Circle | Devpost

Inspiration

Our inspiration comes from the intersection of human error and the bureaucratically heavy healthcare system that we live in which often leaves vital healthcare workers overworked and underprovided for. As a result Doctors often suffer from diagnosis fatigue, confirmation bias, and missing less apparent symptoms in the information overflow

We found our inspiration by exploring the mundane at first. Many of us spend at least an hour in the whole process of booking an appointment, traveling to the hospital and completing patient intake(which can take up to ~20 minutes--that's before the nurse). As a result, we wanted to explore efficiency and improving speed. However our further exploration and decisions truly amplified the effect that we saw--Our conversations with Phil gave us light into how doctors often are looking for a very specific criteria or tell-tale sign to make a diagnosis however that specific symptom may not be apparent. As a result, in the information overflow that a patient often provides, even if they have essential data--it's not communicated to the doctor.

What it does

Circle aims to reduce misdiagnosis and systemic fatigue by closing communication gaps from patient to doctor and doctor to patient and reducing the repetitive mind-numbing work that talented physicians are forced to do. It consists of a set of an AI assisted solutions which

Automate the scribe process that nurses play and integrate with smart intelligence and QAGeneration to pre-explore possible symptoms and ask critical follow up questions.
Analyze the transcript of patient intake to identify concerning symptoms, possible missed details, and summarize primary health essentially creating a quick and comprehensive EHR(solving a secondary issue of the lack of intercommunication between healthcare systems)
Offer interactive AI that can answer questions about the Patient Intake to the doctor.

How we built it

We used LLMs solutions with prompt engineering and integrating with knowledge sources to help create smart intelligence that analyzes document transcripts and user input in realtime to provide the most useful analysis. React and Material UI made up the frontend implementation whilst Flask API was used to createe a responsive server.

Challenges we ran into

We had issues with integration and time constraints given the 24 time period and having to learn new techniques on the spot.

Accomplishments that we're proud of

We are proud to have been able to build a full-circle solution in exactly 24 hours. This solution takes the problem and eliminates redundancies. We are also proud to have significantly improved communication between doctors and their patients by allowing them to communicate through a chatbot that not only remembers everything that the patient says, but is also able to comprehend and inform doctors about the patients that they have.