MediScribe.ai

Transcription of an example conversation

Inspiration

MediScribe.ai is an initiative to bridge the gap between doctors, patients, insurance providers, and pharmacies through one central platform. It provides patients with a transcription of medical appointments to ensure they understand medical terminology, can refer to an AI-generated summary of their visit, and understand the next steps in their health journey. By promoting transparency between providers and patients, MediScribe.ai ensures patients receive the highest quality of healthcare.

What it does

Our program uses Deepgram to transcribe conversations between doctors and patients during appointments to address the disconnect and lack of transparency often experienced by patients in healthcare. It then uses the GPT-4 API to provide the patient with an organized summary of the appointment along with future action items.

How we built it

First, we used Deepgram to transcribe audio recorded in real time simulating conversations between patients and doctors. The transcription documents the patient's personalized care treatment and is then summarized in concise, understandable terminology through an API call to ChatGPT-4. In order to see that our system is recording the conversation we have included a real-time transcript on our web interface.

Challenges we ran into

Since we don't have very much frontend experience, it was difficult for us to troubleshoot errors when trying to generate waveform analysis for recorded audio and design our website. Another challenge we encountered was that Deepgram would record sentences but due to differences in the discrete time representation of the audio and real time transcript, we do often see words spoken into the microphone repeated. Although this didn't affect ChatGPT-4's summary, it makes the transcript more clunky and hard to read.

Accomplishments that we're proud of

We're proud that we successfully used Deepgram to transcribe conversations accurately in real-time. We also successfully integrated ChatGPT-4 and generated accurate summaries of conversations. Lastly, we have an intuitive front-end design for our program in which users can access our transcription services, our documentation, and our GitHub.

What we learned

We are proud to have learned JavaScript and improved on our frontend skills, which none of us had much experience in. In addition, we learned how to integrate APIs into our application for a seamless experience for users.

What's next for MediScribe.ai

In the future, we hope to expand MediScribe.ai into a centralized interface for doctors, pharmacies, and insurance providers to bridge the disconnect between medical providers and the patient. In addition, we hope to use the Google Calendar API to add health appointments to the patient's calendar. We also hope to convert our program into an app that is accessible to all people enabling people to focus on their health.