Machine Learning to Diagnose Complications of Diabetes
Summary: A recent manuscript in Journal of Diabetes Science & Technology (2025) from Scheideman et al. surveys how machine learning (ML) is being used to detect and predict complications of diabetes—including foot ulcers, retinopathy, nephropathy, autonomic dysfunction, and neuropathy. The review highlights both existing successes and current limitations in bringing AI/ML into clinical workflows.
Key Highlights:
- Deep learning models are already in use for retinopathy screening through fundus images; future systems are pushing toward multimodal risk prediction and smartphone-based deployment.
- For diabetic foot ulcers, thermal imaging combined with convolutional neural networks is showing promise in earlier risk detection than conventional clinical examination.
- Wearables, ECG, and corneal imaging are being explored to detect peripheral and autonomic neuropathy earlier.
- Challenges remain: dataset diversity, labeling quality, external validation, explainability, and seamless integration into clinical care.
Read the full review on DiabeticFootOnline
Keywords:
machine learning,
diabetic foot ulcer,
retinopathy,
nephropathy,
wearable sensors,
thermal imaging,
Scheideman,
David G. Armstrong