Development of A Knowledge-Based Image Reporting System

Langlotz, Curtis

Abstract

We have constructed a prototype structured reporting system that replaces conventional dictation and transcription for medical image reporting. Key design features include reporting speed, generation of graphical reports, structured storage of imaging findings, and the use of existing lexicons. Pilot timing data suggest that a radiologist using our system can generate a report more rapidly than with conventional reporting methods or speech recognition systems. Support is sought to refine the prototype, and to conduct feasibility testing.
The specific aims are (1) to augment and refine a hierarchical test lexicon of imaging terms (2) to develop methods for the representation and reporting of imaging findings and their logical relationships, (3) to evaluate the system's performance on clinical imaging reports, and (4) to assess the completeness of existing lexicons for imaging. The project is supported by a Technical Advisory Committee comprised of experts in key methodological areas. A successful Phase I will lead (in Phase II) to testing of the approach for cross-sectional imaging, to adoption, augmentation, and use of an existing lexicon, to construction of real-time decision support techniques based on the knowledge base of imaging findings generated by the system, and to system evaluation in a clinical setting.

Proposed Commercial Applications

Our structured reporting system is appealing to radiologists because it speeds up the reporting process compared to conventional dictation/transcription or speech recognition. In addition, the system eliminates the costs, delays, inaccuracies, and other organizational problems associated with transcription services, thereby improving patient care. Therefore, time-efficient, speech-augmented structured reporting systems will likely capture a significant segment of the $1.2 billion annual market for radiology transcription services.