Published in:
01-04-2019 | Original Paper
Anterior segment ultrasound biomicroscopy image analysis using ImageJ software: Intra-observer repeatability and inter-observer agreement
Published in: International Ophthalmology | Issue 4/2019
Login to get accessAbstract
Purpose
In this novel study, we demonstrate a standardized imaging and measurement protocol of anterior segment (AS) structures with reliability analysis using ultrasound biomicroscopy (UBM) and ImageJ software.
Methods
Ten pediatric and young adult patients undergoing examination under anesthesia for AS pathology were imaged using UBM. Four trained observers analyzed 20 images using ImageJ. Forty-five structural parameters were measured. Those that relied on the trabecular-iris angle (TIA) as a reference landmark were labeled TIA-dependent (TD) and all others were labeled non-TIA dependent (NTD). Intra-observer repeatability (IOR) and inter-observer agreement (IOA) of measurements were determined using coefficient of variation (CV) and intra-class correlation (ICC) followed by assessment of Bland–Altman plots (BAP) for each pair of observers, respectively.
Results
For NTD parameters, non-ciliary body (CB) related measurements showed CV range 0.60–16.22% and ICC range 0.84–0.89, whereas CB-related parameters showed CV range 2.86–23.40% and ICC range 0.29–0.92. For TD parameters, parameters < 2 degrees removed from reference showed CV range 0.02–5.40% and ICC range 0.89–1.00, whereas parameters > 1 degree removed showed CV range 0.63–27.44% and ICC range 0.22–1.00. No systematic proportional bias was detected by BAPs.
Conclusions
Preplaced landmarks yielded good IOR and IOA in quantitative assessment of AS structures that were NTD and non-CB-related or less removed from the reference. CB-related NTD measurements varied greatly in IOR and IOA, indicating protocol modifications or CB qualitative assessments needed to improve accuracy. Variability in TD measurements increased the further removed from the reference, which supports implementation of a reliable reference landmark to minimize variation.