While visual field measurements continue to be a primary tool in the diagnosis and management of glaucoma, a poor reproducibility
and overall patient experience continues to motivate the need for more objective and patientfriendly structural measurements.
Spectraldomain optical coherence tomography (SDOCT) is a promising modality for providing such parameters, with retinal nerve
fibre layer (RNFL) thicknesses, ganglion cell layer (GCL) thicknesses and cuptodisc ratio measurements already being used.
Nevertheless, with the limited correlation of current structural parameters with functional parameters, there is a continued
interest in obtaining better structural parameters.
Thus, in a collaborative effort at The University of Iowa (also including glaucoma specialist Dr Young Kwon, retinal specialist
and engineer Dr Michael AbrÓmoff, and engineers Dr Milan Sonka, Dr Kyungmoo Lee and Dr Meindert Niemeijer), we have recently
been examining the use of nerve bundle patterns from SD-OCT. As a first step in this direction, we have demonstrated that
a colourcoded correlation map resembling expected nerve fibre bundle patterns emerges through a structure–structure correlative
analysis of the GCL and RNFL from SDOCT across subjects.1 The work, published in Investigative Ophthalmology & Visual Science,1 first involves using our three-dimensional image analysis methodology to obtain regional thickness measurements of the GCL
in 66 macular locations and regional thickness measurements of the RNFL in 12 peripapillary wedge locations from each subject.
Then, the resulting colour-coded correlation map is obtained by colour-coding each macular GCL region with the most correlative
peripapillary wedge region.
Using 3D image analysis to obtain regional GCL and RNFL thickness measurements
To perform a structure–structure correlative analysis,1 we first obtain the thickness of the GCL in 66 macular grid regions and the thickness of the RNFL in 12 peripapillary wedge
regions from the SDOCT volumes of each subject. In particular, the boundaries of the layers themselves are automatically detected
within the SDOCT volumes using our previously reported threedimensional graph-theoretic approach.2,3 The regions of interest are defined by first aligning (i.e., registering) the macula-centred volume to the opticnerveheadcentred
volume of each eye. After registration, the precise location of 66 macular grid regions and 12 peripapillary wedge regions
can be defined, using the position of the fovea and automatically segmented neural canal opening4 for guidance (Figure 1, top). Mean layer thicknesses are then computed within each macular and peripapillary region for
each eye. Note that as the process of obtaining regional thickness measurements depends on a number of underlying ophthalmic
image analysis algorithms, interested readers are encouraged to refer to a recent review paper5 for a broad overview of the field.
Figure 1: Correlating regional macular ganglion cell layer (GCL) thickness measurements with regional peripapillary retinal
nerve fibre layer (RNFL) thickness measurements. (Top) Definition of 66 macular GCL grid regions and 12 peripapillary RNFL
wedge regions for each eye (illustrated on macula-centred and optic-nerve-head-centred SD-OCT projection images) using fovea
and neural canal opening for placement. (Bottom) Example plot of the RNFL thickness in the wedge region indicated with a star
versus the GCL thickness of the grid region indicated with a circle across all 57 subjects.