AI-Powered DIY Screening System for Diabetic Retinopathy

AI-powered Low-cost Portable Fundus Camera to Deliver Diabetic Retinopathy Screening At Primary Health Care Setting: a Pragmatic Trial

A pragmatic trial will be conducted in two representative clinics in each of the three types of targeted settings. It will be run for 3 months in each clinic to complete data collection of up to 100 patients in SPGC and Optometry Clinics, 200 at GPGC and 200 in GOPC in total. All the subjects will conduct a DIY screening, physician consultation, survey with questionnaires and a phone interview three months after their baseline assessment. This study will assess automated screening in terms of success rate of the DIY system without active assistant help, accuracy, screening rate and detection rate, adherence to referral and experience of participants, as well as cost-effectiveness in real-world settings.

Study Overview

• Status: Recruiting
• Conditions: Diabetic Retinopathy
• Intervention / Treatment: Diagnostic test: AI-powered fundus camera screening

• Study Type: Interventional
• Enrollment (Estimated): 500
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Mingguang He; Phone Number: 85234002795; Email: mingguang.he@polyu.edu.hk
• Study Contact Backup: Name: Yanxian Chen; Phone Number: 85227666111; Email: yan-xian.chen@polyu.edu.hk

Study Locations

• Hong Kong
  • Hong Kong, Hong Kong
    • Recruiting
    • SPGC and Optometry Clinics
    • Contact: He Prof.; Phone Number: 85234002795; Email: mingguang.he@polyu.edu.hk

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: Yes

Description

Inclusion Criteria:

1. subjects who are 50 years of age or older, or subjects who are 18 years of age or older and have diabetes, and
2. have not undertaken an eye examination in the previous 12 months (elevated risk of undiagnosed disease).
3. Provide written informed consent.

Exclusion Criteria:

1. individuals with physical disabilities that prevent the use of a fundus camera, or
2. individuals with speech impairments who are unable to complete telephone follow-ups.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Screening
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Screening; All the participants will undertake an AI-powered fundus camera screening, physician/optometrist consultation, survey with questionnaires and a phone interview three months after their baseline assessment.	Diagnostic test: AI-powered fundus camera screening; All the participants will undertake an AI-powered fundus camera screening, physician/optometrist consultation, survey with questionnaires and a phone interview three months after their baseline assessment.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Success Rate without Active Assistant Help	Success rate of using the DIY screening booth help will be determined by the number of successful screening runs without active assistant help divided by the total number of screening runs.	3 months

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Assessment of Consumer Acceptability	Assessment of consumer acceptability will be determined by the positive response rate/screening rate (i.e. the total number of eligible patients who agree to screening (numerator) divided by the total number of eligible individuals (denominator)).	3 months
Diagnostic Accuracy	All retinal images will be re-graded by two independent ophthalmologists, with any disagreements adjudicated by a third ophthalmologist (gold standard). This grading result will be considered as gold standard reference for the diagnostic accuracy of the automated screening. The indicators for diagnostic accuracy include: sensitivity.	3 months
Disease Detection Rates	The detection rate is defined as the proportion of newly diagnosed DR cases divided by the total number of eligible individuals.	3 months
Adherence to Referral	Patient adherence to referral will be measured as the proportion of patients who attend an ophthalmology service among the total number of referred patients.	3 months
Technical Feasibility: Quality of Image Acquisition	The investigators will use a standardized checklist to assess the quality of images acquired during the study. This checklist will include criteria resolution, clarity, and completeness of the images. A sample of images will be reviewed by a panel of experts to ensure consistency and reliability in the assessment.	3 months
Cost-effectiveness of DIY Screening	The cost-effective analyses will involve using the incremental cost-effectiveness ratio (ICER) as the key indicator to identify whether DIY screening model is a good investment.	3 months
Assessment of Consumer Satisfaction	The responses to the 5-point Likert scale question will be analysed using the "document variable statistics" function in MAXQDA software. Data from the open-ended questionnaire will be analysed thematically. All themed information will be shared for review by project steering committee. The Likert scale ranges from 1 to 5. The content includes 5 options, including strongly disagree, disagree, neural, agree, strongly agree. The higher scores (e.g., 4 or 5) indicate a stronger agreement or a more positive attitude towards the statement, while lower scores (e.g., 1 or 2) indicate disagreement or a more negative attitude.	3 months
Diagnostic Accuracy	All retinal images will be re-graded by two independent ophthalmologists, with any disagreements adjudicated by a third ophthalmologist (gold standard). This grading result will be considered as gold standard reference for the diagnostic accuracy of the automated screening. The indicators for diagnostic accuracy include: specificity.	3 months
Diagnostic Accuracy	All retinal images will be re-graded by two independent ophthalmologists, with any disagreements adjudicated by a third ophthalmologist (gold standard). This grading result will be considered as gold standard reference for the diagnostic accuracy of the automated screening. The indicators for diagnostic accuracy include: accuracy.	3 months
Diagnostic Accuracy	All retinal images will be re-graded by two independent ophthalmologists, with any disagreements adjudicated by a third ophthalmologist (gold standard). This grading result will be considered as gold standard reference for the diagnostic accuracy of the automated screening. The indicators for diagnostic accuracy include: positive/negative predictive values.	3 months
Diagnostic Accuracy	All retinal images will be re-graded by two independent ophthalmologists, with any disagreements adjudicated by a third ophthalmologist (gold standard). This grading result will be considered as gold standard reference for the diagnostic accuracy of the automated screening. The indicators for diagnostic accuracy include: area under curve.	3 months
Technical Feasibility:The duration of Clinic Flow	The investigators will conduct time-motion studies to evaluate the impact of the new imaging process on clinic flow. This will track the time taken for image acquisition, processing, and reporting and record the total time spent.	3 months
Technical Feasibility:Reporting and Impact on Clinic Flow	The investigators will involve track any delays or disruptions in the clinic schedule.	3 months
Technical Feasibility: The type of technical error	The investigators will maintain a log of technical errors encountered during the image acquisition process. This log will include the type of error.	3 months
Technical Feasibility: Number of Recorded Technical Errors	The investigators will maintain a log of technical errors encountered during the image acquisition process. This log will record the frequency of each type of technical errors.	3 months
Technical Feasibility: Remedies for technical errors	The investigators will maintain a log of technical errors encountered during the image acquisition process. This log will document any corrective actions taken in response to technical errors. Regular review meetings will be held to analyze these errors and implement strategies to minimize their occurrence.	3 months

Collaborators and Investigators

• Sponsor: The Hong Kong Polytechnic University

Study record dates

Study Major Dates

• Study Start (Actual): March 1, 2025
• Primary Completion (Estimated): June 30, 2025
• Study Completion (Estimated): July 31, 2025

Study Registration Dates

• First Submitted: February 7, 2025
• First Submitted That Met QC Criteria: March 18, 2025
• First Posted (Actual): March 25, 2025

Study Record Updates

• Last Update Posted (Actual): March 25, 2025
• Last Update Submitted That Met QC Criteria: March 18, 2025
• Last Verified: February 1, 2025

More Information

Terms related to this study

• Keywords: AI; diabetic retinopathy; screening system
• Additional Relevant MeSH Terms: Endocrine System Diseases; Vascular Diseases; Cardiovascular Diseases; Diabetes Mellitus; Eye Diseases; Diabetic Angiopathies; Diabetes Complications; Retinal Diseases; Diabetic Retinopathy
• Other Study ID Numbers: HSEARS20231218003

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No