Logistic Regression Analysis for Glaucoma Diagnosis Using Stratus Optical Coherence Tomography

Glaucoma detection model by exploiting multi-region and multi-scan-pattern OCT images with dynamical region score

Currently, deep learning-based methods have achieved success in glaucoma detection. However, most models focus on OCT images captured by a single scan pattern within a given region, holding the high risk of the omission of valuable features in the remaining regions or scan patterns. Therefore, we proposed a multi-region and multi-scan-pattern fusion model to address this issue. Our proposed model exploits comprehensive OCT images from three fundus anatomical regions (macular, middle, and optic nerve head regions) being captured by four scan patterns (radial, volume, single-line, and circular scan patterns). Moreover, to enhance the efficacy of integrating features across various scan patterns within a region and multiple regional features, we employed an attention multi-scan fusion module and an attention multi-region fusion module that auto-assign contribution to distinct scan-pattern features and region features adapting to characters of different samples, respectively. To alleviate the absence of available datasets, we have collected a specific dataset (MRMSG-OCT) comprising OCT images captured by four scan patterns from three regions. The experimental results and visualized feature maps both demonstrate that our proposed model achieves superior performance against the single scan-pattern models and single region-based models. Moreover, compared with the average fusion strategy, our proposed fusion modules yield superior performance, particularly reversing the performance degradation observed in some models relying on fixed weights, validating the efficacy of the proposed dynamic region scores adapted to different samples. Moreover, the derived region contribution scores enhance the interpretability of the model and offer an overview of the model's decision-making process, assisting ophthalmologists in prioritizing regions with heightened scores and increasing efficiency in clinical practice.

Comparison of Optical Coherence Tomography Structural Parameters for Diagnosis of Glaucoma in High Myopia

Importance

Diagnosis of glaucoma in highly myopic eyes is challenging. This study compared the glaucoma detection utility of various optical coherence tomography (OCT) parameters for high myopia.

Objective

To compare the diagnostic accuracy of single OCT parameters, the University of North Carolina (UNC) OCT Index, and the temporal raphe sign for discrimination of glaucoma in patients with high myopia.

Design, Setting, and Participants

This was a retrospective cross-sectional study conducted from January 1, 2014, and January 1, 2022. Participants with high myopia (axial length ≥26.0 mm or spherical equivalent ≤-6 diopters) plus glaucoma and participants with high myopia without glaucoma were recruited from a single tertiary hospital in South Korea.

Exposures

Macular ganglion cell-inner plexiform layer (GCIPL) thickness, peripapillary retinal nerve fiber layer (RNFL) thickness, and optic nerve head (ONH) parameters were measured in each participant. The UNC OCT scores and the temporal raphe sign were checked to compare diagnostic utility. Decision tree analysis with single OCT parameters, the UNC OCT Index, and the temporal raphe sign were also applied.

Main outcome and Measures

Area under the receiver operating characteristic curve (AUROC).

Results

A total of 132 individuals with high myopia and glaucoma (mean [SD] age, 50.0 [11.7] years; 78 male [59.1%]) along with 142 individuals with high myopia without glaucoma (mean [SD] age, 50.0 [11.3] years; 79 female [55.6%]) were included in the study. The AUROC of the UNC OCT Index was 0.891 (95% CI, 0.848-0.925). The AUROC of temporal raphe sign positivity was 0.922 (95% CI, 0.883-0.950). The best single OCT parameter was inferotemporal GCIPL thickness (AUROC, 0.951; 95% CI, 0.918-0.973), and its AUROC difference from the UNC OCT Index, temporal raphe sign, mean RNFL thickness, and ONH rim area was 0.060 (95% CI, 0.016-0.103; P = .007); 0.029 (95% CI, -0.009 to 0.068; P = .13), 0.022 (95% CI, -0.012-0.055; P = .21), and 0.075 (95% CI, 0.031-0.118; P < .001), respectively.

Conclusions and Relevance

Results of this cross-sectional study suggest that in discriminating glaucomatous eyes in patients with high myopia, inferotemporal GCIPL thickness yielded the highest AUROC value. The RNFL thickness and GCIPL thickness parameters may play a greater role in glaucoma diagnosis than the ONH parameters in high myopia.

