Biology subjects

Parrish, T. B.

Publications and source records attributed to Parrish, T. B..

3 recordsLinked to original sources

bioRxiv2022

Effects of variability in manually contoured spinal cord masks on fMRI co-registration and interpretation

Functional magnetic resonance imaging (fMRI) of the human spinal cord (SC) is a unique non-invasive method for characterizing neurovascular responses to stimuli. Group-analysis of SC fMRI data involves co-registration of subject-level data to standard space, which requires manual masking of the cord and may result in bias of group-level SC fMRI results. To test this, we examined variability in SC masks drawn in fMRI data from 21 healthy participants from a completed study mapping responses to sensory stimuli of the C7 dermatome. Masks were drawn on temporal mean functional image by eight raters with varying levels of neuroimaging experience, and the rater from the original study acted as a reference. Spatial agreement between rater and reference masks was measured using the Dice Similarity Coefficient, and the influence of rater and dataset was examined using ANOVA. Each raters masks were used to register functional data to the PAM50 template. Gray matter-white matter signal contrast of registered functional data was used to evaluate the spatial normalization accuracy across raters. Subject- and group-level analyses of activation during left- and right-sided sensory stimuli were performed for each raters co-registered data. Agreement with the reference SC mask was associated with both rater (F(7,140) = 32.12, P < 2x10-16,{eta} 2 = 0.29) and dataset (F(20,140) = 20.58, P < 2x10-16,{eta} 2 = 0.53). Dataset variations may reflect image quality metrics: the ratio between the signal intensity of spinal cord voxels and surrounding cerebrospinal fluid was correlated with DSC results (p<0.001). As predicted, variability in the manually-drawn masks influenced spatial normalization, and GM:WM contrast in the registered data showed significant effects of rater and dataset (rater: F(8,160) = 23.57, P < 2x10-16,{eta} 2 = 0.24; dataset: F(20,160) = 22.00, P < 2x10-16,{eta} 2 = 0.56). Registration differences propagated into subject-level activation maps which showed rater-dependent agreement with the reference. Although group-level activation maps differed between raters, no systematic bias was identified. Increasing consistency in manual contouring of spinal cord fMRI data improved co-registration and inter-rater agreement in activation mapping, however our results suggest that improvements in image acquisition and post-processing are also critical to address.

neuroscience↗

bioRxiv2021

Analyzing Brain Morphology in Alzheimer's Disease Using Discriminative and Generative Spiral Networks

Several patterns of atrophy have been identified and strongly related to Alzheimers disease (AD) pathology and its progression. Morphological changes in brain shape have been identified up to ten years before clinical diagnoses of AD, making its early detection more relevant. We propose novel geometric deep learning frameworks for the analysis of brain shape in the context of neurodegeneration caused by AD. Our deep neural networks learn low-dimensional shape descriptors of multiple neuroanatomical structures, instead of handcrafted features for each structure. A discriminative network using spiral convolution on 3D meshes is constructed for the in-vivo binary classification of AD from healthy controls (HCs) using a fast and efficient "spiral" convolution operator on 3D triangular mesh surfaces of human brain subcortical structures extracted from T1-weighted magnetic resonance imaging (MRI). Our network architecture consists of modular learning blocks using residual connections to improve overall classifier performance. In this work: (1) a discriminative network is used to analyze the efficacy of disease classification using input data from multiple brain structures and compared to using a single hemisphere or a single structure. It also outperforms prior work using spectral graph convolution on the same the same tasks, as well as alternative methods that operate on intermediate point cloud representations of 3D shapes. (2) Additionally, visual interpretations for regions on the surface of brain structures that are associated to true positive AD predictions are generated and fall in accordance with the current reports on the structural localization of pathological changes associated to AD. (3) A conditional generative network is also implemented to analyze the effects of phenotypic priors given to the model (i.e. AD diagnosis) in generating subcortical structures. The generated surface meshes by our model indicate learned morphological differences in the presence of AD that agrees with the current literature on patterns of atrophy associated to the disease. In particular, our inference results demonstrate an overall reduction in subcortical mesh volume and surface area in the presence of AD, especially in the hippocampus. The low-dimensional shape descriptors obtained by our generative model are also evaluated in our discriminative baseline comparisons versus our discriminative network and the alternative shape-based approaches.

neuroscience↗

bioRxiv2020

Novel age-dependent cortico-subcortical morphologic interactions predict fluid intelligence: A multi-cohort geometric deep learning study.

Brain structure is tightly coupled with brain functions, but it remains unclear how cognition is related to brain morphology, and what is consistent across neurodevelopment. In this work, we developed graph convolutional neural networks (gCNNs) to predict Fluid Intelligence (Gf) from shapes of cortical ribbons and subcortical structures. T1-weighted MRIs from two independent cohorts, the Human Connectome Project (HCP; age: 28.81{+/-}3.70) and the Adolescent Brain Cognitive Development Study (ABCD; age: 9.93{+/-}0.62) were independently analyzed. Cortical and subcortical surfaces were extracted and modeled as surface meshes. Three gCNNs were trained and evaluated using six-fold nested cross-validation. Overall, combining cortical and subcortical surfaces yielded the best predictions on both HCP (R=0.454) and ABCD datasets (R=0.314), and outperformed the current literature. Across both datasets, the morphometry of the amygdala and hippocampus, along with temporal, parietal and cingulate cortex consistently drove the prediction of Gf, suggesting a novel reframing of the morphometry underlying Gf.

neuroscience↗