Cross-modality transfer learning with knowledge infusion for diabetic retinopathy grading

Background

Ultra-wide-field (UWF) fundus photography represents an emerging retinal imaging technique offering a broader field of view, thus enhancing its utility in screening and diagnosing various eye diseases, notably diabetic retinopathy (DR). However, the application of computer-aided diagnosis for DR using UWF images confronts two major challenges. The first challenge arises from the limited availability of labeled UWF data, making it daunting to train diagnostic models due to the high cost associated with manual annotation of medical images. Secondly, existing models' performance requires enhancement due to the absence of prior knowledge to guide the learning process.

Purpose

By leveraging extensively annotated datasets within the field, which encompass large-scale, high-quality color fundus image datasets annotated at either image-level or pixel-level, our objective is to transfer knowledge from these datasets to our target domain through unsupervised domain adaptation.

Methods

Our approach presents a robust model for assessing the severity of diabetic retinopathy (DR) by leveraging unsupervised lesion-aware domain adaptation in ultra-wide-field (UWF) images. Furthermore, to harness the wealth of detailed annotations in publicly available color fundus image datasets, we integrate an adversarial lesion map generator. This generator supplements the grading model by incorporating auxiliary lesion information, drawing inspiration from the clinical methodology of evaluating DR severity by identifying and quantifying associated lesions.

Results

We conducted both quantitative and qualitative evaluations of our proposed method. In particular, among the six representative DR grading methods, our approach achieved an accuracy (ACC) of 68.18% and a precision (pre) of 67.43%. Additionally, we conducted extensive experiments in ablation studies to validate the effectiveness of each component of our proposed method.

Conclusion

In conclusion, our method not only improves the accuracy of DR grading, but also enhances the interpretability of the results, providing clinicians with a reliable DR grading scheme.

Design of ultra-wide-field scanning laser fundus imaging system with cascaded conicoid mirrors

We propose and design a multi-stage cascaded scanning laser ophthalmoscope (SLO) for ultra-wide field (UWF), which uses conicoid mirrors, constructed by conjugation of pupil plane. The vergence uniformity and the angular magnification of a cascaded conicoid mirrors (CCM) system are analyzed recursively and optimized preliminarily to achieve high quality imaging with UWF, and the optimal system with the model eye are obtained by simulation and optimization. Two-stage and three-stage cascaded systems are designed with this method, and the formulas of beam vergence and angular magnification are obtained by theoretical derivation. As compared to the two-stage CCM system, the proposed three-stage cascaded UWF SLO has superior performance in imaging quality. Its average RMS radius of spot diagram is calculated to be 26.372 µm, close to the diffractive limit resolution. The image resolution of human retina can be up to 30 µm with 135° FOV in theory. The three-stage cascaded SLO is more suitable for UWF fundus imaging. This study will be helpful for early screening and accurate diagnosis of various diseases in the peripheral retina.

Prevalence of Diabetic Eye Diseases in American Indians and Alaska Natives (AI/AN) as Identified by the Indian Health Service's National Teleophthalmology Program Using Ultrawide Field Imaging (UWFI)

PURPOSE

Estimates of diabetic eye disease in American Indian and Alaska Natives (AI/AN) vary over time, region, and methods. This article reports recent prevalence of diabetic retinopathy (DR) and diabetic macular edema (DME) in AI/AN served by the Indian Health Services' (IHS) teleophthalmology program, as identified using ultrawide field imaging (UWFI).

METHODS

This was a retrospective analysis of 2016-2019 clinical data (n = 53,900). UWF images were acquired by certified imagers using a validated protocol, and graded by licensed, certified optometrists supervised by an ophthalmologist. Graders evaluated the extent/severity of retinal lesions in comparison to standard photographs. DR lesions predominantly in any peripheral field were considered "predominantly peripheral lesions" (PPL). The analyses calculated prevalence of any DR, any DME, DR and DME severity, sight-threatening disease, and PPL.

RESULTS

Patients averaged 56 years of age with a 68 mmol/mol A1c and 55% had had diabetes for 5+ years. Prevalence of any DR, any DME, and sight-threatening disease was 28.6%, 3.0%, and 3.0%. In patients with mild nonproliferative DR, PPL was seen in 25.3%. PPL suggested a more severe level of DR in 8.7% of patients. DR increased with age. DME decreased with age. Males and patients in the Nashville IHS area had more diabetic eye disease.

CONCLUSION

AI/AN have a high burden of diabetes and its complications. The IHS is resource-constrained, making accurate disease estimates necessary for resource allocation and budget justifications to Congress. These data update the estimates of diabetic eye disease in Indian Country and suggest that UWFI identifies early DR.

