3D Convolutional Neural Network Segmentation of White Matter Tract Masks from MR Diffusion Anisotropy Maps

Pomiecko, Kristofer; Sestili, Carson D.; Fissell, Kate; Pathak, Sudhir Kumar; Okonkwo, David O.; Schneider, Walter

doi:10.1109/isbi.2019.8759575

Cited by 7 publications

(2 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As opposed to the previous deep learning approaches, a few other ones (Wasserthal et al (2018); Pomiecko et al (2019); Li et al (2020)) provide voxel-wise segmentations representing the locations traversed by a given bundle's streamlines.…”

Section: Streamline Clusteringmentioning

confidence: 99%

Filtering in tractography using autoencoders (FINTA)

Legarreta,

Petit,

Rheault

et al. 2020

Preprint

View full text Add to dashboard Cite

Current brain white matter fiber tracking techniques show a number of problems, including: generating large proportions of streamlines that do not accurately describe the underlying anatomy; extracting streamlines that are not supported by the underlying diffusion signal; and under-representing some fiber populations, among others. In this paper, we describe a novel unsupervised learning method to filter streamlines from diffusion MRI tractography, and hence, to obtain more reliable tractograms.We show that a convolutional neural network autoencoder provides a straightforward and elegant way to learn a robust representation of brain streamlines, which can be used to filter undesired samples with a nearest neighbor algorithm. Our method, dubbed FINTA (Filtering in Tractography using Autoencoders) comes with several key advantages: training does not need labeled data, as it uses raw tractograms, it is fast and easily reproducible, it does not rely on the input diffusion MRI data, and thus, does not suffer from domain adaptation issues. We demonstrate the ability of FINTA to discriminate between "plausible" and "implausible" streamlines as well as to recover individual streamline group instances from a raw tractogram, from both synthetic and real human brain diffusion

show abstract

Section: Streamline Clusteringmentioning

confidence: 99%

Filtering in tractography using autoencoders (FINTA)

Legarreta,

Petit,

Rheault

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Applications of segmentation based on the inherently multi-channel dMRI signals include whole-brain GM/WM/Cerebrospinal fluid [30,35], WM regions [19,31], nuclei (cerebellar [15] and thalamic [13]), organs [3,8,26,39], tumors [28] and stroke lesions [4,20]. As an alternative approach to traditional tractography, neural networks can directly segment WM tracts based on dMRI [16,17,21,23] or diffusion orientations [32,33,37,38], from clinical [21] or high-quality [32] datasets. Unfortunately, segmenting WM tracts as volumetric labels does not provide an along-the-tract parameterization which is useful for point-wise analyses of microstructural and geometric features of the tracts.…”

Section: Introductionmentioning

confidence: 99%

Learning Anatomical Segmentations for Tractography from Diffusion MRI

Ewert¹,

Kügler²,

Yendiki³

et al. 2020

Preprint

View full text Add to dashboard Cite

Deep learning approaches for diffusion MRI have so far focused primarily on voxel-based segmentation of lesions or white-matter fiber tracts. A drawback of representing tracts as volumetric labels, rather than sets of streamlines, is that it precludes point-wise analyses of microstructural or geometric features along a tract. Traditional tractography pipelines, which do allow such analyses, can benefit from detailed wholebrain segmentations to guide tract reconstruction. Here, we introduce fast, deep learning-based segmentation of 170 anatomical regions directly on diffusion-weighted MR images, removing the dependency of conventional segmentation methods on T1-weighted images and slow pre-processing pipelines. Working natively in diffusion space avoids non-linear distortions and registration errors across modalities, as well as interpolation artifacts. We demonstrate consistent segmentation results between 0.70 and 0.87 Dice depending on the tissue type. We investigate various combinations of diffusion-derived inputs and show generalization across different numbers of gradient directions. Finally, integrating our approach to provide anatomical priors for tractography pipelines, such as TRACULA, removes hours of pre-processing time and permits processing even in the absence of high-quality T1-weighted scans, without degrading the quality of the resulting tract estimates.

show abstract