By skyforbes Hello!
Testing GPT OCR with medical reports (lab values, structured tables, hospital PDFs). Works okay for simple text, but struggles with:
– tables + units
– handwritten notes
– compressed scan images
I'm currently preprocessing via phone scan apps, adjusting contrast, and splitting pages. Still not great.
Anyone here reverse-engineered a prompt/workflow combo that consistently boosts OCR accuracy for clinical docs?
Not for diagnosis, purely documentation.