A system that recognizes medical lab results from photos
We applied computer vision and machine learning to automate medical document processing
The project
32desk approached us to build a system that could recognize medical lab results from photos and then run analytics on the extracted data. The solution had to automate medical document processing, cut data-entry time, and reduce the risk of human error.
Solution
We assembled a dataset of 10,000 images of medical lab results. To expand the data volume and make the model more robust to varying conditions, we applied data-augmentation techniques: rotation, distortion, noise injection, and others.
Result
The pilot was successfully rolled out across a network of dental clinics nationwide. The system automated lab-result processing, significantly reduced data-entry time, and lowered error rates. Clinics gained a tool for fast analytics and decision-making based on up-to-date medical data, which improved patient care and the productivity of medical staff.
