A survey of text mining approaches, techniques, and tools on discharge summaries

Abstract

The discharge summary contains voluminous information regarding the patient like history, symptoms, investigations, treatment, medication, etc. Though the discharge summary has a general structured way of representation, it is still not structured in a way that clinical systems can process. Different natural language processing (NLP) and machine learning techniques have been explored on the discharge summaries to extract various interesting information. Text mining techniques have been carried out in public and private discharge summaries. This survey discusses different tasks performed on discharge summaries and the existing tools which have been explored. The major dataset which has been used in existing research is also discussed. A common outline of system architectures on discharge summaries across various researches is explored. Major challenges in extracting information from discharge summaries are also detailed.