Interobserver repeatability of heidelberg retinal flowmetry using pixel-by-pixel analysis

To determine the interobserver reproducibility of Heidelberg retinal flowmeter (HRF) blood flow measurements using independently selected study areas for pixel-by-pixel analysis.

Blood flow measurements were performed on 257 scans from 15 patients, 14 of whom had glaucoma or ocular hypertension. HRF was used to record capillary perfusion in a 2560x640 mum area of the supratemporal peripapillary region and pixel-by-pixel analysis was performed from an area adjacent to the optic disc with a minimum of 1600 pixels. Each observer independently selected the area for analysis. The percentage of pixels with <1 arbitrary unit of flow (no flow) and 10, 25, 50, 75, and 90th percentiles of flow values was calculated. Interobserver variability was assessed by estimating the intraclass correlation coefficient (ICC) and its 95% confidence interval. Bland-Altman plots of the difference between the 2 physicians versus the average of the 2 physicians for each outcome were created.

ICC was 0.79 (range: 0.74 to 0.83) for mean flow values. For 0, 10, 25, 50, 75, and 90th percentiles of flow, the ICC was 0.67 (0.60 to 0.73), 0.74 (0.68 to 0.79), 0.82 (0.78 to 0.86), 0.85 (0.82 to 0.88), 0.85 (0.81 to 0.88), and 0.77 (0.72 to 0.82), respectively. Zero flow pixels had a nonsignificant mean difference between observers (P=0.542), whereas the remainder of the flow values demonstrated significant mean differences.

This study demonstrates that independent observers can review high-quality HRF scans and may produce different absolute values while retaining strong consistency of agreement when independently selecting areas for analysis using the pixel-by-pixel method.