Corteχ

Corteχ (Core-tech) is designed to streamline patient selection for clinical trials by processing complex clinical trial criteria and matching patients based on their data. This application leverages natural language processing and a rules engine to assess patient suitability, providing scores and exclusion reasons based on specified inclusion and exclusion criteria.

Corteχ was designed in 24 hours for the Nucleate Pittburgh 2024 Biohackathon.

Features

• Natural Language Rule Parsing: Input clinical trial criteria in natural language, and the app translates it into structured rules that the engine can understand.
• Patient Scoring and Exclusion: Each patient is assigned a score based on how well they meet the inclusion criteria, with mandatory exclusion rules applied as needed.
• Logs and Records: Exclusion reasons and scores are saved in a log folder for reference.
• Customizable Criteria: Supports age, gender, medication, and preexisting condition criteria, with a fallback option for unsupported conditions.

Prerequisites

• Python 3.7+
• Required Python libraries:
  • pandas
  • customtkinter
  • tkinter
  • anthropic (for rule building through Claude API)
  • PIL (for image handling)
• Anthropic API Key: Store your key in a secret.txt file in the project root.

Installation

1. Clone this repository:

git clone https://github.com/yourusername/clinical-trial-patient-selector.git
cd clinical-trial-patient-selector

2. Install dependencies:

pip install -r requirements.txt

3. Add your Anthropic API Key:

• Place your API key in a file named secret.txt in the project root:

YOUR_ANTHROPIC_API_KEY

Usage

1. Start the Application: Run the main file:

python main.py

2. Select Patient Data File:

• Click "Select Patient CSV File" and choose a CSV file containing patient data.
• Note: The CSV file should contain columns including age, gender, prescriptions, and icd9_codes.

3. Enter Clinical Trial Criteria:

• Describe the inclusion and exclusion criteria for the trial in natural language. For example:

Patients aged 58-70 with no history of heart disease, who are currently not on insulin.

4. Run the Selection:

• Click "Run" to start the patient evaluation.
• The output will display patients who meet the criteria, sorted by their suitability score.

5. Review Results:

• Logs are saved in the /logs folder, including exclusion reasons and patient scores.
• The application will show the percentage of patients who met the criteria.

File Structure

clinical-trial-patient-selector/
│
├── main.py                # Main application file
├── secret.txt             # API key file for Anthropic
├── requirements.txt       # Required Python packages
├── logs/                  # Folder for exclusion reasons and patient scores
├── images/
│   └── cortex_logo.png    # Logo for the application
└── README.md              # Project documentation

Example Patient CSV Format

Your CSV file should contain the following columns for accurate parsing:

subject_id	first_name	last_name	age	gender	prescriptions	icd9_codes
1234	John	Doe	65	M	["insulin", "calcium gluconate"]	["401.9", "250.00"]

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• Anthropic API for rule building.
• CustomTkinter for the modern Tkinter widgets.
• All contributors who helped with testing and development. (Speicifically, Myra Haider, Liv Toft, Juhi Gupta, and Cleo Chen)

Contact

For questions or suggestions, please open an issue or reach out to the repository maintainer, @Jonpot.