Robert Grzeszczuk

The spatial location of EEG electrodes: locating the best-fitting sphere relative to cortical anatomy

The location of the international 10-20 system electrode positions and 14 fiducial landmarks are described in cartesian coordinates (+/- 1.4 mm average accuracy). Six replications were obtained on 3 separate days from 4 normal subjects, who were compared to each other with a best-fit sphere algorithm. Test-retest reliability depended on the electrode position: the parasagittal electrodes were associated with greater measurement errors (maximum 7 mm) than midline locations. Location variability due to head shape was greatest in the temporal region, averaging 5 mm from the mean. For each subjects electrode locations a best-fitting sphere was determined (79-87 mm radius, 6% average error). A surface-fitting algorithm was used to transfer the electrode locations and best-fitting sphere to MR images of the brain and scalp. The center of the best-fitting sphere coincided with the floor of the third ventricle 5 mm anterior to the posterior commissure. The melding of EEG electrode location information with brain anatomy provides an empirical basis for associating hypothetical equivalent dipole locations with their anatomical substrates.

Volumetric visualization of anatomy for treatment planning

PURPOSE Delineation of volumes of interest for three-dimensional (3D) treatment planning is usually performed by contouring on two-dimensional sections. We explore the usage of segmentation-free volumetric rendering of the three-dimensional image data set for tumor and normal tissue visualization. METHODS AND MATERIALS Standard treatment planning computed tomography (CT) studies, with typically 5 to 10 mm slice thickness, and spiral CT studies with 3 mm slice thickness were used. The data were visualized using locally developed volume-rendering software. Similar to the method of Drebin et al., CT voxels are automatically assigned an opacity and other visual properties (e.g., color) based on a probabilistic classification into tissue types. Using volumetric compositing, a projection into the opacity-weighted volume is produced. Depth cueing, perspective, and gradient-based shading are incorporated to achieve realistic images. Unlike surface-rendered displays, no hand segmentation is required to produce detailed renditions of skin, muscle, or bony anatomy. By suitable manipulation of the opacity map, tissue classes can be made transparent, revealing muscle, vessels, or bone, for example. Manually supervised tissue masking allows irrelevant tissues overlying tumors or other structures of interest to be removed. RESULTS Very high-quality renditions are produced in from 5 s to 1 min on midrange computer workstations. In the pelvis, an anteroposterior (AP) volume rendered view from a typical planning CT scan clearly shows the skin and bony anatomy. A muscle opacity map permits clear visualization of the superficial thigh muscles, femoral veins, and arteries. Lymph nodes are seen in the femoral triangle. When overlying muscle and bone are cut away, the prostate, seminal vessels, bladder, and rectum are seen in 3D perspective. Similar results are obtained for thorax and for head and neck scans. CONCLUSION Volumetric visualization of anatomy is useful in treatment planning, because 3D views can be generated without the need for segmentation. When relationships among anatomical structures, rather than geometric models of them, are important, volume rendering presents advantages. The presented algorithm is readily adaptable to distributed parallel implementation on a network of heterogeneous workstations.

Retrospective fusion of radiographic and MR data for localization of subdural electrodes

Prior to epilepsy surgery, subdural electrodes are often implanted and monitored for a few days to identify the focus of abnormal electrical activity. During the implantation and subsequent brain resection, there may be uncertainty about the exact location of the electrodes with respect to features of brain anatomy such as specific gyral convolutions or lesions. In experiments with a phantom and patients, implanted electrodes were imaged with multiplanar skull radiographs (or CT scans). After retrospective registration with pre-implantation MR data, the electrodes were mapped from these studies onto an MR-derived three-dimensional brain model. The resulting multimodality displays showed the relationship of the electrodes to brain anatomy. In one patient the position of each electrode with respect to a metabolic lesion was also displayed by mapping preimplantation PET data onto the same brain model. This new display of electrode positions may strengthen the interpretation of subdural electrical recordings and thereby reduce uncertainty in planning the resection of epileptic tissue.

Accuracy of object depiction and opacity transfer function optimization in CT volume-rendered images.

Volume rendering is a visualization technique that has important applications in diagnostic radiology and radiotherapy. A methodology is presented for (a) evaluation of the quantitative accuracy of representation of known objects in volume-rendered scenes and for (b) optimization of the opacity transfer function to achieve the most accurate representation of a particular structure. Results using this methodology are shown for structures representing each of the major tissue interfaces and are discussed.

Real-Time Merging of Visible Surfaces for Display and Segmentation

A technique has been developed which combines the high image quality of volumetric rendering with the interactivity associated with surface rendering by texture mapping pre-rendered images onto depth map-based partial models of the surface. An original real-time algorithm is presented which allows us to combine multiple overlapping visible surfaces free of occlusion artifacts associated with poorly estimated fragments. In virtual exploration, decoupling the rendering process from display can result in improved interactivity and navigability without a corresponding loss of image quality. Visible surfaces can also be used as an aid in segmentation for applications requiring object models and to easily combine volumetrically defined structures with purely geometric information (e.g., computed isodose surfaces).

Multimodality intracranial angiography: Registration and integration of conventional angiograms and MRA data

MRA and conventional angiographic data sets are registered in order to produce an integrated 3-D model of intracranial vasculature with high spatial and temporal resolution. The correlation is achieved with help of a 3-D digitizer and a retrospective surface matching program.

Segmenting images by simulated annealing

A new image segmentation paradigm is presented in which a contour is deformed stochastically until it fits around an object of interest. The evolution of the contour is controlled by a simulated annealing process which causes the contour to settle into the minimum of an empirically derived “energy” function.

Computer-assisted Treatment Of Brain Lesions With The Aid Of 3-13 Images

A frameless stereotaxic device can make retrospective use of medical images in order to guide treatment of brain lesions. For example, computer-simulated incision contous, drawn on 3-D brain images derived from MRI data, can be eansfened lxeoperatively to the patient’s head. Radiotherapy play. drawn on CT images. can be used to steer radiation beams over a pabent. The.se applications will be illustrated with experimental data from phantoms, volunteers. and patients.

Distributed parallel volume rendering applied to virtual endoscopy

Improved realism in virtual endoscopy may be achieved by using volume rather than surface rendering, which preserves grayscale information and fine detail from the original CT dataset. Since volume rendering is computationally intensive, parallelization of the calculations is attractive. A distributed parallel rendering algorithm has been applied to calculation of volume rendered cine sequences in a few seconds per frame when using a modest number of readily available workstations.