SimpleHealth

Inspiration

Medical doctors have a difficult time finding the right patients for clinical trials. A primary driver for this problem is not having enough patients, but rather finding the right patients through noisy unstructured patient data.

With recent advancements in creating "clinically-aware" LLMs, a software for doctors to find the signal from the noise of unstructured patient data for trial recruitment seemed compelling.

What it does

Our platform allows doctors to input specific requirements for their clinical trial and allows doctors to find the appropriate patients. Next, doctors can actually further investigate leads generated from the system through directly contacting patients. If a patient seems fit, the doctor can update their system by accepting or rejecting patients. The LLM algorithm dynamically approves based on manual annotations set by patients.

How we built it

We used BeautifulSoup4 to web-scrape Mayo Clinic for information regarding each disease in it's database. We then use this information to train GPT-4.0 to classify a patient given a string of unstructured patient notes. Front-end was built with React and connected to the backend using FastAPI.

Built With

• fastapi