Multi-parametric Magnetic Resonance Imaging (MRI) in Prostate Cancer

• Main Author: Langer, Deanna Lyn
• Other Authors: Trachtenberg, John
• Language: en_ca
• Published: 2010
• Subjects: magnetic resonance imaging; prostate cancer; 0760
• Online Access: http://hdl.handle.net/1807/24802

Prostate cancer is extremely prevalent, with shifting patient demographics leading to an increasing number of men balancing treatment efficacy with associated side-effects. Non-invasive characterization of disease – useful for guiding biopsy, to monitor disease progression during active surveillance, or for treatment planning of focal therapies – could have a significant impact on patient management. Through its excellent anatomic imaging capabilities and its ability to characterize physiologic properties, magnetic resonance imaging (MRI) has the potential to fulfill clinical goals; however, further improvements are necessary to maximize accuracy and impact. Thus, this thesis presents: 1) the development of a multi-parametric model to combine parameters derived from measurement of T2 relaxation, diffusion weighted imaging, and dynamic contrast-enhanced MRI to improve the discrimination between normal and malignant peripheral zone tissue; 2) determination of the impact that the presence of normal tissue within regions of tumour has on the measurement of apparent diffusion coefficient (ADC) and T2 relaxation in the peripheral zone; and 3) relationships between MRI measurement and underlying prostate tissue composition. A common patient cohort was used for all studies, with prostate cancer patients having in vivo MRI prior to prostatectomy followed by whole-mount histologic sectioning of the surgical specimens, facilitating the use of pathology as a gold-standard for all analyses. In the first study, the optimal multi-parametric model combines ADC, T2, and volume transfer constant (Ktrans) to yield the probability of malignancy for each voxel. Performance of the model is better than each single parameter, but not significantly so compared to ADC. The second study demonstrates that there is no difference in ADC and T2 between tumours containing significant portions of normal tissue and the surrounding normal tissue itself, indicating that full characterization of prostate cancer with MRI may be limited. Finally, by determining relationships between MRI parameters and tissue characteristics, the third study suggests mechanisms driving MR image appearance in the prostate, including the visualization of cancer. Taken together, this thesis presents potential improvements to prostate cancer imaging, and provides further insight into the interplay between the underlying histology and MRI.