Functional MR Imaging in prostate radiotherapy - relationship with prostate histology

Abstract

The treatment of prostate cancer with radiotherapy might be improved by either increasing the radiation dose on the most important tumor areas with focal boosting or by reducing the dose on healthy prostate tissue by means of focal therapy. In both scenarios, selection of the tumor areas is a critical step. As local recurrences often originate at the side of the dominant tumor lesion, especially this region is important. Second criterion is the presence of aggressive characteristics like high Gleason score and high cell density. Multiparametric magnetic resonance imaging (MRI) is the modality of choice for the visualization of tumor areas. In this thesis, focus was on dynamic contrast-enhanced MRI and diffusion-weighted imaging. The main goal of this thesis was to understand the histological background of functional MRI of the prostate, to come to more robust, MRI-based voxel-wise prostate tumor delineation. To this end, we created a logistic regression model based on functional MRI parameters to predict tumor presence on a voxel level. Validation of the model on pathology specimen showed that the performance of the model was good, although not perfect. Correlation with prostate histology revealed in correctly detected tumor voxels high cell density, high microvessel density and a trend towards higher Gleason scores. On the other hand, non-detected tumor voxels had cell densities and microvessel densities resembling normal tissue. This indicates that the detection of voxels depends on their histological characteristics. Additionally, larger tumors volumes were associated with lower apparent diffusion coefficient values and higher expression levels of hypoxia-inducible factor 2α.These larger tumours are detected more frequently and determine outcome after radiotherapy. Altogether these results indicate that the clinically most important tumour voxels, which are likely in need of higher radiation doses, are detected accurately. In conclusion, this thesis addresses several aspects regarding prostate tumor voxel detection and delineation, in relation to histological characteristics. Relevant tumor voxels, most in need of aggressive treatment, can be detected with high accuracy. The obtained knowledge of histological characteristics of detected and missed tumor voxels is of relevance when in the future the treatment of prostate cancer becomes more personalized with focal boost and focal therapy.