The ability of the GDx Nerve Fibre Analyser neural network to diagnose glaucoma

Artificial intelligence in glaucoma: opportunities, challenges, and future directions

Artificial intelligence (AI) has shown excellent diagnostic performance in detecting various complex problems related to many areas of healthcare including ophthalmology. AI diagnostic systems developed from fundus images have become state-of-the-art tools in diagnosing retinal conditions and glaucoma as well as other ocular diseases. However, designing and implementing AI models using large imaging data is challenging. In this study, we review different machine learning (ML) and deep learning (DL) techniques applied to multiple modalities of retinal data, such as fundus images and visual fields for glaucoma detection, progression assessment, staging and so on. We summarize findings and provide several taxonomies to help the reader understand the evolution of conventional and emerging AI models in glaucoma. We discuss opportunities and challenges facing AI application in glaucoma and highlight some key themes from the existing literature that may help to explore future studies. Our goal in this systematic review is to help readers and researchers to understand critical aspects of AI related to glaucoma as well as determine the necessary steps and requirements for the successful development of AI models in glaucoma.

Machine Learning Techniques in Clinical Vision Sciences

This review presents and discusses the contribution of machine learning techniques for diagnosis and disease monitoring in the context of clinical vision science. Many ocular diseases leading to blindness can be halted or delayed when detected and treated at its earliest stages. With the recent developments in diagnostic devices, imaging and genomics, new sources of data for early disease detection and patients' management are now available. Machine learning techniques emerged in the biomedical sciences as clinical decision-support techniques to improve sensitivity and specificity of disease detection and monitoring, increasing objectively the clinical decision-making process. This manuscript presents a review in multimodal ocular disease diagnosis and monitoring based on machine learning approaches. In the first section, the technical issues related to the different machine learning approaches will be present. Machine learning techniques are used to automatically recognize complex patterns in a given dataset. These techniques allows creating homogeneous groups (unsupervised learning), or creating a classifier predicting group membership of new cases (supervised learning), when a group label is available for each case. To ensure a good performance of the machine learning techniques in a given dataset, all possible sources of bias should be removed or minimized. For that, the representativeness of the input dataset for the true population should be confirmed, the noise should be removed, the missing data should be treated and the data dimensionally (i.e., the number of parameters/features and the number of cases in the dataset) should be adjusted. The application of machine learning techniques in ocular disease diagnosis and monitoring will be presented and discussed in the second section of this manuscript. To show the clinical benefits of machine learning in clinical vision sciences, several examples will be presented in glaucoma, age-related macular degeneration, and diabetic retinopathy, these ocular pathologies being the major causes of irreversible visual impairment.

Predicting glaucomatous progression in glaucoma suspect eyes using relevance vector machine classifiers for combined structural and functional measurements

PURPOSE

The goal of this study was to determine if glaucomatous progression in suspect eyes can be predicted from baseline confocal scanning laser ophthalmoscope (CSLO) and standard automated perimetry (SAP) measurements analyzed with relevance vector machine (RVM) classifiers.

METHODS

Two hundred sixty-four eyes of 193 participants were included. All eyes had normal SAP results at baseline with five or more SAP tests over time. Eyes were labeled progressed (n = 47) or stable (n = 217) during follow-up based on SAP Guided Progression Analysis or serial stereophotograph assessment. Baseline CSLO-measured topographic parameters (n = 117) and baseline total deviation values from the 24-2 SAP test-grid (n = 52) were selected from each eye. Ten-fold cross-validation was used to train and test RVMs using the CSLO and SAP features. Receiver operating characteristic (ROC) curve areas were calculated using full and optimized feature sets. ROC curve results from RVM analyses of CSLO, SAP, and CSLO and SAP combined were compared to CSLO and SAP global indices (Glaucoma Probability Score, mean deviation and pattern standard deviation).

RESULTS

The areas under the ROC curves (AUROCs) for RVMs trained on optimized feature sets of CSLO parameters, SAP parameters, and CSLO and SAP parameters combined were 0.640, 0.762, and 0.805, respectively. AUROCs for CSLO Glaucoma Probability Score, SAP mean deviation (MD), and SAP pattern standard deviation (PSD) were 0.517, 0.513, and 0.620, respectively. No CSLO or SAP global indices discriminated between baseline measurements from progressed and stable eyes better than chance.

CONCLUSIONS

In our sample, RVM analyses of baseline CSLO and SAP measurements could identify eyes that showed future glaucomatous progression with a higher accuracy than the CSLO and SAP global indices. (ClinicalTrials.gov numbers, NCT00221897, NCT00221923.).

Integration and fusion of standard automated perimetry and optical coherence tomography data for improved automated glaucoma diagnostics

Background

The performance of glaucoma diagnostic systems could be conceivably improved by the integration of functional and structural test measurements that provide relevant and complementary information for reaching a diagnosis. The purpose of this study was to investigate the performance of data fusion methods and techniques for simple combination of Standard Automated Perimetry (SAP) and Optical Coherence Tomography (OCT) data for the diagnosis of glaucoma using Artificial Neural Networks (ANNs).

Methods

Humphrey 24-2 SITA standard SAP and StratusOCT tests were prospectively collected from a randomly selected population of 125 healthy persons and 135 patients with glaucomatous optic nerve heads and used as input for the ANNs. We tested commercially available standard parameters as well as novel ones (fused OCT and SAP data) that exploit the spatial relationship between visual field areas and sectors of the OCT peripapillary scan circle. We evaluated the performance of these SAP and OCT derived parameters both separately and in combination.

