AI Assignment 11: Disease Prediction Expert System

Name: Debabrata Mondal
Roll Number: 002111001124

Live Demo

URL: Web GUI
Source code: GitHub

Abstract

This project includes an implementation for an expert system for medical disease diagnosis incorporating the following methodologies:

Rule-based systems
Knowledge-based systems
Database methodology
Inference engines
System-user interaction

It includes a web GUI where users are presented with a set of symptoms to choose from. The user can optionally select more symptoms. A limited set of diseases are supported at this time. The knowledge base inside a Prolog program is used to find the matching set of diseases

Project Description

The project includes two separate resources:

Web application
Server

The web application serves as the GUI for user interaction. All the processing is done at the server which is hosted on Linode. The Nextjs web application is hosted on Vercel.

The server has access to a Prolog knowledge base and can execute queries against that. That is how it tries to match user-selected symptoms to diseases.

This project uses a bitmasking approach developed in collaboration with Aditya Mayukh Som to generate a dynamic set of questions based on user's previous selections. Each symptom is assigned a number based on the number of diseases where it prevails. The mask of a disease is computed as the bitwise OR of the masks of all the symptoms associated with that disease.

When the user selects their initial symptoms, these symptoms are converted into their respective masks and a new mask is created by computing the bitwise OR of all the selected symptoms. Bitwise AND operations are then performed with all diseases. If the result equals the newly created mask, additional questions about the other symptoms of that disease are presented. This approach creates a chain of questions that aids in precise disease prediction.

Tech Stack

Frontend: Nextjs, TypeScript

Backend: Python, FastAPI, Prolog

Attributions