Applications of Natural Language Processing for the Management of Stroke Disorders: Scoping Review

Area of Science:

• Medical Informatics
• Computational Linguistics
• Neurology

Background:

• Natural Language Processing (NLP) advances are increasing medical applications, especially for stroke emergencies.
• Understanding accumulated experience in NLP for stroke is crucial in this evolving field.

Purpose of the Study:

• To conduct a 10-year scoping review of NLP applications in stroke emergency management.
• To identify state-of-the-art techniques, application contexts, and software tools.

Main Methods:

• Scopus and Medline databases were searched using keywords 'natural language processing' and 'stroke'.
• Data on study phases, contexts, data types, methods, and software were extracted and analyzed.
• Multiple correspondence analysis was used to explore relationships between categories.

Main Results:

• Twenty-nine papers were reviewed, mostly cohort studies on ischemic stroke from the last two years.
• NLP primarily assists in diagnosis and outcome prognosis, using diagnostic reports and medical image annotations.
• Machine learning with simpler NLP methods and ontologies is common, with a rise in deep learning techniques.

Conclusions:

• NLP applications in stroke mirror broader AI trends, focusing on clinical process improvement over rehabilitation.
• Deep learning, particularly Bidirectional Encoder Representations from Transformers (BERT), is state-of-the-art for NLP in stroke.
• There's a growing emphasis on processing medical image annotations using advanced NLP.