LLM + CvT Analysis of Echocardiograms

Presentation demo of the specific algorithms in the project

Inspiration

My mom is a cardiac sonographer, and I noticed that she transcribes various values manually from an echocardiogram whenever she makes a scan. I thought to myself that if she messed up with that task, the patient's outcome would be wrong, and I decided to build a system that can cross-check her values with AI.

After building this part, I then went back into contemplation, and I noticed that the cardiologist at the clinic handled many cases at once. Due to this, I added a functionality to my app that uses AI to diagnose the patient. There is also the opportunity to perform a Vector Search, which gets all the cases similar in cardiac structure to the current patient. This offers a clinician as much data as possible to make a diagnosis.

How I built it

I built this project with ML in Python. I then developed a web-based application demonstrated in my video. Unfortunately, there isn't a public demo for this project, due to the massive server costs of running the CVT and LLM algorithms. In the future, I hope to get this down to make a free public access beta.