Ultra-wide-field fundus photography compared to ophthalmoscopy in diagnosing and classifying major retinal diseases

To analyze the performance of ultra-wide-field (UWF) fundus photography compared with ophthalmoscopy in identifying and classifying retinal diseases. Patients examined for presumed major retinal disorders were consecutively enrolled. Each patient underwent indirect ophthalmoscopic evaluation, with scleral depression and/or fundus biomicroscopy, when clinically indicated, and mydriatic UWF fundus imaging by means of CLARUS 500™ fundus camera. Each eye was classified by a clinical grader and two image graders in the following groups: normal retina, diabetic retinopathy, vascular abnormalities, macular degenerations and dystrophies, retinal and choroidal tumors, peripheral degenerative lesions and retinal detachment and myopic alterations. 7024 eyes of new patients were included. The inter-grader agreement for images classification was perfect (kappa = 0.998, 95% Confidence Interval (95%CI) = 0.997-0.999), as the two methods concordance for retinal diseases diagnosis (kappa = 0.997, 95%CI = 0.996-0.999) without statistically significant difference. UWF fundus imaging might be an alternative to ophthalmoscopy, since it allows to accurately classify major retinal diseases, widening the range of disorders possibly diagnosed with teleophthalmology. Although the clinician should be aware of the possibility that a minority of the most peripheral lesions may be not entirely visualized, it might be considered a first line diagnostic modality, in the context of a full ophthalmological examination.

Comparison of different methods of retinal imaging for the screening of diabetic retinopathy: a systematic review

BACKGROUND

Screening for diabetic retinopathy (DR) is recommended to detect sight-threatening complications prior to visual loss. Early Treatment Diabetic Retinopathy Study (ETDRS) seven standard field (7SF) retinal imaging has traditionally been regarded the gold standard for DR classification, but other methods are often preferred clinically. The purpose of this systematic review was to determine whether 7SF is the most optimal screening method for DR grading, or if similar results can be achieved by other methods using a smaller field of view (<7SF) or ultra-wide field (UWF) imaging.

METHODS

Based on Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines, two independent reviewers initially identified 7167 publications in PubMed, Cochrane and Embase databases. Of these, 16 publications were included based on predefined inclusion criteria.

RESULTS

7SF was used as reference standard in 12 studies (compared with < 7SF in five studies and UWF in seven studies), and four studies compared other reference standards. Compared to 7SF, studies using < 7SF and UWF images both reported of similar agreement. A lower rate of ungradable images was reported for mydriatic and non-mydriatic UWF as compared to non-mydriatic < 7SF modalities.

CONCLUSION

Retinal imaging of <7SF and UWF both provide acceptable performance compared to 7SF. Given the time-consuming nature of the latter, these methods could be reasonable options in DR screening, even though a high number of ungradable images in non-mydriatic < 7SF may pose a clinical challenge.

Ultra-wide field retinal imaging: A wider clinical perspective

The peripheral retina is affected in a variety of retinal disorders. Traditional fundus cameras capture only a part of the fundus even when montaging techniques are used. Ultra-wide field imaging enables us to delve into the retinal periphery in greater detail. It not only facilitates assessing color images of the fundus, but also fluorescein angiography, indocyanine green angiography, fundus autofluorescence, and red and green free images. In this review, a literature search using the keywords "ultra-widefield imaging", "widefield imaging", and "peripheral retinal imaging" in English and non-English languages was done and the relevant articles were included. Ultra-wide field imaging has made new observations in the normal population as well as in eyes with retinal disorders including vascular diseases, degenerative diseases, uveitis, age-related macular degeneration, retinal and choroidal tumors and hereditary retinal dystrophies. This review aims to describe the utility of ultra-wide field imaging in various retinal disorders.

Recent trends in vision impairment certifications in England and Wales

BACKGROUND

The Certificate of Visual Impairment (CVI) provides essential data for preventable sight loss indicators as part of the Public Health Outcomes Framework (PHOF) published annually by the Department of Health. Trends in CVI certification rates can provide information on the effectiveness of current services and treatments and may be used to guide allocation of resources, and is the only such indicator within ophthalmology. This study aimed to compare recent trends in new vision impairment certifications in 2017/18 against prior baseline data in England and document trends in new certifications in Wales.

METHODS

PHOF data from 2010/11 and 2017/18 were examined with respect to preventable sight loss indicators: age-related macular degeneration (AMD) (Indicator E12a), glaucoma (Indicator E12b), diabetic eye disease (Indicator E12c) as well as the total numbers of certifications (Indicator E12d).

RESULTS

In 2017/18, the rate of new CVI certifications was 41 per 100,000 population which has reduced from 43/100,000 in 2010/11 in England. Certifications for AMD reduced from 132/100,000 in 2010/11 to 107/100,000 in 2017-18. Certifications for glaucoma have remained stable at 13/100,000 in 2017/8. Certifications for diabetic eye disease have declined from 4/100,000 in 2010/11 to 3/ 100,000 in 2017/18. The number of vision impaired individuals that each Clinical Commissioning Group (CCG) has to support varies from 8 to 82 per 100,000 population.

CONCLUSIONS

There has been a significant decrease in the rate of all CVI certifications particularly from AMD and diabetic retinopathy. However, maintaining this will require changes in the way care is delivered as the elderly population, which is at greatest risk of preventable sight loss, is projected to increase by 50% over the next 20 years. Inherited retinal diseases are now the leading cause of sight loss in the working age population. CVI data demonstrate the need for CCGs to tailor their investment in ophthalmic services to the needs of their specific patient populations. It is important that all ophthalmologists continue to provide accurate CVI data in order to help support the future equitable allocation of adequate resources to reduce avoidable vision loss.