Multiview Feature Learning With Multiatlas-Based Functional Connectivity Networks for MCI Diagnosis

Abstract:

Resting-state functional magnetic resonance imaging (rs-fMRI) functional connectivity (FC) networks can diagnose Alzheimer’s disease and its prodromal stage, mild cognitive impairment (MCI). FC is estimated as a temporal correlation of regional mean rs-fMRI signals between any two brain regions parcellated with a brain atlas. Most studies use a brain atlas for all subjects. FC networks always ignore subject-specific information, especially brain parcellation. FC networks based on a single atlas may not be able to reveal the complex differences between normal controls and disease-affected patients due to atlas bias, similar to the drawback of “single view” versus “multiview” learning in medical image-based classification. We propose multiview feature learning with multiatlas-based FC networks to improve MCI diagnosis. A three-step transformation generates multiple individually specified atlases from the standard automated anatomical labeling template, from which exemplars are selected. Based on these preselected atlas exemplars, multiple FC networks are created to provide multiple FC network-based feature representations for each subject. We create a multitask learning algorithm for joint feature selection from multiple FC networks. Multiatlas-based MCI diagnosis uses a support vector machine classifier and selected features. The proposed method is extensively compared to other methods, including the single-atlas-based method. Our method significantly improves MCI classification, showing promise in brain connectome-based individualized brain disease diagnosis.

Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.