Interactions across emotional, cognitive and subcortical motor networks underlying freezing of gait

Functional connectivity-based classification of rapid eye movement sleep behavior disorder

BACKGROUND

Isolated rapid eye movement sleep behavior disorder (iRBD) is a clinically important parasomnia syndrome preceding α-synucleinopathies, thereby prompting us to develop methods for evaluating latent brain states in iRBD. Resting-state functional magnetic resonance imaging combined with a machine learning-based classification technology may help us achieve this purpose.

METHODS

We developed a machine learning-based classifier using functional connectivity to classify 55 patients with iRBD and 97 healthy elderly controls (HC). Selecting 55 HCs randomly from the HC dataset 100 times, we conducted a classification of iRBD and HC for each sampling, using functional connectivity. Random forest ranked the importance of functional connectivity, which was subsequently used for classification with logistic regression and a support vector machine. We also conducted correlation analysis of the selected functional connectivity with subclinical variations in motor and non-motor functions in the iRBDs.

RESULTS

Mean classification performance using logistic regression was 0.649 for accuracy, 0.659 for precision, 0.662 for recall, 0.645 for f1 score, and 0.707 for the area under the receiver operating characteristic curve (p < 0.001 for all). The result was similar in the support vector machine. The classifier used functional connectivity information from nine connectivities across the motor and somatosensory areas, parietal cortex, temporal cortex, thalamus, and cerebellum. Inter-individual variations in functional connectivity were correlated with the subclinical motor and non-motor symptoms of iRBD patients.

CONCLUSIONS

Machine learning-based classifiers using functional connectivity may be useful to evaluate latent brain states in iRBD.

An Explainable Geometric-Weighted Graph Attention Network for Identifying Functional Networks Associated with Gait Impairment

One of the hallmark symptoms of Parkinson's Disease (PD) is the progressive loss of postural reflexes, which eventually leads to gait difficulties and balance problems. Identifying disruptions in brain function associated with gait impairment could be crucial in better understanding PD motor progression, thus advancing the development of more effective and personalized therapeutics. In this work, we present an explainable, geometric, weighted-graph attention neural network (xGW-GAT) to identify functional networks predictive of the progression of gait difficulties in individuals with PD. xGW-GAT predicts the multi-class gait impairment on the MDS-Unified PD Rating Scale (MDS-UPDRS). Our computational- and data-efficient model represents functional connectomes as symmetric positive definite (SPD) matrices on a Riemannian manifold to explicitly encode pairwise interactions of entire connectomes, based on which we learn an attention mask yielding individual- and group-level explainability. Applied to our resting-state functional MRI (rs-fMRI) dataset of individuals with PD, xGW-GAT identifies functional connectivity patterns associated with gait impairment in PD and offers interpretable explanations of functional subnetworks associated with motor impairment. Our model successfully outperforms several existing methods while simultaneously revealing clinically-relevant connectivity patterns. The source code is available at https://github.com/favour-nerrise/xGW-GAT.

An Explainable Geometric-Weighted Graph Attention Network for Identifying Functional Networks Associated with Gait Impairment

One of the hallmark symptoms of Parkinson's Disease (PD) is the progressive loss of postural reflexes, which eventually leads to gait difficulties and balance problems. Identifying disruptions in brain function associated with gait impairment could be crucial in better understanding PD motor progression, thus advancing the development of more effective and personalized therapeutics. In this work, we present an explainable, geometric, weighted-graph attention neural network (xGW-GAT) to identify functional networks predictive of the progression of gait difficulties in individuals with PD. xGW-GAT predicts the multi-class gait impairment on the MDS-Unified PD Rating Scale (MDS-UPDRS). Our computational- and data-efficient model represents functional connectomes as symmetric positive definite (SPD) matrices on a Riemannian manifold to explicitly encode pairwise interactions of entire connectomes, based on which we learn an attention mask yielding individual- and group-level explain-ability. Applied to our resting-state functional MRI (rs-fMRI) dataset of individuals with PD, xGW-GAT identifies functional connectivity patterns associated with gait impairment in PD and offers interpretable explanations of functional subnetworks associated with motor impairment. Our model successfully outperforms several existing methods while simultaneously revealing clinically-relevant connectivity patterns. The source code is available at https://github.com/favour-nerrise/xGW-GAT.