Clinical Context-Aware Biomedical Text Summarization Using Deep Neural Network: Model Development and Validation.

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Artificial Intelligence

Background:

• Traditional automatic text summarization (ATS) struggles with clinical context and evidence quality in biomedical data.
• Deep learning models show promise but are underexplored for medical text summarization.

Purpose of the Study:

• To develop a novel framework, Biomed-Summarizer, for quality-aware, context-enabled summarization of biomedical text.
• To improve the extraction of precise, succinct, and coherent information for clinical decision-making.

Main Methods:

• Developed a deep neural network for quality recognition of biomedical studies.
• Utilized a bidirectional LSTM recurrent neural network for clinical context-aware PICO (Patient/Problem, Intervention, Comparison, Outcome) identification.
• Employed Jaccard similarity with semantic enrichment from medical ontologies for query-PICO matching.

Main Results:

• The quality recognition model achieved 95.41% accuracy.
• The clinical context-aware classifier reached 93% accuracy, outperforming traditional machine learning.
• Semantic similarity improved by 8.9% over baseline, with expert evaluations indicating satisfactory automated summarization.

Conclusions:

• Biomed-Summarizer achieves high accuracy in automatic text summarization for biomedical literature.
• The framework facilitates seamless curation of research evidence for clinical decision-making.