A hierarchical cluster analysis of normal-tension glaucoma using spectral-domain optical coherence tomography parameters

Classifying glaucoma exclusively with OCT: comparison of three clustering algorithms derived from machine learning

BACKGROUND/AIMS

To objectively classify eyes as either healthy or glaucoma based exclusively on data provided by peripapillary retinal nerve fiber layer (pRNFL) and ganglion cell-inner plexiform (GCIPL) measurements derived from spectral-domain optical coherence tomography (SD-OCT) using machine learning algorithms.

METHODS

Three clustering methods (k-means, hierarchical cluster analysis -HCA- and model-based clustering-MBC-) were used separately to classify a training sample of 109 eyes as either healthy or glaucomatous using solely 13 SD-OCT parameters: pRNFL average and sector thicknesses and GCIPL average and minimum values together with the six macular wedge-shaped regions. Then, the best-performing algorithm was applied to an independent test sample of 102 eyes to derive close estimates of its actual performance (external validation).

RESULTS

In the training sample, accuracy was 91.7% for MBC, 81.7% for k-means and 78.9% for HCA (p value = 0.02). The best MBC model was that in which subgroups were allowed to have variable volume and shape and equal orientation. The MBC algorithm in the independent test sample correctly classified 98 out of 102 cases for an overall accuracy of 96.1% (95% CI, 92.3-99.8%), with a sensitivity of 94.3 and 100% specificity. The accuracy for pRNFL was 92.2% (95% CI, 86.9-97.4%) and for GCIPL 98.0% (95% CI, 95.3-100%).

CONCLUSIONS

Clustering algorithms in general (and MBC in particular) seem promising methods to help discriminate between healthy and glaucomatous eyes using exclusively SD-OCT-derived parameters. Understanding the relative merits of one method over others may also provide insights into the nature of the disease.

Cluster analysis reveals patterns of age-related change in anterior chamber depth for gender and ethnicity: clinical implications

Purpose

To identify patterns of age‐, gender‐ and refractive‐ related changes in Scheimpflug‐based anterior chamber depth across the central 8 mm of chamber width, to derive normative models, potentially useful for angle closure disease diagnosis.

Methods

This was a retrospective, cross‐sectional study. Scheimpflug photography was used to obtain anterior chamber depth measurements at 57 points across the central 8 mm of the chamber width from one eye of each healthy subject (male Caucasians (n = 189), female Caucasians (n = 186), male Asians (n = 165) and female Asians (n = 181)). Sliding window and nonlinear regression analysis was used to identify the age‐related changes in chamber depth. Hierarchical cluster analysis was used to identify test locations with statistically identical age‐related shifts, which were used to perform age‐correction for all subjects, resulting in normative distributions of chamber depth across the chamber width. The model was examined with and without the contribution of spherical equivalent refractive error.

Results

Distinct clusters, demonstrating statistically indistinguishable age‐related changes of chamber depth over time, were identified. These age‐related changes followed a nonlinear regression (fifth or sixth order polynomial). Females tended to have a greater rate of chamber depth shallowing. Incorporating refractive error into the model produced minimal changes to the fit relative to the ground truth. Comparisons with cut‐offs for angle closure from the literature showed that ageing alone was insufficient for identifying angle closure disease.

Conclusions

Age‐, ethnicity‐ and gender‐related differences need to be acknowledged in order to utilise anterior chamber depth data for angle closure disease diagnosis correctly. Ageing alone does not adequately account for the angle closure disease process.

Progression patterns of normal-tension glaucoma groups classified by hierarchical cluster analysis

OBJECTIVES

To investigate differences in progression patterns of normal-tension glaucoma (NTG) patients in three clusters classified by hierarchical cluster analysis (HCA).

MATERIALS AND METHODS

In a retrospective study, 200 eyes of NTG patients classified by HCA in 2015 who were followed up to the current date were evaluated. Peripapillary retinal nerve fibre layer (RNFL) thicknesses were measured by Cirrus HD-OCT and progression rate was calculated by trend analysis (Guided Progression Analysis [GPA]). VF progression rate was evaluated by linear regression analysis of mean deviation (MD). Progression patterns of three clusters were compared by histograms.

RESULTS

In total, 153 eyes of 153 patients were followed up. Mean observation period was 5 years. RNFL reduction rate was -0.83 μm/year in cluster 1, which showed early glaucomatous damage in previous reports; -0.45 μm/year in cluster 2, which showed moderate glaucomatous damage; and -0.36 μm/year in cluster 3, which showed young and myopic glaucomatous damage. The progression pattern of cluster 3 showed a double-peak distribution; RNFL reduction rate was 0.11 μm/year in the non-progressive group and -1.07 μm/year in the progressive group.

CONCLUSION

The progression patterns were different among three NTG groups that were divided by HCA. In particular, the group of young and myopic eyes showed a mixture of two different patterns.

Modeling Changes in Corneal Parameters With Age: Implications for Corneal Disease Detection

PURPOSE

To apply computational methods to model normal age-related changes in corneal parameters and to establish their association with demographic factors, thereby providing a framework for improved detection of subclinical corneal ectasia (SCE).

