From Pixels to Paragraphs: Extraction of Text from Medical Reports using LLMs, Systematic Literature Review

Large language models are emerging very rapidly in applications covering nearly all fields including healthcare. In healthcare, these are commonly used for image to text, information extraction and generation and finding results from the medical notes and images. In spite of their extraordinary applications in healthcare, a standard and comprehensive systematic literature review (SLR) in this domain is lacking. However, there are quite a few systematic literature reviews in the field of large language models (LLMs) approaching the medical field. For that reason, this paper evaluates the literature to gather the relevant studies regarding LLMs in the medical field. The SLR process including a detail search was utilized to evaluate the research papers published between 2018 to date i.e. 2024, and a total of 82 studies, after rigorous screening, were included in this SLR. To wisely categorize these studies, consequently, we employed specific quality assessment criteria. Out of these 82 significant studies, only 4% were recognized to have utilized standard SLR guidelines. The results were presented in the form of charts and tables covering all the important aspects of meta data related to these selected studies. The quality assessment in screening these studies is also carried out in a very meticulous manner. Finally, an extensive discussion is carried out on formulated research questions in the methodology section and the suitable answers are discussed there. This paper therefore fulfills the gap for the scholars working on LLMs in the medical domain.