This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Genital / Reproductive system male, Oncology, Pelvis, MR, MR-Functional imaging, MR-Diffusion/Perfusion, Observer performance, Localisation, Neoplasia, Cancer
Authors:
M.-R. Kwon, C. K. Kim, J. J. Park, J. G. Kim, J. Lee, S. Y. Yang, M. Seong; Seoul/KR
DOI:
10.1594/ecr2017/C-2874
Results
The patient and cancer characteristics are summarized in Table 1.
Table 2 presents the visual assessment results on DWI1000 and DWI1500 for detecting the CSCs for the both readers (Figures 1‒3).
The sensitivity,
specificity and diagnostic accuracy for the experienced reader were the same between DWI1000 and DWI1500 (p> 0.05).
The sensitivity for the less-experienced reader was greater at DWI1500 (80%) than at DWI1000 (76.7%),
which was not significant (p > 0.05).
The specificity for the less-experienced reader was greater at DWI1000 (50%) than at DWI1500 (25%),
which was not significant (p > 0.05).
The diagnostic accuracy for the less-experienced reader was greater at DWI1500 (77.6%) than at DWI1000 (75.5%),
which was not significant (p > 0.05).
For PI-RADS scoring in all patients,
the inter-modality agreement between DWI1000 and DWI1500 was excellent in both experienced and less-experienced readers (κ = 0.905 and 0.880).
The inter-reader agreement for DWI1000 and DWI1500 was excellent in both readers (κ = 0.911 and 0.877).
For each PZ and TZ scoring,
the inter-modality agreements between DWI1000 and DWI1500 were excellent in both readers (all κ ≥ 0.871),
and inter-reader agreements for DWI1000 and DWI1500 were excellent (all κ ≥ 0.878).
For predicting the PZ CSCs,
the AUC of DWI1000 and DWI1500 was 0.714 and 0.753 for reader 1 and 0.690 and 0.639 for reader 2,
respectively,
which were not significantly different (p = 0.12 and 0.45).