Results

The diagnostic accuracy from a combination of fused SAP and OCT data (95.39%) was higher than that of the best conventional parameters of either instrument, i.e. SAP Glaucoma Hemifield Test (p < 0.001) and OCT Retinal Nerve Fiber Layer Thickness ≥ 1 quadrant (p = 0.031). Fused OCT and combined fused OCT and SAP data provided similar Area under the Receiver Operating Characteristic Curve (AROC) values of 0.978 that were significantly larger (p = 0.047) compared to ANNs using SAP parameters alone (AROC = 0.945). On the other hand, ANNs based on the OCT parameters (AROC = 0.970) did not perform significantly worse than the ANNs based on the fused or combined forms of input data. The use of fused input increased the number of tests that were correctly classified by both SAP and OCT based ANNs.

Conclusions

Compared to the use of SAP parameters, input from the combination of fused OCT and SAP parameters, and from fused OCT data, significantly increased the performance of ANNs. Integrating parameters by including a priori relevant information through data fusion may improve ANN classification accuracy compared to currently available methods.

The impact of retardance pattern variability on nerve fiber layer measurements over time using GDx with variable and enhanced corneal compensation

PURPOSE

To examine the impact of retardance pattern variability on retinal nerve fiber layer (RNFL) measurements over time using scanning laser polarimetry with variable (GDxVCC) and enhanced corneal compensation (GDxECC; both by Carl Zeiss Meditec, Inc., Dublin, CA).

METHODS

Glaucoma suspect and glaucomatous eyes with 4 years of follow-up participating in the Advanced Imaging in Glaucoma Study were prospectively enrolled. All eyes underwent standard automated perimetry (SAP), GDxVCC, and GDxECC imaging every 6 months. SAP progression was determined with point-wise linear regression analysis of SAP sensitivity values. Typical scan score (TSS) values were extracted as a measure of retardance image quality; an atypical retardation pattern (ARP) was defined as TSS < 80. TSS fluctuation over time was measured using three parameters: change in TSS from baseline, absolute difference (maximum minus minimum TSS value), and TSS variance. Linear mixed-effects models that accommodated the association between the two eyes were constructed to evaluate the relationship between change in TSS and RNFL thickness over time.

RESULTS

Eighty-six eyes (51 suspected glaucoma, 35 glaucomatous) of 45 patients were enrolled. Twenty (23.3%) eyes demonstrated SAP progression. There was significantly greater fluctuation in TSS over time with GDxVCC compared with GDxECC as measured by absolute difference (18.40 ± 15.35 units vs. 2.50 ± 4.69 units; P < 0.001), TSS variance (59.63 ± 87.27 units vs. 3.82 ± 9.63 units, P < 0.001), and change in TSS from baseline (-0.83 ± 11.2 vs. 0.25 ± 2.9, P = 0.01). The change in TSS over time significantly (P = 0.006) influenced the TSNIT average RNFL thickness when measured by GDxVCC but not by GDxECC.

CONCLUSIONS

Longitudinal images obtained with GDxECC have significantly less variability in TSS and retardance patterns and have reduced bias produced by ARP on RNFL progression assessment.

Comparing rates of retinal nerve fibre layer loss with GDxECC using different methods of visual-field progression

BACKGROUND

This prospective analysis was designed to examine the rate of RNFL loss using scanning laser polarimetry (GDx enhanced corneal compensation (GDxECC)) in progressing versus non-progressing eyes using various methods to define functional progression.

METHODS

Glaucoma suspect and glaucomatous eyes with ≥3 years of follow-up participating in the Advanced Imaging for Glaucoma Study were enrolled. All eyes underwent standard automated perimetry (SAP) and GDxECC imaging every 6 months. The annual rate of RNFL loss with GDxECC was calculated using linear regression analysis. Functional progression was determined using the Early Manifest Glaucoma Trial (EMGT) criterion, SAP Visual Field Index (VFI) and Progressor software.

RESULTS

Fifty-three eyes (30 glaucoma suspect, 23 glaucoma) of 53 patients (mean age 64.5±10.7 years, range 42-79) were enrolled. Eighteen eyes (40%) demonstrated SAP progression during the follow-up period using the Progressor criterion, 10 eyes (18.9%) using the VFI criterion, and 3 eyes (5.7%) using the EMGT criterion. The annual rate (μm/year) of mean RNFL loss was significantly greater (p<0.05) in progressing versus non-progressing eyes using Progressor (-1.24±0.99 vs -0.18±0.49), EMGT (-1.95±0.99 vs -0.46±0.78) and VFI (-1.11±0.64 vs -0.41±0.85) criteria.

CONCLUSION

Despite differences in the criteria used to judge functional progression, progressing eyes have a significantly greater rate of RNFL loss measured using GDxECC as compared with non-progressing eyes.

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.

Assessing visual field clustering schemes using machine learning classifiers in standard perimetry

PURPOSE

To compare machine learning classifiers trained on three clustering schemes to determine whether distinguishing healthy eyes from those with glaucomatous optic neuropathy (GON) can be optimized by training with clustered data.

METHODS

Two machine learning classifiers-quadratic discriminant analysis (QDA) and support vector machines with Gaussian kernel (SVMg)-were trained separately using standard perimetry data from the Diagnostic Innovations in Glaucoma Study (DIGS), clustered using three clustering schemes on a training data set (123 eyes/123 glaucoma patients with GON; 135 eyes/135 normal control subjects). Trained classifiers were then applied to an independent data set containing 69 eyes of 69 glaucoma patients with early visual field loss and 83 eyes of 83 normal control subjects. Two control conditions were included: unclustered data and a random assignment of locations to clusters.

