Advancing Machine Learning for Neuroimaging Through Topology-Aware Signal Processing

The proposed work plan is to develop machine learning techniques to work directly with graph signal processing techniques in the context of neuroimaging. The goal is to leverage information in the form of the topology of the signal sensors or measurement locations to refine the neural signal representations in order to improve the statistical power of tests for distinguishing differences between conditions or stimuli. The project’s scope includes the formulation, mathematical and statistical analysis, and initial validation of the proposed methodology.

Outcomes

• NER'21: Hassan Baker, A. J. Brockmeier “Local and Sparse Linear Causal Models for fMRI Resting-State Signals” Full-text (IEEE) | Poster PDF
• Machine Learning for Health (ML4H): Hassan Baker, A. J. Brockmeier “Exploring latent networks in resting-state fMRI using voxel-to-voxel causal modeling feature selection” Paper, Poster
• Hassan Baker’s poster presentation at 2022 DARWIN Computing Symposium, Poster 1.5 minute description
  

Acknowledgements

Research was carried out with the support of the University of Delaware Research Foundation. Hassan Baker acknowledges support from the Gore Engineering Graduate Fellowship in Spring 2021. This research was supported in part through the use of Data Science Institute (DSI) computational resources at the University of Delaware, specifically the DARWIN system. The DARWIN computing project at the University of Delaware is supported by the National Science Foundation under Grant No. OAC-1919839. The authors thank the support from the University of Delaware IT Research Computing Group.