Impact of Artifacts From Optical Coherence Tomography Retinal Nerve Fiber Layer and Macula Scans on Detection of Glaucoma Progression

The AI revolution in glaucoma: Bridging challenges with opportunities

Recent advancements in artificial intelligence (AI) herald transformative potentials for reshaping glaucoma clinical management, improving screening efficacy, sharpening diagnosis precision, and refining the detection of disease progression. However, incorporating AI into healthcare usages faces significant hurdles in terms of developing algorithms and putting them into practice. When creating algorithms, issues arise due to the intensive effort required to label data, inconsistent diagnostic standards, and a lack of thorough testing, which often limits the algorithms' widespread applicability. Additionally, the "black box" nature of AI algorithms may cause doctors to be wary or skeptical. When it comes to using these tools, challenges include dealing with lower-quality images in real situations and the systems' limited ability to work well with diverse ethnic groups and different diagnostic equipment. Looking ahead, new developments aim to protect data privacy through federated learning paradigms, improving algorithm generalizability by diversifying input data modalities, and augmenting datasets with synthetic imagery. The integration of smartphones appears promising for using AI algorithms in both clinical and non-clinical settings. Furthermore, bringing in large language models (LLMs) to act as interactive tool in medicine may signify a significant change in how healthcare will be delivered in the future. By navigating through these challenges and leveraging on these as opportunities, the field of glaucoma AI will not only have improved algorithmic accuracy and optimized data integration but also a paradigmatic shift towards enhanced clinical acceptance and a transformative improvement in glaucoma care.

Comparison of Retinal Nerve Fiber Layer and Ganglion Cell Complex Rates of Change in Patients With Moderate to Advanced Glaucoma

PURPOSE

To compare ganglion cell complex (GCC) and retinal nerve fiber layer (RNFL) rates of change (RoC) in eyes with central or moderate to advanced glaucoma.

DESIGN

Prospective cohort study.

PARTICIPANTS

A total of 918 matched macular and RNFL OCT scan pairs from 109 eyes (109 patients) enrolled in the Advanced Glaucoma Progression Study with ≥2 years of follow-up and ≥4 OCT scans.

METHODS

We exported GCC and RNFL thickness measurements in 49 central macular superpixels and 12 RNFL clock-hour sectors, respectively. We applied our latest Bayesian hierarchical longitudinal model to estimate population and subject-specific baseline thickness (intercepts) and rates of change (RoC) in macular superpixels and RNFL sectors. Global RNFL and GCC RoC were analyzed in a single bivariate longitudinal model to properly compare them accounting for the correlation between their RoC.

MAIN OUTCOME MEASURES

Proportion of significant negative (deteriorating) and positive (improving) RoC expressed in μm/year. Standardized RoC were calculated by dividing RoC by the corresponding population SD. Analyses were repeated in eyes with visual field mean deviation (MD) ≤-6 and > -6 dB.

RESULTS

Average (SD) 24-2 visual field MD and follow-up length were -8.6 (6.3) dB and 4.2 (0.5) years, respectively. Global RNFL RoC (-0.70 µm/year) were faster than GCC (-0.44 µm/year) (P < .001); corresponding normalized RoC were not significantly different (P = .052). In bivariate analysis, patients with a significant negative global RNFL RoC (n = 63, 57%) or GCC (n = 56, 51%) frequently did so for both outcomes (n = 49, 45%). The average proportion of significantly decreasing RNFL sectors within an eye was 30.7% in eyes with MD > -6 dB compared to 20.5% in those with MD ≤ -6 dB (P = .014); the proportions for GCC superpixels were 21.1% versus 18.7%, respectively (P = .63).

CONCLUSIONS

Both GCC and RNFL measures can detect structural progression in glaucoma patients with central damage or moderate to advanced glaucoma. The clinical utility of RNFL imaging decreases with worsening severity of glaucoma.

Artifacts in OCT Retinal Nerve Fiber Layer Imaging in Patients with Boston Keratoprosthesis Type 1

PURPOSE

To determine the clinical utility of OCT retinal nerve fiber layer (OCT RNFL) imaging for glaucoma evaluation in patients with Boston keratoprosthesis type 1 (KPro) by investigating imaging artifacts.

DESIGN

Case-control study.

SUBJECTS

Patients with KPro and without KPro (controls) matched for age, gender, and glaucoma diagnosis.

