An OCR Post-Correction Approach Using Deep Learning for Processing Medical Reports

According to a recent Deloitte study, the COVID-19 pandemic continues place huge strain on global health care sector. Covid-19 has also catalysed digital transformation across sector for improving operational efficiencies. As result, amount of digitally stored patient data such as discharge letters, scan images, test results or free text entries by doctors grown significantly. In 2020, 2314 exabytes medical was generated globally. This does not conform generic structure and is mostly in form unstructured scanned paper documents part patient’s reports. digitised using Optical Character Recognition (OCR) process. A key challenge here that accuracy OCR process varies due inability current engines correctly transcribe handwritten which may be skewed, obscured illegible. compounded fact processed comprised specific terminologies do necessarily general language lexicons. The proposed work uses deep neural network based self-supervised pre-training technique: Robustly Optimized Bidirectional Encoder Representations from Transformers (RoBERTa) can learn predict hidden (masked) sections fill gaps non-transcribable parts being processed. Evaluating method domain-specific datasets include real documents, shows significantly reduced word error rate demonstrating effectiveness approach.