Evaluation of University of North Carolina OCT Index for Diagnosis of Early Glaucoma

PURPOSE

To evaluate the diagnostic performance of the University of North Carolina (UNC) OCT Index based on Cirrus high-definition OCT to discriminate early glaucomatous eyes from normal eyes in clinical practice.

DESIGN

Evaluation of diagnostic test or technology.

PARTICIPANTS

Ninety-eight patients with early glaucoma and 98 age-matched normal subjects.

METHODS

Macular ganglion cell-inner plexiform layer (GCIPL) thickness, peripapillary retinal nerve fiber layer (RNFL) thickness, and optic nerve head parameters were measured in each subject. The measurements were run through the UNC OCT algorithm to compare their diagnostic abilities.

MAIN OUTCOME MEASURES

Area under the curve (AUC) of the receiver operating characteristic and sensitivity at 95% specificity.

RESULTS

The AUC of the UNC OCT Index was 0.974. The best AUCs of the single parameters were those of the minimum GCIPL (0.926) of the macular GCIPL, average RNFL (0.916) of the peripapillary RNFL, and rim area (0.964) of the optic nerve head. The AUC of the UNC OCT Index was significantly greater than those of the minimum GCIPL and average RNFL (all P values < 0.05), and also outperformed the rim area. The sensitivity value of the UNC OCT Index (90.8) was greater than that of single OCT parameters (minimum GCIPL, 42.9; average RNFL, 64.3; rim area, 84.7) at 95% specificity.

CONCLUSIONS

The diagnostic performance of the UNC OCT Index in discriminating early glaucomatous eyes from normal eyes is high and exceeds the best optic nerve head, peripapillary RNFL, and macular GCIPL parameters in clinical practice.

A new diagnostic model of primary open angle glaucoma based on FD-OCT parameters

AIM

To build a clinical diagnostic model of primary open angle glaucoma (POAG) using the normal probability chart of frequency-domain optical coherence tomography (FD-OCT).

METHODS

This is a cross-sectional study. Total 133 eyes from 133 healthy subjects and 99 eyes from 99 early POAG patients were included in the study. The retinal nerve fibre layer (RNFL) thickness parameters of optic nerve head (ONH) and RNFL3.45 scan were measured in one randomly selected eye of each subject using RTVue-100 FD-OCT. Then, we used these parameters to establish the diagnostic models. Four different diagnostic models based on two different area partition strategies on ONH and RNFL3.45 parameters, including ONH traditional area partition model (ONH-T), ONH new area partition model (ONH-N), RNFL3.45 traditional area partition model (RNFL3.45-T) and RNFL3.45 new area partition model (RNFL3.45-N), were built and tested by cross-validation.

RESULTS

The new area partition models had higher area under the receiver operating characteristic (AROC; ONH-N: 0.990; RNFL3.45-N: 0.939) than corresponding traditional area partition models (ONH-T: 0.979; RNFL3.45-T: 0.881). There was no statistical difference among AROC of ONH-T, ONH-N, and RNFL3.45-N. Nevertheless, ONH-N was the simplest model.

CONCLUSION

The new area partition models had higher diagnostic accuracy than corresponding traditional area partition models, which can improve the diagnostic ability of early POAG. In particular, the simplest ONH-N diagnostic model may be convenient for clinical application.

Validation of the UNC OCT Index for the Diagnosis of Early Glaucoma

Purpose

To independently validate the performance of the University of North Carolina Optical Coherence Tomography (UNC OCT) Index in diagnosing and predicting early glaucoma.

Methods

Data of 118 normal subjects (118 eyes) and 96 subjects (96 eyes) with early glaucoma defined as visual field mean deviation (MD) greater than -4 decibels (dB), aged 40 to 80 years, and who were enrolled in the Full-Threshold Testing Size III, V, VI comparison study were used in this study. CIRRUS OCT average and quadrants' retinal nerve fiber layer (RNFL); optic disc vertical cup-to-disc ratio (VCDR), cup-to-disc area ratio, and rim area; and average, minimum, and six sectoral ganglion cell-inner plexiform layer (GCIPL) measurements were run through the UNC OCT Index algorithm. Area under the receiver operating characteristic curve (AUC) and sensitivities at 95% and 99% specificity were calculated and compared between single parameters and the UNC OCT Index.

