[Show abstract][Hide abstract] ABSTRACT: PurposeExpertise in viewing medical images is thought to be due to the ability to process holistic image information. Eye care clinicians can inspect photographs of the retina to search for signs of disease. However, they commonly also view the eye in vivo using the restricted view of a slit lamp, which removes the potential benefits of holistic processing. We investigated how expert and novice clinicians inspect the fundus using these two methods.Methods
Twenty clinicians (10 experienced, 10 novices) examined 64 photographs of human retinae. Each participant viewed half of the images as fundus photographs while having their eye position recorded. The other half were viewed via a simple slit lamp simulation, whereby a computer mouse was used to control the position of a viewing window that revealed the underlying fundus photograph.ResultsExperienced clinicians made decisions significantly faster than novices, with faster decision-making when viewing the fundus photograph compared to via the slit lamp simulation. The distribution of inspection was similar, although novices spent longer examining the optic nerve head than other regions. Experienced clinicians showed significantly earlier inspection of the optic nerve head when it was judged to be unhealthy.Conclusions
Our results support the idea that experienced eyecare clinicians use holistic image information, if available, when inspecting the fundus. This was particularly prominent for the optic nerve head region, which was the region that novices spent most of their time examining. Holistic processing benefits were only present in experts’ free-viewing fundus photographs; the limited field of view from the slit lamp disrupts such global image benefits.
[Show abstract][Hide abstract] ABSTRACT: To develop a perimetric test strategy, Structure Estimation of Minimum Uncertainty (SEMU), that uses structural information to drive stimulus choices.
Structure Estimation of Minimum Uncertainty uses retinal nerve fiber layer (RNFL) thickness data as measured by optical coherence tomography to predict perimetric sensitivity. This prediction is used to set suprathreshold levels that then alter a prior probability distribution of the final test output. Using computer simulation, we studied SEMU's performance under three different patient error response conditions: No Error, Typical False Positive errors, and Extremely Unreliable patients. In experiment 1, SEMU was compared with an existing suprathreshold cum thresholding combination test procedure, Estimation of Minimum Uncertainty (EMU), on single visual field locations. We used these results to finalize SEMU parameters. In experiment 2, SEMU was compared with full threshold (FT) on 163 glaucomatous visual fields.
On individual locations, SEMU has similar accuracy to EMU, but is, on average, one presentation faster than EMU. For the typical false-positive error condition, SEMU has significantly lower error compared with FT (SEMU average 0.33 dB lower; p < 0.001) and the 90% measured sensitivity range for SEMU is also smaller than that for FT. For unreliable patients, however, FT has lower mean and SD of error. Structure Estimation of Minimum Uncertainty makes significantly fewer presentations than FT (1.08 presentation on average fewer in a typical false-positive condition; p < 0.001). Assuming that a location in the field is marked abnormal if it falls below the 5th percentile of normal, SEMU has a false-positive rate of less than 10% for all error conditions compared with FT's rate of 20% or more.
On average, simulations show that using RNFL information to guide stimulus placement in a perimetric test procedure maintains accuracy, improves precision, and decreases test duration for patients with less than 15% false-positive rates.
Optometry and vision science: official publication of the American Academy of Optometry 11/2014; 92(1). DOI:10.1097/OPX.0000000000000447 · 1.60 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: This study aimed to determine if response times gathered during perimetry can be exploited within a thresholding algorithm to improve the speed and accuracy of the test. Frequency of Seeing (FoS) curves were measured at 24 locations across the central 30 degrees of the visual field of 10 subjects using a Method of Constant Stimuli, with response times recorded for each presentation. Spatial locations were interleaved, and built up over multiple 5-minute blocks, in order to mimic the attentional conditions of clinical perimetry. FoS curves were fitted to each participant's data for each location, and response times derived as a function of distance-from-threshold normalised to the slope of each FoS curve. This data was then used to derive a function for the probability of observing response times given the distance-from-threshold, and to seed simulations of a new test procedure (BURTO) that exploited the probability function for stimulus placement. Test time and error were then simulated for patients with various false response rates. When compared with a ZEST algorithm, simulations revealed that BURTO was about one presentation per location faster than ZEST, on average, while sacrificing less precision and bias in threshold estimates than simply terminating the ZEST earlier. Despite response times varying considerably for a given individual and their thresholds, response times can be exploited to reduce the number of presentations required in a visual field test without loss of accuracy.
Vision research 05/2014; 101. DOI:10.1016/j.visres.2014.04.013 · 1.82 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Purpose:
We have developed customized maps that relate visual field and optic nerve head (ONH) regions according to individual anatomy. In this study, we aimed to determine feasible map resolution for research use, and to make a principled recommendation of sector size for clinical applications.
