Explainability of Methods for Critical Information Extraction From Clinical Documents: A survey of representative works
Tin Kam Ho, Yen-Fu Luo, Rodrigo Capobianco Guido · IEEE Signal Processing Magazine · 2022
The accumulation and integration of electronic health records have opened up opportunities for their new uses over the course of a patient’s care. In artificial intelligence (AI), natural language processing (NLP) methods that can extract important information from clinical documents are gaining success. For clinicians to consider using automatically extracted information in their decision making, the information should be explainable to support a reasoning process that is traceable and understandable. We review the state of the art in several types of information extraction methods designed for use in health care. Our focus is on comparing the approaches based on the explainability of the extracted results. We highlight the advantages and challenges faced by the adopted methods in conveying support for the extracted information.