Inspiration
We were inspired by the deep communication gap that exists between people with progressive neurological conditions and their doctors. We understood that the typical model of occasional, 15-minute visits is not representative of the everyday experience of having a condition such as Parkinson's or Multiple Sclerosis. Patients' self-reports of data are frequently incomplete or subjective because of memory problems and the natural variability of symptoms. We set out to design a solution that would be able to serve as a continuous, objective data stream, allowing patients to better describe their day-to-day experiences and supporting providers in being able to provide truly personalized care.
What it does
MycoMind is an exhaustive, interactive tool that generates a live "digital twin" of a patient's neurological well-being. The initiative is comprised of two principal elements: The Patient App: An easy-to-use smartphone app that simplifies data collection and makes it fun. It includes a gamified symptom log, brief cognitive games to objectively score mental and motor function, and passive synchronization from wearable devices to monitor measurements such as sleep and activity. The Provider Dashboard: An elegant web-based dashboard that combines the patient's information into actionable AI-driven summaries. The system detects key trends and anomalies that would otherwise go unnoticed, enabling providers to have more directed and effective interactions with their patients during consultations.
How we built it
MycoMind was developed as a full-stack web application. The patient app and provider dashboard front-ends were built with React, which enabled us to build a dynamic and responsive user experience. Tailwind CSS was employed for quick and repetitive styling. The backend is based on a serverless architecture with Firebase, which offers a Firestore database for synchronization and real-time data storage. This was key to having provider dashboards refresh in real-time as patients enter new information. The foundation of our project, the AI analysis, was developed utilizing the Gemini API. We created a system to input patient data into the model, asking it to look for patterns and create the brief summaries for providers.
Challenges we ran into
One of our biggest challenges was translating subjective patient experiences into quantifiable data that the AI could effectively analyze. A user's description of a "bad day with brain fog" is very different from a numerical score. We tackled this by designing a hybrid system that combines qualitative input (e.g., voice-to-text logging) with objective metrics from our cognitive mini-games. Another challenge was designing a clean and intuitive UI/UX that would be convenient for patients who may have fine motor control or memory impairments. We iterated our design several times, reducing the workflow to a few taps and employing large, legible graphical objects.
Accomplishments that we're proud of
Most of what we are most proud of was successfully developing a functional proof-of-concept for the AI analysis engine. We were able to show that the Gemini API can take a stream of real-world, heterogeneous data and create a real insight, human-readable summary that could save a physician hours of time. We're also happy with the clean and easy-to-use user interface we implemented. Building an app that isn't simply working but actually useful and usable by its intended audience seemed like a big success.
What we learned
This hackathon educated us on the vast potential of AI as a bridge in healthcare, and not a substitute for human touch. We learned that machine learning's real strength lies in enhancing human capabilities, in this instance enabling doctors to get a better picture of a patient's life in between visits. We also got some first-hand experience with serverless architectures and real-time data syncing, an important piece of any scalable healthcare application.
What's next for MycoMind
We'll next widen the scope of MycoMind. We intend to: Add modules for more neurological disorders. Integrate with additinal wearable devices to create a more comprehensive data set. Create a secure messaging component to allow direct patient-provider communication. Investigate more advanced predictive analytics to notify providers of possible health emergencies prior to their onset, such as an impending symptom flare or a danger of hospitalization.
Log in or sign up for Devpost to join the conversation.