license: mit
DiabetesDeepInsight-CSV
A comprehensive, multi-source CSV collection for Type 2 Diabetes prediction, combining clinical indicators and retinopathy features. Ideal for researchers and practitioners in medical AI and data science.
🚀 Highlights
- Multi-Dataset Fusion: Integrates Pima Indians, BRFSS surveys, and Retinopathy Debrecen—over 300,000 records in total.
- Clinical & Retinopathy Features: Blood tests, demographics, lifestyle factors, and retinal image–derived biomarkers.
- Balanced & Stratified: Includes 50/50 splits, three-class→binary conversions, and curated train/test splits.
- Plug‐and‐Play CSVs: Ready for immediate ingestion with popular ML frameworks (scikit-learn, XGBoost, PyTorch).
📂 Included Files
diabetes.csv
Pima Indians Diabetes Database (8 clinical features + Outcome).diabetes_data_upload.csv
Alternate Pima format, ensuring consistency for cross-validation.diabetes_binary_health_indicators_BRFSS2015.csv
CDC BRFSS 2015 health survey (binary diabetes flag, demographics, labs).diabetes_binary_5050split_health_indicators_BRFSS2015.csv
Balanced 50/50 subset of BRFSS 2015 (equal cases/controls).diabetes_012_health_indicators_BRFSS2015.csv
BRFSS 2015 three-class (“No”, “Pre-diabetes”, “Diabetes”) converted to binary.Retinopathy_Debrecen.csv
Tabular features from EyePACS retinal exams (0/1 retinopathy → proxy for diabetes).diabetic_data.csv
(Optional) 130-US Hospitals clinical records—can be extended for readmission or ICD-9-based diabetes flags.
✨ Key Features
Rich Clinical Indicators
- Age, BMI, blood pressure, insulin, lipid profiles, lifestyle habits (smoking, activity), etc.
Retinopathy-Derived Biomarkers
- Vessel diameter, hemorrhage counts, texture features—ideal for image-to-CSV pipelines.
Preprocessed & Label-Aligned
- Unified
Outcomecolumn (0 = No Diabetes, 1 = Type 2 Diabetes) across all CSVs.
- Unified
Unlock deeper insights and achieve >95% accuracy with integrated clinical & retinopathy features! 🎉