Classification of Functional Networks Poster
Download as PPTX, PDF1 like158 views
1) The document discusses developing computational fingerprints to define functional brain networks and noise in fMRI data using spatial and temporal features. 2) Features distinguishing noise are dominated by temporal features, while both spatial and temporal features are used to classify networks and noise. 3) More work is needed, including using sub-networks, to better diagnose neurological disorders from fMRI data.
Classification of Functional Networks Poster
- 1. • Noisy components can be computationally defined using spatial and temporal features • Features to distinguish noise are dominated by temporal features • More work is needed using sub-networks to diagnose disorder Preprocessing Tool displays spatial map, time-course, and frequency distribution to assist user in defining a standard for a component type Functional MRI Developing Fingerprints to Computationally Define Functional Brain Networks and Noise V.Sochat, Biomedical Informatics, Stanford University School of Medicine, Stanford CA N = 818 (700) N=46 (1472) N=40 (1478) N=67 (1451) N=48 (1470) Introduction Realign / Reslice Motion Correction Segmentation Smoothing Filtering Normalization ICA n x m n x n n x m Bad Good Bad 740 121 Good 78 579 N = 1518 Networks spatial maps and timecourses DATA INDEPENDENT COMPONENT ANALYSIS SPATIAL AND TEMPORAL FEATURES Methods STANDARD DEVELOPMENT Results Discussion and Conclusion FUNCTIONAL NETWORK AND NOISE FINGERPRINTS EVALUATION OF CLASSIFIERS Lasso L1 constrained linear regression selects features to distinguish real from noisy components (N=1518) with cross validation accuracies of .8689, .9834, .9808, .9675, and .9695 respectively. ALL NOISE EYEBALLS HEAD MOTION WHITE MATTER PARIETO OCCIPITAL CORTEX • Independent Component Analysis ICA is a data-driven method to decompose functional neuroimaging data into independent components. • The decomposed independent components encompass a mix of true neural signal, machine artifact, motion, and physiological noise that are typically visually distinguished. • Neurological disorders are beginning to be understood based on aberrant brain structure and function on the single network level. • Methods to computationally define noise and networks are necessary to automatically filter large publicly available datasets and identify patterns of fMRI that distinguish disorder. 111 Temporal Features signal metrics, peaks, kurtosis, skewness, entropy, amplitudes, power bands, HPSD, auto correlation etc. 135 Spatial Features Regional activation, matter types, kurtosis, entropy, skewness, degree of clustering • 53 resting BOLD functional magnetic resonance imaging data-sets • 24 Healthy Control / 29 Schizophrenia PRELIMINARY WORK WITH UNSUPERVISED CLUSTERING OF SUBNETWORKS • 8739 subnetworks extracted with higher dimensionality ICA, filtered to 3184 • Unsupervised clustering of filtered networks reveals new type of noise