DCE-MRI has become an important protocol in mpMRI analysis of prostate cancer and it has been quantified typically using pharmaco-kinetic modelling and the estimated parameters are then used with other approaches (machine learning or deep learning (DL)) to characterize/discriminate tumor tissue against healthy tissue. However, it is not clear if applying DL to the DCE-MRI time series directly is beneficial for prostate cancer detection. Hence, we propose a DL based method to differentiate prostate tumor from healthy tissues at the voxel level using raw arbitrary signal DCE time-series itself. Overall, DL based tumor characterization provided similar detectability for prostate tumor when compared to Ktrans and ve maps. We also evaluated differences in tumor characterization when contrast agent concentration time-curves were used instead of arbitrary signal curves and found them to provide similar detectability.
This abstract and the presentation materials are available to members only; a login is required.