Geometric Deep Learning of the Human Connectome Project Multimodal Cortical Parcellation

AbstractUnderstanding the topographic heterogeneity of cortical organisation is an essential step towards precision modelling of neuropsychiatric disorders. While many cortical parcellation schemes have been proposed, few attempt to model inter-subject variability. For those that do, most have been proposed for high-resolution research quality data, without exploration of how well they generalise to clinical quality scans. In this paper, we benchmark and ensemble four different geometric deep learning models on the task of learning the Human Connectome Project (HCP) multimodal cortical parcellation. We employ Monte Carlo dropout to investigate model uncertainty with a view to propagate these labels to new datasets. Models achieved an overall Dice overlap ratio of >0.85 ± 0.02. Regions with the highest mean and lowest variance included V1 and areas within the parietal lobe, and regions with the lowest mean and highest variance included areas within the medial frontal lobe, lateral occipital pole and insula. Qualitatively, our results suggest that more work is needed before geometric deep learning methods are capable of fully capturing atypical cortical topographies such as those seen in area 55b. However, information about topographic variability between participants was encoded in vertex-wise uncertainty maps, suggesting a potential avenue for projection of this multimodal parcellation to new datasets with limited functional MRI, such as the UK Biobank.

Download Full-text

FOD-Net: A Deep Learning Method for Fiber Orientation Distribution Angular Super Resolution

10.1101/2021.01.17.427042 ◽

2021 ◽

Author(s):

Rui Zeng ◽

Jinglei Lv ◽

He Wang ◽

Luping Zhou ◽

Michael Barnett ◽

...

Keyword(s):

Deep Learning ◽

Angular Resolution ◽

Super Resolution ◽

Orientation Distribution ◽

Quality Data ◽

High Angular Resolution ◽

Clinical Protocols ◽

High Quality ◽

Human Connectome Project ◽

Fiber Orientation Distribution

ABSTRACTMapping the human connectome using fibre-tracking permits the study of brain connectivity and yields new insights into neuroscience. However, reliable connectome reconstruction using diffusion magnetic resonance imaging (dMRI) data acquired by widely available clinical protocols remains challenging, thus limiting the connectome/tractography clinical applications. Here we develop fibre orientation distribution (FOD) network (FOD-Net), a deep-learning-based framework for FOD angular super-resolution. Our method enhances the angular resolution of FOD images computed from common clinical-quality dMRI data, to obtain FODs with quality comparable to those produced from advanced research scanners. Super-resolved FOD images enable superior tractography and structural connectome reconstruction from clinical protocols. The method was trained and tested with high-quality data from the Human Connectome Project (HCP) and further validated with a local clinical 3.0T scanner. Using this method, we improve the angular resolution of FOD images acquired with typical single-shell low-angular-resolution dMRI data (e.g., 32 directions, b=1000 s/mm2) to approximate the quality of FODs derived from time-consuming, multi-shell high-angular-resolution dMRI research protocols. We also demonstrate tractography improvement, removing spurious connections and bridging missing connections. We further demonstrate that connectomes reconstructed by super-resolved FOD achieve comparable results to those obtained with more advanced dMRI acquisition protocols, on both HCP and clinical 3T data. Advances in deep-learning approaches used in FOD-Net facilitate the generation of high quality tractography/connectome analysis from existing clinical MRI environments.

Download Full-text

Structure can predict function in the human brain: a graph neural network deep learning model of functional connectivity and centrality based on structural connectivity

Brain Structure and Function ◽

10.1007/s00429-021-02403-8 ◽

2021 ◽

Author(s):

Josh Neudorf ◽

Shaylyn Kress ◽

Ron Borowsky

Keyword(s):

