Multimodal neuroimaging of locus coeruleus-default mode network connectivity for predicting dexmedetomidine response
Yongqiong Tao1,2, Yonghong Zhou3, Yongtao Tang2
1Department of Radiology, Jinshan Hospital of Fudan University, Shanghai, China.
Objectives:
To explore whether pre-treatment functional connectivity between the locus coeruleus (LC) and default mode network (DMN), nodal graph-theoretical features, and gray matter volume can identify chronic insomnia patients who respond poorly to dexmedetomidine patient-controlled sleep therapy (PCSL).
Methods:
Chronic insomnia patients underwent PCSL with sub-anesthetic dexmedetomidine and were classified as responders or non-responders based on changes in the Pittsburgh Sleep Quality Index. All patients and matched good sleeper received pre-treatment resting-state functional Magnetic Resonance Imaging (MRI) and structural MRI. Functional connectivity, graph metrics, and gray matter volumes of LC and DMN regions were extracted. Predictors were selected using univariate analysis and Least Absolute Shrinkage and Selection Operator regression (LASSO), then used in a multivariate logistic model. Model performance was assessed via Receiver Operating Characteristic (ROC), calibration, decision curve analysis, and bootstrap validation.
Results:
No gray matter volume differences were found between responders and non-responders except increased right hippocampal volume in non-responders versus controls. Non-responders showed widespread increased LC-DMN connectivity pre-treatment, especially between right LC and left lateral temporal cortex versus responders. Graph metrics of key nodes (temporoparietal junction, lateral temporal cortex) were significantly reduced in non-responders. Three imaging features were included in the model, achieving an AUC of 0.816 with good calibration and clinical utility.
Conclusion:
Pre-treatment LC-DMN functional and network topology differences may underlie variable responses to dexmedetomidine PCSL in chronic insomnia. The multimodal imaging model effectively predicts treatment sensitivity and may guide personalized interventions.
