Discovering body site and severity modifiers in clinical texts

Dligach, Dmitriy; Bethard, Steven; Becker, Lee B.; Miller, Timothy A.; Savova, Guergana

doi:10.1136/amiajnl-2013-001766

Cited by 24 publications

(16 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These feature types have been successful for us on other relation extraction tasks[27] and we found they were surprisingly effective on the development set. This is a potentially valuable representation for learning whether a candidate is coreferent at all, but surface features like tokens may overfit to the training domain, so we designed experiments attempting to measure this.…”

Section: Methodsmentioning

confidence: 87%

Towards generalizable entity-centric clinical coreference resolution

Miller

Dligach

Bethard

et al. 2017

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

Objective This work investigates the problem of clinical coreference resolution in a model that explicitly tracks entities, and aims to measure the performance of that model in both traditional in-domain train/test splits and cross-domain experiments that measure the generalizability of learned models. Methods The two methods we compare are a baseline mention-pair coreference system that operates over pairs of mentions with best-first conflict resolution and a mention-synchronous system that incrementally builds coreference chains. We develop new features that incorporate distributional semantics, discourse features, and entity attributes. We use two new coreference datasets with similar annotation guidelines – the THYME colon cancer dataset and the DeepPhe breast cancer dataset. Results The mention-synchronous system performs similarly on in-domain data but performs much better on new data. Part of speech tag features prove superior in feature generalizability experiments over other word representations. Our methods show generalization improvement but there is still a performance gap when testing in new domains. Discussion Generalizability of clinical NLP systems is important and under-studied, so future work should attempt to perform cross-domain and cross-institution evaluations and explicitly develop features and training regimens that favor generalizability. A performance-optimized version of the mention-synchronous system will be included in the open source Apache cTAKES software.

show abstract

Section: Methodsmentioning

confidence: 87%

Towards generalizable entity-centric clinical coreference resolution

Miller

Dligach

Bethard

et al. 2017

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

show abstract

“…A number of researchers have investigated NLP methods to automatically recognize and encode the concepts conveyed in medical texts [27,29,31,34–37,48,52,54–60]. Most of those NLP systems have been created as general purpose systems that attempt to extract diseases and findings that are commonly discussed in a medical texts.…”

Section: Discussionmentioning

confidence: 99%

Using automatically extracted information from mammography reports for decision-support

Bozkurt

Gimenez

Burnside

et al. 2016

Journal of Biomedical Informatics

View full text Add to dashboard Cite

Objective To evaluate a system we developed that connects natural language processing (NLP) for information extraction from narrative text mammography reports with a Bayesian network for decision-support about breast cancer diagnosis. The ultimate goal of this system is to provide decision support as part of the workflow of producing the radiology report. Materials and methods We built a system that uses an NLP information extraction system (which extract BI-RADS descriptors and clinical information from mammography reports) to provide the necessary inputs to a Bayesian network (BN) decision support system (DSS) that estimates lesion malignancy from BI-RADS descriptors. We used this integrated system to predict diagnosis of breast cancer from radiology text reports and evaluated it with a reference standard of 300 mammography reports. We collected two different outputs from the DSS: (1) the probability of malignancy and (2) the BI-RADS final assessment category. Since NLP may produce imperfect inputs to the DSS, we compared the difference between using perfect (“reference standard”) structured inputs to the DSS (“RS-DSS”) vs NLP-derived inputs (“NLP-DSS”) on the output of the DSS using the concordance correlation coefficient. We measured the classification accuracy of the BI-RADS final assessment category when using NLP-DSS, compared with the ground truth category established by the radiologist. Results The NLP-DSS and RS-DSS had closely matched probabilities, with a mean paired difference of 0.004 ± 0.025. The concordance correlation of these paired measures was 0.95. The accuracy of the NLP-DSS to predict the correct BI-RADS final assessment category was 97.58%. Conclusion The accuracy of the information extracted from mammography reports using the NLP system was sufficient to provide accurate DSS results. We believe our system could ultimately reduce the variation in practice in mammography related to assessment of malignant lesions and improve management decisions.

show abstract

“…negative mentions (eg, “no history of angina”). 3–5 Machine-learning algorithms that use both structured and NLP-extracted narrative EHR data to identify classes of patients have been recently shown to work robustly across multiple institutions. 6 …”

Section: Appendix Table A1mentioning

confidence: 99%

Merging Electronic Health Record Data and Genomics for Cardiovascular Research

Hall¹,

Ryan²,

Bray³

et al. 2016

Circ Cardiovasc Genet

View full text Add to dashboard Cite

The process of scientific discovery is rapidly evolving. The funding climate has influenced a favorable shift in scientific discovery toward the use of existing resources such as the electronic health record. The electronic health record enables long-term outlooks on human health and disease, in conjunction with multidimensional phenotypes that include laboratory data, images, vital signs, and other clinical information. Initial work has confirmed the utility of the electronic health record for understanding mechanisms and patterns of variability in disease susceptibility, disease evolution, and drug responses. The addition of biobanks and genomic data to the information contained in the electronic health record has been demonstrated. The purpose of this statement is to discuss the current challenges in and the potential for merging electronic health record data and genomics for cardiovascular research.

show abstract

Discovering body site and severity modifiers in clinical texts

Cited by 24 publications

References 26 publications

Towards generalizable entity-centric clinical coreference resolution

Towards generalizable entity-centric clinical coreference resolution

Using automatically extracted information from mammography reports for decision-support

Merging Electronic Health Record Data and Genomics for Cardiovascular Research

Contact Info

Product

Resources

About