A study to assess whether a system of automated image software can be developed to read retinal images taken without dilating the pupils in patients attending diabetic eye screening

ISRCTN ISRCTN14412723
DOI https://doi.org/10.1186/ISRCTN14412723
IRAS number 275896
Secondary identifying numbers Version 2.1, 24 March 2021, IRAS 275896, CPMS 44499
Submission date
04/08/2021
Registration date
09/02/2022
Last edited
11/01/2024
Recruitment status
No longer recruiting
Overall study status
Completed
Condition category
Eye Diseases
Prospectively registered
Protocol
Statistical analysis plan
Results
Individual participant data
Record updated in last year

Plain English summary of protocol

Background and study aims
Diabetes affects 4.7 million people in the UK. A common complication is diabetic retinopathy (DR), which is damage to the blood vessels in the eye. Screening for the early detection of DR has contributed to a reduction in blindness due to diabetes in the UK. This is carried out each year by using photography of the back of the eye (the retina). In England and Wales, the pupils are dilated with eye drops and then two 45-degree photographs of each eye are taken which gives a combined photographic width of 60 degrees of retina.
Scanning confocal ophthalmoscopes (SCO) are cameras that use low-powered laser or LED light to scan across the retina without the need for a bright flash and the width of one photograph varies from 60 - 200 degrees. This means only one photo of each eye is needed and pupil dilation may not be required. This would be an advantage as pupil dilation takes time, can be uncomfortable, and blurs the vision for several hours, meaning most people need to be accompanied to screening appointments and can’t drive afterwards. Better accuracy, shorter appointment times, and the elimination of eye drops could increase the number of people who attend eye screening and DR could be caught earlier which may reduce the risk of vision loss.
At present, it is unknown whether white light LED or laser light is better for detection of DR, or if a wider photo might pick up more disease in the wider retina but may pick up less in the centre. Both might have an effect on referral to hospital eye clinics. This study will assess the accuracy of two new cameras when screening for any eye disease caused by diabetes and show whether the new cameras are as good as the current system or offer improved detection.
A company called RetinaScan Ltd have developed automated analysis software to detect diabetic retinopathy in the standard (two-field digital) images and they wish to extend this work to include images from wide-field scanning confocal ophthalmoscopes. To do this they need to have images of the back of the eye to develop and test their computer programmes.
Finding out whether these cameras can capture images of adequate quality without using drops to dilate the pupils is an important secondary objective of this study. The screening programme in Scotland takes images without dilating the pupils and only does so if the images are of poor quality. This is around one third still needing to have drops. Fewer patients could require dilation with these new wide-field cameras, they provide a much wider field of image, and using computers to read the images could improve the service.

Who can participate?
People with diabetes who meet the inclusion criteria for the National Diabetes Eye Screening Programme in the UK. These will be people with diabetes over the age of 12 years (except for those under 16 years attending their first screening appointment) who give their consent to take part.

What does the study involve?
People attending their routine screening appointment have two extra photographs taken of each eye by each of the cameras before pupil dilation, followed by the normal screening procedure. If the image quality is poor they will be asked to have a further photograph taken of each eye with the new cameras after their pupils are dilated. The images and the results of the screening for diabetic eye disease will shared with the company developing the computer programme. The researchers will also combine the results with another study called CONCORDIA 1 to see which camera provides the best photographs for screening. Participants will be asked for their opinion on having their eyes examined on the existing and new cameras. This is a single appointment visit and no study follow up is required.

What are the risks?
The extra images will add an extra 30 minutes to a participant’s appointment time but this was felt to be acceptable providing clear information is given and the extra parking costs reimbursed. There are no anticipated risks of taking part in the study or any direct benefits to the study population. The benefits may be in the future if these devices are considered an improvement and are adopted into the National Screening Programme.

