Detecting Technical Image Quality in Radiology Reports

Image interpretation accuracy is critical to ensure optimal care, yet many diagnostic reports contain expressions of uncertainty often due to shortcomings in technical quality among other factors. While radiologists will usually attempt to interpret images and render a diagnosis even if the imaging quality is suboptimal, often the details related to any quality concerns are dictated into the report. Despite imaging exam quality being an import factor for accurate image interpretation, there is a significant knowledge gap in terms of understanding the nature and frequency of technical limitations mentioned in radiology reports. To address some of these limitations, in this research we developed algorithms to automatically detect a broad spectrum of acquisition-related quality concerns using a dataset containing 1,210,858 exams. There was some type of a quality concern mentioned in 2.4% of exams with motion being the most frequent