Identifying neuroimaging biomarkers for psychogenic erectile dysfunction by fusing multi-level brain information: A resting-state functional magnetic resonance imaging study.

Background: Psychogenic erectile dysfunction (pED) patients who are under their 40s in China consist of a major component of erectile dysfunction. Existing neuroimaging studies have demonstrated that pED is a functional disorder with aberrant neural representations on the local level, the regional level, and the global level, respectively. Therefore, it is reasonable to incorporate brain information from all these levels simultaneously into consideration when identifying neuroimaging biomarkers for pED. However, no such endeavors have been made in previous studies to fully disclose the central mechanism of pED.

Method: To incorporate multi-level brain features to fully explore the neural representation of pED, a novel machine learning framework was proposed in the current study. Specifically, we used amplitude of low-frequency fluctuation, regional homogeneity, and degree centrality as indices for local, regional, and global brain activity, respectively. A fully data-driven method, that is, support vector machine (SVM) with recursive feature elimination analyses, was used to investigate discriminative brain map between 48 pED patients and 39 healthy control subjects for resting state functional magnetic resonance imaging (rs-fMRI) data.

Results: By fusing multi-level brain features, our method led to a superb classification accuracy of 95.12% between two groups. Interestingly, the right anterior cingulate gyrus and the left precuneus showed abnormal representations at different levels simultaneously in pED patients, which also explicated highest discriminative power between groups. Moreover, the right insular, the left fusiform gyrus, the right inferior temporal gyrus, the right superior frontal gyrus, the right precentral gyrus, the bilateral parahippocampal gyrus, and the bilateral inferior frontal gyrus were discriminative for pED. Also, correlation analysis explicated that several core brain regions were associated with the clinical manifestations in pED patients.

Conclusion: This is one of the first study investigating brain alterations on different levels simultaneously in pED patients. Our results suggested that pED involves multi-level aberrant brain representations in multi-dimensional neurobehavioral components, which closely interrelated with cognitive and psychosocial factors, that is, attention, appraisal, emotion, and sensorimotor. Our findings are likely to help foster new insights into the pathophysiological mechanisms of pED and the aberrant brain regions may serve as potential therapeutic targets for targeted therapy for brain.