Inspiration: Medical information is often scattered across doctor’s visits, test results, medications and personal notes. This can make it difficult for patients to remember their own health history and difficult for healthcare providers to reconstruct it quickly.
HealthThread was built on a simple premise: what if you could take a patient’s health history and organize it into one clear, chronological story?
How it works: HealthThread is an AI-assisted tool that extracts key information like symptoms, medications, tests, diagnoses, treatments, and dates from fragmented health information. It then puts together these events into a chronological timeline for the patient to view, amend and verify before creating a brief summary for a healthcare provider.
It is not intended to be a substitute for diagnosis, or to recommend treatment, but to facilitate understanding, management and communication of existing health information.
How it was built: The proposed prototype is built on a simple web-based architecture, with frontend developed in HTML, CSS, and JavaScript, backend in Python and Flask, SQLite for basic storage, and an LLM API for extracting structured health events from text.
What I learned: The project demonstrated the importance of patient engagement when using AI with sensitive health data. Rather than accepting an AI-generated summary as-is, HealthThread adds in a patient verification step between the AI extraction and the final clinician summary.
Challenges: The main challenges are dealing with incomplete or contradictory medical information, preventing mistakes in extracted dates or medical information, protecting sensitive health information, and making sure the tool is clearly understood as an organizational tool and not a medical decision-making tool.
Log in or sign up for Devpost to join the conversation.