RESULTS

Areas under the receiver operating characteristic (ROC) curve ranged from 0.85 (SVMg, thresholds clustered by Glaucoma Hemifield Test sectors) to 0.92 (QDA, thresholds clustered by Garway-Heath mapping) for the training data set. Use of clustered data showed no significant optimization of sensitivity over use of unclustered data, and no single clustering method resulted in significantly higher performance in the independent data set. Sensitivities tended to be higher with QDA than with SVMg, regardless of specificity cutoff and clustering

METHOD

CONCLUSIONS

QDA performed better with the early glaucoma data set than did the SVMg. Clustering may be advantageous when data-dimension reduction is needed-for example, when combining field results with other high-dimensional data (e.g., structural imaging data)-but it is not necessary for visual field data alone.

Trained artificial neural network for glaucoma diagnosis using visual field data: a comparison with conventional algorithms

PURPOSE

To evaluate and confirm the performance of an artificial neural network (ANN) trained to recognize glaucomatous visual field defects, and compare its diagnostic accuracy with that of other algorithms proposed for the detection of visual field loss.

METHODS

SITA Standard 30-2 visual fields, from 100 glaucoma patients and 116 healthy participants, formed the data set. Our ANN was a previously described fully trained network using scored pattern deviation probability maps as input data. Its diagnostic accuracy was compared to that of the Glaucoma Hemifield Test, the Pattern Standard Deviation index at the P<5% and <1%, and also to a technique based on the recognizing clusters of significantly depressed test points.

RESULTS

The included tests had early to moderate visual field loss (median MD=-6.16 dB). ANN achieved a sensitivity of 93% at a specificity level of 94% with an area under the receiver operating characteristic curve of 0.984. Glaucoma Hemifield Test attained a sensitivity of 92% at 91% specificity. Pattern Standard Deviation, with a cut off level at P<5% had a sensitivity of 89% with a specificity of 93%, whereas at P<1% the sensitivity and specificity was 72% and 97%, respectively. The cluster algorithm yielded a sensitivity of 95% and a specificity of 82%.

CONCLUSIONS

The high diagnostic performance of our ANN based on refined input visual field data was confirmed in this independent sample. Its diagnostic accuracy was slightly to considerably better than that of the compared algorithms. The results indicate the large potential for ANN as an important clinical glaucoma diagnostic tool.

Study of patients with ocular hypertension with scanning laser polarimetry and short-wavelength automatic perimetry

AIMS

To compare and correlate retinal nerve fiber layer (RNFL) measurements obtained by scanning laser polarimetry (SLP) with defects detected by short-wavelength automatic perimetry (SWAP) in eyes with ocular hypertension (OHT).

METHODS

SLP and SWAP were performed in 96 eyes of 48 consecutive patients with OHT.

RESULTS

Twenty-five eyes (26%) had SWAP visual field defects. Twenty-seven eyes (28.1%) had abnormal RNFL evaluation defined by the GDx neural network ('number' > 29). Fourteen eyes of 10 patients (14.5%) had abnormal RNFL evaluation and SWAP visual field defects. RNFL thickness measurements were significantly reduced in eyes with abnormal SWAP. A weak but statistically significant correlation between the 'number' and pattern standard deviation (r = 0.3, p = 0.006) and the corrected pattern standard deviation (r = 0.3, p = 0.007) in SWAP was found. Areas of abnormal RNFL thickness corresponded to the localization of the SWAP visual field defects in corrected pattern deviation plots in 10 of the 14 eyes with defects in both tests.

CONCLUSIONS

SWAP visual field defects frequently coexist and correspond with abnormalities of RNFL detected by SLP in eyes with OHT. In certain eyes, however, the two methods detect different glaucoma properties.

A comparison of HRT II and GDx imaging for glaucoma detection in a primary care eye clinic setting

PURPOSE

To evaluate the performance of the HRT II (Heidelberg retinal tomograph) and GDx (glaucoma detection) retinal nerve fibre analyzer in GDx when used in the primary care eye clinic setting for glaucoma screening.

PATIENTS AND METHODS

The study was prospective, cross-sectional, and hospital-based. One-hundred and twelve patients, 59 women and 53 men with a mean age of 57.8 years (range 18-85 years), had consecutive HRT II disc imaging and GDx retinal nerve fiber layer analysis. The Moorfield's regression classification and the 'GDx number' were used to predict the likelihood of glaucoma. A separate clinician, masked to the instrument results determined a definitive diagnosis, based on clinical examination. The extent of agreement between instrument prediction and the clinician diagnosis of glaucoma was examined by generating sensitivity and specificity tables.

RESULTS

The HRT II had a sensitivity of 0.79 (95% CI: 0.60-0.92) and a specificity of 0.70 (95% CI: 0.60-0.78). The positive predictive value of the HRT II was 0.43 (95% CI: 0.29-0.57). Using a GDx number of 50 as 'cutoff' for glaucoma detection, the GDx had a sensitivity of 0.80 (95% CI: 0.59-0.93) and a specificity of 0.72 (95% CI: 0.61-0.80), with a positive predictive value of 0.43 (95% CI: 0.28-0.59).

CONCLUSIONS

For glaucoma detection, neither the HRT II nor the GDx are effective as stand-alone screening devices in the primary care setting.

Influence of optic disc size on parameters of retinal nerve fiber analysis with laser scanning polarimetry

PURPOSE

The aim of the study was to evaluate the influence of optic disc size on the variables of laser scanning polarimetry (GDx).

PATIENTS AND METHODS

One hundred and nineteen healthy controls and 161 patients with ocular hypertension (OHT) received detailed ophthalmologic investigation with respect to glaucoma including retinal nerve fiber analysis with GDx (Version 3.0.05x1; Laser Diagnostic Technologies Europe). Optic disc size was measured with planimetry using 15 degrees optic disc photographs. With respect to frequency of optic disc size in the normal population patients were divided in quartiles of equal sample size.