Results

Mean age was 60.1 ± 11.0 years for normal subjects and 66.5 ± 8.1 years for glaucoma patients (P < 0.001). MD was 0.29 ± 1.04 dB and -1.30 ± 1.35 dB in normal and glaucomatous eyes (P < 0.001), respectively. The AUC of the UNC OCT Index was 0.96. The best single metrics when compared to the UNC OCT Index were VCDR (0.93, P = 0.054), average RNFL (0.92, P = 0.014), and minimum GCIPL (0.91, P = 0.009). The sensitivities at 95% and 99% specificity were 85.4% and 76.0% (UNC OCT Index), 71.9% and 62.5% (VCDR, all P < 0.001), 64.6% and 53.1% (average RNFL, all P < 0.001), and 66.7% and 58.3% (minimum GCIPL, all P < 0.001), respectively.

Conclusions

The findings confirm that the UNC OCT Index may provide improved diagnostic perforce over that of single OCT parameters and may be a good tool for detection of early glaucoma.

Translational Relevance

The UNC OCT Index algorithm may be incorporated easily into routine clinical practice and be useful for detecting early glaucoma.

Clinical Use of an Optical Coherence Tomography Linear Discriminant Function for Differentiating Glaucoma From Normal Eyes

PURPOSE

To determine and validate the diagnostic ability of a linear discriminant function (LDF) based on retinal nerve fiber layer (RNFL) and ganglion cell-inner plexiform layer (GCIPL) thickness obtained using high-definition optical coherence tomography (Cirrus HD-OCT) for discriminating between healthy controls and early glaucoma subjects.

METHODS

We prospectively selected 214 healthy controls and 152 glaucoma subjects (teaching set) and another independent sample of 86 healthy controls and 71 glaucoma subjects (validating set). Two scans, including 1 macular and 1 peripapillary RNFL scan, were obtained. After calculating the LDF in the teaching set using the binary logistic regression analysis, receiver operating characteristic curves were plotted and compared between the OCT-provided parameters and LDF in the validating set.

RESULTS

The proposed LDF was 16.529-(0.132×superior RNFL)-(0.064×inferior RNFL)+(0.039×12 o'clock RNFL)+(0.038×1 o'clock RNFL)+(0.084×superior GCIPL)-(0.144×minimum GCIPL). The highest area under the receiver operating characteristic (AUROC) curve was obtained for LDF in both sets (AUROC=0.95 and 0.96). In the validating set, the LDF showed significantly higher AUROC than the best RNFL (inferior RNFL=0.91) and GCIPL parameter (minimum GCIPL=0.88). The LDF yielded a sensitivity of 93.0% at a fixed specificity of 85.0%.

CONCLUSIONS

The LDF showed better diagnostic ability for differentiating between healthy and early glaucoma subjects than individual OCT parameters. A classification algorithm based on the LDF can be used in the OCT analysis for glaucoma diagnosis.

Cross-sectional study: Does combining optical coherence tomography measurements using the 'Random Forest' decision tree classifier improve the prediction of the presence of perimetric deterioration in glaucoma suspects?

OBJECTIVES

To develop a classifier to predict the presence of visual field (VF) deterioration in glaucoma suspects based on optical coherence tomography (OCT) measurements using the machine learning method known as the 'Random Forest' algorithm.

DESIGN

Case-control study.

PARTICIPANTS

293 eyes of 179 participants with open angle glaucoma (OAG) or suspected OAG.

INTERVENTIONS

Spectral domain OCT (Topcon 3D OCT-2000) and perimetry (Humphrey Field Analyser, 24-2 or 30-2 SITA standard) measurements were conducted in all of the participants. VF damage (Ocular Hypertension Treatment Study criteria (2002)) was used as a 'gold-standard' to classify glaucomatous eyes. The 'Random Forest' method was then used to analyse the relationship between the presence/absence of glaucomatous VF damage and the following variables: age, gender, right or left eye, axial length plus 237 different OCT measurements.

MAIN OUTCOME MEASURES

The area under the receiver operating characteristic curve (AROC) was then derived using the probability of glaucoma as suggested by the proportion of votes in the Random Forest classifier. For comparison, five AROCs were derived based on: (1) macular retinal nerve fibre layer (m-RNFL) alone; (2) circumpapillary (cp-RNFL) alone; (3) ganglion cell layer and inner plexiform layer (GCL+IPL) alone; (4) rim area alone and (5) a decision tree method using the same variables as the Random Forest algorithm.

