Optical Character Recognition (OCR) with Machine Learning
Every day, we OCR tens of thousands of pages of medical records, making them easily searchable for our clients. Also, OCR is the foundation of our value-added services such as ReChron, our chronological ordering program. High-quality OCR helps our clients and our software to make the right decisions.
The key challenge has been low-quality records provided by the record custodians. Often, records are faxed over multiple times, and multiple copies are made and scanned before they reach us. When you run traditional OCR on these records, the outcome is less than optimal, rendering the text layer on the PDF file practically useless. The situation gets even worse when handwriting is present on pages.
To solve this issue, we introduced machine learning into the OCR process with incredible success. Even low-quality pages with font sizes below 8 are properly recognized now. Though handwriting is still a challenge, we have made great advancements in that area as well.
The new OCR process requires significantly more computing resources than the traditional one, but thanks to YoCierge’s containerized architecture, this does not result in any significant delay in delivering the records.