METHODS

The most recent Cirrus OCT RNFL scan from 1 eye was categorized as having good signal strength (SS; ≥ 6 out of 10) or poor SS (< 6). Those with good SS were analyzed by 2 independent reviewers for artifacts. Images with good SS and no artifacts affecting the scanning circle were considered useful for glaucoma evaluation.

MAIN OUTCOME MEASURES

The incidence of poor SS and artifacts in OCT RNFL images; patient characteristics associated with useful scans.

RESULTS

Sixty-five patients with KPro and 75 controls were included; 89.2% of KPro patients and 89.3% of control subjects had glaucoma (P = 0.98). Forty percent of KPro patients and 5.3% of controls had poor SS (P < 0.001). The proportion of images with either poor SS or artifacts was similar in KPro (76.9%) vs. controls (72.0%, P = 0.51). The most common artifacts in both groups were missing data (43.6%, 53.2%, respectively, P = 0.32) and motion artifact (25.6%, 19.7%, respectively, P = 0.47). Images were useful for glaucoma evaluation in 43.1% of KPro patients and in 69.3% of controls (P = 0.002). In the KPro group, patients with useful OCT scans, compared with those without, had better visual acuity (0.4 ± 0.3 vs. 0.9 ± 0.7 logarithm of the minimum angle of resolution, P = 0.004), and did not have congenital corneal pathologies (0.0% vs. 24.3%, P = 0.008). A multivariate analysis showed that KPro patients with older age had higher odds of useful OCT images (odds ratio, 1.05; P = 0.03). Among KPro patients with useful OCT scans, retinal nerve fiber layer thickness correlated with observed cup-to-disc ratio (Pearson correlation: r = -0.42, P = 0.03).

CONCLUSIONS

The rate of OCT RNFL images with either poor signal strength or artifacts in the KPro and control population was comparable. In patients with KPro, where intraocular pressure measurements are difficult and glaucoma is highly prevalent and often severe, OCT RNFL imaging can be useful for glaucoma evaluation.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Quantitative analysis of deep learning-based denoising model efficacy on optical coherence tomography images with different noise levels

BACKGROUND

To quantitatively evaluate the effectiveness of the Noise2Noise (N2N) model, a deep learning (DL)-based noise reduction algorithm, on enhanced depth imaging-optical coherence tomography (EDI-OCT) images with different noise levels.

METHODS

The study included 30 subfoveal EDI-OCT images averaged with 100 frames from 30 healthy participants. Artificial Gaussian noise at 25.00, 50.00, and 75.00 standard deviations were added to the averaged (original) images, and the images were grouped as 25N, 50N, and 75N. Afterward, noise-added images were denoised with the N2N model and grouped as 25dN, 50dN, and 75dN, according to previous noise levels. The choroidal vascularity index (CVI) and deep choroidal contrast-to-noise ratio (CNR) were calculated for all images, and noise-added and denoised images were compared with the original images. The structural similarity of the noise-added and denoised images to the original images was assessed by the Multi-Scale Structural Similarity Index (MS-SSI).

RESULTS

The CVI and CNR parameters of the original images (68.08 ± 2.47 %, and 9.71 ± 2.80) did not differ from the only 25dN images (67.97 ± 2.34 % and 8.50 ± 2.43) (p:1.000, and p:0.062, respectively). Noise reduction improved the MS-SSI at each noise level (p < 0.001). However, the highest MS-SSI was achieved in 25dN images.

CONCLUSIONS

The DL-based N2N denoising model can be used effectively for images with low noise levels, but at increasing noise levels, this model may be insufficient to provide both the original structural features of the choroid and structural similarity to the original image.

Artifact Correction in Retinal Nerve Fiber Layer Thickness Maps Using Deep Learning and Its Clinical Utility in Glaucoma

Purpose

Correcting retinal nerve fiber layer thickness (RNFLT) artifacts in glaucoma with deep learning and evaluate its clinical usefulness.

Methods

We included 24,257 patients with optical coherence tomography and reliable visual field (VF) measurements within 30 days and 3,233 patients with reliable VF series of at least five measurements over ≥4 years. The artifacts are defined as RNFLT less than the known floor value of 50 µm. We selected 27,319 high-quality RNFLT maps with an artifact ratio (AR) of <2% as the ground truth. We created pseudo-artifacts from 21,722 low-quality RNFLT maps with AR of >5% and superimposed them on high-quality RNFLT maps to predict the artifact-free ground truth. We evaluated the impact of artifact correction on the structure-function relationship and progression forecasting.

