Artificial Intelligence for Detecting Retinal Diseases

Completed

• Conditions: Retinal Diseases; Artificial Intelligence
• Interventions: Diagnostic Test: Retinal diseases diagnosed by artificial intelligence algorithm

• Registration Number: NCT04678375
• Lead Sponsor: Beijing Tongren Hospital

Brief Summary

The objective of this study is to apply an artificial intelligence algorithm to diagnose multi retinal diseases from fundus photography. The effectiveness and accuracy of this algorithm was evaluated by sensitivity, specificity, positive predictive value, negative predictive value, and area under curve.

Detailed Description

The objective of this study is to apply an artificial intelligence algorithm to diagnose referral diabetes retinopathy, referral age-related macular degeneration, referral possible glaucoma, pathological myopia, retinal vein occlusion, macular hole, macular epiretinal membrane, hypertensive retinopathy, myelinated fibers, retinitis pigmentosa and other retinal lesions from fundus photography. The effectiveness and accuracy of this algorithm was evaluated by sensitivity, specificity, positive predictive value, negative predictive value, area under curve, and F1 score.

Study Timeline

• Status Verified: 2018-06
• Study Start: 2018-06-01
• Primary Completion: 2020-06-30
• Study Completion: 2020-10-01
• Study First Submitted: 2020-12-16
• Study First Submitted QC: 2020-12-16
• Study First Posted: 2020-12-21
• Last Update Submitted: 2021-04-12
• Last Update Posted: 2021-04-15

Recruitment & Eligibility

• Status: COMPLETED
• Sex: All
• Target Recruitment: 1000000

Inclusion Criteria

• fundus photography around 45° field which covers optic disc and macula
• complete identification information

Exclusion Criteria

• insufficient information for diagnosis.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
Retinal diseases diagnosed by artificial intelligence algorithm	Retinal diseases diagnosed by artificial intelligence algorithm	Retinal diseases diagnosed by artificial intelligence algorithm

Primary Outcome Measures

Name	Time	Method
F1 score	1 week	We used F1 score to examine the ability of this artificial intelligence algorism recognition and classification of retinal diseases.
Area under curve	1 week	We used the receiver operating characteristic (ROC) curve and area under curve to examine the ability of this artificial intelligence algorism recognition and classification of retinal diseases.
Sensitivity and specificity	1 week	We used sensitivity and specificity to examine the ability of this artificial intelligence algorism recognition and classification of retinal diseases.
Positive predictive value, negative predictive value	1 week	We used positive predictive value and negative predictive value to examine the ability of this artificial intelligence algorism recognition and classification of retinal diseases.

Secondary Outcome Measures

Name	Time	Method
Systemic biomarkers and diseases	1 week	Using medical records as the gold standard, we test the accuracy of this artificial intelligence algorism recognition and classification of systemic biomarkers and diseases: age, sex, blood pressure, blood hemoglobin, cardiovascular diseases, thyroid function and kidney function.

Trial Locations

Locations (1)

Wen-Bin Wei; 🇨🇳 Beijing, Beijing, China