Ocular microvascular complications in diabetic retinopathy: insights from machine learning
Authors' conclusion
The findings from this study provide valuable insights into the microvascular involvement of DM and DR, facilitating the development of early detection methods and intervention strategies.
Study context
- Population
- Eyes of healthy individuals and of patients with diabetes mellitus (DM) without diabetic retinopathy (DR) or with mild or moderate non-proliferative DR, imaged with optical coherence tomography angiography (OCTA)
- Exposure
- OCTA parameters across healthy, DM no DR, mild DR and moderate DR eyes
- Healthyn = 196 eyes
- DM no DRn = 120 eyes
- Mild DRn = 71 eyes
- Moderate DRn = 66 eyes
- Outcome
- Importance (SHAP values) of OCTA parameters in four classification tasks: DM no DR vs healthy, mild DR vs DM no DR, moderate DR vs mild DR, any DR vs no DR
Methodological detailsnon-parametric machine learning, SHapley Additive exPlanations
- Non-parametric machine learning
- SHapley Additive exPlanations
Main finding
As reported in the abstract
Importance of OCTA microvascular parameters in classification
- Large choriocapillaris flow deficits were most important for healthy vs DM no DR, and became less important in mild or moderate DR
- Superficial microvasculature important for healthy vs DM no DR and mild vs moderate DR, but not for DM no DR vs mild DR, where deep microvasculature plays an important role
- Foveal avascular zone metric generally less affected, but its involvement increased with worsening DR
Observational data: associations, not effects.
Original publication
Ocular microvascular complications in diabetic retinopathy: insights from machine learning. BMJ Open Diabetes Res Care. Published 2 January 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →