MedTech — Cancer QoL Prediction
ML EngineerImpact
- Built ML pipeline predicting Quality of Life (SF-36) in Head & Neck Cancer patients with R² = 0.88 using XGBoost
- Engineered 142 features from 226 patient records combining clinical TNM staging with Ayurvedic Prakriti analysis
- Designed feature engineering pipeline parsing TNM staging, one-hot encoding cancer sites, and encoding Vata/Pitta/Kapha body types
- Conducted 3-model comparison (LinearRegression, RandomForest, XGBoost) with comprehensive evaluation and feature importance analysis
Skills
Project Abstract & Details
A machine learning project predicting Quality of Life (SF-36 scores) in Head & Neck Cancer patients by combining clinical/demographic data with Ayurvedic Prakriti (Vata/Pitta/Kapha) body constitution assessment. Engineered 142 features from 226 patient records including TNM staging, cancer site encoding, and 36 SF-36 quality-of-life indicators. Achieved R² = 0.88 using XGBoost regression with RandomForest baseline.
Key Features
XGBoost Regression
R² = 0.88 predicting 8 SF-36 quality-of-life domains — outperforms RandomForest (R²=0.78) and LinearRegression (R²=0.62)
142-Feature Engineering
Engineered from 226 patient records combining TNM staging, cancer sites, demographics, and Ayurvedic body types
Ayurvedic Integration
Vata/Pitta/Kapha Prakriti body constitution encoded as categorical features alongside clinical indicators
3-Model Comparison
Systematic evaluation of LinearRegression, RandomForest, and XGBoost with hyperparameter tuning and feature importance analysis
SF-36 Domain Analysis
Predicting all 8 quality-of-life domains: Physical Functioning, Role Physical, Bodily Pain, General Health, Vitality, Social Functioning, Role Emotional, Mental Health
Clinical ML Pipeline
End-to-end from raw patient questionnaires to actionable QoL predictions with feature importance for clinical interpretability
Architecture
Data pipeline loads 226 patient records with clinical (TNM staging, cancer site, age, gender), demographic, and Ayurvedic Prakriti (Vata/Pitta/Kapha) assessments. Feature engineering parses TNM staging (T1-T4, N0-N3, M0-M1) into ordinal features, one-hot encodes cancer sites (oral cavity, oropharynx, larynx, hypopharynx), and encodes Prakriti types. Target variables are 8 SF-36 domain scores. Three regression models compared: LinearRegression (baseline), RandomForest (100 estimators), and XGBoost (n_estimators=200, max_depth=6, learning_rate=0.1) with 5-fold cross-validation.
Architecture diagram placeholder