HealthME

We reduce mental tax for surgeons pre-operation

Comment

The Problem That Sparked Our Idea

When paramedics arrive on a scene and find an unresponsive patient, they are essentially flying blind. A sudden drop in oxygen or a flatlining monitor raises immediate questions: Does the patient have diabetes? Are they allergic to epinephrine? Do they have a pacemaker? Unfortunately, while local hospitals possess this critical medical history, complex HIPAA regulations and fragmented, state-by-state Health Information Exchange (HIE) policies keep this data locked away from EMS crews. We built this platform to bridge that gap, giving first responders immediate, secure access to life-saving patient context without violating data privacy laws.

The Core Solution

Think of our product as an intelligent compliance bridge. It interfaces with standard EMR systems, cleans the data, and acts as a strict filter that only exposes the specific patient information permitted by regional HIE and HIPAA regulations. This ensures emergency personnel get the exact insights they need seamlessly, right when every second counts.

Key Features

  • Secure Record Management: Powered by PostgreSQL, Next.js, and React, this feature gives clinicians a consolidated, highly visual dashboard of a patient’s EMR data, including history, vitals, and insurance.
  • Advanced Medical Imaging: Using Cloudinary and custom parsing algorithms, the platform supports the secure upload, rendering, and interaction of complex files like X-rays, CAT scans, and DICOM images.
  • AI Clinical Search: We implemented an intelligent search function that quickly extracts vital patient details from dense records, letting doctors focus on the patient rather than digging through files.
  • Interactive 3D Diagnostics: Built with React Three Fiber, this feature generates 3D anatomical models to instantly highlight patient emergencies and past conditions for rapid visual processing.
  • Intelligent Note Processing: By combining CedarOS for high-fidelity voice-to-text with Google Gemini and the Maestra Agent Framework, the system automatically structures, categorizes, and organizes raw clinical audio notes.
  • Location-Based Prescriptions: Integrates GPS data to help physicians quickly locate the nearest relevant pharmacies for their patients.
  • Granular Access Control: We utilized NextAuth to build a robust, HIPAA-compliant permission system that strictly governs who can view specific pieces of patient data.

System Architecture & Stack

  • The Frontend Foundation: Built as a multi-tenant SaaS, we used Next.js, React, and TypeScript. The UI is styled with Tailwind CSS—featuring custom geometric designs and monospace fonts—and utilizes Radix UI for accessibility alongside Framer Motion for fluid transitions and theme switching.
  • AI and Visualization Engine: We optimized Google Gemini 2.0 Flash to extract structured demographics, allergies, and care plans from messy medical texts. For the visuals, Three.js and React Three Fiber power our interactive 3D models, blending transparent skeletal views with selectable block-model organs, all set against a dynamic, parallax blood-cell background.
  • Data and Security: A PostgreSQL database managed via Prisma ORM handles over 25 distinct medical data models. The system supports multi-tenancy and role-based access. Authentication is driven by NextAuth.js and Resend for magic links, complete with an emergency "break-glass" access protocol and strict audit logging.
  • API and State Management: The backend features over 15 distinct REST API endpoints built via Next.js routes. The frontend ensures real-time reliability using debounced calls, optimistic UI rendering, and an 800+ term medical mapping database for accurate injury logging.

Roadblocks We Overcame

  • Taming AI Outputs: We needed our AI transcription to be flawless, even in noisy environments. The biggest hurdle was getting the AI to return consistently formatted data. By iterating heavily on our multi-step prompt pipeline, we successfully forced the model to output reliable, programmatic data.
  • Perfecting the 3D Anatomy: We wanted a quick visual reference for doctors, but standard 2D diagrams fell short. Transitioning to a 3D model in React Three Fiber was tricky; simple rectangular blocks lacked clarity, but highly detailed skeletons caused performance drops and visual clutter. We ultimately engineered a hybrid model that perfectly balanced performance, clarity, and responsiveness.
  • Navigating HIPAA Compliance: Sharing data with EMS under the "minimum necessary" rule was incredibly tough. Our initial approach of applying a blanket filter to hide data ended up breaking essential medical workflows. We pivoted to a highly granular, role-based access control system. Now, hospital admins dictate exact field-level permissions, and every data request generates a secure audit trail.

Our Biggest Wins

  • Sleek, Functional Design: We are incredibly proud of the visual clarity we achieved. The custom 3D models and cohesive color schemes resulted in a UI that is not just highly functional and lightning-fast, but genuinely beautiful to look at.
  • Meaningful AI Application: We didn't just use AI as a gimmick. Implementing the Gemini API allowed us to build highly accurate tools that genuinely streamline workflows for doctors and EMS, potentially saving lives by cutting down administrative bloat.
  • Bulletproof Authentication: Navigating the legal minefield of HIPAA was daunting, but our NextAuth implementation resulted in a remarkably secure, policy-compliant permission system that protects patients while empowering first responders.

What's Next for the Platform

  • Infrastructure Scaling: To make a real impact, we need to handle massive, enterprise-level datasets. Our next technical hurdle is expanding our backend architecture to support seamless integration across multiple, large-scale healthcare providers.
  • Strategic Corporate Partnerships: Widespread adoption requires industry backing. We aim to secure corporate sponsorships and partnerships with major healthcare networks to deploy this technology into active medical settings.
  • Global Accessibility: We want to optimize and simplify the underlying technology so it can be deployed in developing nations, giving under-resourced healthcare providers access to vital, life-saving patient data.

Built With

Share this project:

Updates