Glaucoma is an optic neuropathy characterized by progressive loss of retinal ganglion cells (RGCs) and their axons, which comprise the retinal nerve fiber layer (RNFL). This structural damage respects the anatomic pattern of the RNFL, and the horizontal raphe, affects RGCs in the central and peripheral retina, is consistent with morphologic changes in the neuroretinal rim and the lamina cribrosa, and can be detected by optical coherence tomography (OCT). Glaucomatous visual functional deficits typically are consistent with the arcuate nature of the RNFL damage. There is widespread agreement that in the absence of other ocular pathology, the presence of spatially consistent functional and structural optic nerve damage is a hallmark of the disease.1,2
LIMITS OF THE CLINICAL OPTIC NERVE EXAMINATION
Intraobserver and interobserver variability limit the application of clinical assessment of the optic nerve alone for glaucoma diagnosis in clinical care and research,3 particularly in early disease. To address this deficiency in clinical research, large, longitudinal, prospective glaucoma studies often employ the use of Optic Disc Reading Centers. Optic Disc Reading Centers, utilizing certified technicians or other personnel, such as used in the Ocular Hypertension Treatment Study (OHTS)4 and the African Descent and Evaluation Study (ADAGES),5 set criteria for photograph quality, employ common features of the disease that are reenforced by education, training and testing of graders, and offer adjudication when disagreements occur. The improved consistency of reading centers allows for standardization of study enrollment but does not assure accurate diagnosis. For example, the OHTS Confocal Laser Ophthalmoscopy Ancillary Study clearly demonstrated that many subjects had optic nerve damage at the time of study entry even though they met the criteria for a nonglaucomatous clinical nerve appearance by photography and as determined by certified Optic Disc Reading Center personnel.6,7
Limits of Perimetry
Anchored by this historical perspective, the limitations surrounding clinical evaluation of the optic nerve alone and its impact on patient care and research paradigms fostered an emphasis on perimetry to aid in disease evaluation. While ubiquitous in the modern care paradigm, automated perimetry, as currently performed, is often insensitive to early disease, especially in the central visual field (VF),8–24 and prone to numerous forms of testing error and human inconsistency.25 A classic example of this problem was noted in the OHTS, in which a single abnormal field test meeting the endpoint criterion was not confirmed on repeat testing 86% of the time, and of those with endpoint confirmation, 50% reverted to a normal examination on the third test; hence the need for 3 successive and abnormal VF tests in the OHTS trial to achieve a VF endpoint determination.26,27 Perimetry is also hampered by a patient learning effect, which limits its usefulness as a stand-alone diagnostic and screening tool,28 with unacceptably high false-positive rates for the latter.29
Limits of Current Combined Structure-Function Approaches
Despite these drawbacks in both traditional clinical optic nerve structural examination and functional testing (automated perimetry), numerous systems for defining glaucoma have been suggested and employed for screening, diagnosis, and longitudinal monitoring. Most of these older definitions required both structural and functional evidence of glaucomatous disease based upon the clinical optic nerve examination and perimetric data, despite the aforementioned weaknesses of these measures. For example, before the widespread availability of high-resolution imaging, Foster et al30 utilized a combination of cup-disc ratio and perimetry in studies of glaucoma prevalence.
Though the glaucoma community has long accepted spatially consistent structural and functional damage as a reference standard for glaucoma diagnosis,10,18,31,32 to date, there is clearly no agreed-upon system to serve as a gold standard for diagnosis of glaucomatous optic neuropathy or detection of progression. In a recent survey of a large number of international, self-identified glaucoma specialists, Iyer et al33 confirmed this unmet need and a desire to develop a more objective reference standard for glaucoma research. Using a crowd-sourcing methodology and a variety of databases, they subsequently reported that using OCT or VF information alone for diagnosis had inadequate specificity and that the best trade-off was at 77% sensitivity at 98% specificity when spatially consistent damage consisting of an OCT quadrant and a corresponding abnormal glaucoma hemifield test were used as the diagnostic criteria.34 The relatively poor sensitivity is not surprising as previous work has reported on the limitations of summary, metric data from OCT images18,19,32,35–40 and have documented the reasons for their poor performance,37,38,41–45 including that OCT and VF metrics underestimate the degree of agreement between structural and functional damage.
