Prostate cancer kills hundreds of thousands of men each year and early disease detection and effective treatment solutions constitute significant areas of research in the medical domain. Multi-parametric MRI (mpMRI) is a standardly used technique to noninvasively image prostate tissue and it is the recommended imaging modality for various automated prostate cancer detection studies, some of them in the newly developed field of radiogenomics, which combines imaging features (radiomics) with genomic information to characterise the tumor phenotype. As part of a larger radiogenomics project, this study focuses on the radiomics features, specifically investigating the impact of different feature extraction settings on the final parametric maps and feature statistics. Three feature statistics were determined to be of poor robustness when feature extraction settings were modified and alternative feature statistics were recommended, with a focus on not compromising the discriminative power of the feature statistics set.
Evaluation of multiparametric MRI feature extraction settings influence for prostate cancer detection
Bărbulescu, R. (Author). 15 Jun 2022
Student thesis: Bachelor