Measurement variability in fovea-ONH distance and angle was estimated from 10 repeat OCT scans of 10 healthy people. Errors in estimating axial length from refractive error were determined from published data. Structure-function maps were generated, and customized to varied clinically-plausible anatomical parameters. For each parameter set (n = 210), 200 maps were generated by sampling from measurement/estimation error distributions. Mapped 1° sectors at each visual field location from each parameter set were normalized to difference from their mean. Variation (90% ranges) in normalized mapped sectors represents the precision of individualized maps.
Standard deviations of repeated measures of fovea-ONH distance and angle were 61 μm and 0.97° (coefficients of variation 1.3% and 12.0%, respectively). Neither measure varied systematically with mean (Spearmans's ρ = 0.26, P = 0.47 for distance, ρ = -0.31, P = 0.39 for angle). Variation (90% ranges) in normalized mapped sectors varied across the visual field and ranged from 3° to 18° when axial length was measured accurately, and from 6° to 32° when axial length was estimated from refractive error.
The 90% ranges represent the minimum feasible ONH sector size at each visual field location. For clinical use an easily interpretable scheme of 30° sectors is suggested.
[Show abstract][Hide abstract] ABSTRACT: Purpose:
We evaluated variability and conviction in tracing paths of retinal nerve fiber bundles (RNFBs) in retinal images, and compared traced paths to a computational model that produces anatomically-customized structure-function maps.
Ten retinal images were overlaid with 24-2 visual field locations. Eight clinicians and 6 naïve observers traced RNFBs from each location to the optic nerve head (ONH), recording their best estimate and certain range of insertion. Three clinicians and 2 naïve observers traced RNFBs in 3 images, 3 times, 7 to 19 days apart. The model predicted 10° ONH sectors relating to each location. Variability and repeatability in best estimates, certain range width, and differences between best estimates and model-predictions were evaluated.
Median between-observer variability in best estimates was 27° (interquartile range [IQR] 20°-38°) for clinicians and 33° (IQR 22°-50°) for naïve observers. Median certain range width was 30° (IQR 14°-45°) for clinicians and 75° (IQR 45°-180°) for naïve observers. Median repeatability was 10° (IQR 5°-20°) for clinicians and 15° (IQR 10°-29°) for naïve observers. All measures were worse further from the ONH. Systematic differences between model predictions and best estimates were negligible; median absolute differences were 17° (IQR 9°-30°) for clinicians and 20° (IQR 10°-36°) for naïve observers. Larger departures from the model coincided with greater variability in tracing.
Concordance between the model and RNFB tracing was good, and greatest where tracing variability was lowest. When RNFB tracing is used for structure-function mapping, variability should be considered.
[Show abstract][Hide abstract] ABSTRACT: Citation: Denniss J, McKendrick AM, Turpin A. Towards patient-tailored perimetry: automated perimetry can be improved by seeding procedures with patient-specific structural in-formation. Trans Vis Sci Tech. 2013; 2(4):3, http://tvstjournal.org/doi/full/ 10.1167/tvst.2.4.3, doi:10.1167/tvst. 2.4.3 Purpose: To explore the performance of patient-specific prior information, for example, from structural imaging, in improving perimetric procedures. Methods: Computer simulation was used to determine the error distribution and presentation count for Structure–Zippy Estimation by Sequential Testing (ZEST), a Bayesian procedure with prior distribution centered on a threshold prediction from structure. Structure-ZEST (SZEST) was trialled for single locations with combinations of true and predicted thresholds between 1 to 35 dB, and compared with a standard procedure with variability similar to Swedish Interactive Thresholding Algorithm (SITA) (Full-Threshold, FT). Clinical tests of glaucomatous visual fields (n ¼ 163, median mean deviation À1.8 dB, 90% range þ2.1 to À22.6 dB) were also compared between techniques. Results: For single locations, SZEST typically outperformed FT when structural predictions were within 6 9 dB of true sensitivity, depending on response errors. In damaged locations, mean absolute error was 0.5 to 1.8 dB lower, SD of threshold estimates was 1.2 to 1.5 dB lower, and 2 to 4 (29%–41%) fewer presentations were made for SZEST. Gains were smaller across whole visual fields (SZEST, mean absolute error: 0.5 to 1.2 dB lower, threshold estimate SD: 0.3 to 0.8 dB lower, 1 [17%] fewer presentation). The 90% retest limits of SZEST were median 1 to 3 dB narrower and more consistent (interquartile range 2–8 dB narrower) across the dynamic range than those for FT. Conclusion: Seeding Bayesian perimetric procedures with structural measurements can reduce test variability of perimetry in glaucoma, despite imprecise structural predictions of threshold. Translational Relevance: Structural data can reduce the variability of current perimetric techniques. A strong structure–function relationship is not necessary, however, structure must predict function within 69 dB for gains to be realized.