RESULTS

The AROC from the combined Random Forest classifier (0.90) was significantly larger than the AROCs based on individual measurements of m-RNFL (0.86), cp-RNFL (0.77), GCL+IPL (0.80), rim area (0.78) and the decision tree method (0.75; p<0.05).

CONCLUSIONS

Evaluating OCT measurements using the Random Forest method provides an accurate prediction of the presence of perimetric deterioration in glaucoma suspects.

Diagnostic capability of spectral-domain optical coherence tomography for glaucoma

PURPOSE

To determine the diagnostic capability of spectral-domain optical coherence tomography in glaucoma patients with visual field defects.

DESIGN

Prospective, cross-sectional study.

METHODS

SETTINGS

Participants were recruited from a university hospital clinic.

STUDY POPULATION

One eye of 85 normal subjects and 61 glaucoma patients with average visual field mean deviation of -9.61 ± 8.76 dB was selected randomly for the study. A subgroup of the glaucoma patients with early visual field defects was calculated separately.

OBSERVATION PROCEDURES

Spectralis optical coherence tomography (Heidelberg Engineering, Inc) circular scans were performed to obtain peripapillary retinal nerve fiber layer (RNFL) thicknesses. The RNFL diagnostic parameters based on the normative database were used alone or in combination for identifying glaucomatous RNFL thinning.

MAIN OUTCOME MEASURES

To evaluate diagnostic performance, calculations included areas under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio.

RESULTS

Overall RNFL thickness had the highest area under the receiver operating characteristic curve values: 0.952 for all patients and 0.895 for the early glaucoma subgroup. For all patients, the highest sensitivity (98.4%; 95% confidence interval, 96.3% to 100%) was achieved by using 2 criteria: ≥ 1 RNFL sectors being abnormal at the < 5% level and overall classification of borderline or outside normal limits, with specificities of 88.9% (95% confidence interval, 84.0% to 94.0%) and 87.1% (95% confidence interval, 81.6% to 92.5%), respectively, for these 2 criteria.

CONCLUSIONS

Statistical parameters for evaluating the diagnostic performance of the Spectralis spectral-domain optical coherence tomography were good for early perimetric glaucoma and were excellent for moderately advanced perimetric glaucoma.

Glaucoma diagnostic accuracy of ganglion cell-inner plexiform layer thickness: comparison with nerve fiber layer and optic nerve head

PURPOSE

To determine the diagnostic performance of macular ganglion cell-inner plexiform layer (GCIPL) thickness measured with the Cirrus high-definition optical coherence tomography (HD-OCT) ganglion cell analysis (GCA) algorithm (Carl Zeiss Meditec, Dublin, CA) to discriminate normal eyes and eyes with early glaucoma and to compare it with that of peripapillary retinal nerve fiber layer (RNFL) thickness and optic nerve head (ONH) measurements.

DESIGN

Evaluation of diagnostic test or technology.

PARTICIPANTS

Fifty-eight patients with early glaucoma and 99 age-matched normal subjects.

METHODS

Macular GCIPL and peripapillary RNFL thicknesses and ONH parameters were measured in each participant, and their diagnostic abilities were compared.

MAIN OUTCOME MEASURES

Area under the curve (AUC) of the receiver operating characteristic.

RESULTS

The GCIPL parameters with the best AUCs were the minimum (0.959), inferotemporal (0.956), average (0.935), superotemporal (0.919), and inferior sector (0.918). There were no significant differences between these AUCs and those of inferior quadrant (0.939), average (0.936), and superior quadrant RNFL (0.933); vertical cup-to-disc diameter ratio (0.962); cup-to-disc area ratio (0.933); and rim area (0.910), all P>0.05.

CONCLUSIONS

The ability of macular GCIPL parameters to discriminate normal eyes and eyes with early glaucoma is high and comparable to that of the best peripapillary RNFL and ONH parameters.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found after the references.

Application of optical coherence tomography in glaucoma suspect eyes

PURPOSE

  The aim of the study was to compare the optical coherence tomography (OCT) parameters of the optic nerve head (ONH) and retinal nerve fibre layer (RNFL) and to identify which measurements are best able to differentiate between normal and glaucoma suspect eyes.

