. Conventionally, segmentation is performed on T1-weighted MRI scans, due to the strong soft-tissue contrast. In this work, we report on a comparative study of automated, learning-based brain segmentation on various other contrasts of MRI and also computed tomography (CT) scans and investigate the anatomical soft-tissue information contained in these imaging modalities. A large database of in total 853 MRI/CT brain scans enables us to train convolutional neural networks (CNNs) for segmentation. We benchmark the CNN performance on four different imaging modalities and 27 anatomical substructures. For each modality we train a separate CNN based on a common architecture. We find average Dice scores of 86.7 ± 4.1% (T1-weighted MRI), 81.9 ± 6.7% (fluid-attenuated inversion recovery MRI), 80.8 ± 6.6% (diffusion-weighted MRI) and 80.7 ± 8.2% (CT), respectively. The performance is assessed relative to labels obtained using the widely-adopted FreeSurfer software package. The segmentation pipeline uses dropout sampling to identify corrupted input scans or low-quality segmentations. Full segmentation of 3D volumes with more than 2 million voxels requires <1s of processing time on a graphical processing unit.
Highlights
Anatomical segmentation of magnetic resonance imaging (MRI) or computed tomography (CT) scans is important for clinical diagnostics and scientific research
Quantitative volumetric measures of anatomical structures can be derived from accurate segmentation labels, which can be used to identify and monitor the progression of degenerative diseases, such as Alzheimer’s disease, which is characterized by atrophy of the hippocampus and the medial temporal lobe [1], Huntington disease, which results in athrophy of the striatum [2], and frontotemporal lobar degeneration, which causes atrophy of the frontal and temporal lobes [3]
For CT scans, we find that the segmentation performance is strongly structure-dependent: The low signal contrast between gray and white matter limits to some extent the accuracy of the segmentation, especially of the gray matter regions
Summary
Anatomical segmentation of magnetic resonance imaging (MRI) or computed tomography (CT) scans is important for clinical diagnostics and scientific research. An alternative approach is to automatize segmentation, which sparked the development of various segmentation software packages. In brain imaging these include e.g., FreeSurfer [4], BrainSuite [5], FSL [6], and ANTS [7]. These tools apply a set of complex transformations and thresholding procedures to the input volume [8] and are typically tailored toward T1-weighted scans.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
