Modern imaging technology enables measurement of optic nerve head (ONH) structure with increasingly high resolution. Yet, the color of the neuroretinal rim is one important aspect of the traditional ophthalmoscopic evaluation of the ONH that is not quantified by most current devices.
Multispectral/hyperspectral imaging techniques measure wavelength-specific differences in light absorption, which, when applied to the ONH, can be related to levels of hemoglobin
1 and to visual function in glaucoma.
2 Gonzalez de la Rosa et al.
3 should be commended on their attempt to adapt readily available red-green-blue fundus photography for estimating hemoglobin levels in the ONH. Despite the substantial overlap in spectral sensitivity between the color channels of typical digital camera sensors, the authors appear to have been successful in quantifying “redness” of ONH tissue, which may relate to hemoglobin levels. This technique therefore has considerable clinical appeal. However, if it is to be useful as a diagnostic tool it must add information to that already available from existing devices. This could be by identifying regions of neuroretinal rim that, while structurally normal, may be abnormal in their hemoglobin content.
Unfortunately, the image analysis carried out by Gonzalez de la Rosa et al.
3 failed to take into account the largest signal difference in their images, that between the neuroretinal rim and optic cup tissue. As is clearly apparent in their Figure 2, the difference in redness between the cup and the neuroretinal rim far exceeds any differences between regions of the neuroretinal rim. As such, when the pixel hemoglobin values are averaged over areas of the ONH that contain both neuroretinal rim and cup tissue (Fig. 3), by far the largest contributor to the measurement will be the area of optic cup included. It is therefore likely that the analysis simply represents an alternative, color-based method for the measurement of cup/disc ratio.
That the most diagnostically useful hemoglobin estimates came from sectors around the border of the cup in the superior and inferior poles of the ONH further supports this hypothesis because these would be the areas of the ONH most likely affected by cup enlargement due to glaucoma (for review see Broadway et al.
4 ). Further support can be found in the high diagnostic agreement and correlation between hemoglobin estimates and vertical cup/disc ratio in Tables 3 and 4. Since the studied populations appear easily separable on the basis of cup/disc ratio (Table 1), the high diagnostic ability of the hemoglobin estimates is not surprising, especially in the context of a receiver operating characteristic area of 0.95 for vertical cup/disc ratio (Table 2).
In our recent study of multispectral imaging of the ONH in glaucoma we used the Heidelberg Retina Tomograph (Heidelberg Engineering, Germany) to segment the cup region according to surface depth and isolated the neuroretinal rim tissue on this basis.
2 While we concede that this method is imperfect in separating the two tissues, it allows for a more reasonable analysis of where this technique may be useful, that is in identifying differences in light absorption/hemoglobin level between different regions of the neuroretinal rim.
I reiterate the appeal of the Laguna ONhE in its use of readily available technology, and in its possible applicability to existing series of color photographs. However, I urge the authors to investigate their measurements on images from which the optic cup has been removed so that the potential of the technique to add useful clinical information may be better evaluated.