No abstract

Accurate assessment of regional myocardial function after infarction is a challenging task in biplanar X-ray cineangiography. An estimate of motion is given by tracking contours of the left ventricle (LV) in projection images.In conventional clinical practice, LV wall motion analysis is done in 2D, based on end-diastolic and end-systolic contours. Our work, by contrast, presents novel methods for analyzing left ventricular motion in 4D, making possible the evaluation of myocardial function in any region of the ventricle rather than at artificial boundary areas in the projection image.4D LV models are reconstructed from a biplanar image sequence. The centerline and radial method, originally developed for analysis in 2D, are extended to 3D for assessment of cardiac motion based on our 4D ventricular model. Color-encoded ventricle models and 2D maps of heart wall dynamics are developed for visualization. Results obtained by using the methods presented herein are in correspondence with clinical findings. IntroductionInterventional X-ray angiography is state of the art in diagnosis and treatment of cardiovascular diseases. To diagnose viability of myocardium after infarction, for instance, X-ray images of the left ventricle are acquired with biplanar cineangiography equipment. A major drawback of the underlying imaging modality is the loss of 3D information, as objects are projected onto a 2D image plane. Consequently, only motion of the outer ventricular shape is visible in the images.In clinical practice, quantitative left ventricle analysis is based on end-diastolic and end-systolic contours. For wall motion analysis (WMA) in 2D, several methods have been developed in the past [1,2,3]. But as no 3D information is available, myocardial function is only assessed at the ventricular boundary area projected to the image. This work, however, illustrates how myocardial activity can be evaluated in 4D based on 3D models of the LV evolving in time. 2.Material and methods DataAngiography data is recorded with a Siemens BICOR system, capturing 8-bit gray level images of size 512x512 pixels at a frame rate of 25fps. A 4D model of the contracting ventricle, reconstructed from a biplanar image sequence, provides the basis for quantitative analysis of LV wall motion. LV model reconstructionAccuracy of X-ray image acquisition is affected by pincushion distortion and spiral distortion [4]. Thus, distortion correction is performed prior to segmentation and reconstruction, based on an approach using control points obtained from a reference image. For semi-automatic segmentation, statistical shape models of the LV are deduced from a set of expert-segmented images [5]. Principle Component Analysis (PCA) of the shape model determines a circumferential region of interest where the contour of the LV is most likely detected.A 3D LV contour model is reconstructed slice-by-slice from biplanar images. The algorithm [6] utilizes contour information obtained from segmentation to define the location of X-ray attenuation profiles of the ventricle...