A new approach to 3D reconstruction in echocardiography based on non-cubic voxels

Reconstruction of 3D objects from 2D data is a common problem in many different fields and especially medical imaging, e.g. CT and NMR. Most reconstruction methods presented so far, first define a three-dimensional matrix representing the final volume. In this volume the measured 2D data sets are placed. Then the object is reconstructed by filling the “empty space” using different types of interpolation. The results obtained this way are satisfactory, if the 2D data sets are parallel to each other, i.e. if each measured data point can be identified with one element of the 3D matrix. This corresponds to identifying each element of the matrix with one cubic voxel in space. However in some applications the acquired data does not come in parallel slices. In this case the measured points have to be fitted into cubic voxels by interpolation. This however introduces some non-quantifiable error. Furthermore it is not very efficient in terms of CPU time and use of memory. This paper describes a new approach to 3D reconstruction based on non-cubic voxels. It is well adapted to the specific properties of our data acquisition system in echocardiography but can be used with any acquisition system that is based on a rotation