Comparison Of Blue-Light Autofluorescence and Ultrawidefield Green-Light Autofluorescence for Assessing Geographic Atrophy

PURPOSE

The goal of this study was to evaluate and compare the inter-modality and inter-reader agreement of manual and semiautomated GA (Geographic Atrophy) area measurements in eyes with GA due to age-related macular degeneration (AMD) using conventional blue and ultrawidefield (UWF) green light fundus autofluorescence (FAF) systems.

METHODS

FAF images of eyes with GA were obtained during a single visit using both the Spectralis HRA+OCT2 device and the Optos California device. Images were exported for masked analysis by two independent masked graders. The area of the GA lesion(s) was segmented and quantified (mm2) with a fully manual approach where the lesions were outlined using Optos Advance and Heidelberg Eye Explorer (HEYEX) software. In addition, for the Heidelberg blue FAF images, GA lesions were also measured using the instrument's semi-automated software (Region Finder 2.6.4). For comparison between modalities/grading method, the mean values of the two graders were used. Intraclass correlation coefficients (ICC) were computed to judge the agreement between graders.

RESULTS

72 eyes of 50 patients were included in this study. There was nearly perfect agreement between graders for the measurement of GA area for all three modalities (Intraclass Correlation coefficient = 0.996 for manual Optos Advance, 0.996 for manual Heidelberg HEYEX, 0.995 for Heidelberg Region Finder). The measurement of GA area was strongly correlated between modalities, with Spearman correlation coefficients of 0.985 (p < 0.001) between manual Heidelberg and manual Optos, 0.991 (p < 0.001) for Region Finder versus manual Heidelberg, and 0.985 (p < 0.001) for Region Finder versus manual Optos. The absolute mean area differences between the Heidelberg manual vs Region Finder, manual Optos vs Region Finder, and manual Optos vs manual Heidelberg were 1.61 mm2 (p<0.001), 0.90 mm2 (p<0.001), and 0.71 mm2 (p<0.001), respectively.

CONCLUSIONS

We observed excellent inter-reader agreement for measurement of GA using either 30-degree blue FAF or UWF green FAF, establishing the reliability of UWF imaging for macular GA assessment. While the absolute measurements between devices were strongly correlated, they differed significantly, highlighting the importance of using the same device for a given patient for the duration of a study.

Performance of Automated Machine Learning for Diabetic Retinopathy Image Classification from Multi-field Handheld Retinal Images

PURPOSE

To create and validate code-free automated deep learning models (AutoML) for diabetic retinopathy (DR) classification from handheld retinal images.

DESIGN

Prospective development and validation of AutoML models for DR image classification.

PARTICIPANTS

A total of 17 829 deidentified retinal images from 3566 eyes with diabetes, acquired using handheld retinal cameras in a community-based DR screening program.

METHODS

AutoML models were generated based on previously acquired 5-field (macula-centered, disc-centered, superior, inferior, and temporal macula) handheld retinal images. Each individual image was labeled using the International DR and diabetic macular edema (DME) Classification Scale by 4 certified graders at a centralized reading center under oversight by a senior retina specialist. Images for model development were split 8-1-1 for training, optimization, and testing to detect referable DR ([refDR], defined as moderate nonproliferative DR or worse or any level of DME). Internal validation was performed using a published image set from the same patient population (N = 450 images from 225 eyes). External validation was performed using a publicly available retinal imaging data set from the Asia Pacific Tele-Ophthalmology Society (N = 3662 images).

MAIN OUTCOME MEASURES

Area under the precision-recall curve (AUPRC), sensitivity (SN), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), accuracy, and F1 scores.