In contrast, we have previously demonstrated that the spatially consistent agreement between abnormal structural (OCT) and functional (VF) damage is excellent, but only if local glaucomatous OCT abnormalities on both RNFL and ganglion cell layer (GCL) deviation maps are compared with those seen with both 24-2 and 10-2 testing.32,46 While this suggests a better definition/reference standard than the ones based upon metrics, it requires 24-2 and 10-2 VFs as well as OCT. Regardless, we believe that the existing literature supports the notion that it is time to move beyond global and summary metrics and that inclusion of more localized OCT information could aid in the evolution of an improved method for detecting glaucomatous optic neuropathy and a future reference standard.
The Advantages of an OCT-based Method for Identifying Glaucomatous Damage
We choose to focus on an OCT-based system for a variety of reasons. First, compared with imaging, the use of perimetry as a stand-alone diagnostic or screening modality is limited because of its psychophysical nature, learning curve, longer test duration, and poorer reproducibility, as well as poor patient acceptance. The strength of functional assessment is greatest when used in conjunction with structural testing.32,46 Second, in addition to high intraobserver and interobserver variability, clinical examination of the optic nerve is limited by normal physiological variability in optic nerve shape, size, insertion, and topography. The advent of high-resolution imaging and the increasing time constraints placed on clinical practice have further eroded careful physician attentiveness to the clinical examination of the optic nerve. Third, and most importantly, spectral-domain and swept-source OCT technologies and analytic programs are now able to provide information about RNFL and macular GCL anatomy that extend well beyond the evaluative ability of any clinician, despite the limitations imposed by a variety of ocular conditions (ie, ocular surface disorders, opaque media, high refractive error) that can limit image acquisition. In addition, the OCT examination is faster, more acceptable to patients, and more likely to be able to be applied as a screening tool than perimetry. In fact, we have demonstrated that perimetry does not necessarily aid in glaucoma diagnosis when OCT is used for structural examination of the optic nerve.47
Given these factors, we hypothesized that by utilizing the wealth of information available in OCT imaging, it is possible to provide a more precise definition and method for the detection of glaucomatous eyes in individuals with VF loss and narrow the group of individuals in whom the diagnosis remains uncertain.
THE DEVELOPMENT OF AN OCT-DRIVEN METHOD FOR DETECTING GLAUCOMA
To test our assumption, an iterative, Delphi-like approach was used to generate a consensus-based, OCT-driven glaucoma definition and diagnostic method by members of the glaucoma division at the Department of Ophthalmology at the Edward S. Harkness Eye Institute at Columbia University Irving Medical Center and Vagelos College of Physicians and Surgeons. Members of the panel met virtually and were surveyed to gain consensus. This resulted in a formalized, methodological approach which included a pilot study in which graders using our method could separate healthy eyes from those with the clear disease.
The pilot study confirmed the consenus opinion of the utility of sequentially presenting elements of the OCT report in Figure 1. In particular, it was agreed that individual RNFL and GCL probability plots (5 and 6 in Fig. 1) followed by the combined set of imaging RNFL and GCL thickness maps (3 and 4), and circumpapillary B-scan (1) provided important and significant information, enhanced uniformity of approach among graders, and was consistent with clinical practice.
This method resulted in an OCT-based structural definition of glaucoma (Table 1), a method for OCT evaluation (Table 2), and an OCT Decision Tree (Fig. 2). Details of the use of this method can be found in Hood et al.48
TABLE 1 -
OCT-driven Glaucoma Definition
||Level of Importance
||An OCT RNFL defect is present as determined by the RNFL probability map
||Located in temporal half of the disc An abnormal region or pattern that respects the anatomy of the RNFL (often seen as an “arcuate” defect in the superior and inferior quadrants) In the temporal quadrant, an RNFL defect may be localized to the parapapillary region
||An OCT GCL defect is present which provides structural corroboration of the RNFL probability map defect
||The OCT GCL defect must be in the same hemifield as the RNFL defect The OCT GCL defect should overlap or be consistent with the RNFL defect Most often appears temporal to fixation
||2a or 2b must be present
||OCT evidence of corroborative RNFL and/or GCL defects on the GCL and RNFL thickness maps, circumpapillary B-scan or RNFL NSTIN plot
||Clinical examination of the optic nerve and retina eliminates other causes of RNFL or GCL defects
||Examples include, but are not limited to, retinal or retinovascular disease, optic disc drusen, nonglaucomatous optic atrophy, and developmental anomalies of the optic nerve complex (disc anatomy, vasculature, parapapillary region)
GCL indicates ganglion cell layer; OCT, optical coherence tomography; RNFL, retinal nerve fiber layer.
