Abstract #3913

Spectral Entropy: A Useful Engineered Feature for Classification of fMRI Data Quality?

Christopher O'Grady^1,2,3, Steve Patterson⁴, Alessandro Guida², James Rioux², Antonina Omisade⁵, Javeria Hashmi^3,6,7, and Steven Beyea^6,8,9

¹Department of Research, Nova Scotia Health Authority, Halifax, NS, Canada, ²Biomedical Translational Imaging Centre, Halifax, NS, Canada, ³Brain Networks & Neurophysiology Lab, Halifax, NS, Canada, ⁴Nova Scotia Department of Health and Wellness, Halifax, NS, Canada, ⁵Acquired Brain Injury, Nova Scotia Health Authority, Halifax, NS, Canada, ⁶Department of Physics and Atmospheric Science, Dalhousie University, Halifax, NS, Canada, ⁷Department of Medical Neuroscience, Dalhousie University, Halifax, NS, Canada, ⁸School of Biomedical Engineering, Dalhousie University, Halifax, NS, Canada, ⁹Department of Diagnostic Radiology, Dalhousie University, Halifax, NS, Canada

This work evaluates the use of regularized spectral entropy (rSpecEn) as a potentially useful engineered feature for the classification of fMRI scan quality in real time. rSpecEn relies on the frequency sparsity of task-based fMRI signals, and a data-driven regularization method is employed to counteract unavoidable noise. Comparison to known measures of quality are used, as well as the use of fMRI scans with intentionally degraded quality. Sensitivity to fMRI data quality measures and an exploration of high accuracy scan quality classification are presented in this feasibility study.

This abstract and the presentation materials are available to members only; a login is required.

Join Here