Volumetric Segmentation of Cardiac Images

Title: Volumetric Image Segmentation Invention: The present invention is a semi-automated segmentation method and system which significantly reduces the amount of time and expertise required by an operator and which improves accuracy and repeatability. While the inventors have implemented this approach specifically for segmentation of right ventricles in 4D images, this invention is adaptable to a wide range of other medical imaging and other application areas. The user only selects a few landmarks, rather than entire contours, and then the remaining segmentation and related analytics are performed automatically. By minimizing user input and eliminating the need for training data, UA’s new invention can reduce the time by an order of magnitude and, more importantly, shift much of the workload - from radiologists to technicians in the case of the current cardiac implementation. Background: There are a number of limitations of current segmentation algorithms when dealing with complex images over time. Dynamic programming can manage some of these shortcomings but, in turn, this approach creates new challenges including the need to conduct training for each segmentation analysis as well as managing errors introduced with multi-point shape intersections. Current cardiac segmentation techniques require the user to manually outline the ventricle, for example, and then use training data to mimic the movement throughout the cardiac phase. They also use shape models, which reduce accuracy and, in particular, the ability to segment pathological cases.  Applications: Right Ventricle and other medical image segmentationOther time-series image segmentation Advantages: Shift segmentation input to techniciansReduce input time by 10xImproved accuracy and repeatabilityEnhanced analytics Licensing Manager: John Geikler(520) 626-4605JohnG@tla.arizona.edu

USA