Functional connectome fingerprint of sleep quality in insomnia patients: Individualized out-of-sample prediction using machine learning

Abstract

ObjectivesInsomnia disorder has been reclassified into short-term/acute and chronic subtypes based on recent etiological advances. However, understanding the similarities and differences in the neural mechanisms underlying the two subtypes and accurately predicting the sleep quality remain challenging. MethodsUsing 29 short-term/acute insomnia participants and 44 chronic insomnia participants, we used whole-brain regional functional connectivity strength to predict unseen individuals’ Pittsburgh sleep quality index (PSQI), applying the multivariate relevance vector regression method. Evaluated using both leave-one-out and 10-fold cross-validation, the pattern of whole-brain regional functional connectivity strength significantly predicted an unseen individual’s PSQI in both datasets. ResultsThere were both similarities and differences in the regions that contributed the most to PSQI prediction between the two groups. Further functional connectivity analysis suggested that between-network connectivity was re-organized between short-term/acute insomnia and chronic insomnia. ConclusionsThe present study may have clinical value by informing the prediction of sleep quality and providing novel insights into the neural basis underlying the heterogeneity of insomnia.