Neural Network ◽

Deep Learning ◽

Functional Connectivity ◽

Brain Research ◽

Structural Connectivity ◽

Model Performance ◽

Future Research ◽

Individual Level ◽

Human Connectome Project ◽

Deep Learning Model

AbstractAlthough functional connectivity and associated graph theory measures (e.g., centrality; how centrally important to the network a region is) are widely used in brain research, the full extent to which these functional measures are related to the underlying structural connectivity is not yet fully understood. Graph neural network deep learning methods have not yet been applied for this purpose, and offer an ideal model architecture for working with connectivity data given their ability to capture and maintain inherent network structure. Here, we applied this model to predict functional connectivity from structural connectivity in a sample of 998 participants from the Human Connectome Project. Our results showed that the graph neural network accounted for 89% of the variance in mean functional connectivity, 56% of the variance in individual-level functional connectivity, 99% of the variance in mean functional centrality, and 81% of the variance in individual-level functional centrality. These results represent an important finding that functional centrality can be robustly predicted from structural connectivity. Regions of particular importance to the model's performance as determined through lesioning are discussed, whereby regions with higher centrality have a higher impact on model performance. Future research on models of patient, demographic, or behavioural data can also benefit from this graph neural network method as it is ideally-suited for depicting connectivity and centrality in brain networks. These results have set a new benchmark for prediction of functional connectivity from structural connectivity, and models like this may ultimately lead to a way to predict functional connectivity in individuals who are unable to do fMRI tasks (e.g., non-responsive patients).

Download Full-text

Biological Faithfulness is Unnecessary for Machine Learning

University of Western Ontario Medical Journal ◽

10.5206/uwomj.v87i2.1134 ◽

2019 ◽

Vol 87 (2) ◽

pp. 27-29

Author(s):

Meagan Wiederman

Keyword(s):

Artificial Intelligence ◽

Machine Learning ◽

Deep Learning ◽

Feature Detection ◽

Black Box ◽

Compression Algorithms ◽

Recent Success ◽

Human Connectome Project ◽

Definition Of ◽

The Brain

Artificial intelligence (AI) is the ability of any device to take an input, like that of its environment, and work to achieve a desired output. Some advancements in AI have focused n replicating the human brain in machinery. This is being made possible by the human connectome project: an initiative to map all the connections between neurons within the brain. A full replication of the thinking brain would inherently create something that could be argued to be a thinking machine. However, it is more interesting to question whether a non-biologically faithful AI could be considered as a thinking machine. Under Turing’s definition of ‘thinking’, a machine which can be mistaken as human when responding in writing from a “black box,” where they can not be viewed, can be said to pass for thinking. Backpropagation is an error minimizing algorithm to program AI for feature detection with no biological counterpart which is prevalent in AI. The recent success of backpropagation demonstrates that biological faithfulness is not required for deep learning or ‘thought’ in a machine. Backpropagation has been used in medical imaging compression algorithms and in pharmacological modelling.

Download Full-text

A-GHN: Attention-based Fusion of Multiple GraphHeat Networks for Structural to Functional Brain Mapping

10.1101/2021.08.12.456134 ◽

2021 ◽

Author(s):

SUBBA REDDY OOTA ◽

Archi Yadav ◽

Arpita Dash ◽

Surampudi Bapi Raju ◽

Avinash Sharma

Keyword(s):

Deep Learning ◽

Functional Connectivity ◽

Complex Dynamics ◽

Brain Activity ◽

Structural Connectivity ◽

Reaction Diffusion ◽

Complex Nature ◽

Relational Structures ◽

Human Connectome Project ◽

The Brain

Over the last decade, there has been growing interest in learning the mapping from structural connectivity (SC) to functional connectivity (FC) of the brain. The spontaneous fluctuations of the brain activity during the resting-state as captured by functional MRI (rsfMRI) contain rich non-stationary dynamics over a relatively fixed structural connectome. Among the modeling approaches, graph diffusion-based methods with single and multiple diffusion kernels approximating static or dynamic functional connectivity have shown promise in predicting the FC given the SC. However, these methods are computationally expensive, not scalable, and fail to capture the complex dynamics underlying the whole process. Recently, deep learning methods such as GraphHeat networks along with graph diffusion have been shown to handle complex relational structures while preserving global information. In this paper, we propose a novel attention-based fusion of multiple GraphHeat networks (A-GHN) for mapping SC-FC. A-GHN enables us to model multiple heat kernel diffusion over the brain graph for approximating the complex Reaction Diffusion phenomenon. We argue that the proposed deep learning method overcomes the scalability and computational inefficiency issues but can still learn the SC-FC mapping successfully. Training and testing were done using the rsfMRI data of 100 participants from the human connectome project (HCP), and the results establish the viability of the proposed model. Furthermore, experiments demonstrate that A-GHN outperforms the existing methods in learning the complex nature of human brain function.

