Quantitative evaluation of short-term resting-state brain networks for primary insomnia diagnosis

Chi Zhang,Lina Sun,Shuang Ge,Yi Chang,Mingyan Jin,Yang Xiao,Hanbing Gao,Lin Wang,Fengyu Cong

DOI: https://doi.org/10.1016/j.bspc.2022.103498

IF: 5.1

2022-04-01

Biomedical Signal Processing and Control

Abstract:Primary insomnia (PI) manifesting as insufficient and non-restorative sleep disturbs the function of central nervous system. Electroencephalogram (EEG), as a technique of recording the electrical signals of the brain, has demonstrated potential to access and quantify PI. However, most existing EEG indices rely on time–frequency analysis and separate channels, which limits its clinical application. In this study, we propose a novel quantitative evaluation method by introducing spatial information from resting-state brain networks of insomniacs to make rapid diagnosis implementable. To suppress false positive observations of coupling attributed to signal spread, the connections were binarized based on an adaptive threshold technology so that the statistical network characteristics were extracted automatically to form a comprehensive measurement index. The clinical experiments proved that the specificity of PI brain networks could be quantified objectively by the comprehensive index in the resting state. PI specificity showed consistency across the connectivity estimated in time (Pearson Correlation Coefficient, PCC), phase (Phase Lag Index, PLI) and frequency (Granger Causality, GC) domains. All the three kinds of connectivity revealed the significant difference between the PI patients and normal subjects (PCC: p = 0.0021, PLI: p = 0.0071, GC: p = 0.0142). The strong connectivity of PI consistent with clinical rating scale indicates the hyperarousal of PI brain. It is difficult to achieve normal inhibition, so it consumes more resources in the resting state. Further, bidirectional long short-term memory (Bi-LSTM) network was applied to classify the healthy status (normal or PI) automatically and achieved 85% accuracy and 90% sensitivity, which demonstrated its potential for clinical diagnosis.

engineering, biomedical