Lectures and seminars Sfo Epi Seminar in Biostatistics: Development of a multiparametric MR classifier for prostate cancer
Welcome to attend the SfoEpi Seminar Series in Biostatistics!
Title: Development of a multiparametric MR classifier for prostate cancer
Joseph Koopmeiners, Mayo Professor and Head, Division of Biostatistics, School of Public Health, University of Minnesota
Multiparametric magnetic resonance imaging (mpMRI) represents a powerful tool for developing a non-invasive, user-independent tool for the detection of prostate cancer. Previously, our group showed that a voxel-wise classifier that combined quantitative mpMRI parameters from multiple modalities resulted in better classification of prostate cancer than any single parameter, alone. While these results are promising, there are several unique features of the mpMRI data that can be leveraged to improve classification accuracy. First, the prostate has a unique anatomical structure, whereby it can be segmented into multiple zones, with the largest being the central gland and peripheral zone. Both the mpMRI parameters and the prevalence of cancer are associated with region, and leveraging this information can potentially improve classification. Second, there is substantial spatial correlation in both the mpMRI parameters and the voxel-wise cancer status, but developing a classifier that leverages this information is challenging due to the high dimensionality of the data. Finally, while data are available on the voxel-level, voxel-wise predictions must be translated into contiguous lesions for clinical use in imaging-guided biopsy and focal therapy. In this talk, I will discuss the development of quantitative methods that allow us to leverage the unique features of our data to improve prostate cancer classification and localization. First, I will discuss the development of a voxel-wise prostate cancer classifier that accounts for the anatomical structure of the prostate and spatial correlation in the data. I will then discuss a novel Bayesian functional spatial partitioning algorithm that uses functional estimation tools to estimate smooth boundary curves of arbitrary shape. I will illustrate our methods through an application to a data set consisting of mpMRI of the prostate co-registered to pathology.
If you have any questions, please contact Erin Gabriel at Erin.Gabriel@ki.se