This AI directory is a a free and useful resource that the College has created for clinicians, commissioners, vendors, researchers and anyone with an interest in AI.
Introducing the AI directory
This is a free and useful resource that the College has created for clinicians, commissioners, vendors, researchers and anyone with an interest in artificial intelligence (AI).
We have developed the directory because we believe that AI has the potential to have a profound impact on the science and practice of ophthalmology.
Comprising information published by vendors and academic sources, the AI directory shows at-a-glance artificial intelligence as a medical device (AIaMD) tools that have contributed to the specialty in the UK. It does not recommend specific vendors or their products, or draw any comparisons between devices.*
Why do we need an AI directory?
The use and influence of AI will continue to expand as the technology develops and more AIaMD tools are adopted for clinical, research and administrative use. Our AI directory will capture this development and growth. Currently, no AIaMD tools have been rolled out across NHS ophthalmology departments: their use is localised or has contributed to studies and research, and some tools have been tested in pilot form. It is strongly anticipated that in time devices will become patient-facing.
How will ophthalmology lead the way in AI adoption?
Ophthalmology is an image-focused specialty, and so AI is playing an increasingly important role in diagnostics. Datasets such as multiple retinal scans that ophthalmologists rely on for making a diagnosis and developing a treatment plan readily lend themselves to the training of AI algorithms.
Addressing concerns about the role of AI in the science of ophthalmology
As we outline in our AI position statement, AI will not replace doctors but powerful time-saving tools will support efficiency improvements. Globally, AI has the potential to offer health providers scalable solutions for screening and diagnosis, which will speed up patient waiting times and streamline referral to treatment pathways.
AI directory
Welcome to the AI directory for ophthalmology. Click on the AIaMD tools to see more details.
Product version: 2.1
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Diabetic Retinopathy Screening
Intended use (such as in diagnosis of DR, AMD, Glaucoma): Diagnosis of DR
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): CE 0413 marking as a class IIb medical device in the European Union
Application
Function (such as in the detection of a specific eye disease): The EyeArt AI Eye Screening System provides in-clinic, real-time diabetic retinopathy (DR) screening for primary care practices, diabetes centres and optometric offices. It allows physicians to quickly and accurately identify referable DR patients during a diabetic patient’s regular exam.
Integration (Technologies such as DICOM and PACS used to manage and access medical images): EyeArt is indicated for use with Canon CR-2 AF, Canon CR-2 Plus AF, and Topcon NW400 cameras.
PACS vendor(s): Integration with existing electronic health record (EHR) systems and picture archival and communication systems (PACS)
Project stage (such as pilot, clinical usage): Use in clinical screening.
NHS setting: DESPs (Diabetic Eye Screening Programmes) in North East London (NEL), South East London (SEL) and Gloucestershire (GS).
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Diabetic Retinopathy Screening
Intended use (such as in diagnosis of DR, AMD, Glaucoma):Detection but not diagnosis of DR
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): class IIa medical device in the European Union
Application
Function (such as in the detection of a specific eye disease): Retmarker is a mathematical algorithm that automatically analyses colour fundus photographs and identifies diabetics in risk of progression of Diabetic Retinopathy. One study evaluating RetmarkerDR identified that its sensitivity was 73.0% for any retinopathy, 85.0% for referable retinopathy and 97.9% for proliferative retinopathy when compared to the results of human grading.
Integration (Technologies such as DICOM and PACS used to manage and access medical images):State-of-art automated fundus camera from leading provider, Visual Acuity measurement software, Laptop with configured secure internet connection to send data.
PACS vendor(s): The AI algorithms rely on either a standalone web application running on a Windows PC with an internet connection or a PACS/EHR system integration.
Project stage (such as pilot, clinical usage): Clinical and research screening.
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Diabetic Retinopathy Screening
Intended use (such as in diagnosis of DR, AMD, Glaucoma): Diagnosis of DR
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): CE marked for detection of RDR, EU MDR Class II Regulatory Approval
Application
Function (such as in the detection of a specific eye disease): Medios AI enables automated, offline screening for DR, providing healthcare providers with a quick and reliable diagnostic tool. Medios AI sensitivity was 100% for Referable DR and the specificity was 92% for No DR, in a prospective, opportunistic screening trial on 250 diabetic patients. It has an offline capability.
Subspecialty (such as Glaucoma, AMD, DR): DR
Usage
Setting: Research and in clinical settings outside of the UK
Integration (Technologies such as DICOM and PACS used to manage and access medical images):Devices upload images and patient data to a cloud platform. The web platform transmits patient data to EMRs and PACS servers using API integrations or DICOM / HL7 gateway.
PACS vendor(s): Medios DR can be an integrated part of the Fundus on Phone.
Project stage (such as pilot, clinical usage): Research screening, and clinical screening outside of the UK.
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Software for the interpretation of OCT images as part of the management of geographic atrophy (GA), the dry form of AMD.