Results

The mean absolute error and Pearson correlation of the artifact correction were 9.89 µm and 0.90 (P < 0.001), respectively. Artifact correction improved R2 for VF prediction in RNFLT maps with AR of >10% and AR of >20% up to 0.03 and 0.04 (P < 0.001), respectively. Artifact correction improved (P < 0.05) the AUC for progression prediction in RNFLT maps with AR of ≤10%, >10%, and >20%: (1) total deviation pointwise progression: 0.68 to 0.69, 0.62 to 0.63, and 0.62 to 0.64; and (2) mean deviation fast progression: 0.67 to 0.68, 0.54 to 0.60, and 0.45 to 0.56.

Conclusions

Artifact correction for RNFLTs improves VF and progression prediction in glaucoma.

Translational Relevance

Our model improves clinical usability of RNFLT maps with artifacts.

Segmentation Errors and Off-Center Artifacts in SS-OCT: Insight from a Population-Based Imaging Study

PURPOSE

To evaluate the frequency and associated factors of artifacts in swept-source optical coherence tomography (SS-OCT) imaging.

METHODS

This was a population-based cross-sectional study. Individuals aged 35 years or older, residing in the Yuexiu district of Guangzhou, China, were recruited by random cluster sampling. Nearly half of the participants were randomly selected for SS-OCT imaging centered on the optic nerve head. Six types of artifacts in the peripapillary choroidal layer and retinal nerve fiber layer (RNFL) were graded and identified. Univariate and multivariate logistic regression analyses were used to investigate the association between the presence of artifacts and clinical characteristics.

RESULTS

Out of the 616 eligible individuals who underwent SS-OCT imaging, 18.3% and 13.6% of subjects exhibited at least one artifact in peripapillary RNFL (pRNFL) and peripapillary choroidal thickness (pCT) measurements, respectively, with posterior segmentation error and off-center artifact ranked as the most common artifacts. The presence of artifacts was significantly associated with age (odds ratio [OR], 1.03; 95% confidence interval [CI], 1.01-1.06; p = .003), refractive error (OR, 0.80; 95% CI, 0.71-0.89; p < .001), and signal strength (OR, 0.95; 95% CI, 0.90-0.997; p = .039) in pRNFL thickness measurement. Similarly, the presence of artifacts in pCT measurement was significantly associated with age (OR, 1.05; 95% CI, 1.03-1.08; p < .001), and refractive error (OR, 0.76; 95% CI, 0.68-0.86; p < .001).

CONCLUSION

Nearly one-fifth of the eyes were noted with at least one artifact in the population-scale SS-OCT study. Age was a risk factor for the presence of artifacts and should be considered in clinical settings.

On Machine Learning in Clinical Interpretation of Retinal Diseases Using OCT Images

Optical coherence tomography (OCT) is a noninvasive imaging technique that provides high-resolution cross-sectional retina images, enabling ophthalmologists to gather crucial information for diagnosing various retinal diseases. Despite its benefits, manual analysis of OCT images is time-consuming and heavily dependent on the personal experience of the analyst. This paper focuses on using machine learning to analyse OCT images in the clinical interpretation of retinal diseases. The complexity of understanding the biomarkers present in OCT images has been a challenge for many researchers, particularly those from nonclinical disciplines. This paper aims to provide an overview of the current state-of-the-art OCT image processing techniques, including image denoising and layer segmentation. It also highlights the potential of machine learning algorithms to automate the analysis of OCT images, reducing time consumption and improving diagnostic accuracy. Using machine learning in OCT image analysis can mitigate the limitations of manual analysis methods and provide a more reliable and objective approach to diagnosing retinal diseases. This paper will be of interest to ophthalmologists, researchers, and data scientists working in the field of retinal disease diagnosis and machine learning. By presenting the latest advancements in OCT image analysis using machine learning, this paper will contribute to the ongoing efforts to improve the diagnostic accuracy of retinal diseases.

