A visualization tool for improved assessment and follow-up with ultrasound

Ultrasound has been shown to be a powerful imaging tool for the evaluation of musculoskeletal disease (e.g. rheumatoid arthritis or carpal tunnel syndrome). Power Doppler imaging in particular plays an important role in treatment monitoring. However it is also known that ultrasound is operator dependent and has low intra-and inter-observer repeatability. The aim of this work was to reduce the uncertainty of the follow-up process of a typical musculoskeletal exam by ensuring that the same anatomy is visualized at baseline and follow-up. A visualization tool, which receives a continuous stream of ultrasound data, has been implemented in Qt. The streamed images are compared on the fly against a pre-loaded baseline image. To compensate for in plane translation and rotation differences, 2D rigid registration is performed. Normalized mutual information was used as a similarity metric and is color coded from red (0%) to green (100%) to show the amount of similarity among the current images. Several display alternatives (e.g. semi-transparent overlay, color coded overlay) have been implemented and tested. Optional anisotropic filtering can be applied to both images before registration to reduce the impact of the speckle noise on the final result. Two healthy volunteers and a phantom have been used for validation purposes. Both the reference and follow-up ultrasound data have been acquired with a Vivid E9 scanner, with an ML6-15 probe (GE Vingmed Ultrasound). At the same time the position of the probe with regards to a reference frame was tracked with an accurate optical position sensor (Polaris, NDI). It was shown that the highest similarity scores corresponded to the lowest translation offsets and angle differences. The average offset and angle errors for the patient data were: [0.68 +/- 0.97, 0.47 +/- 0.79, -2.28 +/- 1.38] mm, [2.92 +/- 2.35, -0.17 +/- 1.2, 1.24 +/- 2.38]degrees. A novel method for real-time fusion of ultrasound images aimed at musculoskeletal disease monitoring has been presented. Being provided with instant feedback, the user is more confident during the acquisition process.