Inspiration
Doctors' handwriting and complex abbreviations on prescriptions cause confusion and potentially dangerous medication errors for patients. Goal: Create a simple tool to instantly read prescriptions and translate them into clear, easy-to-understand instructions. Motto: Bridging the Gap between clinical notes and patient safety.
What it does
Users take a picture of their prescription. Technology: It uses Optical Character Recognition (OCR) to read the text. Result: It gives the user a simplified explanation: What: Drug name and dosage. How: "Take 1 pill every morning and 1 every night." Warnings: Alerts for key side effects.
How we built it
A simple mobile app interface. OCR: We used a powerful OCR tool (like Google Vision) to digitize the handwriting. AI Core: Natural Language Processing (NLP) takes the digitized text and translates the medical abbreviations (like 'BID' or 'PRN') into full, clear sentences. Security: We focused on patient privacy—we don't save the image or the health data.
Challenges we ran into
Getting the AI to reliably read very poor and varied handwriting was the hardest part. Speed: Making the whole process—from scan to explanation—happen in just a few seconds. Medical Jargon: Teaching the system to correctly translate complex, ambiguous medical shorthand.
Accomplishments that we're proud of
Achieving over 95% accuracy in translating test prescriptions. Building a simple user interface that anyone can use. Successfully integrating a feature that flags potential drug interactions or major warnings.
What we learned
The importance of cleaning up images (pre-processing) before the OCR can read them. How critical it is to test the AI with real-world, messy data (not just perfect examples). That even complex AI projects can be packaged into a very simple user experience.
What's next for MediScan AI - Bridging the Gap
Look into connecting with Pharmacy Information Systems to verify the prescription details. Voice Integration: Add a feature to read the instructions aloud for visually impaired users.
Built With
- csv
- intellij-idea
- json
- nextjs
- python
Log in or sign up for Devpost to join the conversation.