Rule Extraction for Glaucoma Detection with Summary Data from StratusOCT

Detection of Early Glaucomatous Damage: Performance of Summary Statistics From Optical Coherence Tomography and Perimetry

Purpose

To evaluate the diagnostic performance of optical coherence tomography (OCT) and visual field (VF) summary statistics (metrics) that are available in OCT and VF reports.

Methods

OCT disc and macular scans and 24-2 and 10-2 VFs were obtained from 56 healthy control (HC) eyes/participants and 61 eyes/patients with 24-2 mean deviation of better than –6 dB. All metrics were obtained from OCT radial, circle, and posterior pole cube scans and 24-2 and 10-2 VFs. Their diagnostic performances were evaluated, in isolation and in combinations. For specificity, the 56 HC eyes were used. For sensitivity, 40 of the 61 patient eyes were deemed likely glaucomatous based on an automated topographic method that evaluates structure–function (S–F) agreement. Any 1 of these 40 eyes not judged as abnormal by any given metric was considered a false negative.

Results

All single OCT and VF metrics misclassified HCs as glaucomatous and missed likely glaucomatous eyes. The best performing single metric was the temporal inferior thickness of the 3.5-mm circle scan, with 96% specificity and 83% sensitivity. Combinations of OCT–OCT and OCT–VF metrics markedly improved specificity. A newly proposed metric that evaluates structure–structure (S–S) agreement at a hemifield level had the highest accuracy. This S–S metric had 98% specificity and 80% sensitivity.

Conclusions

OCT and VF metrics, single or in combinations, have only moderate sensitivity for eyes with early glaucoma.

Translational Relevance

OCT and VF metrics combinations evaluating S–S or S–F agreement can be highly specific, which is an important implication for clinical and research purposes.

Glaucoma Detection Using Support Vector Machine Method Based on Spectralis OCT

Spectralis optical coherence tomography (OCT) provided more detailed parameters in the peripapillary and macular areas among the OCT machines, but it is not easy to understand the enormous information (114 features) generated from Spectralis OCT in glaucoma assessment. Machine learning methodology has been well-applied in glaucoma detection in recent years and has the ability to process a large amount of information at once. Here we aimed to analyze the diagnostic capability of Spectralis OCT parameters on glaucoma detection using Support Vector Machine (SVM) classification method in our population. Our results showed that applying all OCT features with the SVM method had good capability in the detection of glaucomatous eyes (area under curve (AUC) = 0.82), as well as discriminating normal eyes from early, moderate, or severe glaucomatous eyes (AUC = 0.78, 0.89, and 0.93, respectively). Apart from using all OCT features, the minimum rim width (MRW) may be good feature groups to discriminate early glaucomatous from normal eyes (AUC = 0.78). The combination of peripapillary and macular parameters, including MRW_temporal inferior (TI), MRW_global (G), ganglion cell layer (GCL)_outer temporal (T2), GCL_inner inferior (I1), peripapillary nerve fiber layer thickness (ppNFLT)_temporal superior (TS), and GCL_inner temporal (T1), provided better results (AUC = 0.84). This study showed promise in glaucoma management in the Taiwanese population. However, further validation study is needed to test the performance of our proposed model in the real world.

Digital technology, tele-medicine and artificial intelligence in ophthalmology: A global perspective

The simultaneous maturation of multiple digital and telecommunications technologies in 2020 has created an unprecedented opportunity for ophthalmology to adapt to new models of care using tele-health supported by digital innovations. These digital innovations include artificial intelligence (AI), 5th generation (5G) telecommunication networks and the Internet of Things (IoT), creating an inter-dependent ecosystem offering opportunities to develop new models of eye care addressing the challenges of COVID-19 and beyond. Ophthalmology has thrived in some of these areas partly due to its many image-based investigations. Tele-health and AI provide synchronous solutions to challenges facing ophthalmologists and healthcare providers worldwide. This article reviews how countries across the world have utilised these digital innovations to tackle diabetic retinopathy, retinopathy of prematurity, age-related macular degeneration, glaucoma, refractive error correction, cataract and other anterior segment disorders. The review summarises the digital strategies that countries are developing and discusses technologies that may increasingly enter the clinical workflow and processes of ophthalmologists. Furthermore as countries around the world have initiated a series of escalating containment and mitigation measures during the COVID-19 pandemic, the delivery of eye care services globally has been significantly impacted. As ophthalmic services adapt and form a "new normal", the rapid adoption of some of telehealth and digital innovation during the pandemic is also discussed. Finally, challenges for validation and clinical implementation are considered, as well as recommendations on future directions.

