MediTrust
MediTrust is a Secure, AI powered Web3 Health Record
Created on 5th April 2025
•
MediTrust
MediTrust is a Secure, AI powered Web3 Health Record
The problem MediTrust solves
-
Handwritten Prescription Recognition
- Accurately reads and interprets handwritten doctor's prescriptions
- Eliminates errors from manual data entry and misreading
-
Medical Information Extraction
- Identifies medication names, dosages, and instructions from complex prescriptions
- Structures unorganized prescription text into clear, organized data
-
Easy Digital Record Creation
- Converts paper prescriptions to secure digital format with minimal effort
- Creates standardized medical records from inconsistent prescription formats
-
Patient Safety Enhancement
- Reduces medication errors by clearly presenting dosage instructions
- Highlights important warnings and precautions in a prominent format
-
Medical History Management
- Builds a searchable history of patient prescriptions
- Makes prescription details available to authorized healthcare provider.
Challenges we ran into
Challenges We Ran Into
1. Handwritten Text Recognition Accuracy
Challenge: Handwritten prescriptions vary dramatically in style, clarity, and format, making them extremely difficult to interpret with standard OCR.
Solution: We implemented Google Cloud Vision API's specialized document text detection, which uses advanced ML models specifically trained on handwritten text. We enhanced this with a custom multi-stage parsing algorithm that can identify medications even when the OCR results are imperfect.
2. Medical Terminology Recognition
Challenge: Standard text processing approaches struggled with medical terminology, abbreviations, and specialized notation common in prescriptions.
Solution: We developed domain-specific regex patterns that can identify medication names, dosages, and instructions by understanding the context and format of prescription writing. Our parser recognizes multiple prescription formats (numbered lists, bullet points, and Rx notation).
3. Blockchain Integration for Protected Health Information
Challenge: Storing patient prescription data securely while maintaining HIPAA compliance posed significant technical and ethical concerns.
Solution: We implemented a hybrid approach where the actual prescription image is stored in IPFS with encryption, while the structured data is stored on the blockchain with access controls. This ensures only authorized healthcare providers can access patient prescription information.
5. Parsing Inconsistent Prescription Formats
Challenge: Prescriptions follow no universal standard format, making it extremely difficult to extract information reliably.
Solution: We built a multi-stage parsing pipeline that tries increasingly flexible pattern matching techniques. If structured pattern matching fails, the system falls back to line-by-line analysis looking for key indicators of medication information, ensuring we always extract at least basic information.