Download Full-text

Non-linearity matters: a deep learning solution to generalization of hidden brain patterns across population cohorts

10.1101/2021.03.10.434856 ◽

2021 ◽

Author(s):

Mariam Zabihi ◽

Seyed Mostafa Kia ◽

Thomas Wolfers ◽

Richard Dinga ◽

Alberto Llera ◽

...

Keyword(s):

Deep Learning ◽

Fmri Data ◽

Space Representation ◽

Model Parameters ◽

Independent Dataset ◽

Latent Space ◽

Human Connectome Project ◽

Nonlinear Features ◽

Latent Representations ◽

High Level

AbstractThe increasing number of neuroimaging scans in recent years has facilitated the use of complex nonlinear approaches to analyzing such data. More specifically, deep learning, which has been previously hindered by the curse of dimensionality is now feasible. However, it remains challenging to use these techniques develop reliable biomarkers and find an optimal representation of data that explains the biological underpinnings of the mental disorders Here, we employed a 3-dimensional autoencoder with an architecture designed from the ground up for task-fMRI data. Our study presented a coherent strategy for optimizing model parameters and architecture and a method for visualizing and interpreting the latent space representation. We trained our model with multi-task fMRI data derived from the Human Connectome Project (HCP) that provides whole-brain coverage across a range of cognitive tasks. Next, in a transfer learning setting, we tested the generalization of our latent space on UK Biobank data as an independent dataset. We showed that the model did not only learn salient features such as age but also high-level behavioral characteristics and that this representation was highly generic and generalizable to an independent dataset. Furthermore, we demonstrated that the projection of latent space back into the original space is meaningful and interpretable. Finally, our results show that with careful implementation, nonlinear features can provide complementary information that accessible to purely linear methods. Our results provide an important step toward learning interpretable and generalizable latent representations that link cognition with underlying brain systems.

Download Full-text

The Connectivity–Based Parcellation of The Angular Gyrus: Fiber Dissection And MR Tractography Study

10.21203/rs.3.rs-1228692/v1 ◽

2022 ◽

Author(s):

Fatih Yakar ◽

Pınar Çeltikçi ◽

Yücel Doğruel ◽

Emrah Egemen ◽

Abuzer Güngör

Keyword(s):

Anterior Commissure ◽

Angular Gyrus ◽

Middle Temporal Gyrus ◽

Supramarginal Gyrus ◽

Middle Temporal ◽

Fiber Tracts ◽

Human Connectome Project ◽

Cortical Parcellation ◽

Fiber Dissection ◽

First Time

Abstract The angular gyrus (AG) wraps the posterior end of the superior temporal sulcus (STS), so it is considered as a continuation of the superior/middle temporal gyrus and forms the inferior parietal lobule (IPL) with the supramarginal gyrus (SMG). The AG was functionally divided in the literature, but there is no fiber dissection study in this context. This study divided AG into superior (sAG) and inferior (iAG) parts by focusing on STS. Red blue silicone injected eight human cadaveric cerebrums were dissected via the Klingler method focusing on the AG. White matter (WM) tracts identified during dissection were then reconstructed on the Human Connectome Project 1065 individual template for validation. According to this study, superior longitudinal fasciculus (SLF) II and middle longitudinal fasciculus (MdLF) are associated with sAG; the anterior commissure (AC), optic radiation (OR) with iAG; the arcuate fasciculus (AF), inferior frontooccipital fasciculus (IFOF), and tapetum (Tp) with both parts. In cortical parcellation of AG based on STS, sAG and iAG were found to be associated with different fiber tracts. Although it has been shown in previous studies that there are functionally different subunits with AG parcellation, here, for the first time, different functions of the subunits have been revealed with cadaveric dissection and tractography images.

Download Full-text

Deep Learning Based Segmentation of Brain Tissue from Diffusion MRI

10.1101/2020.07.30.228809 ◽

2020 ◽

Cited By ~ 1

Author(s):

Fan Zhang ◽

Anna Breger ◽

Kang Ik Kevin Cho ◽

Lipeng Ning ◽

Carl-Fredrik Westin ◽

...

