📅 Day 1: Digital Twin Foundations & Healthcare Data
Explain the difference between a digital model, a digital shadow and a digital twin.
Introduce the physical, data, virtual and intelligence layers of a healthcare Digital Twin.
Review patient, organ and disease twins, with examples from chronic disease care.
Explore one prepared dataset and identify its patient features, outcome and limitations.
Discuss how clinical, physiological and wearable data could be combined; use a small prepared example rather than attempting to integrate multiple complex datasets live.
Introduce privacy, representativeness and data-quality considerations.
🛠️ Hands-on
Use Google Colab to inspect a prepared synthetic or openly accessible healthcare dataset.
Create a basic patient-feature table and visualize selected variables.
Sketch a Digital Twin workflow for one defined use case, such as diabetes or cardiovascular risk monitoring.
Hands-on Tools: Google Colab, Python, pandas, matplotlib and a prepared Synthea sample or teaching dataset. Synthea’s generated records are synthetic, not real patient records.
Practical Output: A data profile and a one-page conceptual Digital Twin workflow.
📅 Day 2: Predictive Modeling & Explainable AI
Prepare a small dataset for one prediction task, such as risk classification or outcome estimation.
Explain training and test data, baseline models and why data leakage can distort results.
Build and evaluate one prepared scikit-learn model; focus on interpretation rather than comparing many algorithms.
Review suitable evaluation measures, such as recall, precision and ROC-AUC, and discuss their limits.
Introduce SHAP feature contributions to show how a model produced a prediction.
Emphasize that model output is educational and does not provide a clinical diagnosis.
🛠️ Hands-on
Run a prepared Colab notebook to train a baseline model on the selected dataset.
Review its evaluation results and generate a feature-importance or SHAP visualization.
Discuss which factors influence the model output and what the result cannot establish.
Hands-on Tools: Google Colab, Python, pandas, scikit-learn, matplotlib and SHAP. SHAP provides tools for explaining model predictions.
Practical Output: A baseline model, evaluation summary and interpretable prediction plot.
📅 Day 3: Mini Digital Twin Prototype & Responsible Use
Connect the dataset, prediction model and visualization into a small educational prototype.
Show how a change in selected input features changes a model estimate; explain that this is a model response, not a treatment recommendation.
Discuss monitoring over time and what additional data would be needed for a more realistic patient model.
Introduce interoperability, privacy, security and bias as implementation challenges.
Briefly review emerging topics such as multimodal data, generative AI and federated learning as future directions rather than hands-on build goals.
🛠️ Hands-on
Use a prepared Colab notebook with simple widgets to adjust selected inputs and view the resulting model output.
Create a basic patient-level visualization or dashboard.
Complete a short interpretation template covering the output, limitations, possible bias and human review needed.
Hands-on Tools: Google Colab, Python, pandas, matplotlib and ipywidgets or Plotly.
Practical Output: A small interactive prototype and a concise, responsible-use summary.