Where is the study run from?
Gloucestershire Hospitals NHS Foundation Trust (UK)

When will the study take place?
November 2019 to December 2022

Who is funding the study?
1. Innovate UK
2. RetinaScan Ltd (UK)

Who is the main contact?
Prof. Peter Scanlon
p.scanlon@nhs.net

Contact information

Prof Peter Scanlon
Scientific

GRRG Office, above Oakley Ward
Cheltenham General Hospital
Cheltenham
GL53 7AN
United Kingdom

ORCiD logoORCID ID 0000-0001-8513-710X
Phone +44 (0)3004222851
Email p.scanlon@nhs.net
Mrs Laura Lodge
Public

The Orchard Centre
Gloucestershire Royal Hospital
Great Western road
Gloucester
GL1 3NN
United Kingdom

Phone +44 (0)3004223853
Email laura.lodge1@nhs.net

Study information

Study designNon-randomized controlled trial
Primary study designInterventional
Secondary study designNon randomised study
Study setting(s)Hospital
Study typeScreening
Participant information sheet Not available in web format, please use the contact details to request a participant information sheet
Scientific titleArtificial image detection of diabetic retinopathy in wide-field images and assessment of non-mydriatic image quality of wide-field scanning confocal ophthalmoscopes
Study acronymAIDED
Study objectivesTo assess whether retinal images taken on the Optos California and Zeiss Clarus Scanning ophthalmoscopes can be read using artificial imaging software
Ethics approval(s)

Approved 14/02/2020, South West - Cornwall & Plymouth Research Ethics Committee (Level 3, Block B, Whitefriars, Lewins Mead, Bristol, BS1 2NT, United Kingdom; +44 (0)207 104 8019; cornwallandplymouth.rec@hra.nhs.uk), ref: 20/SW/0009

Health condition(s) or problem(s) studiedDiabetic eye disease
InterventionThe interventions are two different scanning confocal ophthalmoscopes (SCO) devices (the Optos California and Zeiss Clarus Scanning ophthalmoscopes) and the images are compared with the standard digital images taken in the English National Screening Programme. The device order is randomised but not concealed for the operator as this isn’t possible. The random order of the camera use is generated by the software used. The study software randomises the camera order to ensure there is an equal split on which camera is used first for imaging. If a subject is withdrawn from the study either during or post imaging the study software will recalculate required numbers for an even spilt and set the randomisation process accordingly.

The duration of participation in the trial is short, around 30 min per participant, as they will have additional images taken with the new devices before their pupils are dilated and then have their normal screening images taken for comparison. They will be asked to complete a short satisfaction questionnaire and the EuroQoL EQ5D-5L which is part of the data collection for the health economics element in work package 3 of a linked study (CONCORDIA 1 study). If the images on the new device are not readable, they will be re-taken after the pupils are dilated. After this their participation is finished as there is no follow-up for the study.

The retinal images and grading information are compared to support the development of an algorithm for the detection of diabetic retinopathy (DR) on wide-field SCO images that can be tested in a larger-scale study. Images from the devices will be uploaded to a secure NHS server, checked and prepared into ‘reading lists’ for assessors. Image grading will be performed in a masked fashion by expert graders from the Gloucester Reading Centre and the National Screening College (NSC). A dedicated state-of-the-art image grading platform will be used on a secure server. All grading is carried out in line with the English Diabetic Eye Screening programme using trained retinal image graders. Double grading with adjudication will be performed and detection of individual lesions by the different cameras within the standard screening 45-degree fields will be recorded.
Intervention typeDevice
Pharmaceutical study type(s)Not Applicable
PhaseNot Applicable
Drug / device / biological / vaccine name(s)Optos California ophthalmoscope, Zeiss Clarus Scanning ophthalmoscope
Primary outcome measureThe presence or absence of diabetic eye disease in the retinal images taken by the Optos California and Zeiss Clarus Scanning ophthalmoscope as assessed by the grading system used in the England Diabetic Eye screening service at a single timepoint
Secondary outcome measures1. The proportion of those with non-mydriatic images that the screener considers are ungradable and decides that a mydriatic image is required with that SCO device, measured using the grading system used in the England Diabetic Eye screening service at a single timepoint
2. The proportion of those in whom a second mydriatic image is taken that then become gradable, measured using the grading system used in the England Diabetic Eye screening service at a single timepoint
3. Screener and participant's perspectives on the devices:
3.1. Participant feedback: immediately after imaging on the SCO device, participants will be asked to complete a graphical feedback question. Participants will be asked to score the experience of the SCO devices from one of four options: No problem, Okay, Not nice, unpleasant.
3.2. Research imager feedback: as soon as each participant has completed imaging on each SCO device and has left the room, the imager will score the experience of screening the patient on each SCO from one of the following four options: No problem, Okay, Difficult, Very difficult.
4. Time taken to capture and grade images by device type. The time taken to capture the images is recorded automatically via the software used in the screening department. The graders will record the time spent grading each image.
5. Quality of life measured using the EuroQuol EQ-5D 5L at the screening appointment. The researchers will convert the responses to the EQ-5D-5L instrument into health utilities as per national guidelines (https://www.nice.org.uk/about/what-we-do/our-programmes/nice-guidance/technology-appraisal-guidance/eq-5d-5l)
6. Seek consent from participants to link images with their clinical information relating to diabetic care for a future cost-benefit analysis.
Data from outcomes 3-6 will contribute to the CONCORDIA 1 study
Overall study start date01/11/2019
Completion date31/12/2022

