Related Experiment Video
Updated: Jul 22, 2026
05:17
A Reproducible Computerized Method for Quantitation of Capillary Density using Nailfold Capillaroscopy
Published on: October 27, 2015
9.0K
Nailfold capillaroscopy and deep learning in diabetes.
Reema Shah1, Jeremy Petch1,2,3,4, Walter Nelson2,5
1Population Health Research Institute, McMaster University and Hamilton Health Sciences, Hamilton, Ontario, Canada.
Journal of Diabetes
|January 15, 2023
Summary
Nailfold capillary images analyzed with deep learning can help diagnose diabetes and predict cardiovascular events. This technology shows promise for identifying diabetes complications through non-invasive imaging.
Area of Science:
- Ophthalmology
- Cardiology
- Endocrinology
- Medical Imaging
- Artificial Intelligence
Background:
- Diabetes mellitus is a global health concern associated with microvascular and macrovascular complications.
- Nailfold capillaroscopy is a non-invasive technique to visualize microcirculation.
- Early detection of diabetes and its complications is crucial for effective management.
Purpose of the Study:
- To evaluate the diagnostic utility of nailfold capillary images for diabetes detection.
- To assess the potential of machine learning models in predicting diabetes-related complications using these images.
Main Methods:
- Nailfold video capillaroscopy was performed on 120 adult patients.
- Convolutional neural networks (deep learning) were employed to analyze 5236 nailfold images.
- Models were trained and validated to predict diabetes, high glycosylated hemoglobin, cardiovascular events, retinopathy, albuminuria, and hypertension.
Main Results:
- Machine learning models accurately identified diabetes with an AUROC of 0.84.
- Models also predicted a history of cardiovascular events in diabetic patients (AUROC 0.65).
- The study demonstrated high precision recall for diabetes detection (AUPR 0.84).
Conclusions:
- Nailfold imaging combined with machine learning shows potential for diagnosing diabetes.
- This approach may aid in identifying individuals at higher risk for diabetes-related complications.
- This proof-of-concept study highlights a novel, non-invasive diagnostic tool for diabetes management.
