Deep Learning-Based Detection of Early Renal Function Impairment Using Retinal Fundus Images: Model Development and...

Area of Science:

• Ophthalmology
• Nephrology
• Artificial Intelligence

Background:

• Deep learning models analyze retinal images for eye diseases and cardiovascular risks.
• Retinal microvascular and structural changes are linked to impaired kidney function.
• Detecting early renal function impairment via retinal images using deep learning is underexplored.

Purpose of the Study:

• Develop and evaluate a deep learning model for early renal function impairment detection.
• Utilize retinal fundus images as the primary data source.
• Assess model performance in identifying reduced estimated glomerular filtration rate (eGFR).

Main Methods:

• Retrospective study design using color fundus images and renal function tests.
• Deep learning model construction for renal function assessment.
• Definition of early renal function impairment as eGFR <90 mL/min/1.73 m².
• Performance evaluation using receiver operating characteristic curves and area under the curve (AUC).

Main Results:

• Analysis of 25,706 retinal images from 6212 patients.
• The deep learning model achieved an overall AUC of 0.81.
• AUCs improved in subgroups with elevated serum hemoglobin A1c (HbA1c) levels, reaching 0.87 for HbA1c >10%.

Conclusions:

• A deep learning model effectively detects early renal function impairment from retinal fundus images.
• Model accuracy is enhanced in patients with higher serum HbA1c levels.
• Retinal imaging offers a non-invasive approach for monitoring kidney health.