Faculty Bibliography

The visualization of 3D phenomena and the extraction of quantitative information from magnetic resonance (MR) image data require efficient semiautomated or automated segmentation techniques. The application of multivariate statistical classification to the segmentation of dual-echo volume data of the human head into tissue types (grey matter, white matter and fluid spaces) is studied in this paper. Tests of the radiometric variability of tissue classes within the data volume demonstrate the improvement of the image acquisition technology and the suitability of statistical methods to perform brain tissue segmentation. Supervised classification is successfully applied to a study of 16 MR volume images of the human head, illustrating the robustness of this method in segmenting brain (white and grey matter) and cerebrospinal fluid (CSF). To avoid subjective criteria involved in the supervised approach, ISODATA clustering as well as clustering based on nonparametric probability density estimation were tested. Both methods performed well (success rates 93.8% and 87.5%, respectively), indicating that the classification procedure can be completely automated. The reproducibility and reliability of supervised and unsupervised classification were studied by comparing results of segmentation performed by five expert operators. Results suggest that the interoperator and intraoperator variations could be reduced using automated clustering techniques. The accuracy of the volume calculations was quantified by applying the MR imaging and segmentation process to a phantom resembling shape and tissue characteristics of brain tissue. The segmented brain objects are displayed using 3D surface rendering.

The most important problem in the analysis of time sequences is the compensation for artifactual motion. Owing to motion, medical images of the abdominal region do not represent organs with fixed configuration. Analysis of organ function with dynamic contrast medium studies using regions of interest (ROIs) is thus not readily accomplished. Images of the organ of interest need to be registered and corrected prior to a detailed local analysis. We have developed an image analysis scheme that allows the automatic detection of the organ contours, the extraction of the motion parameters per frame, and the registration of images. The complete procedure requires only minimal user interaction and results in a readjusted image sequence, where organs of interest remain fixed. Both a visual analysis of the dynamic behavior of functional properties and a quantitative statistical analysis of signal intensity versus time within local ROIs are considerably facilitated using the corrected series.

Fast MR techniques and the application of water-soluble contrast agents allow the simultaneous examination of renal morphology and the functional aspects of glomerular filtration using bolus injections of Gd-DTPA. Spatial resolution is sufficient to resolve individual renal pyramids, but the quantitative examination of regions of interest (ROIs) is severely impeded by organ movements due to variations of the end-inspiratory position. A new image-processing scheme has been used and tested in 23 normal volunteers and patients. This scheme replaces a tedious frame-by-frame ROI analysis by positional correction of renal regions of all frames of the sequence such that the definition of the regions has to be performed only once. The signal intensities (SIs) of the local regions in each frame are used to compute statistics and to generate curves representing local temporal SI changes due to contrast agent excretion. The success rate of the procedure depends largely on the image quality and on the adherence to a proved acquisition protocol. The present article shows that the combination of MR and robust and reliable image-processing methods can be important for the highly automated analysis of a large number of images acquired as dynamic studies.

Both time-of-flight and phase contrast magnetic resonance angiography images are combined with stationary tissue images to provide data depicting two contrast relationships yielding intrinsic discrimination of brain matter and flowing blood. A computer analysis is based on nearest neighbor segmentation and the connection between anatomical structures to partition the images into different tissue categories: from which, high resolution brain parenchymal and vascular surfaces are constructed and rendered in juxtaposition, aiding in surgical planning.