To assess the independent association between a deep learning-derived retinal biomarker and DF-related amputation in patients with type 2 diabetes.
Approach:
Study Design: A retrospective observational study including 392 individuals with type 2 diabetes, split into training (70%) and validation (30%) sets.
Data Collection: Participants underwent fundus photography and had baseline HbA1c measurements; demographic and clinical variables were reviewed.
Model Evaluation: Model performance was assessed using AUC, cNRI, and IDI, with prespecified sensitivity and specificity thresholds.
Key Findings:
The retinal biomarker showed an incremental association with DF-related amputation (AUC 0.146, 95% CI: 0.046-0.249; cNRI 0.629, 95% CI: 0.184-1.027; IDI 0.062, 95% CI: 0.012-0.110).
The full model achieved an AUC of 0.791.
At a 27% amputation prevalence, sensitivity was 87.5% and specificity was 90.4%.
Interpretation:
A DL-derived retinal biomarker is associated with DF-related amputation risk beyond conventional diabetes variables.
Limitations:
The study is retrospective and conducted at a single center.
