Hybridization of Fuzzy Theory and Nature-Inspired Optimization for Medical Report Summarization

ABSTRACT

It is becoming increasingly challenging to construct a smart medical system because of the large amount of accumulating information in the scientific literature of the biomedical sector. The current scenario reflects progress in a variety of less known regions as a means of extraction for the understanding of prevention and treatment for significant medical diseases such as COVID-19. Recently, many good scientific research publications in the biomedical arena was released using the MEDLINE/PubMed dataset. In the fields of biomedical research and healthcare, assessing these enormous data sets and extracting valuable information is a critical but difficult endeavour. Here, we attempt to retrieve relevant data from openly accessible text materials, like medical reports, journals, articles, papers, and some other research works, in the medical area. hese types of text data undergo first preprocessing in this chapter using sentence tokenization, then stopword removal, stemming operations, and ultimately vectorization using the BioBERT model. Consequently, a structured data is generated to process each report in feature extraction process and then clustered the similar sentences by Fuzzy C-means clustering. Then, using multiple similarity clustering measures and a bi-objective strength measure, defuzzify the clusters and construct the base summaries. To construct the report summary, en ensemble summarising approach has been employed by using Pareto evolutionary algorithm. The method contains two optimization methods(or functions) one dependent on the produced summary size, which is constant, and the other dependent on the IG (i.e., information gain) of the considering base summaries, which is variable. When the process of evolution converges, the strongest chromosomal solution of the ultimate population offers a desired summary report. This approach is used to generate an efficient summary from biomedical reports publically available in the MEDLINE/PubMed dataset, and finally, its performance comparing with a few similar cutting-edge techniques. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.