A support tool for early breast cancer detection, based on the following criteria..
Plenty of information, lack of data
Most clinical information in millions of health records is in unstructured format, ie text.
Mentioned risk factors and symptoms are in visit notes. This means that they are “buried” under 75%-85% unstructured information, being ignored and leading to late diagnosis.
Natural Language Processing
Electronic Health Records. Including text
Natural Language Processing identifies all clinical concepts.
Structured Database through NLP Coded
Recurring queries are run upon structured database.
Patients that comply with criteria are selected.
Doctors receive an alarm and decide if selected patients require additional examination.