Learning white matter subject‐specific segmentation from structural MRI

Abstract

Purpose Mapping brain white matter (WM) is essential for building an understanding of brain anatomy and function. Tractography‐based methods derived from diffusion‐weighted MRI (dMRI) are the principal tools for investigating WM. These procedures rely on time‐consuming dMRI acquisitions that may not always be available, especially for legacy or time‐constrained studies. To address this problem, we aim to generate WM tracts from structural magnetic resonance imaging (MRI) image by deep learning. Methods Following recently proposed innovations in structural anatomical segmentation, we evaluate the feasibility of training multiply spatial localized convolution neural networks to learn context from fixed spatial patches from structural MRI on standard template. We focus on six widely used dMRI tractography algorithms (TractSeg, RecoBundles, XTRACT, Tracula, automated fiber quantification (AFQ), and AFQclipped) and train 125 U‐Net models to learn these techniques from 3870 T1‐weighted images from the Baltimore Longitudinal Study of Aging, the Human Connectome Project S1200 release, and scans acquired at Vanderbilt University. Results The proposed framework identifies fiber bundles with high agreement against tractography‐based pathways with a median Dice coefficient from 0.62 to 0.87 on a test cohort, achieving improved subject‐specific accuracy when compared to population atlas‐based methods. We demonstrate the generalizability of the proposed framework on three externally available datasets. Conclusions We show that patch‐wise convolutional neural network can achieve robust bundle segmentation from T1w. We envision the use of this framework for visualizing the expected course of WM pathways when dMRI is not available.

Automatic segmentation of the core of the acoustic radiation in humans

“…However, enlarging the number of inputs to the neural network has the disadvantage of needing more training data or changing the neural network architecture, which is beyond the scope of this article. Although TractSeg (29) can still be considered state-of-the-art for fiber bundle segmentation, new AI-based segmentation methods have recently been proposed [e.g., (39)(40)(41)(42)]. It is interesting to assess if adapting these methods can yield better results for segmenting the AR.…”

Convolutional-recurrent neural networks approximate diffusion tractography from T1-weighted MRI and associated anatomical context

“…Second, 132 brain regions are computed with the SLANT deep learning framework and grouped into 46 larger regions based on the BrainColor protocol (Huo et al, 2019). Third, 72 WM bundle regions defined by the TractSeg algorithm are computed with the WM learning (WML) framework (Yang et al, 2022; Wasserthal et al, 2018). All the contextual information is one-hot encoded.…”

“…To address this concern, prior studies have sought to skip the tractography step for common tractography-based analyses, like WM bundle analysis and structural connectomics, producing voxel-based WM segmentations or probabilistic atlases that require reworking of existing tractography workflows (Yeh, 2020; Sporns et al, 2005; Yang et al, 2022; Alemán-Gómez et al, 2022). As such, a “plug-and-play” solution to facilitate arbitrary subject-specific streamline-based tractography analyses without dMRI remains elusive.…”

Enhancing hierarchical transformers for whole brain segmentation with intracranial measurements integration

“…Traditionally, researchers utilized atlases for brain segmentation, 7,8 leveraging anatomical priors. With the advent of deep learning in medical imaging, [9][10][11][12][13][14][15][16][17][18] the focus has been shifted, replacing manual features with automatically learned features by computers.…”

“…This transition has demonstrated impressive performance in brain segmentation tasks, prompting extensive efforts in this direction. 7,8,[19][20][21] In particular, Huo et al 1 proposed a tile-based approach, called SLANT, for segmenting 132 brain regions. They partitioned the entire brain into 27 tiles, each processed by distinct U-Nets, 22 and aggregated the ensemble outcomes of these 27 U-Nets to produce the final results.…”