A Machine Learning Approach to Digitize Medical History and Archive in a Standard Format

Abstract

Thanks to the advancement of information technology and the wide adoption of smartphone-based apps, an enormous amount of medical information is being produced worldwide. However, most of the medical records are yet to be standardized. Small clinics in developing countries generate only handwritten medical documents. Our past medical history is not digitized. Machine learning approaches applied to predict disease are quite common. But it will need sufficient past medical records to analyze. However, we do not have past medical records in digital form. This research aims to generate standard Electronic Health Records (EHRs) from paper-based documents. The major research tasks will be to investigate (1) the commonalities and differences of current unstructured paper-based medical documents, (2) the best technology to convert the paper-based documents into unstructured data, and (3) Extracting structured data from the unstructured data, (4) Integrating the structured into EHR databse using FHIR-based or OpenEHR Type System. This will produce standard medical history. Once medical histories are available in a standard format, it will be possible to predict personalized health status more accurately.