Abstract
DATA CONCLUSION
Regions of prostate cancer lesions that are invisible on mpMRI have different histopathology features than visible regions. This may have implications for monitoring during active surveillance and focal treatment strategies.
RESULTS
Small, anterior, and ISUP grade 1-2 lesions (n = 34) were missed more frequent than large, posterior, ISUP grade ≥ 3 lesions (n = 35). Invisible regions on mpMRI had lower tumor density, heterogeneous tumor morphology, and were located in the transition zone. Both T2 and ADC values were higher in "intermediate" compared with "dense" regions (P = 0.002 and < 0.001) and in regions with heterogeneous compared with homogeneous morphology (P < 0.001 and 0.03). Ktrans was not significantly different (P = 0.24 and 0.99).
BACKGROUND
Previous studies have reported tumor volume underestimation with multiparametric (mp)MRI in prostate cancer diagnosis.
FIELD STRENGTH/SEQUENCE
T2 -weighted, diffusion-weighted imaging, T2 mapping, and dynamic contrast-enhanced MRI on two 3T systems and one 1.5T system.
POPULATION
Thirty-four patients with biopsy-proven prostate cancer scheduled for prostatectomy (median 68.7 years).
LEVEL OF EVIDENCE
3 Technical Efficacy: Stage 3 J. Magn. Reson. Imaging 2020;51:1235-1246.
ASSESSMENT
Two readers delineated suspicious lesions on mpMRI. A pathologist delineated the lesions on histopathology. A patient-customized mold enabled the registration of histopathology and MRI. On histopathology we identified mpMRI visible and invisible lesions. Subsequently, within the visible lesions we identified regions that were visible and regions that were invisible on mpMRI. For each lesion and region the following characteristics were determined: size, location, International Society of Urological Pathology (ISUP) grade, and Gleason subpatterns (density [dense/intermediate], tumor morphology [homogeneous/heterogeneous], cribriform growth [yes/no]).
PURPOSE
To investigate why some parts of lesions are not visible on mpMRI by comparing their histopathology features to those of visible regions.
STATISTICAL TESTS
With generalized linear mixed-effect modeling we investigated which features explain why a lesion or a region was invisible on MRI. We compared imaging values (T2 , ADC, and Ktrans ) for these features with one-way analysis of variance (ANOVA).
STUDY TYPE
Retrospective.