Search

Objective: We describe and evaluate the mapping of computerized tomography (CT) terms from 40 hospitals participating in a health information exchange (HIE) to a standard terminology., Methods: Proprietary CT exam terms and corresponding exam frequency data were obtained from 40 participant HIE sites that transmitted radiology data to the HIE from January 2013 through October 2015. These terms were mapped to the Logical Observations Identifiers Names and Codes (LOINC®) terminology using the Regenstrief LOINC mapping assistant (RELMA) beginning in January 2016. Terms without initial LOINC match were submitted to LOINC as new term requests on an ongoing basis. After new LOINC terms were created, proprietary terms without an initial match were reviewed and mapped to these new LOINC terms where appropriate. Content type and token coverage were calculated for the LOINC version at the time of initial mapping (v2.54) and for the most recently released version at the time of our analysis (v2.63). Descriptive analysis was performed to assess for significant differences in content-dependent coverage between the 2 versions., Results: LOINC's content type and token coverages of HIE CT exam terms for version 2.54 were 83% and 95%, respectively. Two-hundred-fifteen new LOINC CT terms were created in the interval between the releases of version 2.54 and 2.63, and content type and token coverages, respectively, increased to 93% and 99% (P < .001)., Conclusion: LOINC's content type coverage of proprietary CT terms across 40 HIE sites was 83% but improved significantly to 93% following new term creation.

Objective: This paper presents an automated system for classifying the results of imaging examinations (CT, MRI, positron emission tomography) into reportable and non-reportable cancer cases. This system is part of an industrial-strength processing pipeline built to extract content from radiology reports for use in the Victorian Cancer Registry., Materials and Methods: In addition to traditional supervised learning methods such as conditional random fields and support vector machines, active learning (AL) approaches were investigated to optimize training production and further improve classification performance. The project involved two pilot sites in Victoria, Australia (Lake Imaging (Ballarat) and Peter MacCallum Cancer Centre (Melbourne)) and, in collaboration with the NSW Central Registry, one pilot site at Westmead Hospital (Sydney)., Results: The reportability classifier performance achieved 98.25% sensitivity and 96.14% specificity on the cancer registry's held-out test set. Up to 92% of training data needed for supervised machine learning can be saved by AL., Discussion: AL is a promising method for optimizing the supervised training production used in classification of radiology reports. When an AL strategy is applied during the data selection process, the cost of manual classification can be reduced significantly., Conclusions: The most important practical application of the reportability classifier is that it can dramatically reduce human effort in identifying relevant reports from the large imaging pool for further investigation of cancer. The classifier is built on a large real-world dataset and can achieve high performance in filtering relevant reports to support cancer registries., (Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.)