CERC: an interactive content extraction, recognition, and construction tool for clinical and biomedical text

Area of Science:

• Computational linguistics
• Biomedical informatics
• Machine learning

Background:

• Automated summarization of scientific literature and patient records is crucial for clinical decision-making and precision medicine.
• Existing methods often rely on single relevance indicators, lack visualization capabilities, and do not personalize to user interests.
• The Content Extraction, Recognition, and Construction (CERC) system integrates machine learning, visualization, and domain knowledge for extracting key information from clinical and biomedical texts.

Purpose of the Study:

• To develop an interactive system (CERC) for automated summarization of clinical and biomedical text.
• To introduce a novel sentence-ranking framework, Multi-Indicator Text Summarization (MINTS), for extractive summarization.
• To improve the accuracy and utility of automated text summarization in healthcare settings.

Main Methods:

• Developed the CERC system, combining machine learning, visualization, and domain knowledge.
• Implemented the MINTS algorithm using random forests and multiple importance indicators for sentence ranking.
• Utilized a controlled vocabulary from MeSH, SNOMED-CT, and PubTator for term identification and weighted term frequency-inverse document frequency (TF-IDF) scores.

Main Results:

• The random forests model achieved 87.5% accuracy in classifying sentences.
• MINTS demonstrated superior performance over single-indicator methods, achieving higher ROUGE scores (ROUGE-1, ROUGE-2, ROUGE-SU4) (p < 0.01).
• The CERC system, including its automatic language translator and customizable pipeline, can be integrated into clinical decision support systems.

Conclusions:

• The CERC system provides a web-based tool for extracting salient information from clinical and biomedical text.
• The MINTS algorithm outperforms existing single-characteristic methods for text summarization.
• CERC facilitates early medical risk detection and supports data-driven, evidence-based patient care decisions.