Connectivity‐based parcellation of normal and anatomically distorted human cerebral cortex

Abstract

For over a century, neuroscientists have been working toward parcellating the human cortex into distinct neurobiological regions. Modern technologies offer many parcellation methods for healthy cortices acquired through magnetic resonance imaging. However, these methods are suboptimal for personalized neurosurgical application given that pathology and resection distort the cerebrum. We sought to overcome this problem by developing a novel connectivity-based parcellation approach that can be applied at the single-subject level. Utilizing normative diffusion data, we first developed a machine-learning (ML) classifier to learn the typical structural connectivity patterns of healthy subjects. Specifically, the Glasser HCP atlas was utilized as a prior to calculate the streamline connectivity between each voxel and each parcel of the atlas. Using the resultant feature vector, we determined the parcel identity of each voxel in neurosurgical patients (n = 40) and thereby iteratively adjusted the prior. This approach enabled us to create patient-specific maps independent of brain shape and pathological distortion. The supervised ML classifier reparcellated an average of 2.65% of cortical voxels across a healthy dataset (n = 178) and an average of 5.5% in neurosurgical patients. Our patient dataset consisted of subjects with supratentorial infiltrating gliomas operated on by the senior author who then assessed the validity and practical utility of the re-parcellated diffusion data. We demonstrate a rapid and effective ML parcellation approach to parcellation of the human cortex during anatomical distortion. Our approach overcomes limitations of indiscriminately applying atlas-based registration from healthy subjects by employing a voxel-wise connectivity approach based on individual data.

Connectomic disturbances underlying insomnia disorder and predictors of treatment response

“…This includes the following steps: (1) the diffusion image is resliced to ensure isotropic voxels, (2) motion correction is performed using a rigid body alignment, (2) slices with excess movement (defined as DVARS > 2 sigma from the mean slice) are eliminated, (3) the T1 image is skull stripped using a convolutional neural net (CNN). This is inverted and aligned to the DWI image using a rigid alignment, which is then used as a mask to skull strip the DWI image, (4) gradient distortion correction is performed using a diffeomorphic warping method which aims to locally similarize the DWI and T1 images, (5) eddy current correction is performed, (6) the fiber response function is estimated and the diffusion tensors are calculated using constrained spherical deconvolution, (7) deterministic tractography is performed with random seeding, usually creating about 300,000 streamlines per brain ( Doyen et al, 2022 ).…”

“…Since we sought to minimize the effects of gyral variation, we created a machine learning based, subject specific version of the Human Connectome Project Multimodal Parcelation (HCP-MMP1) atlas (Glasser et al, 2016) based on diffusion tractography structural connectivity, which we have described elsewhere (Doyen et al, 2022). In short, this was created by 10.3389/fnhum.2022.960350 training a machine learning model on a separate cohort of 200 normal adult subjects by first processing T1 and DT images as above.…”

Machine Learning Decomposition of the Anatomy of Neuropsychological Deficit in Alzheimer’s Disease and Mild Cognitive Impairment

“…In order to minimize the effects of gyral variation, we used a machine-learning based, subject specific version of the Human Connectome Project Multimodal Parcelation (HCP-MMP1) atlas ( Glasser et al, 2016 ) generated based on each subject’s structural connectivity, which has been described elsewhere ( Doyen et al, 2022 ). Figure 1 demonstrates the steps in creation of this personalized atlas.…”

Use of machine learning to identify functional connectivity changes in a clinical cohort of patients at risk for dementia

“…In order to minimize this, we adopted a machine-learning based method to create subject-specific versions of the Human Connectome Project Multimodal Parcellation (HCP-MMP1) atlas ( Glasser et al, 2016 ). While this method is described in detail elsewhere ( Doyen et al, 2021a ), briefly, a machine learning model was trained using tractography data from 178 healthy controls obtained from the SchizConnect database, preprocessed as above, in order for it to learn the structural connectivity pattern between voxels included within the 379 parcels of the HCP-MMP1 atlas. The same unaltered atlas was then warped onto each brain in the study sample and the trained machine learning model was applied to each subject to re-appoint voxels located at the endpoint of tractography streamlines to their most likely warped parcellation based on the structural connectivity feature vectors.…”