Computation Core
William "Sandy" Wells, Ph.D., Core PI
Cycle II 2010-2015
Objectives
The overall goal of the NCIGT Computation Core is to develop technology that facilitates the integration of spatial information for image-guided therapy design and guidance. To this end we will pursue research and development of computer algorithms that are aimed at the integration of pre- and intra-operative images, and other information, for the guidance of prostate brachytherapy, neurosurgery, and external beam radiation therapy. In addition, we will serve as a software engineering resource for the NCIGT and collaborators.
Neurosurgical navigation systems have reduced the risk of complications from surgery and have allowed surgeons to remove tumors that were once considered inoperable. Recent neurosurgical navigation systems enable surgeons, during a procedure, to align pre-operative neuro-images, e.g., MRI, with the current configuration of the patient in the procedure. Via that alignment, the surgeon can more accurately determine the location and extent of a tumor during a procedure. These navigation systems frequently employ hand-held probes that may be positioned in the operative field for interactive exploration of surgical margins in reference to the planning images. Analogous navigational methods have the potential to provide similar benefits in brachytherapy of the prostate that is guided by intraoperative trans-rectal ultrasound (TRUS) images and in other radiation therapy (RT) applications.
Many techniques used to align pre- and intra-operative images have difficulties when tissue deformations (e.g., brain shift or prostate deformation from endorectal MRI coil) occur. Typically, fiducials, landmarks, and rigid-body transforms are used to align pre- and intraoperative images. Those methods can be easily and rapidly initialized, monitored, and updated during a surgical procedure, and they provide sufficient accuracy in many surgical situations. However, they have a limited ability to account for deformations that occur in patients’ tissues during, for example, craniotomies or prostate biopsies. Deformations commonly arise from tumor resection, from the use of various instruments and devices, from gravitational effects on the organ, and from the use of hyperosmotic drugs. The limited ability to account for these deformations results in uncertainties concerning the location and extent of a tumor. As a result, the surgical procedure may not reduce the tumor bulk as much as desired and may unnecessarily affect healthy adjacent tissues.
Specific Aims
Specific Aim 1: MR to Ultrasound Registration for Neurosurgery and Prostate Brachytherapy.
Specific Aim 2: CT- Xray Registration Technology for Radiation Therapy.
Specific Aim 3: Slicer Engineering.
NCIGT Publications
- Sundaram P., Wells III W.M., Mulkern R.V., Bubrick E.J., Bromfield E.B., Münch M., Orbach D.B. Fast Human Brain Magnetic Resonance Responses Associated with Epileptiform Spikes. Magn Reson Med. 2010 Dec;64(6):1728-38. PMID: 20806355.
- Risholm P., Pieper S., Samset E., Wells III W.M. Summarizing and Visualizing Registration Uncertainty in Non-Rigid Registration. Int Conf Med Image Comput Comput Assist Interv. 2010;13(Pt 2):554-561. PMID: 20879359. PMCID: PMC2976974.
NCIGT Resources
- Risholm P., Pieper S., Samset E., Wells III W.M. Pre- and Post-operative MRI Dataset with Resection. SPL Brain Resection Data 2010 September;
- 3D Slicer
Cycle I 2005-2010
Objectives
The Computation Core provides the computing infrastructure advancements required by the NCIGT Cores that will then extend to the IGT community. Its primary focuses are: engineering 3D Slicer, the NCIGT's surgical navigational platform; high performance computing, the improvement of the efficiency and effectiveness of medical image analysis areas involved in image guided therapy; and spatial neuroinformatics, the discovery of and improvement on information processing requirements of the NCIGT's Neurosurgery Core.
Projects
Some of the NCIGT's projects within which the Computation Core provides leadership and central input are: