Eur Radiol. 2018 Mar 29. doi: 10.1007/s00330-018-5365-7.
Li L1,2, Fan W1,2, Li J1,2, Li Q3, Wang J4, Fan Y5, Ye T1,2, Guo J4, Li S4, Zhang Y4, Cheng Y4, Tang Y4, Zeng H4, Yang L6,7, Zhu Z8.
Abstract
OBJECTIVES:
To investigate the cerebral structural changes related to venous erectile dysfunction (VED) and the relationship of these changes to clinical symptoms and disorder duration and distinguish patients with VED from healthy controls using a machine learning classification.
METHODS:
45 VED patients and 50 healthy controls were included. Voxel-based morphometry (VBM), tract-based spatial statistics (TBSS) and correlation analyses of VED patients and clinical variables were performed. The machine learning classification method was adopted to confirm its effectiveness in distinguishing VED patients from healthy controls.
RESULTS:
Compared to healthy control subjects, VED patients showed significantly decreased cortical volumes in the left postcentral gyrus and precentral gyrus, while only the right middle temporal gyrus showed a significant increase in cortical volume. Increased axial diffusivity (AD), radial diffusivity (RD) and mean diffusivity (MD) values were observed in widespread brain regions. Certain regions of these alterations related to VED patients showed significant correlations with clinical symptoms and disorder durations. Machine learning analyses discriminated patients from controls with overall accuracy 96.7%, sensitivity 93.3% and specificity 99.0%.
CONCLUSIONS:
Cortical volume and white matter (WM) microstructural changes were observed in VED patients, and showed significant correlations with clinical symptoms and dysfunction durations. Various DTI-derived indices of some brain regions could be regarded as reliable discriminating features between VED patients and healthy control subjects, as shown by machine learning analyses.
KEY POINTS:
• Multimodal magnetic resonance imaging helps clinicians to assess patients with VED. • VED patients show cerebral structural alterations related to their clinical symptoms. • Machine learning analyses discriminated VED patients from controls with an excellent performance. • Machine learning classification provided a preliminary demonstration of DTI’s clinical use.
KEYWORDS:
Machine-learning classification; Multimode magnetic resonance imaging; TBSS; VBM; Venous erectile dysfunction
PMID: 29600478
DOI: 10.1007/s00330-018-5365-7
;