Eligibility

Participant type(s)Patient
Age groupMixed
SexBoth
Target number of participants1000
Total final enrolment1013
Key inclusion criteria1. People with diabetes giving informed consent
2. Those meeting the inclusion criteria for the English National Diabetic eye screening programme (DESPs) who are people with diabetes over the age of 12 years except for those under 16 years attending their first screening appointment
Key exclusion criteria1. People in whom it is not possible to take retinal images
2. People or parents unable or unwilling to give informed consent
3. Those with eye disease that might affect the interpretation of DR levels
4. Children under 16 years attending for their first retinal screening appointment, which will automatically exclude 12 year old children as this is when they are first invited
Date of first enrolment11/08/2020
Date of final enrolment06/07/2021

Locations

Countries of recruitment

  • England
  • United Kingdom

Study participating centre

Gloucestershire Royal Hospital
The Orchard Centre
Great Western Road
Gloucester
GL1 3NN
United Kingdom

Sponsor information

Gloucestershire Hospitals NHS Foundation Trust
Hospital/treatment centre

Leadon House
Gloucestershire Royal Hospital
Great Western Road
Gloucester
GL1 3NN
England
United Kingdom

Phone +44 (0)3004225460
Email chantal.sunter1@nhs.net
Website https://www.gloshospitals.nhs.uk
ROR logo "ROR" https://ror.org/04mw34986

Funders

Funder type

Government

Innovate UK
Government organisation / National government
Alternative name(s)
innovateuk
Location
United Kingdom
Retinascan

No information available

Results and Publications

Intention to publish date30/09/2024
Individual participant data (IPD) Intention to shareYes
IPD sharing plan summaryAvailable on request
Publication and dissemination planThe researchers will link into wider networks of people with diabetes, locally and nationally through Diabetes UK, Facebook groups and other social media to disseminate the findings to the public and through peer-reviewed journals and conferences to the clinical community. The researchers are happy to consider sharing documentation on request; some may not be available due to commercial confidentiality.
IPD sharing planSharing data will be considered on a case by case basis. All data including retinal images will be anonymised. The contact for this would be the Chief investigator Prof. Peter Scanlon (p.scanlon@nhs.net).

Study outputs

Output type Details Date created Date added Peer reviewed? Patient-facing?
HRA research summary 28/06/2023 No No

Editorial Notes

11/01/2024: The following changes were made:
1. The secondary study design was changed from Randomised cross over trial to Non randomised study and the change was reflected in the study design field.
2. The intention to publish date was changed from 31/12/2023 to 30/09/2024.
11/08/2022: The overall trial end date has been changed from 31/07/2021 to 31/12/2022 and the plain English summary was updated accordingly.
02/03/2022: Internal review.
24/08/2021: Trial's existence confirmed by the South West - Cornwall & Plymouth Research Ethics Committee.