Development of the Vision Ontology.

Development of the Vision Ontology. Poster No.: B-0141 Congress: ECR 2013 Type: Scientific Paper Authors: D. J. Vining 1, U. Salem 1, C. Popovici 2, L. Jiang 3, C. DURAN 4, A. Pitici 2, I. Aghenitei 2, M. Jurca 2, R. Rosu 2 ; 1 Houston, TX/US, 2 Chapel Hill, NC/US, 3 Beijing/CN, 4 ISTANBUL/TR Keywords: DOI: Computer applications, ehealth, Management, PACS, RIS, Teleradiology, Computer Applications-Detection, diagnosis, Structured reporting, Epidemiology 10.1594/ecr2013/B-0141 Any information contained in this pdf file is automatically generated from digital material submitted to EPOS by third parties in the form of scientific presentations. References to any names, marks, products, or services of third parties or hypertext links to thirdparty sites or information are provided solely as a convenience to you and do not in any way constitute or imply ECR's endorsement, sponsorship or recommendation of the third party, information, product or service. ECR is not responsible for the content of these pages and does not make any representations regarding the content or accuracy of material in this file. As per copyright regulations, any unauthorised use of the material or parts thereof as well as commercial reproduction or multiple distribution by any traditional or electronically based reproduction/publication method ist strictly prohibited. You agree to defend, indemnify, and hold ECR harmless from and against any and all claims, damages, costs, and expenses, including attorneys' fees, arising from or related to your use of these pages. Please note: Links to movies, ppt slideshows and any other multimedia files are not available in the pdf version of presentations. www.myesr.org Page 1 of 18

Purpose We developed a structured reporting system, called ViSion, which allows a radiologist to capture key images and tag those images with diagnostic information to generate a multimedia structured report (Figure 1) (1). The operation of the ViSion system is the topic of another EPOS poster, "A ViSion for Global Multimedia Structured Reporting." Fig. 1: Example of a ViSion structured report. Key images are tagged with anatomy and pathology terms. Each diagnosis is assigned a priority value that can be used for automatic notification of critical results. References: Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, University of Texas MD Anderson Cancer Center - Houston/US The ontology (i.e., a controlled vocabulary with defined relationships among the terms) that is used to tag images in ViSion is managed with an integrated software authoring tool which is the subject of this exhibit. The tagging of images with this ontology can either be performed automatically by applying natural language processing to dictated descriptions, or manually by having the radiologist use pull-down menus in the ViSion system. The ViSion ontology also supports electronic billing, translation of reports to multiple languages, critical results notification, and data mining. Page 2 of 18

Images for this section: Fig. 1: Example of a ViSion structured report. Key images are tagged with anatomy and pathology terms. Each diagnosis is assigned a priority value that can be used for automatic notification of critical results. Page 3 of 18

Methods and Materials Although multiple medical ontologies currently exist (e.g., RadLex, SNOMED, ICD-10), we found it necessary to develop an integrated ontology for the ViSion structured reporting system to function optimally (2-3). The ViSion ontology consists of a collection of anatomy, pathology, and secondary characteristic terms that are organized into hierarchies (i.e., tree structures) to define anatomical locations, radiological observations and diagnoses for each anatomical location, and secondary descriptors that provide more detail for each observation or diagnosis. Observations and diagnoses in the ViSion system are created by pairing anatomy and pathology terms (e.g., Colon:Polyp). Radiologists have many ways of saying the same thing (4). Thus, the ViSion ontology provides a means to manage synonyms and homonyms associated with preferred terms. The anatomy terms are organized in a hierarchy that describes anatomical locations from head-to-toe and are divided into seven main body sections (Figure 2): Head Neck Chest Abdomen Pelvis Upper extremity Lower extremity Page 4 of 18

Fig. 2: ViSion's anatomical hierarchy lists locations among seven body sections. References: Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, University of Texas MD Anderson Cancer Center - Houston/US Each body section contains a list of the major organs found in that section followed by "common organ systems" shared by all sections: Artery Vein Lymphatic Skeleton Muscle Soft tissues Each organ is subdivided when appropriate. For example, the colon is divided into the anus, rectum, sigmoid colon, descending colon, splenic flexure of colon, transverse colon, hepatic flexure of colon, ascending colon, ileocecal valve, cecum, and appendix. Page 5 of 18

Every anatomical term is paired to a hierarchy of pathology terms that define the radiological observations and diagnoses that are applicable to that anatomy (Figure 3). The first level of each pathology tree is divided into the following categories: Normal Inflammation Neoplasm Trauma Systemic Function Observation Needs definition Fig. 3: Example of the pathology hierarchy associated with the anatomy term "Colon." Each pathology tree is organized with first-level divisions of Normal, Inflammation, Neoplasm, Trauma, Systemic, Function, Observation, and Requires definition. The last entry allows a user to recommend a term for inclusion into the ViSion ontology. References: Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, University of Texas MD Anderson Cancer Center - Houston/US The pathology terms that are contained in each category are specific to a particular anatomical location. The last entry, "Needs definition," allows for users to recommend terms that are not contained in the ViSion ontology. Page 6 of 18

The combining of anatomy and pathology terms creates "observations" and "diagnoses." Each observation/diagnosis can be described in further detail with "secondary characteristics." For example, the secondary characteristics associated with "Colon:Polyp" would include the following: Shape: Pedunculated Sessile Mass C-RADS: C-RADS 0 C-RADS 1 C-RADS 2 C-RADS 3 C-RADS 4 The ViSion ontology was constructed in English but the anatomy, pathology, and secondary characteristic terms have been translated to German, Chinese, Turkish, French, Arabic, and Spanish (Figures 4-5). Page 7 of 18