DESIGN

Cross-sectional study.

METHODS

One hundred seventeen healthy participants were enrolled from Centre for Eye Health (Sydney, Australia). Corneal thickness (CT), front surface sagittal curvature (FSSC), and back surface sagittal curvature (BSSC) measurements were extracted from 57 corneal locations from 1 eye per participant using the Pentacam HR. Cluster analyses were performed to identify locations demonstrating similar variations with age. Age-related changes were modeled using polynomial regression with sliding window methods, and model accuracy was verified with Bland-Altman comparisons. Pearson correlations were applied to examine the impacts of demographic factors.

RESULTS

Concentric cluster patterns were observed for CT and FSSC but not for BSSC. Sliding window analyses were best fit with quartic and cubic regression models for CT and FSSC/BSSC, respectively. CT and FSSC sliding window models had narrower 95% limits of agreement compared with decade-based models (0.015 mm vs 0.017 mm and 0.14 mm vs 0.27 mm, respectively), but were wider for BSSC than decade-based models (0.73 mm vs 0.54 mm). Significant correlations were observed between CT and astigmatism (P = .02-.049) and FSSC and BSSC and gender (P = <.001-.049).

CONCLUSIONS

The developed models robustly described aging variations in CT and FSSC; however, other mechanisms appear to contribute to variations in BSSC. These findings and the identified correlations provide a framework that can be applied to future model development and establishment of normal databases to facilitate SCE detection.

Development of a Spatial Model of Age-Related Change in the Macular Ganglion Cell Layer to Predict Function From Structural Changes

PURPOSE

To develop location-specific models of normal, age-related changes in the macular ganglion cell layer (GCL) from optical coherence tomography (OCT). Using these OCT-derived models, we predicted visual field (VF) sensitivities and compared these results to actual VF sensitivities.

DESIGN

Retrospective cohort study.

METHODS

Single eyes of 254 normal participants were retrospectively enrolled from the Centre for Eye Health (Sydney, Australia). Macular GCL measurements were obtained using Spectralis OCT. Cluster algorithms were performed to identify spatial patterns demonstrating similar age-related change. Quadratic and linear regression models were subsequently used to characterize age-related GCL decline. Forty participants underwent additional testing with Humphrey VFs, and 95% prediction intervals were calculated to measure the predictive ability of structure-function models incorporating cluster-based pooling, age correction, and consideration of spatial summation.

RESULTS

Quadratic GCL regression models provided a superior fit (P value <.0001-.0066), establishing that GCL decline commences in the late 30s across the macula. The equivalent linear rates of GCL decline showed eccentricity-dependent variation (0.13 μm/yr centrally vs 0.06 μm/yr peripherally); however, average, normalized GCL loss per year was consistent across the 64 macular measurement locations at 0.26%. The 95% prediction intervals describing predicted VF sensitivities were significantly narrower across all cluster-based structure-function models (3.79-4.99 dB) compared with models without clustering applied (5.66-6.73 dB, P < .0001).

CONCLUSIONS

Combining spatial clustering with age-correction based on regression models allowed the development of robust models describing GCL changes with age. The resultant superior predictive ability of VF sensitivity from ganglion cell measurements may be applied to future models of disease development to improve detection of early macular GCL pathology.

Pattern Recognition Analysis Reveals Unique Contrast Sensitivity Isocontours Using Static Perimetry Thresholds Across the Visual Field

Purpose

To determine the locus of test locations that exhibit statistically similar age-related decline in sensitivity to light increments and age-corrected contrast sensitivity isocontours (CSIs) across the central visual field (VF). We compared these CSIs with test point clusters used by the Glaucoma Hemifield Test (GHT).

Methods

Sixty healthy observers underwent testing on the Humphrey Field Analyzer 30-2 test grid using Goldmann (G) stimulus sizes I-V. Age-correction factors for GI-V were determined using linear regression analysis. Pattern recognition analysis was used to cluster test locations across the VF exhibiting equal age-related sensitivity decline (age-related CSIs), and points of equal age-corrected sensitivity (age-corrected CSIs) for GI-V.

Results

There was a small but significant test size–dependent sensitivity decline with age, with smaller stimuli declining more rapidly. Age-related decline in sensitivity was more rapid in the periphery. A greater number of unique age-related CSIs was revealed when using smaller stimuli, particularly in the mid-periphery. Cluster analysis of age-corrected sensitivity thresholds revealed unique CSIs for GI-V, with smaller stimuli having a greater number of unique clusters. Zones examined by the GHT consisted of test locations that did not necessarily belong to the same CSI, particularly in the periphery.

Conclusions

Cluster analysis reveals statistically significant groups of test locations within the 30-2 test grid exhibiting the same age-related decline. CSIs facilitate pooling of sensitivities to reduce the variability of individual test locations. These CSIs could guide future structure-function and alternate hemifield asymmetry analyses by comparing matched areas of similar sensitivity signatures.