High interpretable machine learning classifier for early glaucoma diagnosis

AIM

To develop a classifier for differentiating between healthy and early stage glaucoma eyes based on peripapillary retinal nerve fiber layer (RNFL) thicknesses measured with optical coherence tomography (OCT), using machine learning algorithms with a high interpretability.

METHODS

Ninety patients with early glaucoma and 85 healthy eyes were included. Early glaucoma eyes showed a visual field (VF) defect with mean deviation >-6.00 dB and characteristic glaucomatous morphology. RNFL thickness in every quadrant, clock-hour and average thickness were used to feed machine learning algorithms. Cluster analysis was conducted to detect and exclude outliers. Tree gradient boosting algorithms were used to calculate the importance of parameters on the classifier and to check the relation between their values and its impact on the classifier. Parameters with the lowest importance were excluded and a weighted decision tree analysis was applied to obtain an interpretable classifier. Area under the ROC curve (AUC), accuracy and generalization ability of the model were estimated using cross validation techniques.

RESULTS

Average and 7 clock-hour RNFL thicknesses were the parameters with the highest importance. Correlation between parameter values and impact on classification displayed a stepped pattern for average thickness. Decision tree model revealed that average thickness lower than 82 µm was a high predictor for early glaucoma. Model scores had AUC of 0.953 (95%CI: 0.903-0998), with an accuracy of 89%.

CONCLUSION

Gradient boosting methods provide accurate and highly interpretable classifiers to discriminate between early glaucoma and healthy eyes. Average and 7-hour RNFL thicknesses have the best discriminant power.

The potential application of artificial intelligence for diagnosis and management of glaucoma in adults

BACKGROUND

Glaucoma is the most frequent cause of irreversible blindness worldwide. There is no cure, but early detection and treatment can slow the progression and prevent loss of vision. It has been suggested that artificial intelligence (AI) has potential application for detection and management of glaucoma.

SOURCES OF DATA

This literature review is based on articles published in peer-reviewed journals.

AREAS OF AGREEMENT

There have been significant advances in both AI and imaging techniques that are able to identify the early signs of glaucomatous damage. Machine and deep learning algorithms show capabilities equivalent to human experts, if not superior.

AREAS OF CONTROVERSY

Concerns that the increased reliance on AI may lead to deskilling of clinicians.

GROWING POINTS

AI has potential to be used in virtual review clinics, telemedicine and as a training tool for junior doctors. Unsupervised AI techniques offer the potential of uncovering currently unrecognized patterns of disease. If this promise is fulfilled, AI may then be of use in challenging cases or where a second opinion is desirable.

AREAS TIMELY FOR DEVELOPING RESEARCH

There is a need to determine the external validity of deep learning algorithms and to better understand how the 'black box' paradigm reaches results.

Artificial intelligence in glaucoma

PURPOSE OF REVIEW

The use of computers has become increasingly relevant to medical decision-making, and artificial intelligence methods have recently demonstrated significant advances in medicine. We therefore provide an overview of current artificial intelligence methods and their applications, to help the practicing ophthalmologist understand their potential impact on glaucoma care.

RECENT FINDINGS

Techniques used in artificial intelligence can successfully analyze and categorize data from visual fields, optic nerve structure [e.g., optical coherence tomography (OCT) and fundus photography], ocular biomechanical properties, and a combination thereof to identify disease severity, determine disease progression, and/or recommend referral for specialized care. Algorithms have become increasingly complex in recent years, utilizing both supervised and unsupervised methods of artificial intelligence. Impressive performance of these algorithms on previously unseen data has been reported, often outperforming standard global indices and expert observers. However, there remains no clearly defined gold standard for determining the presence and severity of glaucoma, which undermines the training of these algorithms. To improve upon existing methodologies, future work must employ more robust definitions of disease, optimize data inputs for artificial intelligence analysis, and improve methods of extracting knowledge from learned results.

SUMMARY

Artificial intelligence has the potential to revolutionize the screening, diagnosis, and classification of glaucoma, both through the automated processing of large data sets, and by earlier detection of new disease patterns. In addition, artificial intelligence holds promise for fundamentally changing research aimed at understanding the development, progression, and treatment of glaucoma, by identifying novel risk factors and by evaluating the importance of existing ones.

Comparison of optic nerve head parameter measurements obtained by time-domain and spectral-domain optical coherence tomography

PURPOSE

To compare the measurements of the optic nerve head (ONH) parameters provided by time-domain (TD) and spectral-domain (SD) optical coherence tomography (OCT).

METHODS

Four ONH parameters were analyzed: optic disc area, rim area, cup-to-disc area ratio (CDR), and vertical cup-to-disc ratio (VCDR). Stratus OCT and Cirrus HD-OCT were used to obtain measurements by TD-OCT and SD-OCT, respectively. Stratus OCT measurements were collected before and after manual correction of the ONH edges.

