Glaucoma is an optic neuropathy characterized by progressive loss of retinal ganglion cells (RGCs) and their axons. Imaging techniques that objectively record and quantify structural changes in the optic nerve head (ONH), peripapillary RNFL, and macula are important tools that complement clinical examination and visual field testing2)3).
Structural changes often precede functional changes (visual field defects)3). It has been reported that detection of structural changes by OCT precedes the onset of visual field defects by about 2 years1). There are three types of imaging techniques2):
Systematic reviews have shown that these techniques have comparable ability to distinguish glaucomatous from normal eyes2). However, abnormal results (outside normal limits) do not necessarily indicate disease2)3). Normative database criteria vary between devices, and results can fall outside normal limits for reasons other than glaucoma.
QCan glaucoma be diagnosed by imaging alone?
A
No. Imaging is an adjunct to clinical diagnosis and should not be used alone to diagnose glaucoma4)5). An OCT result “outside normal limits” may be a false positive; all information including clinical findings and visual field tests must be integrated. The sensitivity and specificity of automated diagnostic programs are reported to be around 80%.
Stereoscopic color fundus photography is established as a qualitative recording method for optic disc appearance 2)3). Red-free illumination is useful for evaluating RNFL defects. Serial photographs can be used to detect changes in the optic disc over time 5).
However, in advanced glaucomatous cupping, there is little neural tissue left to evaluate, making it difficult to identify progressive changes with stereoscopic photographs 2). When the disc shape is bowl-like and blood vessels are sparse, the topography is difficult to discern in photographs, and a slit-lamp sketch is needed as an additional record.
The HRT (Heidelberg Retina Tomograph, Heidelberg Engineering) is a device that scans a diode laser (670 nm) to measure the three-dimensional topography of the optic disc1). It quantifies the surface topography of the disc and is also used to detect changes over time 4).
Moorfields Regression Analysis (MRA) performs a statistical assessment of rim area based on disc area 1). The Glaucoma Probability Score (GPS) does not require a reference plane or operator-defined disc margin and is an automated classification based on machine learning 1).
A limitation of HRT is that the definition of the disc margin on the fundus plane is not based on anatomical reference points. This issue was resolved by the OCT approach using the Bruch’s membrane opening (BMO) as a reference point 1). HRT production ended in the 2020s, and OCT has become mainstream in clinical practice 1).
The GDx (Carl Zeiss Meditec) is a device that measures phase retardation using the birefringence properties of the RNFL1). Microtubules within the axons of the RNFL are the main cause of birefringence, which correlates with RNFL thickness 1). Enhanced Corneal Compensation (ECC) technology corrects corneal birefringence.
However, it has been shown to be inferior to SD-OCT in longitudinal detection of glaucoma1), and its use has been discontinued with the widespread adoption of OCT.
OCT (Optical Coherence Tomography)
Principle: Uses low-coherence interferometry to image the cross-sectional structure of the retina4)5)
TD-OCT: Early OCT. A method that obtains cross-sectional images by stacking A-scans in one axis, with long examination time and low resolution. Currently rarely used.
SD-OCT: Achieves high speed and high resolution through spectral analysis. Enables fast analysis of the optic disc, RNFL, and macula at 26,000 A-scans/second or more. Multiple models exist, such as Cirrus OCT and Spectralis OCT.
SS-OCT: Uses a swept-source laser. Has high penetration depth and is also applied to analysis of the lamina cribrosa and choroid. Examples include the DRI OCT Triton (Topcon).
Key OCT Analysis Parameters
RNFL thickness: Measures the retinal nerve fiber layer thickness in a 3.46 mm circle around the optic disc1). This is the most widely used parameter.
BMO-MRW: Three-dimensionally measures the minimum rim width from the Bruch’s membrane opening 1). Uses anatomically accurate reference points and shows better diagnostic ability than conventional rim area measurement.
GCC/GC-IPL: Measures the thickness of the macular ganglion cell complex (GCC) or ganglion cell-inner plexiform layer (GC-IPL) 6). The floor effect occurs later than with RNFL even in advanced stages 5).
Deviation map: Displays deviations from the normal database for each parameter using color coding. It is recommended to evaluate both quantitative values and deviation maps 5).
Importance of image quality: High-quality baseline images are essential 4). Segmentation errors and artifacts are particularly frequent in highly myopic eyes and tilted optic discs 5).
Inter-device compatibility: Measurements are not interchangeable between different OCT devices 4)5). Use of the same device is mandatory for follow-up.
Limitations of normal database: The composition of the database varies by device. It is necessary to assess whether the age, race, and refractive distribution match the patient 1).
Interpretation in Special Situations
Myopic eyes: RNFL thickness is affected by the degree of myopia1). In myopia, peripapillary atrophy and changes in BMO position affect OCT measurements. Correction for axial length (e.g., Littmann formula) is desirable.
Optic disc size: In large discs, a large C/D ratio may be normal. BMO-MRW is less affected by disc size and shows better diagnostic ability than conventional methods in both large and small discs 1).
Racial differences: Most normal databases are composed of specific races (mainly Caucasians), and false positives and false negatives can occur in different races 1)
Parameter
Measurement site
Advantage
Limitation
RNFL thickness
Peripapillary
Widely validated
Early floor effect
BMO-MRW
Optic disc rim
Less affected by disc size
Limited to specific devices
GCC/GC-IPL
Macula
Late floor effect
Affected by macular diseases
QAre OCT measurements interchangeable between different devices?
