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

Abstract

Background Automated summarization of scientific literature and patient records is essential for enhancing clinical decision-making and facilitating precision medicine. Most existing summarization methods are based on single indicators of relevance, offer limited capabilities for information visualization, and do not account for user specific interests. In this work, we develop an interactive content extraction, recognition, and construction system (CERC) that combines machine learning and visualization techniques with domain knowledge for highlighting and extracting salient information from clinical and biomedical text. Methods A novel sentence-ranking framework multi indicator text summarization, MINTS, is developed for extractive summarization. MINTS uses random forests and multiple indicators of importance for relevance evaluation and ranking of sentences. Indicative summarization is performed using weighted term frequency-inverse document frequency scores of over-represented domain-specific terms. A controlled vocabulary dictionary generated using MeSH, SNOMED-CT, and PubTator is used for determining relevant terms. 35 full-text CRAFT articles were used as the training set. The performance of the MINTS algorithm is evaluated on a test set consisting of the remaining 32 full-text CRAFT articles and 30 clinical case reports using the ROUGE toolkit. Results The random forests model classified sentences as “good” or “bad” with 87.5% accuracy on the test set. Summarization results from the MINTS algorithm achieved higher ROUGE-1, ROUGE-2, and ROUGE-SU4 scores when compared to methods based on single indicators such as term frequency distribution, position, eigenvector centrality (LexRank), and random selection, p < 0.01. The automatic language translator and the customizable information extraction and pre-processing pipeline for EHR demonstrate that CERC can readily be incorporated within clinical decision support systems to improve quality of care and assist in data-driven and evidence-based informed decision making for direct patient care. Conclusions We have developed a web-based summarization and visualization tool, CERC (https://newton.isye.gatech.edu/CERC1/), for extracting salient information from clinical and biomedical text. The system ranks sentences by relevance and includes features that can facilitate early detection of medical risks in a clinical setting. The interactive interface allows users to filter content and edit/save summaries. The evaluation results on two test corpuses show that the newly developed MINTS algorithm outperforms methods based on single characteristics of importance.

Patient Information Summarization in Clinical Settings: Scoping Review

“…Using ontologies to find and relate concepts within textual notes (20/128, 15.6% [ 87 , 92 , 93 ]), the use of Unified Medical Language System (UMLS) extraction tools (6/128, 4.7%) to extract them (eg, [ 94 - 96 ]), or improving the performance of abstractive summarization (2/128, 1.6% [ 76 ])…”

Machine and cognitive intelligence for human health: systematic review

“…In Ref. [ 56 ], a sentence-ranking mechanism adopted random forests and multiple importance indicators for relevance measurement and sentence ranking. In Ref.…”

“…Second, there are scholars focusing on correlation analysis, including: (1) conducting correlation analysis for the individual facial action units to understand the decoupling of these individual features [ 47 ], and (2) demonstrating potential correlations between “a person’s descriptions about wartime experiences in their blogs with the ensuing symptoms or disorders via Focus groups and medical records analysis (p. 6) [ 13 ]”. Third, scholars are also encouraged to: (1) add functionality for real-time image annotation during meetings and make the transition to Internet-based telephone services [ 34 ], (2) adopt a health-related misinformation detection framework to English health misinformation detection [ 86 ], (3) explore repositioning drugs according to semantic relations for varied syndromes, for example, Parkinson’s diseases, Alzheimer’s diseases, and cancers [ 82 ], (4) facilitate the extraction of date and confirmed-case counts [ 77 ], (5) propose approaches for predicting user dropout rate to provide timely interventions accordingly [ 64 ], (6) extend system functionality by providing automatic graph-based summarization of input texts [ 56 ], (7) investigate “clustering solutions with a larger number of clusters or implementing additional features in the cluster analysis to represent other dimensions of participant experience (p. 14)” for richer characterization of participant experiences for personalization [ 61 ], (8) perform more intensive label harmonization using common data model ontologies [ 54 ], and (9) conduct syntactic analysis of natural language questions and test syntactic dependencies’ contribution on confirming previously extracted semantic relationships and detecting unfamiliar relationships [ 30 ]. Other directions include: (1) promoting “comprehensive care by establishing additional applications for home follow-ups and working with the children with the rare inherited disorders and their families (p. 11) [ 95 ]”, (2) developing neural-driven security solutions for multimedia data like color medical images, audios, and videos to be stored in the cloud [ 39 ], (3) updating parameters dynamically [ 96 ], and (4) improving predictors to reduce prediction bias to discover physiological mechanisms of ion channel-targeted conotoxins [ 41 ].…”

“…In terms of model or system evaluation, future efforts need to focus on: (1) performing clinical trials to understand the impact of AI classifiers on skin cancer classification in real-life settings [ 72 ], (2) comparing with other predictive models via tenfold validation [ 98 ], (3) evaluating LSTM’s performance with Glove and Fasttext [ 44 ], (4) performing extrinsic evaluation focusing on the system’s ability “for high-risk findings in patient records and its impact on patient care and clinical decision-making (p. 318) [ 56 ]”, (5) validate summarization strategies with varied types of clinical texts (e.g., operative notes and radiology reports) in patient healthcare settings [ 56 ], (6) understanding how interaction patterns impact treatment benefits in Internet-based interventions [ 61 ], (7) exploring drug and side effect relationship extraction and adverse drug reaction extraction [ 62 ], (8) system assessment from different perspectives (e.g., group size and user profile size) [ 69 ], (9) studying the significance of adopting single variable versus multiple variables by search engines [ 87 ], and (10) conducting biological and clinical experiments to validate utility and effectives [ 54 ].…”

Privacy Preserving Text Document Summarization

“…The logit function depends on the probability of feature occurrences P and it is 0 < Pi < 1. Equation ( 4) and Equation( 5) describes the prediction of the test record label as sensitive or not sensitive where, 0 < Pi < 1 [13].…”