Skip to main content

Model Installation

Prerequisites​

Ensure you have Python 3.8+ installed along with the necessary dependencies. Install them using:

pip install -r requirements.txt

Running the Project​

  1. Download ICD-11 Data

    python icddownloader.py

    This script fetches ICD-11 classifications from an API and stores the data in JSON and CSV format.

  2. Process ICD-11 Data

    python ICD_processing.py

    This script processes and cleans ICD-11 data, preparing it for use in the pipeline.

  3. Split DSM-5 Cases

    python dsmsplit.py

    Extracts and structures DSM-5 case data from a text file into a structured CSV.

  4. Generate Vector Embeddings

    python langchainbuilder.py

    Builds Chroma vector stores for efficient retrieval.

  5. Process Medical Cases with AI

    python app.py

    Runs the main pipeline to generate diagnoses for clinical cases.

  6. Expose LLM-based API

    python api.py

    Provides a Flask API endpoint (/askLLM) that allows querying an LLM.

File Descriptions​

api.py​

  • Implements a Flask-based REST API for querying an LLM using LangChain.
  • Retrieves data using Chroma and LangChain's RAG capabilities.
  • Accepts a JSON request with an input_string and returns the LLM response.

app.py​

  • Reads DSM-5 cases from CSV.
  • Uses Chroma and LLMs to generate diagnostic outputs.
  • Saves results in a structured format.

dsmsplit.py​

  • Extracts individual DSM-5 cases from a text file.
  • Parses cases into structured sections: Introduction, Discussion, Diagnosis.
  • Saves the structured data into a CSV file.

ICD_processing.py​

  • Reads ICD-11 classifications from a CSV.
  • Extracts ICD-11 disorder descriptions from a PDF.
  • Cleans and structures data for use in LangChain-based queries.

icddownloader.py​

  • Fetches ICD-11 classification data from an API.
  • Saves data in JSON and CSV formats for further processing.

langchainbuilder.py​

  • Creates vector embeddings using LangChain’s Chroma module.
  • Loads ICD-11 prompts for efficient retrieval-augmented generation (RAG).

API Usage​

To query the API, send a POST request to http://127.0.0.1:5000/askLLM with JSON data:

{
  "input_string": "Describe the symptoms of schizophrenia."
}

The response will contain the AI-generated diagnosis:

{
  "output_string": "Schizophrenia is characterized by..."
}

Acknowledgments​

This project leverages LangChain, Chroma, and various open medical datasets to improve automated clinical case processing and diagnostics.