Imaging of the posterior segment of the eye with spectral-domain optical coherence tomography (SD-OCT) has revolutionized the field of ophthalmology, especially with regard to disease of the retina and glaucoma. We are now able to image and view, in great detail, the main anatomical structures of interest in glaucoma, namely the optic nerve head (ONH), the retinal nerve fiber layer (RNFL), and posterior pole of the eye. However, many challenges remain to be addressed. The SD-OCT images provide a huge amount of data that need to be processed and not all processing can be or has been fully automated. Resolution of the images is not always optimal and enhancement of the images is highly desirable under certain circumstances. Images from different machines do not exactly match, making inter-device comparison impossible at this point. Some structures of interest are sometimes located more deeply leading to attenuation and inadequate penetration of the laser beam and poor image quality. Retinal vessels can cause significant shadowing behind them with attenuation or obliteration of all structural details of the underlying tissues.
Here are a few challenges that I think need to be addressed so that further significant progress can be achieved in this field.
• Delineating the posterior boundary of the lamina cribrosa, one of the important deep layers of the ONH through which the optic nerve axons pass before forming the optic nerve. Lamina cribrosa is currently considered to be the site of glaucoma damage.
• Adequate segmentation of the macular retinal layers
o Our group is developing our own version of segmentation software that would be applicable to various devices and potentially future technology.
• Modeling the posterior eye for predicting the axonal complement of a given eye (RNFL cross-sectional area)
o None of the current hypotheses can adequately predict the cross-sectional RNFL area (a proxy for the number of retinal ganglion cell axons) in a given human eye. There is significant controversy whether the posterior opening of the eye thru which the axons pass (i.e., the size of the optic nerve head) is a good predictor for the number of axons in a healthy eye.
• Image intensity values as a valid outcome measure and normalization issues
o In addition to thickness measurements derived from OCT images, image intensity values (e.g., RNFL reflectance) have been explored as a potential outcome of interest in glaucoma. However, significant inter-individual variation in layer intensity variations exists and some form of normalization needs to be done although there is no consensus on the best reference layer for this purpose.
• Correcting the RNFL thickness measurements and TSNIT curve for location of the main retinal vascular branches
o The RNFL thickness peaks closely follow the location of retinal major vasculature but there are other factors that affect the location of such peaks
• Image enhancement options before segmentation
Back to Mathematical Challenges in Ophthalmology