RESULTS

The ratio between retinal nerve fiber layer thickness in the superior and inferior areas in relation to the nasal and temporal regions decreases significantly with increasing optic disc size and the difference between the highest and lowest retinal nerve fiber layer thickness decreases significantly with increasing optic disc size. The results of multivariate neural network analysis increased with larger optic disc size in controls as well as in patients with OHT. Linear regression analysis showed an increase of 9 units (the Number) per 1 mm(2) of optic disc size. A Number above 30, which indicates suspected glaucoma, was detected in more than a third of the normal population investigated if the optic disc area was larger than 3.5 mm(2). Overall, patients with OHT had a higher Number than controls (20.5+/-11.5 vs. 18.1+/-10.4; p>0.05), but the difference between the two groups did not reach a significant level.

CONCLUSIONS

Retinal nerve fiber analysis in patients with an optic disc size larger than 3.5 mm(2) should be interpreted carefully; the Number in particular requires corrections for optic disc size.

Discrimination between normal and early glaucomatous eyes with scanning laser polarimeter with fixed and variable corneal compensator settings

PURPOSE

To evaluate the ability of scanning laser polarimetry (SLP) with a fixed corneal polarization compensator (GDx-FCC Nerve Fiber Analyzer) compared to one with a variable one (GDx-VCC) in the discrimination between healthy and early glaucomatous eyes.

METHODS

Forty patients with early glaucomatous visual field defects, having a mean deviation of 3.1-/+1.6 dB and a pattern standard deviation of 3.1-/+0.9 dB, and 40 controls underwent both GDx-FCC and GDx-VCC. One eye per patient was considered. The cut-off point, taken as the value dividing healthy from glaucomatous eyes with highest probability, was determined for each GDx parameter. Linear discriminant functions (LDFs) were separately developed for GDx-FCC and GDx-VCC parameters. Sensitivity, specificity, and area under the receiver operating characteristic curve (AROC) for discriminating between healthy and glaucomatous eyes were calculated for each GDx parameter, both according to the GDx normative database and after the selection of new cut-off points, and for the LDFs.

RESULTS

All software-provided parameters showed low sensitivity and high specificity. The selection of new cut-off points improved the performance of all GDx parameters: VCC parameters performed better than FCC parameters; the largest AROCs were associated with the superior/nasal ratio for the GDx-FCC (0.86) and with the Number for the GDx-VCC (0.87). The LDFs provided an AROC of 0.89 with both the GDx-FCC and the GDx-VCC parameters.

CONCLUSIONS

The GDx-VCC showed a higher ability in the early diagnosis of glaucoma when compared with the GDx-FCC. The individuation of the right cut-off point of selected parameters with both GDx settings performed better than the software-provided parameters, and comparably to the GDx parameters-based LDFs.

Diurnal variation of intraocular pressure and its correlation with retinal nerve fiber analysis in Turkish patients with exfoliation syndrome

PURPOSE

The purpose was to evaluate the diurnal variation (DV) of intraocular pressure (IOP) in patients with exfoliation syndrome (XS), to measure retinal nerve fiber layer (RNFL) thickness by using scanning laser polarimetry, and to compare these measurements with those of normal subjects.

METHODS

Forty-five subjects with XS and 40 healthy, age/sex matched subjects were recruited into the study. A detailed ophthalmologic examination was performed. IOP measurements were obtained at 08:00 am, 12:00 pm, 03:00 pm, and 06:00 pm. The XS group was further divided into DV > or = 5 mmHg and DV < 5 mmHg groups and also according to the existence of IOP fluctuation. The IOP measurements and RNFL thickness measurements were compared between the groups.

RESULTS

The mean IOP value was found to be highest in the morning both in the XS and control groups. IOP showed a gradual decrease from 8.00 am to 6.00 pm in the control group, whereas a second peak at 03:00 pm was observed in the XS group. There was a fluctuation in 53.3% of the XS group, while none of the healthy subjects showed fluctuation. Superior and inferior ratios were statistically lower in XS patients than those in control subjects (p<0.05). Moreover, in patients with XS showing a DV > or = 5 mmHg and/or a fluctuation, the superior ratio, inferior ratio, the number, superior average and superior integral were significantly different (all p values <0.05) from those of control subjects.

CONCLUSIONS

As the XS patients with high diurnal IOP variation and fluctuating pattern of IOP had lower RNFL thickness measurements, it is crucial to follow up these patients by performing scanning laser polarimetry in order to discover any possible glaucomatous damage at an earlier stage than with the use of conventional visual field analysis.

Automated detection of wedge-shaped defects in polarimetric images of the retinal nerve fibre layer

PURPOSE

Automated glaucoma detection in images obtained by scanning laser polarimetry is currently insensitive to local abnormalities, impairing its performance. The purpose of this investigation was to test and validate a recently proposed algorithm for detecting wedge-shaped defects.

METHODS

In all, 31 eyes of healthy subjects and 37 eyes of glaucoma patients were imaged with a GDx. Each image was classified by two experts in one of four classes, depending on how clear any wedge could be identified. The detection algorithm itself aimed at detecting and combining the edges of the wedge. The performance of both the experts and the algorithm were evaluated.

RESULTS

The interobserver correlation, expressed as ICC(3,1), was 0.77. For the clearest cases, the algorithm yielded a sensitivity of 80% at a specificity of 93%, with an area under the ROC of 0.95. Including less obvious cases by the experts resulted in a sensitivity of 55% at a specificity of 95%, with an area under the ROC of 0.89.

CONCLUSIONS

It is possible to automatically detect many wedge-shaped defects at a fairly low rate of false-positives. Any detected wedge defect is presented in a user-friendly way, which may assist the clinician in making a diagnosis.

Corneal birefringence changes after laser assisted in situ keratomileusis and their influence on retinal nerve fibre layer thickness measurement by means of scanning laser polarimetry

AIM

To evaluate changes in corneal polarisation properties and their influence on peripapillary retinal nerve fibre layer (RNFL) thickness measurements after laser assisted in situ keratomileusis (LASIK) by means of scanning laser polarimetry (SLP) with variable corneal polarisation compensator (VCC) in normal white subjects.

