{width="5.674239938757656in"
height="5.596415135608049in"}
Dr. GPT
AI Doctor
Team: Melted Ice cream
-
Praval Pattam
-
Krishna Harsha Mamillapalli
[Introduction 3](#introduction)
[The Problem 3](#the-problem)
[The Solution 3](#the-solution)
[User Guide 4](#user-guide)
[How are we different from our competitors? 6](#how-are-we-different-from-our-competitors)
[Technical Stack 6](#technical-stack)
[Deployment 6](#deployment)
[Challenges Faced: 7](#challenges-faced)
[Next Steps 7](#next-steps)
[Solving for India Hackathon 8](#solving-for-india-hackathon)
Greetings, this is Team Melted Ice cream's submission for the Google X AMD: Solving for India hackathon.
Our project is based on the theme Health-Tech called Dr. GPT, an AI Doctor
A lot of people avoid seeking medical attention, these can be due to various factors but according to a study conducted over one-third of participants (33.3% of 1,369) reported unfavourable evaluations of seeking medical care, such as factors related to physicians, health care organizations, and affective concerns.
These people look to online, friends and other options for a quick diagnosis and over the counter medication. Some of these people google their symptoms to get a basic diagnosis and over the counter medication. Unfortunately, this can be time consuming, can lead them down rabbit holes and often tend to develop Cyberchondria.
Dr. GPT is an AI diagnostic tool, that can provide a quick way to check one's symptoms and is the ideal solution for those who need detailed diagnosis, remedies to alleviate the discomfort and over the counter medication. Since Dr. GPT is on based on an LLM which is trained on a humongous data set that might even be greater than the knowledge possessed by the common man. This makes Dr. GPT better than asking friends and family for remedies and googling.
Dr. GPT takes a user's inputted data based on which it provides the user with their diagnosis in an efficient manner that consumes way less time than consulting a doctor (Although we urge you to consult a certified medical practitioner if symptoms persist). The diagnosis could be specific, or it could give possibilities of what the disease might be. It will also provide you with remedies and suggestions on what to do next. The application also gives handy tips regarding lifestyle changes which are a necessity to lead a healthy life.
- As the application opens, the user must fill in their details in the respective text boxes. These details include Name, height, weight, age.
{width="5.310605861767279in" height="0.8537215660542432in"}
It is mandatory to answer each field.
- The User is then asked to fill in Medical History such as previously diagnosed allergies, diseases by entering them into a text box which appears on selecting the 'Yes' option.
{width="5.348040244969379in" height="1.4972856517935258in"}
- After filling in the medical information, the user can choose one option amongst various possible ailments provided.
{width="4.367361111111111in" height="0.6916666666666667in"}
- On selecting a Broader Category of ailments, a drop-down appears with a list of symptoms going into specifics.
{width="3.9984853455818024in" height="1.3181813210848643in"}
- The user can choose the necessary symptoms and press the submit button.
{width="2.2618307086614173in" height="1.8626323272090988in"}
{width="2.414530839895013in" height="0.5301279527559055in"}
- Once the user submits their response, A dialogue box appears after clicking the submit button with the user's diagnosis.
{width="1.5769225721784776in" height="0.47164698162729657in"}
The average wait time is usually 2 minutes or 120 seconds. So, please be patient.
- Finally, we get the diagnosis in a simple and a very readable format (unlike a doctor's writing)
{width="5.687307524059492in" height="2.358707349081365in"}
Symptom Checker apps such as Ada, WebMD, and others store their data regarding diseases and their respective symptoms in large databases. These large data sets must be updated.
One of the places we put ourselves away and ahead of our competitors is by adopting new technologies like ChatGPT into our application. ChatGPT trained on large data sets diagnoses diseases and can also give remedies to keep the ailment under control. ChatGPT saved a lot of time and money that other apps have put into developing and maintaining their databases.
Symptom checker apps ask many questions, of which few might be unnecessary or totally out of context. We filtered out the required questions and condensed all questions regarding symptoms into a dropdown, thus speeding the diagnosis and saving the user's time.
+---------------------+------------------------------------------------+ | User Interface | Vue.js with Vuetify components | +=====================+================================================+ | API | Flask framework with Python | +---------------------+------------------------------------------------+ | AI/ML | ChatGPT | | | | | | Model: GPT 3.5 Turbo | +---------------------+------------------------------------------------+ | Web Server | Nginx | +---------------------+------------------------------------------------+ | OS | Linux | +---------------------+------------------------------------------------+
Deployed on Google Cloud and AMD instances (Specifically the T2D Series)
{width="3.5666721347331585in" height="3.6808716097987753in"}
The web application has been deployed on Google Compute Engine
Application URL; http://34.143.138.12/
and all the source code has been made available on https://github.com/PravalPattam/DrGPT
Below are the technical challenges faced by us during this project:
-
Getting ChatGPT api working as the OpenAI documentation is not detailed.
-
Learning curve with Google Cloud and Technical stack we are using.
The MVP can be extended to help users as a single stop for all the medical diagnosis. Below are the enhancements for Dr. GPT:
- Show Dr. GPT advise progressively.
Currently the application gets the entire response for GPT API and shows it all together. This makes user to wait for 1 to 2 mins. Instead of this we can extend the application to show the message in chunks. This is similar to typing the message (as done by ChatGPT)
- User personalization
Have a personal page for every user where they can review their current and past advice.
- Health news
Provide latest health news.
- Healthy Living
Provide advice to users for healthy living.
- Gamification
Help users to follow healthy practices through gamification.
- Geolocation
Shows the user the nearest medical specialist using geolocation and gps.
Below are the details for Hackathon
Project Title Dr. GPT, an AI Doctor
Project Description Dr. GPT is an AI diagnostic tool, that can provide a quick way to check one's symptoms and is the ideal solution for those who need detailed diagnosis, remedies to alleviate the discomfort and over the counter medication. Dr. GPT is better than googling or asking friends and family for remedies.
Theme Health-Tech
Google Cloud Web App deployed on Google Compute Engine
AMD instance CPU platform: AMD Milan
Machine type: T2D
Web App URL http://34.143.138.12/
Github URL https://github.com/PravalPattam/DrGPT
Google Drive Link for the https://drive.google.com/drive/folders/1NLMH_lCS7g-_JLYPHu-MXDMn9Bd8NdvK?usp=sharing
Demo video and
Documentation