Graph-Based Disease Prediction in Neuroimaging: Investigating the Impact of Feature Selection.

In biomedical machine learning, data often appear in the form of graphs. Biological systems such as protein interactions and ecological or brain networks are instances of applications that benefit from graph representations. Geometric deep learning is an arising field of techniques that has extended deep neural networks to non-Euclidean domains such as graphs. In particular, graph convolutional neural networks have achieved advanced performance in semi-supervised learning in those domains. Over the last years, these methods have gained traction in neuroscience as they could be the key to a deeper understanding in clinical diagnosis at the systems or network level (for an individual brain but also for across a cohort of subjects). As a proof-of-principle, we study and validate a previous implementation of graph-based semi-supervised classification using a ridge classifier and graph convolutional neural networks. The models are trained on population graphs that integrate imaging and phenotypic information. Our analysis employs neuroimaging data of structural and functional connectivity for prediction of neurodevelopmental and neurodegenerative disorders. Here, we particularly study the effect of different strategies to reduce the dimensionality of the neuroimaging features on the graph nodes on the classification performance.

Mixed anxiety-depressive disorder in Parkinson's disease associated with worse resting state functional response to deep brain stimulation of subthalamic nucleus.

Background: Parkinson's disease (PD), even though generally perceived as a dominantly motor disorder, is associated with a wide range of non-motor symptoms, including mixed anxiety-depressive disorder (MADD). Objectives: The aim of the presented study was to determine whether deep brain stimulation (DBS) of the subthalamic nucleus (STN) brings the functional characteristics of non-motor networks closer to the condition detected in healthy population and whether pre-DBS presence of MADD in PD patients was associated with different reaction to this therapeutic modality. Methods: Resting-state fMRI signature elicited by STN DBS activation and deactivation in 81 PD patients was compared against healthy controls, with the focus on measures of efficiency of information processing and localised subnetwork differences. Results: While all the MRI metrics showed statistically significant differences between PD patients in DBS OFF condition and healthy controls, none were detected in such a comparison against DBS ON condition. Furthermore, in the post-DBS evaluation, PD patients with MADD in the pre-DBS stage showed no differences in depression scales compared to pre-DBS psychiatrically intact PD patients, but still exhibited lower DBS-related connectivity in a subnetwork encompassing anterior and posterior cingulate, dorsolateral prefrontal and medial temporal cortices. Conclusions: STN DBS improved all the metrics of interest towards the healthy state, normalising the resting-state MRI signature of PD. Furthermore, pre-DBS presence of MADD, even though clinically silent at post-DBS MRI acquisition, was associated with lower DBS effect in areas highly relevant for depression. This finding points to a possibly latent nature of post-DBS MADD, calling for caution in further follow-up of these patients.