Principal component, Varimax rotation and cost analysis of volume effects in rectal bleeding in patients treated with 3D-CRT for prostate cancer.

Overview

We investigate the utility of principal component analysis as a tool for obtaining dose-volume combinations related to rectal bleeding after radiotherapy for prostate cancer. A direct implementation of principal component analysis reduces the number of degrees of freedom from the patient's dose-volume histograms that are associated with bleeding. However, when low-variance principal components are strongly correlated to outcome, their interpretation is problematic. A Varimax rotation is employed to aid in interpretability of the low-variance principal components. This procedure brings us closer to finding unique dose-volume combinations related to outcome but reintroduces correlation, requiring analysis of the overlap of information contained in such modes. Finally, we present examples of cost-benefit analyses for candidate dose-volume constraints for use in treatment planning.