Keywords:
Tissue characterisation, Neoplasia, Cancer, Statistics, Contrast agent-intravenous, Computer Applications-Detection, diagnosis, MR-Diffusion/Perfusion, Image manipulation / Reconstruction, Pelvis, Oncology, Computer applications
Authors:
E. Aguado-Sarrió, R. Sanz-Requena, L. Marti-Bonmati, J. M. Prats-Montalbán, A. Ferrer-Riquelme; Valencia/ES
DOI:
10.1594/ecr2018/C-0452
Methods and materials
In this IRB-approved retrospective study, 30 histologically confirmed peripheral prostate tumors were studied with 3T MRI,
including DCE-MR series with high temporal resolution.
Tumor ROIs were manually segmented in areas with pathologic confirmation (15 Gleason-6 and 15 Gleason-7) as delimitated on the T2-weighted/Diffusion-weighted images.
Pharmacokinetic parameters from first (extended Tofts) and second (2-compartment exchange,
adiabatic approximation to tissue homogeneity and distributed parameter) generation models were calculated and combined with the dynamic parameters directly associated to the physiological perfusion phenomena obtained from the latent variable based model MCR-ALS (Figure 1).
PLS-DA with double cross validation was applied to predict the tumor aggressiveness,
assessed by the Gleason score.
This analysis was performed individually (only one biomarker at a time) and globally,
including all the biomarkers considered.
Multiple PLS-DA iterations were calculated to compare the performance of all the different individual and combined approximations.