Inspiration
Personal experiences with current public healthcare crisis - we believe this triage system is the solution to this issue.
What it does
Automated triage system that analyses patient conditions based on healthcare data including x-rays, ECGs, Blood tests... alongside contextual data from real healthcare professionals. It will significantly speed up the process of keeping track of patient data.
How we built it
We utilized Python (Flask) for our backend which connected to a MongoDB instance hosted on an AWS server - we also made a JavaScript (Next.js) dashboard where the doctors can access all their patient's contextual data.
Challenges we ran into
Inconsistencies in AI results - we focused on making this application as trustworthy as possible by making the model compare its hypothesis with Brave Search API images to confirm that what it's looking for resembles the original image. We also lowered the Top-P parameter to reduce the randomness of the model.
Accomplishments that we're proud of
We believe we tailored our UI for ease of use - specifically within the healthcare industry, with features such as Context Notes and self fact-checking, we ensure that we deliver an ethical yet productive system.
What we learned
We learned a lot about the ethics in the healthcare industry and the importance of reliable prompt engineering and prompt results self-assessment. We believe there is a future where AI will have to keep checking itself to rid of hallucinations.
What's next for Sana+
The system will be expanded for open-source use and self-deployment alongside the development of a system where third-party medical analytics can be added on. We also want to expand the context awareness of our model and are looking for ways to do this.
Log in or sign up for Devpost to join the conversation.