RESULTS

Twenty healthy participants and 20 patients with glaucomatous eyes were enrolled. Although manual correction of Stratus OCT measurements reduced the differences compared with Cirrus HD-OCT, the latter measured a smaller mean disc area than Stratus OCT in healthy (2.02±0.31 vs. 2.18±0.29 mm2, P=0.0003) and glaucomatous eyes (1.92±0.35 vs. 2.19±0.38 mm2, P<0.0001). Cirrus HD-OCT measured a smaller rim area than Stratus OCT in healthy (1.31±0.30 vs. 1.56±0.32 mm2, P<0.0001) and glaucomatous eyes (0.80±0.25 vs. 0.97±0.36 mm2, P=0.0052), a higher CDR (0.55±0.11 vs. 0.49±0.11, P<0.0001) and VCDR in healthy eyes (0.55±0.11 vs. 0.49±0.11, P<0.0001), and a higher CDR in glaucomatous eyes (0.74±0.10 vs. 0.58±0.18, P<0.0001). No statistically significant differences were detected for VCDR in glaucomatous eyes. All measured values showed good correlation (r≥0.70). Large 95% limits of agreement between the 2 devices were found for most parameters.

CONCLUSIONS

ONH parameter measurements provided by TD-OCT and SD-OCT show significant differences and cannot be considered interchangeable.

Morphometric analysis and classification of glaucomatous optic neuropathy using radial polynomials

PURPOSE

To quantify the morphological features of the optic nerve head using radial polynomials, to use these morphometric models as the basis for classification of glaucomatous optic neuropathy via an automated decision tree induction algorithm, and to compare these classification results with established procedures.

METHODS

A cohort of patients with high-risk ocular hypertension or early glaucoma (n=179) and a second cohort of normal subjects (n=96) were evaluated for glaucomatous optic neuropathy using stereographic disc photography and confocal scanning laser tomography. Morphological features of the optic nerve head region were modeled from the tomography data using pseudo-Zernike radial polynomials and features derived from these models were used as the basis for classification by a decision tree induction algorithm. Decision tree classification performance was compared with expert classification of stereographic disc photographs and analysis of neural retinal rim thickness by Moorfields Regression Analysis (MRA).

RESULTS

Root mean squared error of the morphometric models decreased asymptotically with additional polynomial coefficients, from 62±0.5 (32 coefficients) to 32±5.7 μm (256 coefficients). Optimal morphometric classification was derived from a subset of 64 total features and had low sensitivity (69%), high specificity (88%), very good accuracy (80%), and area under the receiver operating characteristic curve (AUROC) was 88% (95% confidence interval, 78%-98%). In comparison, MRA classification of the same records had a comparatively poorer sensitivity (55%), but had higher specificity (95%), with similar overall accuracy (78%) and AUROC curve, 83% (95% CI, 70%-96%).

CONCLUSIONS

Pseudo-Zernike radial polynomials provide a mathematically compact and faithful morphological representation of the structural features of the optic nerve head. This morphometric method of glaucomatous optic neuropathy classification has greater sensitivity, and similar overall classification performance (AUROC) when compared with classification by neural retinal rim thickness by MRA in patients with high-risk ocular hypertension and early glaucoma.

Low limit for effective signal strength in the Stratus OCT in imperative low signal strength cases

Purpose

To determine the lowest limit of signal strength that is still effective for accurate analysis of optic coherence tomography (OCT) values, we investigated the reproducibility of OCT scans by signal strength (SS).

Methods

A total of 668 subjects were scanned for measurements of retinal nerve fiber layer (RNFL) thickness using the Stratus OCT twice on the same day. The variability of overall RNFL thickness parameters obtained at different SS was analyzed and compared by repeated-measures of ANOVA and Spearman's correlation coefficient. Values of the intraclass correlation coefficient (ICC) and variability (standard deviation) of RNFL thickness were obtained. The false positive ratio was analyzed.

Results

When SS was 3, the variability of RNFL thickness was significantly different (low ICC, high variability) in comparison to when SS was 4 or greater. Significant negative correlations were observed between variability in RNFL thickness and signal strength. The difference of variability of average RNFL thickness between SS 4 (4.94 µm) and SS 6 (4.41 µm) was 0.53 µm.

Conclusions

Clinically, the difference of variability of average RNFL thickness between SS 4 and SS 6 was quite small. High SS is important, however, when signal strength is low due to uncorrectable factors in patients in need of OCT for glaucoma and retinal disease. Our results suggest that SS 4 is the lowest acceptable limit of signal strength for obtaining reproducible scanning images.