TABLE 2 -
Algorithm Queries and Decision-making
||Using only the RNFL probability map, is there an arcuate-like abnormal region on the RNFL probability map associated with temporal half of disc? Possible responses: YES, NO, UNCERTAIN
||If there is no RNFL probability map defect in the parapapillary region, the response is NO
||If the answer is NO, this is NOT GLAUCOMA This represents a decision endpoint
For a YES or UNCERTAIN response to Question 1, proceed to Question 2
||Using the GCL probability map, is there a topographically corresponding abnormal region on the GCL probability map (largely typically temporal to fixation)? Possible responses YES, NO, or UNCERTAIN
||“Corroborative” means overlapping abnormal GCL and RNFL regions and/or abnormal GCL and RNFL regions that share an arcuate RNFL region in the same hemifield
||If the answers to both Questions 1 and 2 are YES, this is GLAUCOMA*This represents a decision endpoint
For a NO or UNCERTAIN response to Question 2, proceed to Question 3
||Question 3. Is there confirmatory evidence of a RNFL defect on the cpRNFL thickness plot/b-scan, the GCL thickness map &/or RNFL thickness map? Possible responses YES, NO, or UNCERTAIN
||If the answer to Question 3 is YES, then this is GLAUCOMA*This represents a decision endpoint If the answer to Question 3 is NO, then this is NOT GLAUCOMA This represents a decision endpoint
If the answer to Question 3 is UNCERTAIN, then this is an OCT SUSPECT
This represents a decision endpoint
Assuming that the clinical examination of the optic nerve and retina eliminates other causes of RNFL or GCL defects (see Step 3, Table 1
GCL indicates ganglion cell layer; OCT, optical coherence tomography; RNFL, retinal nerve fiber layer.
Key Elements of Our Method
The Columbia University OCT-based method and definition of structural glaucomatous optic neuropathy has a number of key elements. First, the questions in Table 2 utilized in driving assessment of each image or image set reflect clinical practice evaluation of OCT images and are familiar to clinicians and researchers; the algorithmic quality of the method serves to formalize the decision-making process. While we recognize that machine learning or artificial intelligence may offer significant advantages in the future, these programs remain nascent at the current time.49 Our method offers a bridge to a future when artificial intelligence can be utilized to aid in decision-making.
There are 2 other critical elements that impacted our success. First, our method makes use of the precise, repeatable, quantitative information from the OCT scan, including thickness maps and probability plots, to support decision-making. The use of the attendant thickness maps, probability plots, and circumpapillary information is a powerful diagnostic tool, even without being coupled with perimetric data. The probability maps, particularly the RNFL probability map, are key components of this process. In addition to available cross-sectional and longitudinal studies that support the use of RNFL probability map, heat maps generated following artificial intelligence assessment of large patient data sets also point to the parapapillary optic nerve as the key regions that these machine methods rely upon.48,50–52 All told, there is considerable evidence to support the use of the RNFL and macular probability plots as the first steps in our method and key components of our OCT-based glaucoma definition. The GCL and RNFL maps provide complementary and corroborative information of structural glaucomatous damage.
Second, reflecting the ambiguity that sometimes exists in making a diagnosis of glaucoma during cross-sectional assessment (the “glaucoma suspect”), we allowed for uncertainty rather than forcing the grader to choose “Yes” or “No” at each decision point. This approach requires a predetermined method of uniform, sequential analysis of all available and pertinent OCT information, including the actual OCT circumpapillary image, to pragmatically support or oppose the glaucoma diagnosis, with the understanding that “Uncertain” (ie, OCT Suspect) is still an acceptable outcome in those specific circumstances where a diagnosis of glaucoma or healthy cannot not be determined.
