Targeted Prostate Biopsy Using Mathematical Optimization R01-CA104976
While prostate cancer is the second leading cause of death for American men, it cannot, in the majority of cases, be reliably detected. Once elevated prostate specific antigen (PSA) levels are measured, the gold standard for the diagnosis and staging of prostate cancer detection is needle biopsy. To date, however, no mathematically rigorous attempt has been made to precisely determine where needles should be placed to maximize the probability of cancer detection. This collaboration between the NCIGT and the University of Pennsylvania is developing and clinically testing a computer-based methodology for optimal sampling of the prostate during biopsy so that the probability of cancer detection is maximized based on statistics obtained by applying advanced image analysis methodology to whole-mounted sections of radical prostatectomy specimens.
The group has performed significant technical developments along the direction of the optimization of needle positioning, constructed a statistical atlas of prostate cancer, and completed the optimized biopsy system and its validation using cross-validation. Analyses include the
- incorporation of physical constraints separately for transperineal and transrectal biopsies
- incorporation of needle positioning uncertainty
- use of a realistic geometric model of a needle
- use of extrapolation schemes to complete the atlas near the apex and the base where missing data exists due to difficulties with specimen sectioning
The group has also made significant progress towards automated segmentation of transrectal ultrasound and MR prostate images, a critical step. To ultimately achieve the expected biopsy accuracy, researchers need to accurately place needles. This, in turn, necessitates accurate algorithms for extracting the prostate capsule and urethra and for deforming the population-based statistical atlas of prostate cancer onto the patient images, thereby transferring the optimal biopsy plan. Finally, the group has incorporated this software into 3D Slicer to aid clinical intervention. To date, in over 20 MR-guided biopsy cases, the group successfully registered and sampled atlas-based targets. Target registration and sampling was done using automated registration pipeline and software that was developed for visualization by the physician in the bore of the MR scanner. The visualization software was successfully ported to the 3D Slicer, an open source package available for other research groups. Slicer has the ability to display targets labeled Penn1, Penn2, ... Penn7 corresponding to atlas-based targets and to provide the physician with information on the distance from the needle location to the intended target site. It also has the ability to direct the MR scanner to an image in a plane containing the target and needle.
As part of the collaboration, an integrated pipeline for the registration of optimized target sites to imaging obtained in the Brigham operating room during the MR biopsy procedure was created. This involved the installation and testing at the Brigham of registration software developed at the University of Pennsylvania, to map optimized targets onto pre-operative imaging. It also involved automating the use of this preoperative registration in programs developed at the Brigham for pre- to intra-operative image registration. Software for in-bore visualization of marked targets and needle locations during the biopsy procedure was modified to provide feedback to the physician on needle placement accuracy. Using this registration procedure and visualization technique for guidance, we successfully performed optimized targeted biopsies in over a dozen patients, along with standard sextant sampling and sampling based on focal targets seen in preoperative MR.
Publications
- Tempany CMC, Straus S, Hata N, Haker S. MR-guided prostate interventions. J Magn Reson Imaging, 2008 Feb;27(2):356-67. PMID: 18219689.
- Nain D, Haker S, Bobick S, Tannenbaum A. Multiscale 3-D shape representation and segmentation using spherical wavelets. IEEE Trans Med Imaging. 2007 Apr;26(4):598-618. PMID: 17427745.
- Nain D, Haker S, Bobick A, Tannenbaum A. Shape-Driven 3D Segmentation Using Spherical Wavelets. Int Conf Med Image Comput Comput Assist Interv. 2006;9(Pt 1):66-74. PMID: 17354875.