A
They are not interchangeable. Because technical specifications, software, and normal databases differ between OCT devices, direct comparison of measurements is not possible 4)5). For follow-up, it is essential to use the same device and the same protocol.
QHow should OCT results be interpreted in myopic eyes?
A
In myopic eyes, RNFL thickness is affected, so caution is needed 1). Peripapillary atrophy, tilted optic disc, and displacement of the BMO position can cause segmentation errors and artifacts. Application of axial length correction and confirmation of segmentation on B-scan images are recommended.
Detection of structural changes by OCT may precede the onset of visual field defects 3). Studies have shown that OCT has a lead time of about 2 years in detecting visual field damage 1). On the other hand, there are visual field changes without structural progression and structural changes without visual field progression, and the agreement between the two is partial and moderate 4).
Both structural and functional assessments are essential for patient management and should be used complementarily 2)3).
Many commercially available OCT devices are equipped with progression analysis software, enabling quantification of the rate of progression 4)5). However, careful interpretation is needed due to measurement variability and the influence of non-glaucomatous changes (e.g., aging) 4). Since most commercial software does not correct for aging, a statistically significant slope does not necessarily indicate glaucomatous progression 4).
In advanced glaucoma, RNFL thickness reaches a “floor” and further progression is no longer reflected in thickness changes 5). Macular parameters (GCC/GC-IPL) are less susceptible to the floor effect than RNFL thickness, making them useful for evaluating advanced stages 1)5). It has been reported that vessel density on OCT-A may also have a later floor effect than RNFL in advanced stages 1).
Green disease: A condition where OCT comparison with the normal database indicates “within normal limits (green)” but the eye actually has glaucomatous changes. This tends to occur in cases outside the coverage of the normal database (e.g., large optic disc, high myopia, specific ethnicities) 1).
Red disease: A condition where OCT indicates “outside normal limits (red)” but the eye is actually not glaucomatous. This is caused by physiological individual differences or features not included in the normal database (e.g., small optic disc, specific ethnic differences) 1).
These phenomena demonstrate the limitations of OCT statistical classification, and integrated assessment with clinical findings and visual fields is essential 4)5).
QWhat are Green disease and Red disease?
A
These are concepts that illustrate the limitations of color-code classification based on OCT normative databases. Green disease refers to a condition where OCT indicates “normal (green)” but glaucoma is actually present, while Red disease refers to a condition where OCT indicates “abnormal (red)” but the eye is actually normal 1). Both highlight the need to interpret OCT results in conjunction with clinical findings and visual field testing.
OCT-A is a technique that images microvasculature of the retina and optic nerve head without the use of contrast agents 1). In glaucomatous eyes, decreased vessel density in the peripapillary and macular regions has been reported. Reproducibility is good 1), but its clinical role has not yet been established 4).
Choroidal microvascular dropout (MvD) is associated with progressive RNFL thinning in glaucoma with disc hemorrhage 1) and may serve as a predictor for the development of normal-tension glaucoma1). However, changes in vessel density are not specific to glaucoma and have also been reported in hypertension, diabetes, Alzheimer’s disease, and multiple sclerosis1).
PS-OCT is a technique that measures the birefringence properties of the RNFL in three dimensions 1). The arrangement of microtubules within axons is the main source of birefringence, and disruption of microtubules or minor axonal loss may be detected as a decrease in birefringence preceding a reduction in RNFL thickness 1).
It can be added to both SD-OCT and SS-OCT, allowing three-dimensional acquisition of polarization parameters in parallel with conventional OCT reflectivity data 1). Diagnostic ability in early glaucoma is comparable to RNFL thickness, but superiority in very early stages has so far only been demonstrated in animal studies 1).
Visible-Light OCT (VL-OCT): Uses visible light instead of conventional near-infrared light. It may detect wavelength-dependent changes in RNFL reflectivity before changes in RNFL thickness occur. However, challenges for clinical application include patient discomfort and the effect of cataracts 1).
RNFL Optical Texture Analysis (ROTA): A new method that analyzes the microstructural pattern of the RNFL rather than its thickness. It shows excellent accuracy in detecting early damage to the papillomacular bundle. Real clinical data are currently lacking 1).
Lamina cribrosa imaging: SS-OCT and EDI (enhanced depth imaging) have made it possible to evaluate the morphology of the lamina cribrosa (depth, curvature, thickness, defects). Posterior bowing of the lamina cribrosa is associated with the rate of visual field deterioration 1). Lamina cribrosa defects are frequently observed in normal-tension glaucoma1).
Adaptive optics (AO): This technology corrects optical aberrations and enables high-resolution in vivo observation of individual RGCs and lamina cribrosa pores 1). It is expected to be applied to early diagnosis of glaucoma.
Deep learning (CNN) shows high accuracy in detecting glaucoma from fundus photographs and OCT images 1). Development of progression prediction models integrating multiple data modalities (visual field, OCT volume scans, OCT-A) is underway 1).
AI also contributes to improved segmentation accuracy, and is expected to enhance measurement reproducibility. However, challenges remain regarding data privacy, standardization, and algorithm validation 1). As a countermeasure to the black-box nature, development of explainable AI models is required.