METHODS

SLP was performed by means of GDx VCC on 32 eyes of 32 normal subjects who underwent LASIK for ametropia correction. Corneal polarisation axis and magnitude and RNFL thickness were measured before and 8 days after LASIK. RNFL thickness data and corneal polarimetric data of one randomly selected eye per subject were analysed by the Wilcoxon signed ranks test. Correlations between corneal ablation depth, corneal polarimetric changes, and RNFL thickness changes were investigated using Spearman's rho test.

RESULTS

The corneal polarisation axis significantly shifted from 15.1 degrees (17.0 degrees ) to 6.9 degrees (12.9 degrees ) (p = 0.00006) after LASIK and this change showed a strong correlation with corneal ablation depth (rho = -0.7, p = 0.00002). Among GDx parameters, TSNIT, SUP, and SD showed significant changes after LASIK and for SUP and SD these changes were well correlated with the shift in corneal polarisation axis (rho = 0.54, p = 0.03 and rho = 0.45, p = 0.01, respectively). SUP and SD changes were neutralised after compensating for corneal polarimetric changes but not TSNIT changes. NFI, a discriminating parameter, was found to be affected after LASIK only after compensating for corneal polarimetric changes.

CONCLUSIONS

LASIK induces a shift in corneal polarisation axis which is responsible for inaccuracies in RNFL thickness measurements. A customised compensation for corneal polarimetric changes after LASIK allows normalisation of some of the thickness parameters except for TSNIT and NFI.

Scanning laser polarimetry of nerve fiber layer thickness in normal eyes after cataract phacoemulsification and foldable intraocular lens implantation

PURPOSE

To assess the effect of cataract phacoemulsification and intraocular lens (IOL) implantation on retinal nerve fiber layer (RNFL) thickness using scanning laser polarimetry (SLP).

SETTING

Eye Clinic, Trieste University, Italy.

METHODS

Forty-eight eyes were evaluated prospectively the day before and 30 days after cataract phacoemulsification and foldable IOL implantation. In each eye, lens opacity grading according to the Lens Opacities Classification System III (LOCS III), and axial length (AL) measurements were performed. Retinal nerve fiber layer thickness was quantified at baseline by means of SLP and anterior segment birefringence compensation was evaluated acquiring macular retardation map (MRM). Acrylic and silicone IOLs were implanted randomly. After surgery, RNFL thickness was reevaluated, and MRM was reacquired. Macular retardation map pattern variations regarding baseline profile were classified into 3 groups: no variation, bow-tie profile enhancement, or attenuation. Distribution of IOL power, AL, and cataract type in the 3 groups was assessed, as were presurgery and postsurgery SLP parameters with mean values (+/-SD) compared by paired t test.

RESULTS

Twenty-two eyes (Group 1, 45.8%) showed no MRM variation, 14 (Group 2, 29.2%) an enhancement, and 12 (Group 3, 25%) an attenuation. In Group 1, no significant RNFL thickness variation occurred. In Group 2, variation 10% to 15% was measured, whereas thickening a 8% to 15% thinning appeared in Group 3. Variations occurred irrespective of IOL material, AL, or cataract type.

CONCLUSIONS

Cataract surgery with IOL implantation was associated with an MRM profile change and RNFL thickness variations in 54.2% of eyes. Variations are probably related to opacified lens removal. A new baseline SLP reading is mandatory after cataract surgery.

The Groningen Longitudinal Glaucoma Study. I. Baseline sensitivity and specificity of the frequency doubling perimeter and the GDx nerve fibre analyser

PURPOSE

To describe the baseline data of a large cohort of patients included for follow-up with perimetry using the frequency doubling technique (FDT) and with quantification of the retinal nerve fibre layer as assessed by GDx, and to calculate the sensitivity and specificity of both devices from these baseline data.

METHODS

Regular visitors to our glaucoma service were included. All subjects were followed for at least 4 years with FDT in full-threshold mode, GDx and conventional perimetry. Patients were classified as having either glaucoma or suspect glaucoma, according to baseline perimetry results. In addition, a group of healthy subjects was recruited outside the hospital.

RESULTS

A total of 452 glaucoma patients, 423 glaucoma suspects and 237 healthy subjects were incorporated into the analyses. Sensitivities for both FDT and GDx were fixed at 0.90. For the group as a whole, the specificity was 0.81 for FDT, using number of depressed test-points p < 0.01 in the total deviation probability plot with a cut-off point > 1, and 0.78 for GDx, using the Number, with a cut-off point > 29. The area under the receiver operating characteristic (ROC) curve was 0.92 for FDT and 0.94 for GDx. Of the subjects with suspect glaucoma, 75% showed normal FDT test results and 52% showed normal GDx results. Unlike FDT, GDx failed to detect some moderate/severe glaucoma cases.

CONCLUSIONS

The performances of FDT and GDx are approximately equivalent in terms of sensitivity, specificity and area under the ROC curve. In glaucoma suspects, GDx in particular yielded a rather high percentage of positive test results. The majority of these positive test results are presumably false-positive results rather than results indicating preperimetric glaucoma.

Scanning laser polarimetry of retinal nerve fibre layer thickness after laser assisted in situ keratomileusis (LASIK): stability of the values after the third post-LASIK month

BACKGROUND

Monitoring the retinal nerve fibre layer thickness (RNFLT) is essential in the diagnosis and treatment of glaucoma. In a previous study we found that a decrease of the polarimetric RNFLT observed in the early period after laser-assisted in situ keratomileusis (LASIK) disappears or tends to disappear by the third post-LASIK month.

