In this study, the authors’ aim was to investigate a previously developed radiomics-based biparametric magnetic resonance imaging (bpMRI) approach used for the discrimination of clinically significant peripheral zone prostate cancer (PZ csPCa) through the use of multi-center, multi-vendor (McMv) and single-center, single-vendor (ScSv) datasets. Using these datasets, the authors were able to determine that a single-center trained radiomics-based bpMRI model does not generalize to multi-center data. However, multi-center trained radiomics-based bpMRI models do generalize, have equal single-center performance, and perform better on multi-center data.
Key points
- Multi-center radiomics-based bpMRI models generalize to new multi-center and single-center data.
- A single-center, single-vendor radiomics-based bpMRI model lacks multi-center generalization.
- Multi-center developed models match the single-center validation of a single-center model.
- Multi-center developed models outperform single-center models in a multi-center validation.
Authors: Jeroen Bleker, Derya Yakar, Bram van Noort, Dennis Rouw, Igle Jan de Jong, Rudi A. J. O. Dierckx, Thomas C. Kwee & Henkjan Huisman