This repository contains two Python optical character recognition (OCR) scripts for use in Optometry practice.
- Referral/report filing - takes scanned PDF inputs (letters from referring practitioners, including reports, referrals, TCAs, etc) and files by patient name and +/- TCA status.
- Humphrey Field Analyzer (HFA) test filing - takes direct files exported by HFA and sorts tests into files by patient name, DOB, and RE/LE.
The OCR scripts were created to try and automate certain aspect of my day-to-day administrative work in optometry practice. The referral filing script was the initial iteration of this type of script; the HFA filing script was a refined version built at a later date for use on an easier set of input data.
- Broad range of letters can be sorted: the referral OCR script is built to handle referrals and reports from 20+ different ophthalmologists and other health practitioners, in spite of their unique letter layout and language.
- Error handling: where an 'error' is encountered and the patient's name cannot be confidently deduced, the user is alerted and the file is sorted into a specified directory for manual review. This also filters spam input as a by-product.
- Respectable accuracy: The HFA script runs at essentially 100% accuracy as the input data is exported by a computer, resulting in a stable, consistent data set. The referral script boasts over 90% accuracy in the face of incredible input varition, both in scan quality and letter layout.
The key technology used here is the Tesseract OCR - an open source OCR program with impressive capability. The Python/Windows specific version is implemented.
- Python
- Tesseract OCR
- OpenCV for input pre-processing
These instructions will help you set up the OCR scripts to run on your system.
These instructions are for Windows OS. There is virtually zero presence of non-Windows OS in Optomety.
First, clone the repository
git clone git@github.com:daniel-moderiano/ocr.git
Inside the root directory, install the following dependencies:
pip install opencv-python pytesseract pyPdf4 pdf2image img2pdf
Tesseract OCR for Windows is also required. At time of writing, the current version is 5.0. Poppler for Windows is also required, and must be added to PATH.
From here, the input and output directories must be adjusted to your requirements. Then, the scripts can be run directly.
Please note the HFA script requires HFA .tif input and the referral script requires .pdf input.