Healthnet - TigerGraph based Patient Triage Bot

Architecture
Graph schema
Graph database schema
Graph database with medical data loaded
Data mapping done on tiger graph
Medical dataset used with ICD10 codes
Medical dataset used with ICD10 codes
Demo chat transcript
Chatbot configuration steps

Inspiration

Healthnet is a medical chat bot useful for patient triaging and preliminary diagnosis of various medical conditions. It is built on an medical terminology Tigergraph database with several analysis queries. This can also be used to build a huge knowledge graph with lab results etc to help doctors arrive at better diagnosis. It can be used for the following

Better triaging patients based on symptoms and lab results
Diagnosing related diseases and disorders
Referring to specialist doctors based on diagnosis
Presenting precautions
Can be integrated into websites, apps, medical systems etc

More information here - https://docs.google.com/presentation/d/1oBOee-Mcj2RT0FvpesCgmplhn2_-UkAgv3ZQsFHlsx0/edit?usp=sharing

Architechture

healthnet

Goal

The goal is to build a bot that can help with patient triage and preliminary medical diagnosis. It can be used as a intelligent symptom checker or a conversational tool to understand the patients needs better.

Data used

We have used aggregated medical sources about various

Diseases and conditions with ICD10 codes
Symptoms associated with each condition
Severity of conditions
Specialist doctors to be referred for each disease/ condition
Detailed description of diseases, symptoms and diagnosis
Triage severity

We have combined all these data to form a medical triage graph database. We have also written several queries to fetch symptoms related to disease , diseases associated with a particular symptom, what specialist to refer to, ideal next course of triage action etc

Tech Stack

TigerGraph cloud database - we have built a complete medical knowledge graph with various graph analysis queries
RASA chat bot - chat bot which is trained on the graph database model . Has several intents and actions related to medical triage workflow.
The Chatbot is hosted on a simple python HTTP server

How we built it

Medical data from various sources were aggregated from various sources. GraphDatabase was created and populated using pyTigerGraph
All the CSV data files were loaded using various load jobs and then analysis queries were writtern The results of these analysis queries are then accessed real time by the Chatbot using a REST API

Who is this for

This patient triaging chat bot will be a great resource for

all healthcare professional for making better informed decisions about their patients
patients who want to understand better about their symptoms and first line precautions , severity and specialist doctors to meet etc
Hospitals who are looking for smarter/better ways to triage incoming patients
Online medical portals who want to better understand their patients
This also helps anybody stay upto date with precautions for various health disorders

It helps make sense of the ever growing medical information available online and converts them into valuable , easily accessible knowledge graph

What's next for Healthnet

Add data from medical tests and lab results to give personalized diagnosis to patients
Add additional intents and entities to the chatbot to make conversation engaging and informative