[Show abstract][Hide abstract] ABSTRACT: To explore the performance of patient-specific prior information, for example, from structural imaging, in improving perimetric procedures.
Computer simulation was used to determine the error distribution and presentation count for Structure-Zippy Estimation by Sequential Testing (ZEST), a Bayesian procedure with prior distribution centered on a threshold prediction from structure. Structure-ZEST (SZEST) was trialled for single locations with combinations of true and predicted thresholds between 1 to 35 dB, and compared with a standard procedure with variability similar to Swedish Interactive Thresholding Algorithm (SITA) (Full-Threshold, FT). Clinical tests of glaucomatous visual fields (n = 163, median mean deviation -1.8 dB, 90% range +2.1 to -22.6 dB) were also compared between techniques.
For single locations, SZEST typically outperformed FT when structural predictions were within ± 9 dB of true sensitivity, depending on response errors. In damaged locations, mean absolute error was 0.5 to 1.8 dB lower, SD of threshold estimates was 1.2 to 1.5 dB lower, and 2 to 4 (29%-41%) fewer presentations were made for SZEST. Gains were smaller across whole visual fields (SZEST, mean absolute error: 0.5 to 1.2 dB lower, threshold estimate SD: 0.3 to 0.8 dB lower, 1 [17%] fewer presentation). The 90% retest limits of SZEST were median 1 to 3 dB narrower and more consistent (interquartile range 2-8 dB narrower) across the dynamic range than those for FT.
Seeding Bayesian perimetric procedures with structural measurements can reduce test variability of perimetry in glaucoma, despite imprecise structural predictions of threshold.
Structural data can reduce the variability of current perimetric techniques. A strong structure-function relationship is not necessary, however, structure must predict function within ±9 dB for gains to be realized.
[Show abstract][Hide abstract] ABSTRACT: Purpose:
To present a computational model mapping visual field (VF) locations to optic nerve head (ONH) sectors accounting for individual ocular anatomy, and to describe the effects of anatomical variability on maps produced.
A previous model that related retinal locations to ONH sectors was adapted to model eyes with varying axial length, ONH position and ONH dimensions. Maps (n = 11,550) relating VF locations (24-2 pattern, n = 52 non-blind-spot locations) to 1° ONH sectors were generated for a range of clinically plausible anatomical parameters. Infrequently mapped ONH sectors (5%) were discarded for all locations. The influence of anatomical variables on the maps was explored by multiple linear regression.
Across all anatomical variants, for individual VF locations (24-2), total number of mapped 1° ONH sectors ranged from 12 to 90. Forty-one locations varied more than 30°. In five nasal-step locations, mapped ONH sectors were bimodally distributed, mapping to vertically opposite ONH sectors depending on vertical ONH position. Mapped ONH sectors were significantly influenced (P < 0.0002) by axial length, ONH position, and ONH dimensions for 39, 52, and 30 VF locations, respectively. On average across all VF locations, vertical ONH position explained the most variance in mapped ONH sector, followed by horizontal ONH position, axial length, and ONH dimensions.
Relations between ONH sectors and many VF locations are strongly anatomy-dependent. Our model may be used to produce customized maps from VF locations to the ONH in individual eyes where some simple biometric parameters are known.
[Show abstract][Hide abstract] ABSTRACT: To investigate the relationship between neuroretinal rim (NRR) differential light absorption (DLA, a measure of spectral absorption properties) and visual field (VF) sensitivity in primary open-angle glaucoma (POAG).
Patients diagnosed with (n = 22) or suspected of having (n = 7) POAG were imaged with a multispectral system incorporating a modified digital fundus camera, 250-W tungsten-halogen lamp, and fast-tuneable liquid crystal filter. Five images were captured sequentially within 1.0 second at wavelengths selected according to absorption properties of hemoglobin (range, 570-610 nm), and a Beer-Lambert law model was used to produce DLA maps of residual NRR from the images. Patients also underwent VF testing. Differences in NRR DLA in vertically opposing 180° and 45° sectors either side of the horizontal midline were compared with corresponding differences in VF sensitivity on both decibel and linear scales by Spearman's rank correlation.
The decibel VF sensitivity scale showed significant relationships between superior-inferior NRR DLA difference and sensitivity differences between corresponding VF areas in 180° NRR sectors (Spearman ρ = 0.68; P < 0.0001), superior-/inferior-temporal 45° NRR sectors (ρ = 0.57; P < 0.002), and superior-/inferior-nasal 45° NRR sectors (ρ = 0.59; P < 0.001). Using the linear VF sensitivity scale significant relationships were found for 180° NRR sectors (ρ = 0.62; P < 0.0002) and superior-inferior-nasal 45° NRR sectors (ρ = 0.53; P < 0.002). No significant difference was found between correlations using the linear or decibel VF sensitivity scales.
Residual NRR DLA is related to VF sensitivity in POAG. Multispectral imaging may provide clinically important information for the assessment and management of POAG.
