Precursor-induced conditional random fields: connecting separate entities by induction for improved clinical named entity recognition.

BACKGROUND: This paper presents a conditional random fields (CRF) method that enables the capture of specific high-order label transition factors to improve clinical named entity recognition performance. Consecutive clinical entities in a sentence are usually separated from each other, and the textual descriptions in clinical narrative documents frequently indicate causal or posterior relationships that can be used to facilitate clinical named entity recognition. However, the CRF that is generally used for named entity recognition is a first-order model that constrains label transition dependency of adjoining labels under the Markov assumption. METHODS: Based on the first-order structure, our proposed model utilizes non-entity tokens between separated entities as an information transmission medium by applying a label induction method. The model is referred to as precursor-induced CRF because its non-entity state memorizes precursor entity information, and the model's structure allows the precursor entity information to propagate forward through the label sequence. RESULTS: We compared the proposed model with both first- and second-order CRFs in terms of their F1-scores, using two clinical named entity recognition corpora (the i2b2 2012 challenge and the Seoul National University Hospital electronic health record). The proposed model demonstrated better entity recognition performance than both the first- and second-order CRFs and was also more efficient than the higher-order model. CONCLUSION: The proposed precursor-induced CRF which uses non-entity labels as label transition information improves entity recognition F1 score by exploiting long-distance transition factors without exponentially increasing the computational time. In contrast, a conventional second-order CRF model that uses longer distance transition factors showed even worse results than the first-order model and required the longest computation time. Thus, the proposed model could offer a considerable performance improvement over current clinical named entity recognition methods based on the CRF models.

Temporal Segmentation for Capturing Snapshots of Patient Histories in Korean Clinical Narrative.

OBJECTIVES: Clinical discharge summaries provide valuable information about patients' clinical history, which is helpful for the realization of intelligent healthcare applications. The documents tend to take the form of separate segments based on temporal or topical information. If a patient's clinical history can be seen as a consecutive sequence of clinical events, then each temporal segment can be seen as a snapshot, providing a certain clinical context at a specific moment. This study aimed to demonstrate a temporal segmentation method of Korean clinical narratives for identifying textual snapshots of patient history as a proof-of-a-concept. METHODS: Our method uses pattern-based segmentation to approximate human recognition of the temporal or topical shifts in clinical documents. We utilized rheumatic patients' discharge summaries and transformed them into sequences of constituent chunks. We built 97 single pattern functions to denote whether a certain chunk has attributes that indicate that it can be a segment boundary. We manually defined the relationships between the pattern functions to resolve multiple pattern matchings and to make a final decision. RESULTS: The algorithm segmented 30 discharge summaries and processed 1,849 decision points. Three human judges were asked whether they agreed with the algorithm's prediction, and the agreement percentage on the judges' majority opinion was 89.61%. CONCLUSIONS: Although this method is based on manually constructed rules, our findings demonstrate that the proposed algorithm can achieve fairly good segmentation results, and it may be the basis for methodological improvement in the future.

Conditional random fields for clinical named entity recognition: A comparative study using Korean clinical texts.

BACKGROUND: This study demonstrates clinical named entity recognition (NER) methods on the clinical texts of rheumatism patients in South Korea. Despite the recent increase in the adoption rate of the electronic health record (EHR) system in global health institutions, health information technologies for handling and acquisition of information from numerous unstructured texts in the EHR system are still in their developing stages. The aim of this study is to verify the conventional named entity recognition (NER) methods, namely dictionary-lookup-based string matching and conditional random fields (CRFs). METHODS: We selected discharge summaries for 200 rheumatic patients from the EHR system of the Seoul National University Hospital and attempted to identify heterogeneous semantic types present in the clinical notes of each patient's history. RESULTS: CRFs outperform string matching in extracting most semantic types (median F1 = 0.761, minimum = 0.705, maximum = 0.906). String matching is found to be better suited for identifying hospital visit information. The performance of both methods is comparable for identifying medications. The 10-fold cross-validation shows that CRFs had median F1 = 0.811 (minimum = 0.752, maximum = 0.918), and exhibited good performance even when trained with simple features. CONCLUSION: CRFs are a good candidate for implementing clinical NER in Korean clinical narrative documents. Increasing the training data and incorporating sophisticated feature engineering might improve the accuracy of identifying health information, enabling automated patient history summarization in the future.

Maximal pericoronary adipose tissue thickness is associated with hypertension in nonobese patients with acute or chronic illness.

BACKGROUND/AIMS: Recent studies have shown an association of epicardial fat thickness with diabetes and hypertension (HTN) in asymptomatic populations. However, there is lack of information as to whether there is similar association between pericoronary adipose tissue (PAT) and HTN in the patients who have acute or chronic illness. METHODS: This study included 214 nonobese patients hospitalized with acute or chronic noncardiogenic illness. PAT thicknesses were measured from fat tissues surrounding left and right coronary arteries in enhanced, chest computed tomography scans, yielding the maximal PAT value from left and right coronary arteries was used for analysis. Baseline data from hypertensive (n = 81) and normotensive (n = 133) patients were collected and compared. RESULTS: PAT is positively correlated with age (r = 0.377, p <0.001), body mass index (BMI; r = 0.305, p < 0.001), systolic blood pressure (r = 0.216, p = 0.001), and total cholesterol (r = 0.200, p = 0.006). The hypertensive group was older (69.58 ± 11.69 years vs. 60.29 ± 14.98 years), and had higher PAT content (16.30 ± 5.37 mm vs. 13.06 ± 5.58 mm) and BMI (23.14 ± 3.32 kg/m2 vs. 20.96 ± 3.28 kg/m) than the normotensive group (all p < 0.001). Multivariate analysis showed that age (odds ratio [OR], 2.193; p = 0.016), PAT thickness (OR, 1.065; p = 0.041), and BMI (25 ≤ BMI < 30 kg/m2 ; OR, 6.077; p = 0.001) were independent risk factors for HTN. CONCLUSIONS: In nonobese patients with noncardiogenic acute or chronic illness, PAT thickness is independently correlated with HTN, age, and BMI.