Effect of idiopathic epiretinal membrane on macular ganglion cell complex measurement in eyes with glaucoma

Background/objectives

Co-existing idiopathic epiretinal membrane (ERM) and glaucoma complicate the estimation of glaucoma severity via optical coherence tomography (OCT). We investigated the effect of ERM and a new associated parameter, SUKIMA (space between the ERM and retinal surface), on ganglion cell complex (GCC) thickness in eyes with glaucoma, based on a matched comparison of visual field defects.

Subjects/methods

We retrospectively recruited 41 eyes from 34 glaucoma patients with idiopathic ERM and 41 eyes from 41 glaucoma patients without ERM as controls (matched by age, axial length, and mean visual field deviation). The thicknesses of GCC layers [retinal nerve fiber layer (RNFL), ganglion cell layer + inner plexiform layer (GCIPL), and GCC (RNFL + GCIPL)] were measured with swept-source OCT. We investigated the presence of SUKIMA and its effect on GCC measurements.

Results

RNFL, GCIPL, and GCC were thicker in ERM (+) eyes than in control eyes (31.0 ± 12.3 μm vs. 22.7 ± 10.8 μm, 62.6 ± 12.2 μm vs. 53.8 ± 5.9 μm, and 91.8 ± 16.6 μm vs. 76.8 ± 13.3 μm, respectively; P &lt; 0.01). Eyes in the ERM-associated SUKIMA (+) group had thicker GCIPL and GCC than those in the ERM-associated SUKIMA (−) and control groups (P &lt; 0.01).

Conclusion

ERM-associated SUKIMA affects GCC thickness and can result in underestimations of glaucoma severity. We should check for the presence of ERM using a B mode scan as well as check for the SKIMA sign.

Detection of Nonglaucomatous Macula Findings With Ganglion Cell Analysis Printouts vs Full Macular Cube Scans

Importance

Ganglion cell analysis (GCA) of ocular coherence tomography (OCT) imaging is routinely used to detect and monitor glaucomatous damage of the ganglion cell complex in the macula. The GCA printout provides qualitative and quantitative data about the macular ganglion cell-inner plexiform layer and a single B-scan of the retina through the fovea. However, the full macular cube scan, including all 128 B-scans, is available for review. The macular cube scan provides considerable information about nonglaucomatous ocular pathology that may be missed if clinicians review only the GCA printout.

Objective

To determine the frequency and type of nonglaucomatous macular findings that are observable in the full macular cube scan but not the GCA printout.

Design, Setting, and Participants

A retrospective cross-sectional analysis of GCA printouts and full macular cube scans to detect nonglaucomatous macular pathology at a tertiary care academic center. Consecutive patients undergoing ganglion cell complex imaging during routine glaucoma evaluations over a 1-week period in a multi-clinician glaucoma clinic.

Main Outcomes and Measures

The prevalence and type of nonglaucomatous macular pathology visible on the GCA printout or macular cube scan.

Results

Among 105 patients (mean (SD) age, 67 (15.46) years; 63 [60%] female and 42 [40%] male) 201 eyes were imaged (64 [31.7%] with suspected glaucoma, 126 [62.4%] with open-angle glaucoma, 6 [3.0%] with closed-angle glaucoma, and 6 [3.0%] with other glaucoma). GCA printouts and macular cube scans revealed nonglaucomatous macular pathology in 65 eyes (32.2%). Of these, 25 eyes (38.5%) included findings that were not visible on the GCA printout. Of the cases not visible on the printout, 16 eyes (64.0% ) included macular pathology that required further evaluation.

Conclusions and Relevance

The findings indicate that nonglaucomatous macular pathology may be missed based on GCA printouts alone. While it may be beneficial to review the full macular cube to detect potentially vision-threatening disease and ensure proper patient care, this study cannot determine if this missed pathology affects clinical outcomes.

Assessment of Artifacts in Swept-Source Optical Coherence Tomography Angiography for Glaucomatous and Normal Eyes

Purpose

To evaluate the frequency of and identify the factors that influence the artifacts of swept-source optical coherence tomography angiography (SS-OCTA) in glaucomatous and normal eyes.

Methods

Artifacts of OCTA images of open-angle glaucoma (OAG) and normal subjects were analyzed using SS-OCTA. Univariate and multivariate logistic regression analyses were performed to evaluate the association of age, sex, best-corrected visual acuity, axial length (AL), intraocular pressure, presence and severity of OAG, and image quality score (IQS) with the presence of artifacts.

