EMA - Elderly Medical Appointment (Assistant)

Narrowing the communication gap between doctors and elderly patients.

Comment

This is the modified README.md in Markdown format, incorporating all your new sections and content while maintaining the project's accessibility focus.

📝 EMA - Elderly Medical Assistant

Project Overview

EMA (Elderly Medical Assistant) is a cross-platform mobile application designed to bridge the communication gap between healthcare providers and elderly patients, ensuring crucial medical advice is never misunderstood or forgotten.

The app uses Speech-to-Text (STT) to record and transcribe doctor-patient consultations and then leverages a Large Language Model (LLM) to transform complex medical jargon into a simple, easy-to-understand summary written at a primary school reading level.

💡 Inspiration

My personal motivation for building EMA stems from a direct observation of this critical healthcare communication failure:

I saw my mom struggle to understand the doctor's explanation, sometimes forgot the crucial details, and could not explain to us well when she had an appointment with the doctor when I cannot accompany her for the sessions.

This experience highlighted the urgent need for a tool that empowers elderly patients with clear, accessible, and shareable medical information, acting as a reliable memory and communication aid.

⚙️ What it does

EMA provides a seamless, end-to-end solution for managing and understanding medical visits:

  1. Appointment Management: Users can add and view appointments using a simple, calendar-first UI.
  2. Consultation Recording: A single, large button initiates the recording of the doctor-patient conversation.
  3. Intelligent Summarization: The raw audio is transcribed (STT) and immediately processed by the LLM (Groq) to generate a 3-point, simplified summary (Diagnosis, Action Plan, Next Appointment) at a Grade 4 reading level.
  4. Review & Share: The summary can be read, listened to (TTS), or shared easily with family and caregivers.

🏗️ How we built it

Component Technology / Service Implementation Details
Mobile Platform Flutter (Dart) Utilized for cross-platform deployment (iOS/Android) with an emphasis on Accessible UI/UX (High-Contrast, Large Targets).
Architecture Clean Architecture (BLoC/Riverpod) Structured the codebase into Presentation, Domain, and Data layers for maximum testability and maintainability when integrating external APIs.
Summarization (LLM) Groq API (Mixtral/Llama 3) Used a highly-optimized, structured prompting strategy to leverage Groq's low latency and speed up the summary generation process significantly.
Transcription (STT) ElevenLabs Scribe v1 / Groq ASR Integrated a robust STT solution to accurately capture medical terminology for processing.
Local Storage Hive Used for fast, lightweight local persistence of appointment logs and summarized text.

🛑 Challenges we ran into

  • Prompt Engineering for Readability: The biggest challenge was fine-tuning the Groq prompt to consistently produce summaries that were medically accurate while strictly adhering to the Grade 4 (Layman's) reading level without losing crucial context.
  • API Latency Management: While Groq is fast, orchestrating the multi-step process (Audio Upload $\rightarrow$ Transcription $\rightarrow$ Summarization) while providing clear feedback to the user on the "Processing" screen required careful asynchronous state management in Flutter.
  • Accessibility Design: Ensuring the Calendar-first view and Timeline History remained intuitive and easy to navigate for users with potential cognitive or dexterity limitations.

✅ Accomplishments that we're proud of

  • Finished the core functionalities of the app: Successfully implementing the full pipeline from Recording $\rightarrow$ Transcription $\rightarrow$ LLM Summarization $\rightarrow$ Display into a functional mobile UI.
  • Established a high-contrast, accessible UI/UX flow that prioritizes the unique needs of the elderly demographic.
  • Integrating a modern, high-speed LLM service (Groq) to provide near real-time summarization, greatly enhancing the user experience.

🧠 What we learned

  • The full potential of Cursor: Utilizing advanced code generation and editing tools to accelerate development and adhere to Clean Architecture standards.
  • Groq API: Gaining hands-on experience in leveraging an ultra-low latency LLM for complex, real-time agentic tasks (like summarization) in a mobile context.
  • Proper 'vibe-coding' development process: Successfully balancing rapid feature implementation with meticulous attention to security and architectural best practices.

⏭️ What's next for EMA - Elderly Medical Appointment (Assistant)

  • Polish the app and refine the visual design to achieve a polished, production-ready aesthetic.
  • Enhance and expand the app functionality with advanced AI features, such as:
    • AI image recognition to automatically log medicine information and schedule reminders based on the medicine picture (e.g., dosage, frequency).
  • Expand the sharing feature or add user & family account functionality to make the appointment sharing session easier and create a centralized hub for caregiver coordination.

Built With

Share this project:

Updates