Animated 3D lung surface reconstruction from asynchronous MR image sequences based on multiple registration

This work addresses the reconstruction of an animated lung surface by using respiratory patterns determined in temporal sequences of images obtained from Magnetic Resonance (MR). The reconstruction is done without the use of any triggering information and any special gas to enhance the contrast and the images temporal sequences are acquired in free breathing. The real movement of the lung has never been seen directly, as it is totally dependent on its surrounding muscles and collapses without them. Comparing with Computerized Tomography (CT), MR imaging involves longer acquisition times but it is preferable because it does not involve ionizing radiation. The multiple registration is based on the analysis of the intersecting segments among multiple coronal and sagittal sequences, since they are orthogonal to each other. The concepts of 2D spatiotemporal (2DST) image, respiratory function, Hough transform, active countors and temporal registration are used. The multiple registration algorithm used here allows the determination of the lung surface, given a fixed sagittal (coronal) MR image, by initially applying the temporal registration multiple times, then performing a three dimensional intersection points matching and finally reconstructing the lung surface with the points obtained from the previous step. Applying this method to all images of a temporal sequence, an animated 3D lung model is obtained. Some results and conclusions are shown.