BP703T · Core · Semester 7
AI in Clinical Applications
Learn to frame pharmacokinetics, drug-safety, personalised-medicine, healthcare-data, and supervised-learning concepts through transparent fictional clinical-AI teaching simulations—not decision support.
Course objectives
- Introduce AI applications in pharmacology, pharmacokinetics, and drug safety.
- Enable supervised-ML concepts for clinical and pharmacovigilance teaching datasets.
- Develop interpretation skills for predictive-modeling concepts in healthcare.
- Build understanding of real-world healthcare-data analytics concepts.
- Promote responsible and ethical AI use in patient-care settings.
Unit 1: AI in Pharmacokinetics & Dose Optimization
Pharmacokinetic teaching models make Cmax, Tmax, AUC, clearance, concentration-time relationships, regression coefficients, RMSE, and R² assumptions visible. A fictional curve is not a patient model, a dose adjustment, or clinical decision support; linear patterns also cannot substitute for nonlinear-pharmacokinetic reasoning.
Compare fictional oral, IV, and slow-release curves, identify one coefficient or error metric, and list two assumptions that prevent patient-specific use.
Unit 2: AI in Drug Safety & Pharmacovigilance
Drug-safety teaching begins with pharmacovigilance systems, fictional FAERS/EudraVigilance-style fields, frequency and signal-detection concepts, logistic-regression ADR-risk vocabulary, a confusion matrix, and bias/confounding. A teaching signal is a prompt for inspection, not a causal finding, reporting decision, or patient-risk prediction.
Change fictional report quality or threshold language, identify a sensitivity/specificity trade-off, and explain why human pharmacovigilance assessment remains necessary.
Unit 3: AI in Personalized Medicine & Risk Stratification
Precision-medicine concepts can describe patient covariates, therapeutic response, logistic-regression risk language, responder/non-responder classification, evaluation metrics, and ethics. They do not turn a fictional dataset into individualised treatment, disease-risk prediction, or a clinical recommendation.
Explore fictional covariates, describe a responder/non-responder teaching pattern, and state why the output cannot select treatment or assess a person's risk.
Unit 4: AI in Clinical Decision Support & Real-World Data
EHR structure, clinical-decision-support vocabulary, regression outcome concepts, classification risk-scoring language, real-world-data coverage, accountability, interpretability, and clinical-environment limits can be studied with fictional records only. This activity is not a CDSS and provides no risk score, clinical decision, or health-data processing.
Reduce a fictional representation value, identify one EHR or real-world-data limitation, and explain why accountability and interpretation must precede any deployment discussion.
Unit 5: Guided Supervised Learning Project — Pharmaceutical and Clinical Applications
A guided supervised-learning project frames a relevant problem, public-data suitability, predictors and outcomes, preprocessing, linear/logistic-regression concepts, evaluation metrics, interpretation, bias, ethics, and reporting. A browser planning canvas does not select or process a real dataset, train a clinical model, or support deployment.
Use the project canvas to identify a fictional question, predictor/outcome boundary, evaluation plan, limitation, ethical consideration, and no-deployment statement.