Fig. 4: The ViSion report in Fig. 1 translated to Chinese. References: Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, University of Texas MD Anderson Cancer Center - Houston/US Page 8 of 18

Fig. 5: The ViSion report in Fig. 1 translated to Arabic. References: Department of Diagnostic Radiology, University of Texas MD Anderson Cancer Center, University of Texas MD Anderson Cancer Center - Houston/US The translations were performed initially using Google Translate, but the output was reviewed and modified when necessary by radiologists who spoke these languages. It is important to note that the ViSion system only translates the metadata used to tag images and not the entire narrative descriptions generated by radiologists. Each of the radiological observations and diagnoses in the ViSion ontology has been crossed-referenced to other standardized medical ontologies (e.g., RadLex, SNOMED, LOINC, ICD-9-CM, and ICD-10-CM) when those items exist. Finally, each observation and diagnosis in the ViSion system has been prioritized on a 5-point scale to indicate a level of action that needs to be taken by a clinician with regard to a particular finding: 1. Incidental Requires no further action 2. Indeterminate Requires further evaluation with an additional imaging or other diagnostic procedures 3. Important 4. Urgent Should be followed on subsequent imaging examinations Requires attention as soon as possible 5. Life-Threatening Requires immediate action In the ViSion reporting system, these default priority settings can be modified as necessary by the radiologist. When a report is signed that contains Urgent or Life- Threatening findings, the ViSion system sends automatic notification of the critical results to the clinician by email and/or SMS messaging with return receipt verification. Page 9 of 18

Images for this section: Fig. 1: Example of a ViSion structured report. Key images are tagged with anatomy and pathology terms. Each diagnosis is assigned a priority value that can be used for automatic notification of critical results. Page 10 of 18

Fig. 2: ViSion's anatomical hierarchy lists locations among seven body sections. Page 11 of 18

Fig. 3: Example of the pathology hierarchy associated with the anatomy term "Colon." Each pathology tree is organized with first-level divisions of Normal, Inflammation, Neoplasm, Trauma, Systemic, Function, Observation, and Requires definition. The last entry allows a user to recommend a term for inclusion into the ViSion ontology. Page 12 of 18

Fig. 4: The ViSion report in Fig. 1 translated to Chinese. Fig. 5: The ViSion report in Fig. 1 translated to Arabic. Page 13 of 18

Results In September 2012, the ViSion ontology consisted of 906 anatomy terms, 1472 pathology terms, and 1380 secondary characteristic terms. The pairing of anatomy and pathology terms in the hierarchal trees resulted in 5653 observations or diagnoses. In January 2013, the ViSion ontology had expanded to 12,126 observations or diagnoses. Each term has been translated to German, Chinese, Turkish, French, Arabic, and Spanish. Each anatomical location and observation/diagnosis has been cross-referenced to the other major ontologies, including RadLex, SNOMED, LOINC, ICD-9-CM, and ICD-10- CM, for whenever these matches exist. Conclusion ViSion is a multimedia structured reporting system that operates in multiple languages and provides a method to automatically code radiologic observations and diagnoses. The development of an integrated ontology authoring tool has been essential for the construction of anatomy and pathology hierarchies, secondary characteristics, foreign language translations, and prioritization of diagnoses. Images for this section: Page 14 of 18

Page 15 of 18

Fig. 6: ViSion multimedia structured reporting. Page 16 of 18

References 1. Martino A. Sketching a new reality: what will the radiology report of the future look like? http://www.acr.org/news-publications/news/news- Articles/2012/ACR-Bulletin/201203-Rad-Report-of-Future Accessed January 28, 2013. 2. The National Center for Biomedical Ontology. http://www.bioontology.org/ Accessed January 28, 2013. 3. Rubin DL. Creating and curating a terminology for radiology: ontology modeling and analysis. J Digit Imaging 2008; 21:355-362. 4. Sobel JL et al. Information content and clarity of radiologists' reports for chest radiography. Acad Radiol 1996; 3:709-717. Personal Information Disclosure: David J. Vining, MD, is the founder, CEO, and a major stockholder of VisionSR which has an option agreement with the University of Texas MD Anderson Cancer Center to license the ViSion technology for commercialization. David J. Vining, MD, Department of Diagnostic Radiology, UT MD Anderson Cancer Center, Houston, Texas, USA, dvining@mdanderson.org, www.facebook.com/ ViSionReporting Usama Salem, MD, Department of Diagnostic Radiology, UT MD Anderson Cancer Center, Houston, Texas, USA, usalem@mdanderson.org Cristi Popovici, Eloquentix, Inc, Chapel Hill, North Carolina, USA, popovici@eloquentix.com Liming Jiang, MD, Department of Radiology, Chinese Academy of Medical Sciences, Beijing, China, dr_jlm@yahoo.cn Cihan Duran, MD, Department of Diagnostic Radiology, UT MD Anderson Cancer Center, Houston, Texas, USA, cduran1@mdanderson.org Andrea Pitici, Eloquentix, Inc, Chapel Hill, North Carolina, USA, andreea@eloquentix.com Page 17 of 18

Iulian Aghenitei, Eloquentix, Inc, Chapel Hill, North Carolina, USA, iulian@eloquentix.com Mihai Jurca, Eloquentix, Inc, Chapel Hill, North Carolina, USA, mihai@eloquentix.com Radu Rosu, Eloquentix, Inc, Chapel Hill, North Carolina, USA, radu@eloquentix.com Page 18 of 18