Third, while some may criticize our use of semiquantitative evaluation of the scans—as opposed to completely automated summary statistics produced by commercial reports—there is considerable evidence in the literature pointing to (1) the need for a panel of trained graders when evaluating imaging results and (2) the dangers of relying on these summary statistics. Regarding the former and as previously noted, Gordon et al27 showed that 86% of eyes meeting an “objective” definition of glaucoma based on perimetry-derived statistics (Glaucoma Hemifield Test and pattern standard deviation) were not confirmed upon repeat testing, and many more were false positives based on expert review. Similarly, OCT summary statistics values or probabilities have limited repeatability upon retesting,53,54 and Reading Centers may be needed for an ultimate determination of the presence of glaucomatous damage.27 Regarding the latter, a number of limitations of summary statistics, including those from OCT, have been noted by different investigators, including but not limited to segmentation errors, epiretinal membrane artifacts, artifacts induced by the position of main peripapillary blood vessels, and tortion or head-tilt artifacts.36,37,41–45 Our approach takes into account the breadth of information of the scan protocol to avoid these common issues.
There are 4 potential limitations of our approach. First, we recognize that glaucoma diagnostic uncertainty remains despite our technological prowess. Yet absent a blood serum biomarker, a disease-specific genetic test, or an imaging device that can detect glaucomatous apoptotic cell death, no diagnostic modality can be 100% sensitive or specific for all glaucomas. Our method seeks to take advantage of this uncertainty by being flexible and, as uncertainty persists, requires further corroborative evidence to support a glaucoma diagnosis. During this process, we learned that one of the important outcomes of this method is its effect on the category of patients whom we label clinically as “glaucoma suspects” because of a suspicious disc appearance. Our method clearly further whittles down this group, thereby redirecting diagnostic and personnel resources to individuals at great risk of vision impairment. This is critical given that considerable resources are utilized in the longitudinal management of glaucoma suspects with clinically “suspicious” optic nerves; most of these will never go on to develop glaucoma or, if they are surveilled periodically, are followed with too great a frequency. Proper classification of individuals to healthy and glaucomatous groups while reducing the number of glaucoma suspects allows for the better apportioning of limited health care resources. Additional patient demographic, ocular, systemic data will aid in longitudinal management and treatment decisions.
Second, the successful deployment of this method comes with a cost. Working knowledge of OCT images, significant grader training, and the meaning of each portion of the printout is crucial. Our experience thus far suggests that it takes far less time to teach image interpretation than required to teach clinical examination of the optic disc. Given the persistently high intraobserver and interobserver reliability among glaucoma specialists endemic to the examination of the optic disc, in the era of high-resolution OCT, we doubt we will ever be able to reproduce the current results even with experts examining the optic nerve. Clinical examination of the optic disc is still required, but only to detect disc hemorrhage and to eliminate other optic nerve or retinal diseases that can mimic glaucomatous damage on diagnostic testing.
Third, while our approach is not entirely quantitative and semiquantitative image grading is still required (since there is as of yet no automated process to detect the localized RNFL damage that is the structural hallmark glaucomatous optic neuropathy), these interpretative interactions are limited to specific questions, and the diagnosis is determined via the decision tree rather than a global physician impression. In any case, there is a need to provide better training tools for evaluating OCT reports in general and using our method in particular.
Fourth, our definition of the glaucoma suspect is based on OCT imaging alone, and thus the possibility exists that damage may be more obvious on VFs than with OCT in some eyes.32 However, we do not mean to imply that perimetry should not play a role in these patients. Current practice standards require that each of these individuals, along with those who meet our OCT-based diagnosis of glaucoma, undergo functional testing as part of the longitudinal assessment. Further, there are many other reasons, of course, that an individual with an unremarkable OCT examination could be labeled as a glaucoma suspect. Individuals with a genetic predisposition increased risk profile (eg, some forms of ocular hypertension) or positive family history are but a few of these examples.
The purpose of this perspective was to review the rationale for the development of a highly specific, highly sensitive, and intersubjectively verifiable method for the detection of glaucomatous optic neuropathy for clinical care and research-based upon OCT alone. Our goal was not to create the “perfect” method wherein each eye is determined to be healthy or glaucomatous, but rather to standardize the approach to these diagnoses, maximize intraobserver and interobserver agreement, and reduce the number of individuals defined as glaucoma suspects, with the goal of providing a rubric for OCT Reading Centers, clinical research, and patient care. In particular, we hypothesized that by using all of the information available in OCT imaging, it is possible to provide a precise definition and method for the detection of glaucomatous eyes in individuals with VF loss and potentially narrow the group of individuals in whom the diagnosis remains uncertain.
