A Hybrid Reporting Platform for Extended RadLex Coding Combining Structured Reporting Templates and Natural Language Processing

Abstract

Structured reporting is a favorable and sustainable form of reporting in radiology. Among its advantages are better presentation, clearer nomenclature, and higher quality. By using MRRT-compliant templates, the content of the categorized items (e.g., select fields) can be automatically stored in a database, which allows further research and quality analytics based on established ontologies like RadLex® linked to the items. Additionally, it is relevant to provide free-text input for descriptions of findings and impressions in complex imaging studies or for the information included with the clinical referral. So far, however, this unstructured content cannot be categorized. We developed a solution to analyze and code these free-text parts of the templates in our MRRT-compliant reporting platform, using natural language processing (NLP) with RadLex® terms in addition to the already categorized items. The established hybrid reporting concept is working successfully. The NLP tool provides RadLex® codes with modifiers (affirmed, speculated, negated). Radiologists can confirm or reject codes provided by NLP before finalizing the structured report. Furthermore, users can suggest RadLex® codes from free text that is not correctly coded with NLP or can suggest to change the modifier. Analyzing free-text fields took 1.23 s on average. Hybrid reporting enables coding of free-text information in our MRRT-compliant templates and thus increases the amount of categorized data that can be stored in the database. This enhances the possibilities for further analyses, such as correlating clinical information with radiological findings or storing high-quality structured information for machine-learning approaches.

Keywords: Database; Medical informatics; Natural language processing; RadLex; Structured reporting.

Fig. 1

Excerpt from the template for SR of suspected pulmonary artery embolism on CT used in our department. Unstructured clinical information and the clinical question provided by the referring physician were automatically sent from the hospital information system to the reporting platform. Most fields contain structured information, as intraluminal filling defects of the pulmonary arteries (yes or no) or the size of the heart (normal or enlarged). Additional free-text fields enable the input of unstructured additional information that cannot be provided in the form of checkboxes or select-fields. At our institution, we decided to use free-text in the impression section to generate an individualized answer to the clinical question.

Fig. 2

Integration of the NLP function. The “Code” button (red arrow) has been added to our internal reporting platform. Coding can be performed before the structured report is saved and translates all unstructured free-text into RadLex® codes

Fig. 3

NLP sequence diagram showing interactions among the user (radiologist), the Mainz Radiology Reporting Engine (MRRE), and the Empolis Healthcare Analytics Services (HAS) with authentications, coding, and feedback function. All analytics as concept recognition using RadLex® were performed by the HAS

Fig. 4

After the HAS translated the free-text information into structured information, radiologists are able to confirm or reject the generated RadLex® codes displayed below each corresponding free-text field. The user is able to change the modifier and choose among “affirmed,” “speculated,” and “negated.” Furthermore, the radiologist can suggest additional RadLex® codes (as in this case of thrombosis) by typing the term into the search field

Fig.5

Excerpt from a structured report (left) and a dashboard integrated into the MRRE (right) for further analysis (here “low-dose CT of the retroperitoneum”). With hybrid reporting, we are able to analyze structured as well as unstructured information. First, we can use structured data from select fields in the structured reports. Second, by using NLP, we are now able to use initially unstructured data from free-text fields (e.g., clinical information) translated to RadLex® codes. Hybrid reporting provides the basis for in-depth analysis, e.g., the correlation between RadLex-coded clinical information (illustrated in the examples above for urolithiasis) and imaging results.