advertisement

Editors Selection IGR 19-4

Anatomical Structures: Detecting Macular Damage in Glaucoma

Comment by Andrew Tatham on:

76858 Evaluation of a Region-of-Interest Approach for Detecting Progressive Glaucomatous Macular Damage on Optical Coherence Tomography, Wu Z; Weng DSD; Weng DSD; Weng DSD; Thenappan A et al., Translational vision science & technology, 2018; 7: 14

Find related abstracts

There is growing appreciation of the importance of macular imaging in glaucoma, with several studies showing macular changes are common, even in early disease, and that glaucoma may be missed if retinal nerve fibre layer (RNFL) analysis is used alone.^1,2 Macular changes and central vision are also of particular importance for daily function and vision-related quality of life. At present, detection of glaucoma progression is overly reliant on global indices, which fail to take account of localized changes and prior knowledge of patterns of glaucomatous damage. Wu and colleagues hypothesized that accuracy of detection of progression could be improved by evaluating changes in regions of observed or suspected glaucomatous damage, using a region(s)-ofinterest (ROI) approach.³

At present, detection of glaucoma progression is overly reliant on global indices, which fail to take account of localized changes and prior knowledge of patterns of glaucomatous damage

The study compared two methods of identifying ROIs; (1) an automatic approach, which considered ROIs those with significantly lower ganglion cell complex (GCC) thickness compared to normative limits, exceeding a prespecified area (288 um²); and (2) a manual approach, where ROIs for each eye were determined manually using the full wealth of OCT images, including an en-face projection image, macular RNFL and GCC thickness plots, and thickness deviation probability plots. ROI approaches were compared to the more common method of progression analysis; change in global macular GCC thickness.

One of the most important aspects of detecting progression is the ability to differentiate true change from the noise of test-retest variability. The ideal method needs a high signal to noise ratio (SNR)4 and so SNR was chosen as the primary outcome for comparison of approaches. SNR was determined by examining test-retest differences in each ROI and calculating region-specific estimates of variability. Region-specific age-related rates of change in GCC thickness were also calculated using a cohort of healthy eyes. As ROIs are unique to each eye, region-specific estimates of variability and age-related change had to be calculated individually.

The study's central finding was that the manual ROI method had a significantly more negative SNR (-1.03 y^-1) than the automated ROI (-0.91 γ^-1) and global GCC thickness (-0.90 y^-1) methods. A more negative SNR indicates a greater extent of GCC thickness loss relative to age-related change and measurement variability, and therefore the manual ROI was the best method of detecting longitudinal change in macular GCC, with automated ROI and global GCC thickness analysis performing similarly.

Although the study did not evaluate the clinical implications of the different approaches and was limited by use of within session estimates of variability, which are likely to be lower than between-session estimates, it nevertheless demonstrates that detection of progression can be improved by evaluating change in regions of observed or suspected glaucomatous macular damage or 'regions of interest'. The manual approach of identifying ROIs is likely to have been superior to automated identification of ROIs as it was based on careful qualitative evaluation of all the available information from the OCT images. The automated method of ROI identification may have failed to detect regions of genuine damage that remained statistically within normative limits or regions of damage not meeting the predefine minimum area. Although manual definition of regions of interest performed best it has the disadvantage of being more time consuming than automated identification, it is, however, possible that automated methods may be improved using alternative definitions of abnormality.

The study provides further evidence of the importance of qualitative evaluation of OCT images and suggests that trend or event-based analyses of macular GCC are more likely to be effective if performed in regions of observed or suspected glaucomatous macular damage.

• ← Previous
• Next →