Growth and Motion in Three-Dimensional Images

W ITH the widespread use of new and improved scanning techniques for gathering volumetric medical image data, a new field of four-dimensional (4-D) medical image processing is emerging. Already, considerable interest has been devoted to three-dimensional (3-D) motion of living tissue within a short-term window, and some attention has been given to the possibly more challenging long-term studies of 3-D growth processes. Recent advances in scanner hardware have significantly increased both spatial and temporal resolution, and real-time volumetric imaging is becoming more common. This leads to strong interest in 4-D image analysis, since new clinically relevant dynamic parameters can be obtained. Techniques for 4-D image processing becomes also a necessity, since the amount of data gathered in a single patient examination easily becomes too overwhelming for manual inspection. In conjunction with the European Conference on Computer Vision, we organized a workshop on Growth and Motion in 3-D Medical Images, hoping to attract the attention of the computer-vision community to these new challenges. The workshop consisted of an invited lecture by Professor Sven Kreiborg from the School of Dentistry at the University of Copenhagen and peer-reviewed contributions. As a result of the large interest into the subject demonstrated by the success of the workshop, we decided to organize a special journal issue in parallel. 17 authors responded to our call for papers in the summer of 2002, where about half had been presented at the workshop in earlier versions. With the very competent help of our reviewers and the immense help of the editorial staff at the IEEE TRANSACTION ON MEDICAL IMAGING, we selected ten contributions for this special issue, which can roughly be categorized as either motion or growth studies. In medical imaging, short term or motion estimation plays an increasing role, both for diagnosis and treatment. Some medical acquisition techniques directly provide an estimate of the local motion, such as phase contrast magnetic resonance angiography (MRA) or Doppler ultra sound. In other imaging modalities, motion can be retrospectively estimated by acquiring images with high temporal resolution and by utilizing the changes in appearance of the images. Next to motion estimation of studying deformation, motion analysis can also be performed in order to compensate for patient movement during acquisition. The papers in this issue on motion estimation reflect the variety of applications for which motion estimation is important. In the paper by Fatouraee et al.a motion field is directly obtained from the imaging modality using phase contrast MRA. However, the obtained motion field may be corrupted owing to complex flow patterns, and a physics-based model is proposed in order to improve the accuracy in estimating this flow field.