RESULTS

Referable DR was present in 17.3%, 39.1%, and 48.0% of the training set, internal validation, and external validation sets, respectively. The model's AUPRC was 0.995 with a precision and recall of 97% using a score threshold of 0.5. Internal validation showed that SN, SP, PPV, NPV, accuracy, and F1 scores were 0.96 (95% confidence interval [CI], 0.884-0.99), 0.98 (95% CI, 0.937-0.995), 0.96 (95% CI, 0.884-0.99), 0.98 (95% CI, 0.937-0.995), 0.97, and 0.96, respectively. External validation showed that SN, SP, PPV, NPV, accuracy, and F1 scores were 0.94 (95% CI, 0.929-0.951), 0.97 (95% CI, 0.957-0.974), 0.96 (95% CI, 0.952-0.971), 0.95 (95% CI, 0.935-0.956), 0.97, and 0.96, respectively.

CONCLUSIONS

This study demonstrates the accuracy and feasibility of code-free AutoML models for identifying refDR developed using handheld retinal imaging in a community-based screening program. Potentially, the use of AutoML may increase access to machine learning models that may be adapted for specific programs that are guided by the clinical need to rapidly address disparities in health care delivery.

FINANCIAL DISCLOSURE(S)

Proprietary or commercial disclosure may be found after the references.

Imaging in diabetic retinopathy

While the primary method for evaluating diabetic retinopathy involves direct and indirect ophthalmoscopy, various imaging modalities are of significant utility in the screening, evaluation, diagnosis, and treatment of different presentations and manifestations of this disease. This manuscript is a review of the important imaging modalities that are used in diabetic retinopathy, including color fundus photography, fluorescein angiography, B-scan ultrasonography, and optical coherence tomography. The article will provide an overview of these different imaging techniques and how they can be most effectively used in current practice.

Ultra-High Resolution Optical Coherence Tomography Imaging of Unilateral Drusen in a 31 Year Old Woman

We report a case of widespread unilateral drusen in a healthy 31 year old Caucasian woman using multi-modal imaging including ultra-high resolution optical coherence tomography (UHR-OCT). Dilated fundus exam showed multiple drusen-like lesions in the posterior pole without heme or fluid. Fundus auto fluorescence demonstrated hyperautofluorescent at the deposits. Fluorescein angiography revealed mild hyperfluorescence and staining of the lesions. Spectral-domain optical coherence tomography (SD-OCT) OS showed accumulations in the temporal macula at Bruch’s membrane. UHR-OCT provided improved axial resolution compared to the standard 5 μm on the commercial SD-OCT and confirmed the presence of deposits in Bruch’s membrane, consistent with drusen. The retinal layers were draped over the excrescences but did not show any disruption.

Cone photoreceptor definition on adaptive optics retinal imaging

Aims

To quantitatively analyse cone photoreceptor matrices on images captured on an adaptive optics (AO) camera and assess their correlation to well-established parameters in the retinal histology literature.

Methods

High resolution retinal images were acquired from 10 healthy subjects, aged 20–35 years old, using an AO camera (rtx1, Imagine Eyes, France). Left eye images were captured at 5° of retinal eccentricity, temporal to the fovea for consistency. In three subjects, images were also acquired at 0, 2, 3, 5 and 7° retinal eccentricities. Cone photoreceptor density was calculated following manual and automated counting. Inter-photoreceptor distance was also calculated. Voronoi domain and power spectrum analyses were performed for all images.

Results

At 5° eccentricity, the cone density (cones/mm² mean±SD) was 15.3±1.4×10³ (automated) and 13.9±1.0×10³ (manual) and the mean inter-photoreceptor distance was 8.6±0.4 μm. Cone density decreased and inter-photoreceptor distance increased with increasing retinal eccentricity from 2 to 7°. A regular hexagonal cone photoreceptor mosaic pattern was seen at 2, 3 and 5° of retinal eccentricity.

Conclusions

Imaging data acquired from the AO camera match cone density, intercone distance and show the known features of cone photoreceptor distribution in the pericentral retina as reported by histology, namely, decreasing density values from 2 to 7° of eccentricity and the hexagonal packing arrangement. This confirms that AO flood imaging provides reliable estimates of pericentral cone photoreceptor distribution in normal subjects.