Our proposed OCT-based method for the detection of glaucomatous optic neuropathy also has the potential to detect the presence of glaucomatous eyes with VF loss without the need for perimetry. The importance of a perimetry-independent system cannot be overemphasized, not only with respect to disease detection but also cost, time, and patient preference. As expected, patients at the highest risk of blindness (significant VF loss at the time of disease discovery) are readily detected in this system, which machine-based learning will only improve in the future. In the absence of a gold standard such as a direct measure of cell function or genetic determinant for glaucoma, we understand that no system will capture every case of early glaucoma or completely eliminate the need for the “glaucoma suspect” category. However, as all glaucomas, regardless of the primary cause, damage RGCs and their axons, we believe our system will prove to be a highly specific, highly sensitive, and intersubjectively verifiable definition of disease for clinical care and research and provides a solid foundation for disease detection. We test this hypothesis in Hood et al.48
In summary, our method addresses an important unmet need in the field of glaucoma. Continued refinement of this approach will likely lead to improvements in disease detection and have implications for screening and longitudinal management. We believe it is time to embrace this paradigm shift away from our traditional, historical-based perspective of glaucoma definitions and operationalize a more objective, OCT-based nomenclature that can standardize enrollment criteria for clinical research and enhance a clinician’s ability to diagnose and treat disease.
1. Weinreb RN, Greve EL. Glaucoma Diagnosis
, Structure and Function Reports and Consensus Statements of the 1st Global AIGS Consensus Meeting on Structure and Function in the Management of Glaucoma
. Hague, The Netherlands: Kugler Publications; 2004.
2. Weinreb RN, Garway-Heath D, Leung C, et al. Diagnosis of Primary Open Angle Glaucoma The 10th Consensus Report of the World Glaucoma Association. Hague, The Netherlands: Kugler Publications; 2016.
3. Coleman AL, Sommer A, Enger C, et al. Interobserver and intraobserver variability in the detection of glaucomatous progression of the optic disc. J Glaucoma. 1996;5:384–389.
4. Feuer WJ, Parrish RK II, Schiffman JC, et al. The Ocular Hypertension Treatment Study: reproducibility of cup/disk ratio measurements over time at an optic disc reading center. Am J Ophthalmol. 2002;133:19–28.
5. Sample PA, Girkin CA, Zangwill LM, et al. The African Descent and Glaucoma Evaluation Study (ADAGES): design and baseline data. Arch Ophthalmol. 2009;127:1136–1145.
6. Weinreb RN, Zangwill LM, Jain S, et al. Predicting the onset of glaucoma: the confocal scanning laser ophthalmoscopy ancillary study to theOcular Hypertension Treatment Study. Ophthalmology. 2010;117:1674–1683.
7. Zangwill LM, Weinreb RN, Beiser JA, et al. Baseline topographic optic disc measurements are associated with the development of primary open-angle glaucoma: the Confocal Scanning Laser Ophthalmoscopy Ancillary Study to the Ocular Hypertension Treatment Study. Arch Ophthalmol. 2005;123:1188–1197.
8. Schiefer U, Papageorgiou E, Sample PA, et al. Spatial pattern of glaucomatous visual field loss obtained with regionally condensed stimulus arrangements. Invest Ophthalmol Vis Sci. 2010;51:5685–5689.
9. Hood DC, Raza AS, de Moraes CG, et al. Initial arcuate defects within the central 10 degrees in glaucoma. Invest Ophthalmol Vis Sci. 2011;52:940–946.
10. Hood DC, Raza AS, de Moraes CG, et al. Glaucomatous damage of the macula. Prog Retin Eye Res. 2013;32:1–21.
11. Hangai M, Ikeda HO, Akagi T, et al. Paracentral scotoma in glaucoma detected by 10-2 but not by 24-2 perimetry. Jpn J Ophthalmol. 2014;58:188–196.
12. Asaoka R, Iwase A, Hirasawa K, et al. Identifying “preperimetric” glaucoma in standard automated perimetry visual fields. Invest Ophthalmol Vis Sci. 2014;55:7814–7820.
13. Traynis I, De Moraes CG, Raza AS, et al. Prevalence and nature of early glaucomatous defects in the central 10° of the visual field. JAMA Ophthalmol. 2014;132:291–297.
14. Park HY, Hwang BE, Shin HY, et al. Clinical clues to predict the presence of parafoveal scotoma on Humphrey 10-2 visual field using a Humphrey 24-2 visual field. Am J Ophthalmol. 2016;161:150–159.