Results

Images from 4426 subjects were included in the study. At least one type of artifact was present in 24.54% of the images. The most common artifacts were occurrence of motion (705 eyes, 15.93%), followed by defocus (628 eyes, 14.19%), decentration (134 eyes, 3.03%), masking (62 eyes,1.40%), and segmentation errors (23 eyes, 0.52%). Multivariate logistic analyses showed that the presence of OAG (odds ratio [OR] = 2.71; 95% confidence interval [CI], 2.09-3.51; P < 0.001), female sex (OR = 1.34; 95% CI, 1.12-1.61; P = 0.001), longer AL (OR = 1.09; 95% CI, 1.02-1.17; P = 0.017), and IQS < 40 (OR = 3.75; 95% CI, 3.15-4.48; P < 0.001) were significantly associated with higher odds for the presence of any artifact. The IQS had poor performance for detecting artifacts, with an area under the curve of 0.723, sensitivity of 73.04%, and specificity of 62.53%.

Conclusions

OAG eyes had more SS-OCTA image artifacts than normal eyes. IQS is an imperfect tool for identifying artifacts.

Translational Relevance

Special attention should be paid to the effect of artifacts when using SS-OCTA in the clinical setting to assess vascular parameters in patients with glaucoma.

Case Report: Longitudinal Effect of Progressive Epiretinal Membrane on the Retinal Nerve Fiber Layer

SIGNIFICANCE

Epiretinal membrane is a common macular pathology known to cause morphologic changes observed on macular optical coherence tomography (OCT) and retinal nerve fiber layer (RNFL) OCT. However, the longitudinal effect of epiretinal membrane progression on RNFL OCT morphology is not well studied.

PURPOSE

This report documents a case of epiretinal membrane progression with associated quantifiable changes to the RNFL OCT over time.

CASE REPORT

A 63-year-old man initially presented in 2014 with a grade 0 epiretinal membrane in his left eye and low suspicion of glaucoma in both eyes. Over the next 6 years, his left eye's epiretinal membrane gradually worsened. Along with this change, the RNFL OCT started to show areas of adjacent suspected RNFL thickening and thinning compared with baseline per guided progression analysis (GPA). Despite this, clinical suspicion for actual glaucomatous progression was low. Closer retrospective analysis suggested that the RNFL was continuously dragged temporally toward the macula over this period. Because of traction, values such as the angular location, width, and peak thickness of the inferior RNFL bundles changed. This dynamic shift of a typically stationary structure contributed to an inability to rely on the RNFL OCT GPA to correctly stratify concern for glaucomatous progression.

CONCLUSIONS

Both macular and RNFL OCT allow us to observe morphologic changes to the retina caused by epiretinal membrane. Other authors have described this phenomenon, but this case demonstrates the continual change over time, suggestive of a dynamic process that requires continuous awareness and monitoring. Clinicians should be especially aware of this phenomenon when a patient is also suspicious of glaucoma. These RNFL changes can make it more problematic to rely on the OCT GPA to determine early progressive glaucomatous changes to the RNFL.

IMAGING ARTIFACTS IN FLUORESCENCE LIFETIME IMAGING OPHTHALMOSCOPY

PURPOSE

To investigate and quantify the influence of imaging artifacts on retinal fluorescence lifetime (FLIO) values and to provide helpful hints and tricks to avoid imaging artifacts and to improve FLIO image acquisition quality.

METHODS

A systematic analysis of potential parameters influencing FLIO quality and/or fluorescence lifetime values was performed in a prospective systematic experimental imaging study in five eyes of five healthy subjects. For image acquisition, a fluorescence lifetime imaging ophthalmoscope (Heidelberg Engineering) was used. Quantitative analysis of FLIO lifetime changes due to imaging artifacts was performed.

RESULTS

Imaging artifacts with significant influence on fluorescence lifetimes included too short image acquisition time, insufficient illumination, ocular surface problems, and image defocus. Prior use of systemic or topical fluorescein makes analysis of retinal fluorescence lifetimes impossible.

CONCLUSION

Awareness of possible sources of imaging artifacts is important for FLIO image acquisition and analysis. Therefore, standardized imaging and analysis procedure in FLIO is crucial for high-quality image acquisition and the possibility for systematic quantitative fluorescence lifetime analysis.