METHODS

  The study included 27 eyes with ocular hypertension (OHT), 33 eyes with pre-perimetric glaucoma (PG), 30 perimetrically unaffected eyes of patients with glaucoma in the fellow eye (FE) and 58 eyes of age-matched normal volunteers. All subjects underwent a complete eye examination with standard automated perimetry, optic disc photography and OCT imaging. Peripapillary 'fast RNFL thickness scans' and 'fast optic disc scans' were performed with time-domain OCT. The ONH and RNFL parameters were compared among the four study groups. The ONH and RNFL parameters were examined alone and then combined via four linear discriminant functions (LDF): LDF 1, the optimal combination of ONH parameters; LDF 2, the optimal combination of RNFL parameters; LDF 3, the optimal combination of both ONH and RNFL parameters; and LDF 4, the optimal combination of the best 11 parameters. The areas under the receiver operating curves (AUC) and the sensitivity at fixed specificity of at least 80 and 95 per cent were calculated for single parameters and LDF combinations and then compared. The best 11 parameters were selected based on their AUC values.

RESULTS

  Comparative analysis of OCT parameters revealed statistically significant differences in all seven ONH parameters in both PG and FE groups (and only in one ONH measurement in the ocular hypertensive group) when compared with normal eyes. Most of the RNFL parameters demonstrated statistically significant differences in all of the study groups when compared with the control group. The max-min parameter (0.835), inferior quadrant (0.833) and average RNFL thickness (0.829) obtained the highest AUC values in the whole glaucoma suspect group. The rim area had the best diagnostic accuracy among the ONH parameters (AUC = 0.817). The AUC values of the four LDF were: 0.825 (LDF 1), 0.882 (LDF 2), 0.902 (LDF 3) and 0.888 (LDF 4). Statistically significant differences were found between the AUC values of the single best ONH and RNFL parameters and LDF 3 and LDF 4.

CONCLUSIONS

  In the present study, RNFL parameters presented with better discriminatory abilities than ONH parameters in the OHT and FE groups. The ONH parameters demonstrated better diagnostic precision in differentiating between PG and normal eyes. The average RNFL thickness, max-min parameter and inferior quadrant RNFL thickness had the best abilities among single OCT measurements for discriminating between glaucoma suspect (including all ocular hypertensive, PG and FE eyes) and normal eyes. The combination of RNFL parameters only or both ONH and RNFL parameters, using linear discriminant analysis, provided the best classification results, improving the diagnostic accuracy of the instrument.

Machine learning classifiers for glaucoma diagnosis based on classification of retinal nerve fibre layer thickness parameters measured by Stratus OCT

PURPOSE

To compare the performance of two machine learning classifiers (MLCs), artificial neural networks (ANNs) and support vector machines (SVMs), with input based on retinal nerve fibre layer thickness (RNFLT) measurements by optical coherence tomography (OCT), on the diagnosis of glaucoma, and to assess the effects of different input parameters.

METHODS

We analysed Stratus OCT data from 90 healthy persons and 62 glaucoma patients. Performance of MLCs was compared using conventional OCT RNFLT parameters plus novel parameters such as minimum RNFLT values, 10th and 90th percentiles of measured RNFLT, and transformations of A-scan measurements. For each input parameter and MLC, the area under the receiver operating characteristic curve (AROC) was calculated.

RESULTS

There were no statistically significant differences between ANNs and SVMs. The best AROCs for both ANN (0.982, 95%CI: 0.966-0.999) and SVM (0.989, 95% CI: 0.979-1.0) were based on input of transformed A-scan measurements. Our SVM trained on this input performed better than ANNs or SVMs trained on any of the single RNFLT parameters (p < or = 0.038). The performance of ANNs and SVMs trained on minimum thickness values and the 10th and 90th percentiles were at least as good as ANNs and SVMs with input based on the conventional RNFLT parameters.

CONCLUSION

No differences between ANN and SVM were observed in this study. Both MLCs performed very well, with similar diagnostic performance. Input parameters have a larger impact on diagnostic performance than the type of machine classifier. Our results suggest that parameters based on transformed A-scan thickness measurements of the RNFL processed by machine classifiers can improve OCT-based glaucoma diagnosis.