Motion tracked 3D Visualization system for Segmented ROI in medical imaging

Abstract

Computer-aided diagnosis systems are increasingly used in the detection and segmentation of abnormalities in medical imaging. However, in many borderline cases, radiologists and physicians need to analyze the images to confirm the diagnosis. This paper proposes a novel analysis system for visualizing the output of 3D segmented abnormalities and the associated computer diagnoses. We use motion trackers to localize the user's viewing device and estimate its 3D orientation. We then use this tracking data to slice the segmented and labeled volume showing the abnormalities in real-time. The user can switch between the labeled and raw views to analyze the abnormalities' boundaries. To test our visualization, we use our brain tumor segmentation pipeline using MRI scans to produce a labeled image and slice and visualize the tumor from all angles. Our system has a low latency of 0.01 seconds and can identify the viewing device's 3D location and orientation within a maximum of 0.2 cm and 2 degrees of error, respectively. The accuracy of our system depends on the segmentation pipeline. Our proposed system is also useful for medical training and as a visual feedback tool for computer-aided system's researchers and developers. © 2021 Association for Computing Machinery.

Keywords CAD, CT, MRI, ROI