Morphologic perfusion patterns and PI-RADSv2.1 in transition zone prostate cancer

PURPOSE

To evaluate morphologic perfusion patterns in transition zone prostate cancer in multiparametric MRI controlled by in-bore MRI-guided prostate biopsy.

METHODS

Two experienced radiologists evaluated MRI perfusion patterns in consensus from 321 biopsy cores from the transition zone in 141 patients. Transition zone cancer was present in 77 cores in 36 patients. Single early-phase perfusion images were evaluated separately for the presence of a transition zone prostate cancer (consensus tumor early perfusion). The proposed criteria for the perfusion pattern (asymmetry, signal strength, and homogeneity) were rated in consensus for each biopsy position in the presence of the T2w images including the markers of the biopsy trace. We analyzed receiver operating characteristic curves for the PI-RADSv2.1 score and the proposed perfusion pattern.

RESULTS

A logistic regression model with PI-RADSv2.1 and perfusion patterns in early perfusion imaging improved the model fit significantly compared to a model containing only PI-RADSv2.1 (Likelihood Ratio Test, LR = 14.5, p < .001). The AUC was 0.96 for the multiple regression model compared to 0.92 for the PI-RADSv2.1 alone. The evaluation of homogeneity in single early-enhancement images is not inferior compared to the conventional DCE parameter of PI-RADSv2.1 (AUC 0.84 versus 0.83).

CONCLUSION

Morphologic perfusion patterns significantly improve the diagnostic performance of PI-RADSv2.1 in TZ prostate cancer.

Influence of brain tumors on the MR spectra of healthy brain tissue

The neurochemical environment of nontumorous white matter tissue was investigated in 135 single voxel spectra of "healthy" white matter regions of 43 tumor patients and 129 spectra of 52 healthy subjects. Spectra were acquired with short TE and TR values. With the data of tumor patients, it was examined whether differences were caused by the tumor itself or aggressive tumor therapies as confounding factors. Comparing the spectra of both classes, an excellent differentiation was possible based on the metabolite peak of N-acetylaspartate (P ≈ 0) and myoinositol (P < 0.03). The area under curve of the receiver operating characteristic was calculated as 0.86 and 0.62, respectively. With linear discriminant analysis using combinations of integrals, a prediction was possible, whether a spectrum belonged to the patient or the healthy subject class with an overall accuracy above 80%. The confounding factors could be ruled out as source of the differences. The results show strong evidence for an influence of malignant growth on the biochemical environment of nontumorous white matter tissue. Because of the T(1) weighting, the measured differences between both classes were most likely concentration changes interfered by T(1) effects. The underlying processes will be subject of future studies.