PURPOSE

To study the stability of the "recovered" polarimetric retardation values between the third and twelfth month after LASIK.

METHODS

Scanning laser polarimetry (SLP) with the classic GDx Nerve Fiber Analyzer was performed on 13 consecutive healthy subjects with no eye disease who underwent LASIK for ametropia correction. Measurements were performed preoperatively, then at 3 and 12 months postoperatively.

RESULTS

Inferior, temporal and nasal average thickness as well as ellipse average thickness and average thickness showed no difference among the three time points (ANOVA, p > 0.05). Superior average thickness was significantly smaller both at three months (Sheffe test, p =0.008) and 12 months (p =0.006) than before LASIK. However, no difference was seen between the values measured at three months and at 12 months after LASIK (p =0.997). A statistically significant interaction between treatment type (myopic or hyperopic correction) and the change of retardation was found for the superior average thickness (two-way ANOVA, p =0.016). In this quadrant the RNFLT values of the myopic eyes decreased between the baseline and the month 3 measurements but became stable after that; the retardation of the hyperopic eyes remained unchanged throughout.

CONCLUSION

RNFLT measured with the classic GDx device after LASIK shows transient changes probably due to the LASIK-induced alteration of the polarization and the healing process. The polarimetric RNFLT values, however, become stable by the third post-LASIK month, and show no further change until the end of the first year after LASIK. Baseline SLP measurements for long-term glaucoma follow-up can be obtained from the third post-LASIK month onwards.

Retinal nerve fiber layer thickness measurements with scanning laser polarimetry predict glaucomatous visual field loss

PURPOSE

To assess whether baseline retinal nerve fiber layer (RNFL) measurements obtained with a scanning laser polarimeter, the GDx Nerve Fiber Analyzer, (Laser Diagnostic Technologies Inc., San Diego, California) are predictive of development of repeatable glaucomatous visual field damage in glaucoma suspect eyes.

DESIGN

Cohort study.

METHODS

Participants were recruited from the UCSD longitudinal Diagnostic Innovations in Glaucoma Study (DIGS). One eye from each of 160 glaucoma suspects with normal standard automated perimetry (SAP) visual fields at baseline was studied. Study eyes were divided into convert and nonconvert groups based on the development of three consecutive glaucomatous visual fields during follow-up. SLP parameters, IOP, vertical cup disk ratio, stereophotograph assessment as glaucoma or normal, corneal thickness, and visual field indices were included in univariate and multivariate Cox proportional hazards models to determine which SLP RNFL and ocular parameters were predictive of visual field conversion.

RESULTS

Sixteen (10%) eyes developed repeatable visual field damage (converts) and 144 (90%) did not (nonconverts). Mean (95%CI) follow-up time until visual field conversion for convert eyes was 2.7 (1.7, 3.6) years. Mean total follow-up of nonconvert eyes was 3.8 (3.5, 4.1) years. Four out of thirteen examined baseline SLP parameters and baseline SAP Mean Deviation (MD), SAP Pattern Standard Deviation (PSD), and glaucomatous stereophotograph assessment were significant univariate predictors of visual field conversion. In multivariate models adjusted for age, IOP and CCT, SLP parameters inferior ratio, ellipse modulation, and UCSD linear discriminant function (LDF) were significant predictors of visual field conversion. When SAP PSD and stereophotograph assessment were also included in the multivariate model inferior ratio and UCSD LDF remained independently predictive of visual field loss.

CONCLUSIONS

Thinner baseline SLP RNFL measurements were independent predictors of visual field damage. In addition to thinner SLP RNFL measurements, higher baseline SAP PSD, and baseline glaucomatous stereophotograph assessment each contributed to an increased risk of the development of abnormal visual fields in glaucoma suspect patients. SLP RNFL measurements were independently predictive of future visual loss even when age, IOP, CCT, vertical cup disk ratio, and SAP PSD were included in the model.

Imaging in glaucoma

Structural assessment using the imaging technologies discussed herein provides reproducible quantitative measurements of posterior segment ocular structures. These measurements have been found to provide useful data for glaucoma detection in various regions of the posterior segment. Further studies are needed to evaluate the utility of these technologies for pre-perimetric glaucoma detection and for monitoring glaucoma progression over an extended period.

Influence of subfoveal choroidal neovascularisation on macular imaging with scanning laser polarimetry of the retinal nerve fibre layer

PURPOSE

To investigate the influence of subfoveal choroidal neovascularisation (CNV) on macular imaging performed using scanning laser polarimetry (SLP) of the retinal nerve fibre layer.

METHODS

SLP was performed on 22 consecutive patients with angiographically verified CNV, and on 23 healthy control subjects. One eye per subject was evaluated using the GDx Nerve Fibre Analyser. Regularity of the corneal retardation on the macular SLP images was assessed according to three criteria: (1) magnitude of the 'macular ratio', defined as the ratio of mean retardation values along two axes (the axis with the maximum retardation and the perpendicular one, corresponding in healthy eyes to minimum retardation); (2) the values of GDx parameters which are independent of quadrant position (ellipse modulation and ellipse average); and (3) the frequency of the regular 'bow-tie' polarisation pattern.

RESULTS

'Macular ratio' was significantly higher in the CNV group than in the control group (P<<0.001). Ellipse modulation did not differ between the groups, but ellipse average was higher in the CNV group (P=0.016). The variance for each of these two parameters was significantly higher for the CNV group (P<<0.001 for both comparisons). A 'bow-tie' pattern polarisation was seen in 23 of the 23 control eyes, but only in 7 of the 22 CNV eyes (P<<0.001).

CONCLUSION

The results show that CNV influences the macular image obtained with SLP. This suggests that measurements with SLP may be disturbed for eyes with CNV when the customised corneal compensation method, which makes use of the macular retardation image, is employed.

