HealthScribe (a Clinical Note Generator using Generative AI)

A web application that allows users to generate clinical notes from transcribed ASR(Automatic Speech Recognition) data of conversations between doctors and patients using fine-tuned BART model (har1/HealthScribe-Clinical_Note_Generator)

Features

• User-friendly web interface for inputting conversational data obtained through ASR
• Tokenization of input text for processing by the BART model
• Fine-tuned BART model for generating clinical notes from medical conversations
• Presentation of the generated clinical note on the web interface

Project Structure

• dataset/ : Contains modified MTS-Dialog dataset
• model notebook/ : Contains 2 .ipynb notebooks (CliNo__version1 : used to fine-tune BART , CliNo__Pipe : Used to test the fine-tuned model metrics in comparision to base BART)
• venv/ : Virtual Environment
• test.txt : Contains test-case conversations
• templates/ : Contains HTML templates for the web interface.
• .gitignore : Specifies files to be ignored by Git.
• app.py : Main application file for running the web interface.
• requirements.txt : Lists dependencies required for the project.

Tech Stack

• Programming Language: Python
• Web Framework: Flask
• Deep Learning Libraries: Transformers (HuggingFace)
• Original Dataset : MTS-Dialog
• Modified Dataset : har1/MTS_Dialogue-Clinical_Note
• Base Model: facebook/bart-large-cnn
• Fine-Tuned Model: har1/HealthScribe-Clinical_Note_Generator
• Frontend: HTML, CSS
• Platform: Google Colab
• Version Control: Git

Installation

A step by step guide that will tell you how to get the development environment up and running.

First Clone the repo

$ git clone https://github.com/hari-krishnan-88/HealthScribe-Clinical_Note_Generator.git

Change directory to the repo directory

$ cd HealthScribe-Clinical_Note_Generator

Activate the Virtual Environment venv

$ activate

Run the Flask App app.py

$ flask --app app --debug run

Click on the URL (http://127.0.0.1:5000) to open the web interface on your browser

 * Serving Flask app 'app'
 * Debug mode: on
WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on http://127.0.0.1:5000
Press CTRL+C to quit
 * Restarting with stat
 * Debugger is active!
 * Debugger PIN: 127-629-777

Usage

This is how the initial web-interface looks like. A user can input the doctor-patient conversation in the text box and click on submit.

After Clicking the submit button the interface tokenizes the data and sends it to the HuggingFace Model and the output is processed.

Final Output is the Clinical note which is derived from the conversation and parameterised on the basis of :

• Symptoms :
• Diagnosis :
• History of Patient :
• Plan of Action :

Contributers

• Aleena Patani
• Amigashabnam F
• Harikrishnan KC
• Sujith J