Qwerty
Video
Project / product name: GlucoMiner
Link to the project: https://glucominer.byqwerty.cz/
Team leader: Jakub Skopal
Challenge: 4. Efficient Glucometer Data Integration
Problem: Efficient parsing of PDFs form Patient's Self Monitoring using Continuous Glucose Measurement Devices.
Solution: Automated extraction of PDFs into structured JSON data and FHIR observations for automatic import into HIS.
Impact: Near absolute reduction of manual labor needed when processing PDFs submitted by patients.
Feasibility & financials: We found that automated processing of PDFs is certainly feasible at least for the devices currenty in use at IKEM:
dexcom, glooko, libre and medtronic.
We were able to process >99% of the reports.
We have achieved a prototype of standalone service performing the data extraction.
Financials -- implementation of such server for production environment should not exceed 100 hours. Implementation of any future similar PDF given sufficiently large sample should not exceed 20 hours.
What is new about your solution?: Ability to automate data extraction eliminating manual data entry work.
What you have built at the hackathon - text explanation + code (e.g. GitHub link): Node server implementing extraction of data from uploaded PDF as well as command-line utility for the same:
https://github.com/bindworks/hack2024-dia
React application that can be used to guide user submitting PDF reports to the hospital:
https://github.com/bindworks/hack2024-dia-fe
We have parsed all of the provided sample data into JSON and enhanced PDFs for validation:
IKEM Dropbox: Team QWERTY, Challenge 4, CGM data extraction.zip
What you had before the hackathon, please mention open source as well: We had some prior domain knowledge from implementation of register of diabetes patients.
Server uses Poppler for PDF parsing, Frontend uses React, shadcn UI components and Tailwind CSS.
What comes next and what you wish to achieve: Implementation of a standalone server which would be usable by various medical institutions in a secure way. Integration into existing applications -- Zlatokop, IKEM Online, Registry of Diabetic patients and ZScanner.
We want to allow the hospital staff to focus on the patient and not menial data entry work.