DCE-MRI dispersion analysis for quantitative angiogenesis imaging in prostate cancer

Pd Eng Thesis

Turco, S. (2015). DCE-MRI dispersion analysis for quantitative angiogenesis imaging in prostate cancer. Eindhoven: Technische Universiteit Eindhoven. ((Co-)promot.: Massimo Mischi & Hessel Wijkstra). Lees meer: Medialink/Full text



Angiogenesis plays a fundamental role in cancer growth and the process leading to the formation of metastasis. It consists of a chaotic and dense assembly of irregular, tortuous, and leaky microvessels that are intended to feed the tumor. Imaging angiogenesis may therefore provide a powerful tool for cancer detection and accurate localization, which is required for optimal management of the disease, including active surveillance, therapeutic decision-making, focal therapy guidance, and monitoring the response to chemo- and antiangiogenic therapies. To this end, several contrast-enhanced imaging methods have been developed for in-vivo, non-invasive assessment of the vascular changes occurring during tumor angiogenesis. In dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), pharmacokinetic modelling of the contrast leakage from the blood vessels into the tissue is used to address the changes in vascular permeability. However, information about the changes in the microvascular architecture is currently lacking. In this work, two models characterizing the intravascular dispersion kinetics of an extravascular contrast agent are presented and adopted to assess the microvascular architecture. Fitting concentration time curves measured with DCE-MRI by the proposed models leads to the generation of parametric maps of the vascular architecture and of the extravascular leakage, enabling the visualization of cancer neo-angiogenesis. Furthermore, three solutions to improve the computational efficiency of the proposed methods are presented and tested in prostate cancer. The estimation accuracy, precision, repeatability, computation time, and robustness-to-noise of the proposed solutions were tested by dedicated simulations. Eventually, a clinical validation was performed on 15 patients with biopsy-proven prostate cancer. Voxel classification by the selected best solution yielded sensitivity, specificity, and ROC curve area equal to 81.0%, 86.3%, and 0.91, outperforming by over 10% currently used methods. The results motivate further research with a larger dataset to establish the promising role of magnetic resonance dispersion imaging (MRDI) for prostate cancer localization.