Report from ISMRM: Prostate imaging studies weigh diagnostic power of MRI methods

May 7, 2008

A clinical trial presented Monday demonstrated how complicated the quantification of dynamic contrast-enhanced MRI of suspected prostate cancer can be.

A clinical trial presented Monday demonstrated how complicated the quantification of dynamic contrast-enhanced MRI of suspected prostate cancer can be. Investigators hold great hope for DCE-MRI for this application because of the insensitivity of contrast-enhanced T2-weighted imaging and the considerable convenience of performing DCE-MRI without the need for an endorectal coil. Presenter Guang Jia, a research student at Ohio State University, and colleagues under the direction of Dr. Michael Knopp evaluated 27 patients who were awaiting robotic prostatectomy for clinically proven prostate cancer. DCE-MRI was performed on a 3T Philips Achieva scanner using an eight-channel phased-array coil. Several semiquantitative parameters were calculated from the dynamic contrast data and compared with stained histological samples of the removed prostate and seminal vesicles. Regions of interest were drawn to identify cancerous regions, including tumor in the peripheral and transition zone, the noncancerous peripheral zone, and the central gland with benign prostate hyperplasia, muscle, and neurovascular bundle.

The researchers reported the various measures and their corresponding results:

  • Time to maximum signal enhancement. The measure of tmax in the tumor region was significantly shorter than those acquired in noncancerous peripheral zones.
  • Wash-out score (the relative difference between maximum signal enhancement and signal intensity at the end of the dynamic scan). This measure was significantly larger in the tumor region than in the other regions.
  • Maximum enhancement ratio. This was unable differentiate between tumor and the central gland.
  • Three separate measures of area under the curve ROC analysis at 60, 90, and 180 seconds. All three were unable to differentiate between tumor and central gland.

Jia found that all parameters could differentiate tumor from the noncancerous peripheral zone. Tumor perfusion showed faster wash-in, resulting in a short tmax and higher enhancement and producing a larger maximum enhancement ratio, and faster wash-out, leading to a larger wash-out score than noncancerous peripheral zone perfusion.

Only tmax and measures acquired using two different sets of contrast diffusion assumptions, however, could differentiate tumor from central gland. Dr. Sophie Riches from the clinical MR group at Royal Marsden Hospital in the U.K. used basically the same experimental framework on 20 presurgical prostate cancer patients to establish the values for histopathologically defined regions of tumor, normal peripheral zone, and normal central gland tissue for T2-weighted MRI, diffusion-weighted DCE-MRI, and H-1 MR spectroscopy. This elaborate exercise led Riches to conclude that functional changes in tumor tissue are more useful than changes observed in T2-weighted imaging for determining the actual extent of tumor in the prostate gland. Imaging was again performed on a 3T scanner. The functional techniques were unable to distinguish between tumor and central gland, however. Riches attributed this problem to increased variation of all parameters in the central gland compared with the peripheral zone. University of Maryland radiologists also weighed in with their own evaluation comparing imaging results and histopathology to assess the relative merits of 3D MR spectroscopic imaging (MRSI) and DCE-MRI on 14 patients. Imaging was performed on a 1.5T system. The study, performed by Dr. Bao Zhang and presented by senior researcher Dr. Rao Gullapalli, concluded that MRSI and DCE-MRI are able to differentiate prostate tumor from normal tissue but are unable to determine tumor grade. An ROC analysis suggests that MRSI is far more specific than DCE-MRI for detecting prostate cancer (85% versus 60%), but the sensitivity rates of the two are about the same (68% and 65%), Gullapalli said. For more online information, visit Diagnostic Imaging's ISMRM Webcast.