Comparison between fixed-angle and customised corneal-polarisation compensation methods in scanning laser polarimetric measurement of the retinal nerve fibre layer in glaucoma

PURPOSE

To investigate the differences between the results of scanning laser polarimetric (SLP) measurements of the retinal nerve fibre layer thickness (RNFLT) made using two different corneal-polarisation techniques; customised (SLP-C), and fixed-angle (SLP-F) compensations.

METHODS

Both SLP-C and SLP-F were performed on 37 consecutive phakic patients with chronic open-angle glaucoma, and on 14 healthy control subjects. One randomly selected eye per subject was evaluated.

RESULTS

Both SLP-C and SLP-F parameters were able to discriminate between the glaucoma group and the control group, except in the case of the ellipse modulation, which differed significantly between the two groups with SLP-C (P=0.017), but not with SLP-F (P=0.056). When SLP-C and SLP-F values were compared, inferior maximum thickness and ellipse standard deviation were significantly lower with SLP-C in both groups (P<0.05 for each parameter). Superior maximum thickness was significantly lower in glaucoma with SLP-C than with SLP-F (P=0.006) and tended to be lower with SLP-C than with SLP-F in the control group (P=0.053). In the glaucoma group, it was only with SLP-C that a significant (positive) correlation between the superior maximum thickness and the inferior hemifield mean sensitivity (MS) (r=0.653, P<0.001), and between the inferior maximum thickness and the superior hemifield MS (r=0.420, P=0.023) was found. The other global and sectoral SLP parameters showed significant correlation with the corresponding visual field parameters with both techniques.

CONCLUSION

Our findings suggest that SLP measurements with customised compensation provide more realistic results for RNFLT than those made with the conventional fixed-angle compensation.

Influence of LASIK on scanning laser polarimetric measurement of the retinal nerve fibre layer with fixed angle and customised corneal polarisation compensation

BACKGROUND/AIM

Retinal nerve fibre layer thickness (RNFLT), as measured with scanning laser polarimetry using the fixed angle corneal polarisation compensator (SLP-F), has been found to be reduced after uncomplicated laser assisted in situ keratomileusis (LASIK) compared to the pre-LASIK measurement. Since this virtual RNFLT thinning is attributed to the corneal changes induced by the LASIK, the authors investigated whether customised corneal polarisation compensation (SLP-C), which compensates for the actual corneal polarisation during each measurement, can avoid the LASIK induced, virtual changes of the polarimetric RNFLT values.

METHODS

Scanning laser polarimetry using both the SLP-F and SLP-C methods (GDx-Access, software version 5.0) was performed on 15 consecutive healthy subjects with no eye disease who underwent LASIK for ametropia correction. The SLP measurements were performed before the surgery, then on day 1 and day 6 after LASIK. Thickness data from images of one randomly selected eye per subject were analysed using the ANOVA and Scheffe multiple comparison tests.

RESULTS

Superior maximum, inferior maximum, normalised superior area, and normalised inferior area (SLP parameters representing the RNFLT at the superior and inferior poles of the optic nerve head) remained unchanged with SLP-C (ANOVA, p>0.05) but decreased (superior maximum, normalised superior area, Scheffe test, p<0.05) or tended to decrease (inferior maximum) after LASIK, when measured using SLP-F. In contrast, certain other parameters-namely, superior ratio and inferior ratio (representing the ratios between the superior or the inferior sector and the temporal sector), maximal modulation, and ellipse modulation decreased with SLP-C (Scheffe test, p<0.05), but remained stable with SLP-F (ANOVA, p>0.05) after LASIK. Superior to nasal ratio, symmetry of the superior and inferior RNFLT as well as the parameter showing the probability of having glaucoma (called "the number") remained unchanged with both types of corneal compensation (ANOVA, p>0.05). With SLP-C the parameter ellipse average thickness increased after LASIK (Scheffe test, p = 0.021). No parameter value altered between day 1 and day 6 after LASIK, for either method.

CONCLUSION

The results suggest that the LASIK induced decrease of the polarimetric RNFLT, which is consistently detected with polarimeters when using the fixed angle corneal polarisation compensator, is due to alterations of the corneal polarisation. The use of customised corneal polarisation compensation avoids this virtual decrease of the polarimetric RNFLTHowever, our results suggest an increase of the measured retardation in the temporal quadrant of the SLP-C image after LASIK. Since ratios of parameters using the temporal RNFLT in the denominator are important in the polarimetric glaucoma diagnosis algorithm, their decrease as a consequence of using SLP-C needs further investigation.

Sensitivity and specificity of the GDx: clinical judgment of standard printouts versus the number

PURPOSE

The Number is a standard parameter of the GDx that reportedly distinguishes normal and glaucomatous eyes. The authors evaluated the sensitivity and specificity of the Number and examined whether expert clinical judgment of GDx printouts leads to a better separation.

MATERIALS AND METHODS

Two experienced observers judged 800 GDx scans on 400 randomly presented printouts from 200 glaucoma patients and 200 age-matched normal subjects. The diagnosis was based on the symmetry analysis printout and was per patient rather than per eye. The observers assessed sensitivity for all glaucoma patients together, and separately for mild, moderate, and severe glaucoma. Their specificity was determined in the group of normal subjects. The same procedure was performed for the Number, at various critical values.

RESULTS

Both observers discriminated better than the Number. At a critical value of 23, the specificity of the Number was 81.5%, which matched the lowest specificity of the 2 observers: 82.5% and 92.0% for observers 1 and 2, respectively. At these specificities, the sensitivity of the 2 observers and of the Number were 92.0%, 89.5%, and 85.5%, respectively. The sensitivity increased with the severity of glaucoma. The Kappa values for intraobserver agreement were 0.80 and 1.0.

CONCLUSIONS

The Number yielded acceptable sensitivity and specificity values at a critical value of 23 in this test population. However, the clinical judgments of the printouts by both expert observers resulted in a better separation between normal and glaucomatous eyes, particularly in the group with mild glaucoma.

Evaluation of the glaucomatous damage on retinal nerve fiber layer thickness measured by optical coherence tomography

PURPOSE

To evaluate the relationship between visual field and retinal nerve fiber layer (RNFL) thickness measured by optical coherent tomography (OCT) and to assess the diagnostic ability of OCT to distinguish between early glaucomatous or glaucoma-suspect eyes from normal eyes.

DESIGN

Retrospective, non-randomized, cross-sectional study.

METHODS

A total of 160 eyes of 120 normal Japanese adults, 23 eyes of 16 patients with ocular hypertension, 38 eyes of 35 glaucoma-suspect patients, and 237 glaucomatous eyes of 140 glaucoma patients were enrolled in the study. The glaucoma group included 89 early glaucomatous eyes. Thickness of the RNFL around the optic disk was determined with three 3.4-mm diameter circle OCT scans. Average and segmental RNFL thickness values were compared among all groups. The correlation between mean deviation and RNFL thickness in glaucomatous eyes was also analyzed. Receiver operating characteristic (ROC) curve area was calculated to discriminate normal eyes from early glaucomatous or glaucoma-suspect eyes.

RESULTS

A significant relationship existed between the mean deviation and RNFL thickness in all parameters excluding the 3-o'clock area. The average RNFL thickness had the strongest correlation in all parameters (r = -0.729, P <.001). Retinal nerve fiber layer thickness at the 7-o'clock inferotemporal segment had the widest areas under the ROC curves in all parameters for early glaucomatous eyes (0.873).

CONCLUSIONS

Measurement of RNFL thickness by OCT is useful in detecting early RNFL damage. Furthermore, OCT measurements of RNFL thickness may provide clinically relevant information in monitoring glaucomatous changes.

GDx in glaucoma

The purpose of this paper is to review the clinical applications of GDx in the diagnosis and follow-up of glaucoma. The limits and potential of GDx technology are discussed.

Change in retinal nerve fiber layer thickness after laser in situ keratomileusis

PURPOSE

To determine the effect of laser in situ keratomileusis (LASIK) on retinal nerve fiber layer (RNFL) dropout.

SETTING

Nevyas Eye Associates, Philadelphia, Pennsylvania, USA.

METHODS

This prospective consecutive study comprised 120 eyes. The GDx((R)) nerve fiber analyzer (Laser Diagnostic Technologies, Inc.) was used to measure RNFL thickness, an early indicator of glaucomatous damage, preoperatively and 1 week to 4 months and 6 months postoperatively to determine whether any change in apparent RNFL thickness was due to corneal change or to actual RNFL damage and to determine whether apoptotic cascade caused ongoing RNFL dropout. Humphrey visual fields were done in eyes with an abnormal GDx (increase of 20 in GDx number).

RESULTS

The GDx was normal in 89 eyes, including 6 eyes with tilted discs, an abnormal number, and a normal image and 15 eyes with a nonprogressive increase in the GDx number. Three eyes, normal at the interim, were abnormal at 6 months but had a normal visual field. Eleven eyes had a significant increase in the number at 6 months, including 1 eye that showed a possible glaucomatous visual field defect at 9 months and a normal visual field at 14 months.

CONCLUSIONS

There was no evidence of RNFL thinning or glaucomatous damage. A new baseline GDx should be established post LASIK to correct for corneal change.

Influence of post-LASIK corneal healing on scanning laser polarimetric measurement of the retinal nerve fibre layer thickness

AIM

To investigate the influence of laser assisted in situ keratomileusis (LASIK) on the values for retinal nerve fibre layer thickness (RNFLT) as measured with scanning laser polarimetry (SLP) during the healing process of the cornea after LASIK.

METHODS

SLP with the GDx instrument was performed on 20 consecutive healthy subjects without any eye disease undergoing LASIK for ametropia correction. The SLP measurements were performed before the surgery, and at 1 and 3 days, as well as at 3 months, after LASIK. Thickness data from images of one randomly selected eye per subject were analysed using the ANOVA and Duncan multiple comparison tests. Correlation coefficients between RNFLT data and the treatment parameters were also calculated.

RESULTS

Somewhat similar results were found for the different retinal areas. The measured values for superior average RNFLT decreased significantly at all time points compared to the preoperative baseline (p<0.003, Duncan test), but increased significantly between postoperative day 1 and the final visit at 3 months (p=0.025, Duncan test). Inferior average RNFLT in the early postoperative days was significantly smaller than at 3 months after LASIK (p<0.05, Duncan test), and tended to be smaller than at baseline. The thickness values before surgery and at the final visit, however, showed no significant difference (p=0.698, Duncan test) in this region. Ellipse average RNFLT was significantly smaller in the early postoperative days than the baseline value before LASIK. However, the measured value had significantly increased again by the time of the final visit (p<0.02, Duncan test). This value at the final visit showed no difference from the baseline value (p=0.46, Duncan test). The changes in the nasal average and temporal average RNFLT were not statistically significant. No correlation was found between the change in the SLP measured thickness values and central corneal thickness at baseline and its change after surgery, nor with the change in cylindrical correction due to LASIK, or the length of the suction time during surgery (p>0.05 for all correlations).

CONCLUSION

The SLP technique is sensitive to the corneal optical properties, and RNFLT as measured with SLP shows changes after LASIK. Most of these changes, however, diminish with time after surgery, and the values tend to return to the preoperative results during the first 3 months of corneal healing following uncomplicated LASIK. It appears that in uncomplicated cases the transient RNFLT changes are artefacts and do not imply pathological thickness alterations due to LASIK.