EduCare

Inspiration

Healthcare is one of the most crucial and rapidly evolving industries, yet communication gaps between doctors and patients continue to pose a significant challenge. Often, post-appointment documentation is complex, medical jargon-heavy, and difficult for patients to fully understand. Studies reveal that 60-80% of medical information provided by healthcare professionals is forgotten by patients immediately after the consultation, with half of the remembered information recalled incorrectly (Journal of Royal Society of Medicine). Furthermore, the lack of clear, accessible medical records can leave patients confused about their diagnosis, treatment plans, and next steps.

Current solutions available focus solely on the doctor’s convenience, automating voice-to-text to generate reports but leaving patients behind in the process. These reports often remain medical-centric and inaccessible, preventing patients from understanding their health journey effectively. EduCare was born from the need to address these gaps—creating a post-appointment reporting tool that is accurate, doctor-approved, and easily understandable for patients, while also providing concise summaries for doctors to review.

What it does

EduCare transforms post-appointment documentation into concise, doctor-approved summaries and digestible, patient-friendly reports. Here’s how it works:

Voice Transcription: During the appointment, the doctor’s voice notes are transcribed into text in real-time using Deepgram API. Medical Parsing & Coding: The transcribed text is processed through NLTK and a custom parser to extract medical codes (CPT/ICD), symptoms, diagnostic procedures, and medications. Doctor Approval Stage: The summarized medical report is sent to the doctor for review and approval through the app interface, ensuring accuracy and mitigating hallucinations or errors from the AI-generated text. Humanized Patient Report: Once approved, the data is passed through OpenAI API to create a patient-friendly, layman’s version of the report, enriched with explanations. Language Translation: To improve accessibility, the final report can be translated into the patient’s native language. Seamless Delivery: The finalized report is stored in Firestore and sent to the patient via the platform for easy access and reference.

How we built it

The development of EduCare required seamless integration of multiple APIs and technologies to deliver an efficient, error-free experience for both doctors and patients. Here’s a breakdown:

Backend Development:

Deepgram API: Used to transcribe the doctor’s voice into text in real-time. PyAudio: To capture and stream audio input from the microphone to the backend server. NLTK (Natural Language Toolkit): To summarize and process medical text, extracting important elements such as symptoms, diagnoses, and medications. Custom Medical Parser: Converts the raw text into CPT/ICD medical codes, providing structured medical information. OpenAI API: Used at two stages—first, to generate an AI-enhanced summary for doctor review, and second, to produce the final patient-friendly report. Firestore Database: Stores the summaries, parsed codes, and final reports in real-time. Flask Backend: Handles API calls between the frontend and backend services, ensuring smooth data exchange. Frontend Development:

React.js: Builds the web interface for doctors and patients, enabling easy approval and report access. Firebase Firestore Integration: Provides real-time synchronization of approved summaries and final reports. Responsive Design: Ensures that the web app functions seamlessly on both mobile and desktop devices. Workflow Integration:

Approval Mechanism: The app features an approval stage where the doctor reviews and fact-checks the auto-generated summary before it is finalized. API Coordination: We ensured smooth integration across multiple APIs—transcription, summarization, parsing, and translation—so the reports flow seamlessly from creation to approval and delivery.

Challenges we ran into

Building EduCare presented several technical and logistical challenges, including:

Handling Real-time Transcription: Integrating PyAudio and Deepgram API to capture and transcribe audio in real-time required careful coordination between streaming input and API responses. Accuracy of Summaries: Managing the risk of hallucinations or inaccuracies in AI-generated text was crucial. We implemented an approval stage to mitigate errors. Medical Parsing & Coding: Creating a parser that correctly interprets CPT/ICD codes while also generating understandable summaries was a complex task. API Integration & Data Flow: Managing the seamless flow of data between multiple APIs and the frontend-backend systems required extensive testing and debugging. User Experience Design: We had to design an interface that was simple for doctors to approve summaries while also producing clear, digestible reports for patients.

Accomplishments that we're proud of

Empowering Patients: We successfully created a tool that shifts the focus from a doctor-centric workflow to one that empowers patients with clear, understandable medical reports. Accurate, Doctor-approved Reporting: We developed a robust approval mechanism that ensures all reports are accurate and fact-checked before being delivered to patients. Seamless API Integration: We integrated multiple APIs and achieved smooth interoperability between the transcription, summarization, and coding systems. Real-time Updates: With Firestore’s real-time database, we enabled instant report generation and updates, ensuring timely delivery to patients. Language Accessibility: By including a translation feature, we broke language barriers, making healthcare information accessible to non-English-speaking patients.

What we learned

Balancing AI and Human Oversight: We realized the importance of combining AI capabilities with human oversight in sensitive areas like healthcare to ensure accuracy and safety. API Coordination: We gained valuable experience working with multiple APIs and learned how to optimize data flows between the frontend and backend. Healthcare Documentation Needs: We deepened our understanding of healthcare documentation challenges and the need for concise, yet comprehensive reports. User Experience in Healthcare: Designing interfaces for both doctors and patients taught us how to balance simplicity and functionality across different user groups

What's next for EduCare

Expanding Language Support: We plan to add more languages to the translation feature, ensuring that patients from diverse backgrounds can access their medical reports. Mobile App Development: A mobile version of EduCare is in the works, allowing patients and doctors to access reports and summaries on the go. Integration with EHR Systems: We aim to integrate EduCare with existing Electronic Health Record (EHR) systems, streamlining workflows for healthcare providers. Voice Command Features: Implementing voice command functionality will allow doctors to interact with the app hands-free, further enhancing usability. AI-powered Insights: We plan to leverage AI analytics to provide patients with personalized insights and recommendations based on their health data.