Intended use (such as in diagnosis of DR, AMD, Glaucoma): Diagnosis of GA
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): MDR 2017/745 / Class IIa medical device
Application
Function (such as in the detection of a specific eye disease): Software provides ophthalmologists with a PDF report for patients with age-related macular degeneration. Measure and visualise disease activity and therapeutic efficacy across photoreceptor (PR) degeneration (EZ layer loss), retinal pigment epithelium (RPE) loss replacing FAF2, and the PR/RPE overlap providing insight into progression and therapeutic response.
Subspecialty (such as Glaucoma, AMD, DR): AMD
Usage
Setting: Research and clinical outside of UK
Launch date: 2020
Integration (Technologies such as DICOM and PACS used to manage and access medical images): Uses optical coherence tomography (OCT)
NHS setting: Commercial agreement with veonet enabling deployment in UK. Study conducted in private practise in London.
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Software for the interpretation of OCT images as part of the management of nAMD.
Intended use (such as in diagnosis of DR, AMD, Glaucoma): Diagnosis of nAMD
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): MDR 2017/745 / Class IIa medical device
Application
Function (such as in the detection of a specific eye disease): Automatically quantify fluid in nAMD, generate a follow-up chart of fluid dynamics to track individual recurrence profiles and visualise disease progression over time, facilitating disease management.
Subspecialty (such as Glaucoma, AMD, DR): AMD
Usage
Setting: Research
Launch date: 2020
Integration (Technologies such as DICOM and PACS used to manage and access medical images):
PACS vendor(s):
Project stage (such as pilot, clinical usage):
NHS setting: Commercial agreement with veonet enabling future deployment in UK
Product version: 1.1
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Single field 45° colour fundus images
Intended use (such as in diagnosis of DR, AMD, Glaucoma): Diagnosis of DR
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): Class I CE mark
Application
Function (such as in the detection of a specific eye disease): Images triaged for manual grading using a support vector machine based autograder (iGradingM) trained to classify image sets as ungradable or contains any DR. iGradingM (Medalytix Ltd) has been shown to safely identify about 50% of the screening episodes not requiring manual grading.
Subspecialty (such as Glaucoma, AMD, DR): DR
Usage
Setting: Research
Launch date: 2011
Integration (Technologies such as DICOM and PACS used to manage and access medical images):
Modality (such as fundus photography, optical coherence tomography (OCT), and fluorescein angiography): Images captured via fundus cameras and OCT scans
Intended use (such as in diagnosis of DR, AMD, Glaucoma): DR, AMD, Glaucoma
Regulatory approval & class (such as UKCA (UK Conformity Assessed) or CE marks in UK, CE (Conformité Européene) and MDR (Medical Device Regulation) in EU and FDA (Food and Drug Administration) in US): Class I CE mark (self-certified) medical device.
Application
Function: (such as in the detection of a specific eye disease): Detection of DR, AMD, Glaucoma, but not diagnosis. Transmits retinal images through its web-based platform for processing.
Subspecialty (such as Glaucoma, AMD, DR): DR, AMD, Glaucoma
Usage
Setting: Research
Launch date: 2013
Integration (Technologies such as DICOM and PACS used to manage and access medical images): The AI algorithms rely on either a standalone web application running on a Windows PC with an internet connection or a PACS/EHR system integration.
The first iteration of the AI directory contains AIaMD tools that are intended for a clinical or research setting and that are powered by diagnostics and treatment-based AI. AI will also advance the science of ophthalmology by streamlining processes for researchers, such as those who work in an ophthalmic reading centre. These AI-powered tools are deployed for image grading and analysis, with an algorithm able to analyse many images faster and more accurately than the human eye. Other digital health technologies may use AI for administrative purposes. Uses include triaging outpatients for follow-up appointments and automatically writing up patient notes for doctors from voice recordings. Such devices are not AIaMD so have not been considered for this iteration of the Directory. See ‘Machine Learning-enabled Medical Devices: Key Terms and Definitions‘ by the International Medical Device Regulators Forum (IMDRF) for a definition of ‘medical device’.
Further resource – the AI and Digital Regulations Service (AIDRS)
The AI Directory may be of interest to vendors, developers and innovators. The products listed within the directory have been through a procurement process and received regulatory approval. This can be a long and complicated process to navigate. The AI and Digital Regulations Service (AIDRS) is a free resource that is a collaboration between NHS Health Research Authority, NICE, MHRA and CQC. It has separate channels of guidance for both developers and adopters, walking them through every step of the process.
*Disclaimer
The Royal College of Ophthalmologists’ AI directory provides details about AIaMD tools that have been or are being used in NHS ophthalmic research or practice. It does not recommend specific vendors or their products, or draw any comparisons between devices. The College strives to keep the directory accurate and up to date, but it may not be possible to capture every AIaMD as soon as it has been applied in ophthalmology. We cannot guarantee that the resource is free from error or omission. Contact [email protected] with any queries about the AI directory. Vendors seeking inclusion for their product should provide full details about its development and use so that it can be considered for inclusion.
