A unified representation of findings in clinical radiology using the UMLS and DICOM
Introduction
Computer assisted medical activity raises the problem of modelling medical information. Indeed structured modelling using a controlled vocabulary allows the information to be standardized and presented in a clear, organized format. In addition to improving readability, structured modelling allows improved information retrieval and automated analysis for decision-support, research (data mining), evidence based medicine and teaching. Collecting and analyzing findings constitute the basis of medical activity. Creating and standardizing a structured radiology report is already the goal of the work in progress with the DICOM Structured Report (SR). But no unified model of findings does still exist.
The aim of this work is to propose a unified, sharable structured representation of findings integrating the representations of findings in the GAMUTS in Radiology from Reeder and Felson [1], the UMLS®, and the DICOM-SR. Starting from a corpus of findings in natural language, we used a semi-automated method in order to find a generic model (or template) for findings. Our model was evaluated for jaws radiological findings.
Section snippets
The GAMUTS in Radiology
A corpus has to gather information for which the to-build structured representation will have to be used: here typically textbooks or clinical reports. We started from descriptions of radiological findings which are proposed in many books [1], [2], [3]. Most of them have been edited many times and some have been translated in several languages [4]. However, the descriptions of findings remain unchanged (in vocabulary and in structure) in various books from various authors and in various
Building the corpus of findings
The finding corpus contained 504 different phrases, 3481 words, 931 different words.
Mapping the corpus of findings in natural language to the UMLS
Metamap discovered 5046 concepts (1220 different CUIs). After manual cleaning, 1888 concepts (37%) were kept (583 different CUI, i.e. 48%).
Discussion
In this work, we showed that it is possible to represent findings using the UMLS and the DICOM SR formalism with a semi-automated method. The UMLS and the Metamap program helped to find a model to represent the semantic structure of free texts with standardized terms. Metamapping the GAMUTS is a semi automated method to create new Templates and new Context Groups thanks to UMLS Concepts and Semantic Types.
Mapping free text to Metathesaurus concepts allows reusing the semantic structure of the
Conclusion
This study shows that the UMLS should include more technical and concepts pertaining to findings in radiology to be suitable for radiology representation. It also suggests that DICOM could build its new Templates and Context Groups respecting, if possible, UMLS Semantic Types.
The translation of the whole GAMUTS or clinical reports using the UMLS concepts and the DICOM SR relations could help to create or supplement the DCMR Templates and Context Groups in the field of the description of imaging
References (12)
- et al.
GAMUTS in Radiology Comprehensive lists of roentgen differential diagnosis
(2003) Clinical Imaging. An Atlas of Differential Diagnosis
(1988)- et al.
Aids to Radiological Diagnosis
(1995) - et al.
Guide du diagnostic différentiel en radiologie
Vigot
(1989) - et al.
GAMUTS in Radiology Comprehensive Lists of Rroentgen Differential Diagnosis
(1975) - UMLS Knowledge Source Server Version 4.2.3....
Cited by (13)
Using the Unified Medical Language System to Expand the Operative Stress Score – First Use Case
2021, Journal of Surgical ResearchCitation Excerpt :The results of this query are available in (eTable 5) and include numerous glycogen storage, tumor and metabolic, and familial hypercholesteremia related diseases and genes. Investigators interested in exploring the UMLS-Graph may can access it freely on Github (https://github.com/dbmi-pitt/UMLS-Graph), are encouraged to use it expose the ever expanding and updated relationships within the UMLS.18 In this first use case, we created a UMLS-Graph to systematically and efficiently expand the OSS by >400% and validated the accuracy of our added ratings.
Identifying QT prolongation from ECG impressions using a general-purpose Natural Language Processor
2009, International Journal of Medical InformaticsStructured Reporting: Coronary CT Angiography. A White Paper from the American College of Radiology and the North American Society for Cardiovascular Imaging
2008, Journal of the American College of RadiologyCitation Excerpt :The future of structured reporting includes capturing of imaging findings during the imaging process (point-and-click approaches) [12] that may translate directly into text, image-annotated text, and templates with constrained, coded vocabulary. These specific utilities, which can assist with data capture, direct comparison of data, outcomes assessment, and improved patient outcomes, can be facilitated by digital imaging and communication in medicine-structured report mechanisms [13]. The examination briefly describes the specific procedure that was performed.
BROMETH: Methodology to design safe reconfigurable medical robotic systems
2017, International Journal of Medical Robotics and Computer Assisted Surgery