A case-control study of volumetric segmentation of extraocular muscles from MRI images in thyroid eye disease.

PurposeThis study aimed to develop a reproducible manual segmentation method using a computer-assisted technique and to (1) compare extraocular muscle volumes between TED patients and healthy controls and (2) assess the intra- and inter-observer reliability of this method.MethodsA retrospective study of MRI scans was performed on 13 patients previously diagnosed with TED and 16 healthy controls. Image segmentation and the derived muscle volume measurements were performed by two independent observers using the software 3D slicer. The groups compared muscle volumes, and intra- and inter-observer reproducibility was assessed.ResultsA statistically significant difference in muscle volume was demonstrated; the mean medial and lateral recti volumes in the TED group were 1342.75 and 1066.7 mm3, respectively, compared to 787.67 and 764.3 mm3 in the control group. The segmentation protocol appears to have good inter-observer reliability. The intraclass correlation coefficient was good for inter-observer measures of the medial recti and intra-observer measures of both horizontal recti.ConclusionThis volumetric analysis verifies that structural MRI contains useful information regarding muscle volume in patients with TED compared to healthy controls. Building a dataset of segmented MRI volumes will help facilitate the creation of supervised machine-learning models for automated volumetric analysis of the extraocular muscles.