15. Grillo LM, Wang DL, Ramachandran R, et al. The 24-2 visual field test misses central macular damage confirmed by the 10-2 visual field test and optical coherence tomography. Transl Vis Sci Technol. 2016;5:15.
16. Sullivan-Mee M, Karin Tran MT, Pensyl D, et al. Prevalence, features, and severity of glaucomatous visual field loss measured with the 10-2 achromatic threshold visual field test. Am J Ophthalmol. 2016;168:40–51.
17. De Moraes CG, Hood DC, Thenappan A, et al. 24-2 visual fields miss central defects shown on 10-2 tests in glaucoma suspects, ocular hypertensives, and early glaucoma. Ophthalmology. 2017;124:1449–1456.
18. Hood DC. Improving our understanding, and detection, of glaucomatous damage: an approach based upon optical coherence tomography (OCT). Prog Retin Eye Res. 2017;57:46–75.
19. Jung KI, Kim EK, Park CK. Usefulness of frequency doubling technology perimetry 24-2 in glaucoma with parafoveal scotoma. Medicine (Baltimore). 2017;96:e6855.
20. Kim KE, Yoo BW, Jeoung JW, et al. Long-term reproducibility of macular ganglion cell analysis in clinically stable glaucoma patients. Invest Ophthalmol Vis Sci. 2015;56:4857–4864.
21. Tomairek RH, Aboud SA, Hassan M, et al. Studying the role of 10-2 visual field test in different stages of glaucoma. Eur J Ophthalmol. 2020;30:706–713.
22. Hood DC, Thenappan AA, Tsamis E, et al. An evaluation of a new 24-2 metric for detecting early central glaucomatous damage. Am J Ophthalmol. 2021;223:119–128.
23. Roberti G, Manni G, Riva I, et al. Detection of central visual field defects in early glaucomatous eyes: comparison of Humphrey and Octopus perimetry. PLoS One. 2017;12:e0186793.
24. Langerhorst CT, Carenini LL, Bakker D, et al. Measurements for description of very early glaucomatous field defects. Perimetry Update 1996/1997: Proceedings for the XIIth International Perimetric Society Meeting, Wurzberg Germany, June 4-8, 1996. Hague, The Netherlands: Kugler Publications; 1997.
25. Heijl A, Asman P. Pitfalls of automated perimetry in glaucoma diagnosis. Curr Opin Ophthalmol. 1995;6:46–51.
26. Keltner JL, Johnson CA, Quigg JM, et al. Confirmation of visual field abnormalities in the Ocular Hypertension Treatment Study. Ocular Hypertension Treatment Study Group. Arch Ophthalmol. 2000;118:1187–1194.
27. Gordon MO, Higginbotham EJ, Heuer DK, et al. Assessment of the Impact of an Endpoint Committee in the Ocular Hypertension Treatment Study. Am J Ophthalmol. 2019;199:193–199.
28. Heijl A, Lindgren G, Olsson J. The effect of perimetric experience in normal subjects. Arch Ophthalmol. 1989;107:81–86.
29. Weinreb RN, Healey PR, Topouzis F. Glaucoma Screening The 5th Consensus Report of the World Glaucoma Association. Hague, The Netherlands: Kugler Publications; 2008.
30. Foster PJ, Buhrmann R, Quigley HA, et al. The definition and classification of glaucoma in prevalence surveys. Br J Ophthalmol. 2002;86:238–242.
31. Garway-Heath DF, Poinoosawmy D, Fitzke FW, et al. Mapping the visual field to the optic disc in normal tension glaucoma eyes. Ophthalmology. 2000;107:1809–1815.
32. Hood DC, Tsamis E, Bommakanti NK, et al. Structure-function agreement is better than commonly thought in eyes with early glaucoma. Invest Ophthalmol Vis Sci. 2019;60:4241–4248.
33. Iyer J, Vianna JR, Chauhan BC, et al. Toward a new definition of glaucomatous optic neuropathy for clinical research. Curr Opin Ophthalmol. 2020;31:85–90.
34. Iyer JV, Boland MV, Jefferys J, et al. Defining glaucomatous optic neuropathy using objective criteria from structural and functional testing. Br J Ophthalmol. 2021;105:789–793.
