AI-Assisted Decision Support in Clinical Oncology

An educational demonstration of how machine learning can support clinical decision-making in oncology using transparent and interpretable models.

Overview

This project demonstrates an end-to-end workflow for AI-assisted clinical decision support:

1. Narrative to Structured Data - Converting unstructured clinical text into structured variables
2. Predictive Modeling - Building interpretable ML models for:
   • 2-year cancer recurrence risk (Random Forest)
   • Treatment toxicity risk (Logistic Regression)
3. Explainability - Using SHAP values to explain predictions at both global and patient levels
4. Clinical Integration - Presenting results in clinician-friendly formats

Prerequisites

• Python 3.12 or higher

Installation

# Clone the repository
git clone <repository-url>
cd cas_demo

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Running the Notebook

jupyter lab

Then open AI_Clinical_Oncology_Demo.ipynb in the JupyterLab interface.

Project Structure

cas_demo/
├── AI_Clinical_Oncology_Demo.ipynb   # Main demonstration notebook
├── data/
│   └── lung_cancer_dataset_toxicity_survival_interaction.csv  # Synthetic dataset
├── pyproject.toml                    # Project configuration
├── requirements.txt                  # Python dependencies
└── README.md                         # This file

Dependencies

Package	Purpose
jupyterlab	Interactive notebook environment
pandas	Data manipulation and analysis
scikit-learn	Machine learning models
shap	Model explainability
matplotlib	Visualizations
seaborn	Statistical visualizations

Dataset

The project uses a synthetic dataset of ~10,000 lung cancer patients with the following characteristics:

• 24 variables including demographics, tumor characteristics, molecular markers, and outcomes
• Outcomes: 2-year recurrence, 5-year survival, treatment toxicity
• Educational purposes only - does not contain real patient information

Disclaimer

This is an educational demonstration for decision-support exploration only. All predictions should be interpreted by qualified healthcare professionals in the context of the individual patient. The models are trained on synthetic data and should not be used for actual clinical decisions.

License

This project is for educational purposes.