Machine Learning Prediction of Cardiovascular Risk in Type 1 Diabetes Mellitus Using Radiomic Features from...

Area of Science:

• Ophthalmology and Medical Imaging
• Cardiovascular Risk Assessment
• Machine Learning in Healthcare

Background:

• Type 1 diabetes mellitus (T1DM) is associated with an increased risk of cardiovascular (CV) disease.
• Early and accurate CV risk stratification is crucial for timely intervention in T1DM patients.
• Retinal imaging offers a non-invasive window into systemic vascular health.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) algorithm for determining CV risk levels in T1DM patients.
• To utilize multimodal retinal images (color fundus photographs, OCT, OCTA) for CV risk assessment.
• To distinguish between moderate, high, and very high CV risk categories.

Main Methods:

• Cross-sectional analysis of retinal images from T1DM patients.
• Extraction of radiomic features from color fundus photographs, OCT, and OCTA.
• Training ML models using radiomic features alone or combined with clinical data (demographics, systemic, ocular, and blood data).

Main Results:

• Radiomic features alone achieved AUCs of 0.79 for moderate risk and 0.73 for high/very high risk discrimination.
• Combining radiomic features with clinical data improved AUCs to 0.99 for moderate risk and 0.95 for high/very high risk.
• OCT/OCTA metrics with ocular data achieved an AUC of 0.89 for very high CV risk without systemic data.

Conclusions:

• Retinal radiomic features effectively discriminate and classify CV risk categories in T1DM.
• Integration of clinical data significantly enhances the accuracy of CV risk stratification.
• This oculomics approach shows promise for non-invasive cardiovascular risk assessment in T1DM.