[Show abstract][Hide abstract] ABSTRACT: To correlate in vivo spatial and spectral morphologic changes of short- to long-pulse 532 nm Nd:YAG retinal laser lesions using Fourier-domain optical coherence tomography (FD OCT), autofluorescence (AF), fluorescein angiography (FA), and multispectral imaging.
Ten eyes with treatment-naive preproliferative or proliferative diabetic retinopathy were studied. A titration grid of laser burns at 20, 100, and 200 milliseconds was applied to the nasal retina and laser fluence titrated to produce four grades of laser lesion visibility: subvisible (SV), barely visible (BV, light-gray), threshold (TH, gray-white), and suprathreshold (ST, white). The AF, FA, FD-OCT, and multispectral imaging were performed 1 week before laser, and 1 hour, 4 weeks, and 3 and 6 months post-laser. Multispectral imaging measured relative tissue oxygen concentration.
Laser burn visibility and lesion size increased in a linear relationship according to fixed fluence levels. At fixed pulse durations, there was a semilogarithmic increase in lesion size over 6 months. At 20 milliseconds, all grades of laser lesion were reduced significantly in size after 6 months: SV, 51%; BV, 54%; TH, 49%; and ST, 50% (P < 0.001), with retinal pigment epithelial proliferation and photoreceptor infilling. At 20 milliseconds, there was healing of photoreceptor inner segment/outer segment junction layers compared with 100- and 200-millisecond lesions. Significant increases in mean tissue oxygenation (range, four to six units) within the laser titration area and in oxygen concentration across the laser lesions (P < 0.01) were detected at 6 months.
For patients undergoing therapeutic laser, there may be improved tissue oxygenation, higher predictability of burn morphology, and more spatial localization of healing responses of burns at 20 milliseconds compared with longer pulse durations over time.
[Show abstract][Hide abstract] ABSTRACT: to examine the practical feasibility of developing a hyperspectral camera from a Zeiss fundus camera and to illustrate its use in imaging diabetic retinopathy and glaucoma patients.
the original light source of the camera was replaced with an external lamp filtered by a fast tunable liquid-crystal filter. The filtered light was then brought into the camera through an optical fiber. The original film camera was replaced by a digital camera. Images were obtained in normals and patients (primary open angle glaucoma, diabetic retinopathy) recruited at the Manchester Royal Eye Hospital.
a series of eight images were captured across 495- to 720-nm wavelengths, and recording time was less than 1.6s. The light level at the cornea was below the ANSI limits, and patients judged the measurement to be very comfortable. Images were of high quality and were used to generate a pixel-to-pixel oxygenation map of the optic nerve head. Frame alignment is necessary for frame-to-frame comparison but can be achieved through simple methods.
we have developed a hyperspectral camera with high spatial and spectral resolution across the whole visible spectrum that can be adapted from a standard fundus camera. The hyperspectral technique allows wavelength-specific visualization of retinal lesions that may be subvisible using a white light source camera. This hyperspectral technique may facilitate localization of retinal and disc pathology and consequently facilitate the diagnosis and management of retinal disease.
[Show abstract][Hide abstract] ABSTRACT: To describe a software package (Discus) for investigating clinicians' subjective assessment of optic disc damage [diagnostic accuracy in detecting visual field (VF) damage, decision criteria, and agreement with a panel of experts] and to provide reference data from a group of expert observers.
Optic disc images were selected from patients with manifest or suspected glaucoma or ocular hypertension who attended the Manchester Royal Eye Hospital. Eighty images came from eyes without evidence of VF loss in at least four consecutive tests (VF negatives), and 20 images from eyes with repeatable VF loss (VF positives). Software was written to display these images in randomized order, for up to 60 s. Expert observers (n = 12) rated optic disc damage on a 5-point scale (definitely healthy, probably healthy, not sure, probably damaged, and definitely damaged).
Optic disc damage as determined by the expert observers predicted VF loss with less than perfect accuracy (mean area under receiver-operating characteristic curve, 0.78; range, 0.72 to 0.85). When the responses were combined across the panel of experts, the area under receiver-operating characteristic curve reached 0.87, corresponding to a sensitivity of ∼60% at 90% specificity. Although the observers' performances were similar, there were large differences between the criteria they adopted (p < 0.001), even though all observers had been given identical instructions.
Discus provides a simple and rapid means for assessing important aspects of optic disc interpretation. The data from the panel of expert observers provide a reference against which students, trainees, and clinicians may compare themselves. The program and the analyses described in this article are freely accessible from http://www.discusproject.blogspot.com/.
Optometry and vision science: official publication of the American Academy of Optometry 10/2010; 88(1):E93-101. DOI:10.1097/OPX.0b013e3181fc30d2 · 1.60 Impact Factor