Real-time uncertainty visualization for B-mode ultrasound

Abstract

B-mode ultrasound is a very well established imaging modality and is widely used in many of today's clinical routines. However, acquiring good images and interpreting them correctly is a challenging task due to the complex ultrasound image formation process depending on a large number of parameters. To facilitate ultrasound acquisitions, we introduce a novel framework for real-time uncertainty visualization in B-mode images. We compute real-time per-pixel ultrasound Confidence Maps, which we fuse with the original ultrasound image in order to provide the user with an interactive feedback on the quality and credibility of the image. In addition to a standard color overlay mode, primarily intended for educational purposes, we propose two perceptional visualization schemes to be used in clinical practice. Our mapping of uncertainty to chroma uses the perceptionally uniform L*a*b* color space to ensure that the perceived brightness of B-mode ultrasound remains the same. The alternative mapping of uncertainty to fuzziness keeps the B-mode image in its original grayscale domain and locally blurs or sharpens the image based on the uncertainty distribution. An elaborate evaluation of our system and user studies on both medical students and expert sonographers demonstrate the usefulness of our proposed technique. In particular for ultrasound novices, such as medical students, our technique yields powerful visual cues to evaluate the image quality and thereby learn the ultrasound image formation process. Furthermore, seeing the distribution of uncertainty adjust to the transducer positioning in real-time, provides also expert clinicians with a strong visual feedback on their actions. This helps them to optimize the acoustic window and can improve the general clinical value of ultrasound.