Artificial Intelligence Based Analysis of Corneal Confocal Microscopy Images for Diagnosing Peripheral Neuropathy: A...

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• Diabetic peripheral neuropathy (DPN) is a major global health concern, leading to significant morbidity and mortality.
• Early and accurate diagnosis of DPN is crucial for effective management and prevention of complications.

Purpose of the Study:

• To develop and evaluate a deep learning algorithm for classifying peripheral neuropathy (PN) in individuals with diabetes or pre-diabetes.
• To utilize corneal confocal microscopy (CCM) images of the sub-basal nerve plexus for PN detection.

Main Methods:

• A modified ResNet-50 deep learning model was trained for binary classification of PN (PN+) versus no PN (PN-).
• The algorithm was trained, validated, and tested on a dataset of 279 participants with diabetes or pre-diabetes using CCM images.
• Diagnostic performance was assessed using sensitivity, specificity, and area under the curve (AUC), with attribution methods (Grad-CAM) used for interpretability.

Main Results:

• The AI-based deep learning algorithm achieved high diagnostic performance for detecting PN.
• Sensitivity was 0.91 (95%CI: 0.79-1.0), specificity was 0.93 (95%CI: 0.83-1.0), and AUC was 0.95 (95%CI: 0.83-0.99).

Conclusions:

• The developed deep learning algorithm shows excellent diagnostic efficacy for PN using CCM imaging.
• Further large-scale prospective studies are needed to validate the algorithm's real-world diagnostic utility before clinical implementation.