35. Leal-Fonseca M, Rebolleda G, Oblanca N, et al. A comparison of false positives in retinal nerve fiber layer, optic nerve head and macular ganglion cell-inner plexiform layer from two spectral-domain optical coherence tomography devices. Graefes Arch Clin Exp Ophthalmol. 2014;252:321–330.
36. Wang DL, Raza AS, de Moraes CG, et al. Central glaucomatous damage of the macula can be overlooked by conventional OCT retinal nerve fiber layer thickness analyses. Transl Vis Sci Technol. 2015;4:4.
37. Eguia MD, Tsamis E, Zemborain ZZ, et al. Reasons why OCT global circumpapillary retinal nerve fiber layer thickness is a poor measure of glaucomatous progression. Transl Vis Sci Technol. 2020;9:22.
38. Sun A, Tsamis E, Eguia MD, et al. Global optical coherence tomography measures for detecting the progression of glaucoma have fundamental flaws. Eye. 2021;35:2973–2982.
39. Hood DC, De Moraes CG. Challenges to the common clinical paradigm for diagnosis of glaucomatous damage with OCT and visual fields. Invest Ophthalmol Vis Sci. 2018;59:788–791.
40. Hood DC, De Moraes CG. Four questions for every clinician diagnosing and monitoring glaucoma. J Glaucoma. 2018;27:657–664.
41. Liu Y, Simavli H, Que CJ, et al. Patient characteristics associated with artifacts in Spectralis optical coherence tomography imaging of the retinal nerve fiber layer in glaucoma. Am J Ophthalmol. 2015;159:565–576.e2.
42. Ye C, Yu M, Leung CK. Impact of segmentation errors and retinal blood vessels on retinal nerve fibre layer measurements using spectral-domain optical coherence tomography. Acta Ophthalmol. 2016;94:e211–e219.
43. Mansberger SL, Menda SA, Fortune BA, et al. Automated segmentation errors when using optical coherence tomography to measure retinal nerve fiber layer thickness in glaucoma. Am J Ophthalmol. 2017;174:1–8.
44. Miki A, Kumoi M, Usui S, et al. Prevalence and associated factors of segmentation errors in the peripapillary retinal nerve fiber layer and macular ganglion cell complex in spectral-domain optical coherence tomography images. J Glaucoma. 2017;26:995–1000.
45. Suwan Y, Rettig S, Park SC, et al. Effects of circumpapillary retinal nerve fiber layer segmentation error correction on glaucoma diagnosis in myopic eyes. J Glaucoma. 2018;27:971–975.
46. Tsamis E, Bommakanti NK, Sun A, et al. An automated method for assessing topographical structure-function agreement in abnormal glaucomatous regions. Transl Vis Sci Technol. 2020;9:14.
47. Hood DC, De Cuir N, Blumberg DM, et al. A single wide-field OCT protocol can provide compelling information for the diagnosis of early glaucoma. Transl Vis Sci Technol. 2016;5:4.
48. Hood DC, La Bruna S, Tsamis E, et al. Detecting glaucoma with only OCT: Implications for the clinic, research, screening, and AI development. Prog Retin Eye Res. 2022:101052. [Epub ahead of print].
49. Medeiros FA. Deep learning in glaucoma: progress, but still lots to do. Lancet Digit Health. 2019;1:e151–e152.
50. Thompson AC, Jammal AA, Medeiros FA. A review of deep learning for screening, diagnosis, and detection of glaucoma progression. Transl Vis Sci Technol. 2020;9:42.
51. Thakoor KA, Koorathota SC, Hood DC, et al. Robust and interpretable convolutional neural networks to detect glaucoma in optical coherence tomography images. IEEE Trans Biomed Eng. 2020;68:2456–2466.
52. Muhammad H, Fuchs TJ, De Cuir N, et al. Hybrid deep learning on single wide-field optical coherence tomography scans accurately classifies glaucoma suspects. J Glaucoma. 2017;26:1086–1094.
53. Hong S, Kim CY, Lee WS, et al. Reproducibility of peripapillary retinal nerve fiber layer thickness with spectral domain cirrus high-definition optical coherence tomography in normal eyes. Jpn J Ophthalmol. 2010;54:43–47.
54. Mwanza JC, Chang RT, Budenz DL, et al. Reproducibility of peripapillary retinal nerve fiber layer thickness and optic nerve head parameters measured with cirrus HD-OCT in glaucomatous eyes. Invest Ophthalmol Vis Sci. 2010;51:5724–5730.