Repeatability of optic nerve head parameters measured by spectral-domain OCT in healthy eyes

BACKGROUND AND OBJECTIVE

To evaluate the repeatability of optic nerve head (ONH) measurements by spectral-domain optical coherence tomography.

PATIENTS AND METHODS

Three scans were acquired in 32 healthy subjects during one session. Using Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA), the cup-to-disc ratio (CDR), disc area, rim area, cup volume, and horizontal and vertical CDRs were investigated. Repeatability was assessed by the coefficient of variation (COV), the test-retest intrasession variability, and the intraclass correlation coefficient (ICC).

RESULTS

Good repeatability was achieved for all parameters, with a COV of 4.23% or less and ICCs of 0.98 or greater for all measurements. Test-retest intrasession variability was 0.024 for CDR, 0.121 mm(2) for disc area, 0.087 mm(2) for rim area, 0.017 mm(3) for cup volume, and 0.048 for horizontal and vertical CDR.

CONCLUSION

In healthy eyes, Cirrus HD-OCT provides repeatable measurements of ONH parameters.

Biological roles of host defense peptides: lessons from transgenic animals and bioengineered tissues

Host defense peptides (HDPs) have long been recognized as microbicidal agents, but their roles as modulators of innate and adaptive immunity have only more recently been appreciated. The study of transgenic animal and tissue models has provided platforms to improve our understanding of the immune modulatory functions of HDPs. Here, the characterization of transgenic animals or tissue models that over-express and/or are deficient for specific HDPs is reviewed. We also attempt to reconcile this data with evidence from human studies monitoring HDP expression at constitutive levels and/or in conjunction with inflammation, infection models, or disease states. We have excluded activities ascribed to HDPs derived exclusively from in vitro experiments. An appreciation of the way that HDPs promote innate immunity or influence the adaptive immune response is necessary in order to exploit their therapeutic or adjuvant potential and to open new perspectives in understanding the basis of immunity. The potential applications for HDPs are discussed.

Effect of pupil dilation on retinal nerve fibre layer thickness measurements and their repeatability with Cirrus HD-OCT

PURPOSE

To assess whether pupil dilation influences retinal nerve fibre layer (RNFL) thickness measurements provided by spectral-domain optical coherence tomography (SD-OCT) in healthy individuals.

PATIENTS AND METHODS

In this observational case series, carried out in a private clinical practice, 32 eyes of 32 participants were investigated. Using Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA) three individual 200 × 200 cube optic disc scans were obtained before and after pupil dilation. The RNFL thickness was the outcome measure. Coefficient of variation (COV) and test-retest variability were calculated.

RESULTS

Pupil size did not influence RNFL thickness measurements: mean values did not change in any sector (except the 9 o'clock hour) after dilation. Excellent repeatability was achieved both before and after mydriasis. In the former condition, COV ranged between 1.37% (for average RNFL) and 4.46% (for clock hour 2 RNFL) and test-retest variability between 2.17 (for temporal quadrant RNFL) and 9.18 microm (for clock hour 6 RNFL). In the latter condition, COV ranged between 1.36% (for average RNFL) and 4.48% (for clock hour 2 RNFL) and test-retest variability between 2.41 (for average RNFL) and 9.29 microm (for clock hour 6 RNFL). The repeatability was higher than that previously reported for time-domain OCT.

CONCLUSION

In eyes with clear media highly repeatable measurements of the RNFL thickness can be obtained by SD-OCT both before and after mydriasis.

Relationship between retinal nerve fiber layer measurement and signal strength in optical coherence tomography

PURPOSE

To examine the relationship between signal strength and retinal nerve fiber layer (RNFL) thickness measured by optical coherence tomography (OCT).

DESIGN

Observational cross-sectional study.

PARTICIPANTS

Forty normal subjects were recruited.

METHODS

Retinal nerve fiber layer (RNFL) thickness was measured by Stratus OCT (Carl Zeiss Meditec, Dublin, CA). In each eye, the focusing knob was adjusted to obtain 6 images with different signal strengths ranging from 5 to 10. The relationships between signal strength and RNFL thickness were examined using the Spearman correlation coefficient. The differences of RNFL thicknesses were compared with repeated-measures analysis of variance.

MAIN OUTCOME MEASURES

Retinal nerve fiber layer thicknesses measured at different signal strengths.

RESULTS

Significant differences were observed between measurements obtained at signal strength of 10 and those obtained with signal strength of less than 10 at the superior, nasal, and temporal clock hours. RNFL thickness generally increased with the signal strength, with significant correlations found with the total average, superior, and nasal clock hours RNFL thicknesses.

CONCLUSIONS

Optical coherence tomography RNFL measurements vary significantly with signal strength. Obtaining the maximal possible signal strength is recommended for RNFL thickness measurement.