Segmentation of 4D CT bone images by sequential registration

The introduction of 4D image acquisition techniques has made it possible to analyse anatomical motion in vivo. With 4D computed tomography (CT), it is now possible to study the motion of joints leading to a deeper understanding of the role of morphology on joint motion and a better assessment of pathologies. Although 4D CT shows a lot of opportunities, the workload to process these 4D acquisitions has increased dramatically. A major part of this process is segmentation, the delineation of the objects of interest within the image volume. This paper presents an algorithm to accelerate this step by registering the segmentation of one frame onto the others. This results in a fast segmentation of the whole 4D dataset, all identical in shape. We show that the proposed algorithm is able to segment two carpal bones, the trapezoid and the scaphoid, with results close to a manual segmentation in less than 5% of the processing time.