Abstract

Dynamic functional connectivity (dFC) of the brain has been explored for the detection of mild cognitive impairment (MCI), preventing potential development of Alzheimer's disease. Deep learning is widely used method for dFC analysis but is unfortunately computationally expensive and unexplainable. Root mean square value (RMS) of the pairwise Pearson's correlation of the dFC is also proposed but is insufficient for accurate MCI detection. The present study aims at exploring the feasibility of several novel features for dFC analysis, and thus, reliable MCI detection. A public resting-state functional magnetic resonance imaging dataset containing healthy controls (HC), early MCI (eMCI), and late MCI (lMCI) patients was used. In addition to RMS, nine features were extracted from the pairwise Pearson's correlation of the dFC, inducing amplitude-, spectral-, entropy-, and autocorrelation-related features, and time reversibility. A Student's t-test and a least absolute shrinkage and selection operator (LASSO) regression were employed for feature dimension reduction. A SVM was then adopted for two classification objectives: HC vs. lMCI and HC vs. eMCI. Accuracy, sensitivity, specificity, F1-score, and area under the receiver operating characteristic curve were calculated as performance metrics. 6109 out of 66700 features are significantly different between HC and lMCI and 5905 between HC and eMCI. Besides, the proposed features produce excellent classification results for both tasks, outperforming most of the existing methods. This study proposes a novel and general framework for dFC analysis, providing a promising tool for the detection of many neurological brain diseases using different brain signals.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call