Reconfiguration of Dynamic Functional Connectivity States in Patients With Lifelong Premature Ejaculation

• Published in: Frontiers in neuroscience Vol. 15; p. 721236
• Main Authors: Lu, Jiaming, Chen, Qian, Li, Danyan, Zhang, Wen, Xing, Siyan, Wang, Junxia, Zhang, Xin, Liu, Jiani, Qing, Zhao, Dai, Yutian, Zhang, Bing
• Format: Journal Article
• Language: English
• Published: Lausanne Frontiers Research Foundation 13.09.2021; Frontiers Media S.A
• Subjects: Accuracy; Brain mapping; classification; dynamic functional connectivity; Ejaculation; Etiology; frequent transitions; Functional magnetic resonance imaging; Latency; Learning algorithms; Machine learning; Magnetic resonance imaging; Medical imaging; Neural networks; Neuroimaging; Neuropsychology; Neuroscience; Patients; premature ejaculation; Questionnaires; reoccurrence times; Vagina
• ISSN: 1662-453X; 1662-4548; 1662-453X
• DOI: 10.3389/fnins.2021.721236

Purpose: Neuroimaging has demonstrated altered static functional connectivity in patients with premature ejaculation (PE), while studies examining dynamic changes in spontaneous brain activity in PE patients are still lacking. We aimed to explore the reconfiguration of dynamic functional connectivity (DFC) states in lifelong PE (LPE) patients and to distinguish LPE patients from normal controls (NCs) using a machine learning method based on DFC state features. Methods: Thirty-six LPE patients and 23 NCs were recruited. Resting-state functional magnetic resonance imaging (fMRI) data, the clinical rating scores on the Chinese Index of PE (CIPE), and intravaginal ejaculatory latency time (IELT) were collected from each participant. DFC was calculated by the sliding window approach. Finally, the Lagrangian support vector machine (LSVM) classifier was applied to distinguish LPE patients from NCs using the DFC parameters. Two DFC state metrics (reoccurrence times and transition frequencies) were introduced and we assessed the correlations between DFC state metrics and clinical variables, and the accuracy, sensitivity, and specificity of the LSVM classifier. Results: By k-means clustering, four distinct DFC states were identified. The LPE patients showed an increase in the reoccurrence times for state 3 ( p < 0.05, Bonferroni corrected) but a decrease for state 1 ( p < 0.05, Bonferroni corrected) compared to the NCs. Moreover, the LPE patients had significantly less frequent transitions between state 1 and state 4 ( p < 0.05, uncorrected) while more frequent transitions between state 3 and state 4 ( p < 0.05, uncorrected) than the NCs. The reoccurrence times and transition frequencies showed significant associations with the CIPE scores and IELTs. The accuracy, sensitivity, and specificity of the LSVM classifier were 90.35, 87.59, and 85.59%, respectively. Conclusion: LPE patients were more inclined to be in DFC states reinforced intra-network and inter-network connection. These features correlated with clinical syndromes and can classify the LPE patients from NCs. Our results of reconfiguration of DFC states may provide novel insights for the understanding of central etiology underlying LPE, indicate neuroimaging biomarkers for the evaluation of clinical severity of LPE.