Development of an Automated Bone Mineral Density Software Application: Facilitation Radiologic Reporting and Improvement of Accuracy

摘要

The conventional method of bone mineral density (BMD) report production by dictation and transcription is time consuming and prone to error. We developed an automated BMD reporting system based on the raw data from a dual energy X-ray absorptiometry (DXA) scanner for facilitating the report generation. The automated BMD reporting system, a web application, digests the DXA's raw data and automatically generates preliminary reports. In Jan. 2014, 500 examinations were randomized into an automatic group (AG) and a manual group (MG), and the speed of report generation was compared. For evaluation of the accuracy and analysis of errors, 5120 examinations during Jan. 2013 and Dec. 2013 were enrolled retrospectively, and the context of automatically generated reports (AR) was compared with the formal manual reports (MR). The average time spent for report generation in AG and in MG was 264 and 1452 s, respectively (p < 0.001). The accuracy of calculation of T and Z scores in AR is 100 %. The overall accuracy of AR and MR is 98.8 and 93.7 %, respectively (p < 0.001). The mis-categorization rate in AR and MR is 0.039 and 0.273 %, respectively (p = 0.0013). Errors occurred in AR and can be grouped into key-in errors by technicians and need for additional judgements. We constructed an efficient and reliable automated BMD reporting system. It facilitates current clinical service and potentially prevents human errors from technicians, transcriptionists, and radiologists.