Keyword(s):

Deep Learning ◽

Brain Tissue ◽

Diffusion Mri ◽

Image Resolution ◽

Diffusion Kurtosis Imaging ◽

Tissue Segmentation ◽

Anatomical Data ◽

Human Connectome Project ◽

Improve Accuracy ◽

Anatomical Mri

Segmentation of brain tissue types from diffusion MRI (dMRI) is an important task, required for quantification of brain microstructure and for improving tractography. Current dMRI segmentation is mostly based on anatomical MRI (e.g., T1- and T2-weighted) segmentation that is registered to the dMRI space. However, such inter-modality registration is challenging due to more image distortions and lower image resolution in the dMRI data as compared with the anatomical MRI data. In this study, we present a deep learning method that learns tissue segmentation from high-quality imaging datasets from the Human Connectome Project (HCP), where registration of anatomical data to dMRI is more precise. The method is then able to predict a tissue segmentation directly from new dMRI data, including data collected with a different acquisition protocol, without requiring anatomical data and inter-modality registration. We train a convolutional neural network (CNN) to learn a tissue segmentation model using a novel augmented target loss function designed to improve accuracy in regions of tissue boundary. To further improve accuracy, our method adds diffusion kurtosis imaging (DKI) parameters that characterize non-Gaussian water molecule diffusion to the conventional diffusion tensor imaging parameters. The DKI parameters are calculated from the recently proposed mean-kurtosis-curve method that corrects implausible DKI parameter values and provides additional features that discriminate between tissue types. We demonstrate high tissue segmentation accuracy on HCP data, and also when applying the HCP-trained model on dMRI data from a clinical acquisition with lower resolution and fewer gradient directions.

Download Full-text

Computing personalized brain functional networks from fMRI using self-supervised deep learning

10.1101/2021.09.25.461829 ◽

2021 ◽

Author(s):

Hongming Li ◽

Srinivasan Dhivya ◽

Zaixu Cui ◽

Chuanjun Zhuo ◽

Raquel E. Gur ◽

...

Keyword(s):

Deep Learning ◽

Resting State ◽

Brain Function ◽

Resting State Fmri ◽

Fmri Data ◽

Functional Networks ◽

Brain Functional Networks ◽

Human Connectome Project ◽

Functional Homogeneity ◽

Decoder Architecture

ABSTRACTA novel self-supervised deep learning (DL) method is developed for computing bias-free, personalized brain functional networks (FNs) that provide unique opportunities to better understand brain function, behavior, and disease. Specifically, convolutional neural networks with an encoder-decoder architecture are employed to compute personalized FNs from resting-state fMRI data without utilizing any external supervision by optimizing functional homogeneity of personalized FNs in a self-supervised setting. We demonstrate that a DL model trained on fMRI scans from the Human Connectome Project can identify canonical FNs and generalizes well across four different datasets. We further demonstrate that the identified personalized FNs are informative for predicting individual differences in behavior, brain development, and schizophrenia status. Taken together, self-supervised DL allows for rapid, generalizable computation of personalized FNs.

Download Full-text

Deep learning models of cognitive processes constrained by human brain connectomes

10.1101/2021.10.12.464145 ◽

2021 ◽

Author(s):

Yu Zhang ◽

Nicolas et Farrugia ◽

Pierre Bellec

Keyword(s):

Deep Learning ◽

Cognitive Processes ◽

Brain Activity ◽

Neural Substrates ◽

Brain Regions ◽

Human Cognition ◽

Experimental Conditions ◽

Cognitive States ◽

Learning Techniques ◽

Human Connectome Project

Brain decoding aims to infer cognitive states from recordings of brain activity. Current literature has mainly focused on isolated brain regions engaged in specific experimental conditions, but ignored the integrative nature of cognitive processes recruiting distributed brain networks. To tackle this issue, we propose a connectome-based graph neural network to integrate distributed patterns of brain activity in a multiscale manner, ranging from localized brain regions, to a specific brain circuit/network and towards the full brain. We evaluate the decoding model using a large task-fMRI database from the human connectome project. By implementing connectomic constraints and multiscale interactions in deep graph convolutions, the model achieves high accuracy of decoding 21 cognitive states (93%, chancel level: 4.8%) and shows high robustness against adversarial attacks on the graph architecture. Our study bridges human connectomes with deep learning techniques and provides new avenues to study the underlying neural substrates of human cognition at scale.

Download Full-text