Introduction In biological tissues, microscopic motion detected by diffusion-weighted MRI (DW-MRI) includes both diffusion of water molecules in tissue (influenced by the structural obstacles) and perfusion i.e. microcirculation of blood in the capillary network. To date, most clinical DW-MRI studies have employed a monoexponential model for data analysis, which produces a single parameter – the apparent diffusion coefficient (ADC) – for assessing diffusion characteristics whose value correlate with Gleason scores in prostate cancer.¹ The intravoxel incoherent motion (IVIM) model introduce a fast diffusing component in the signal due to perfusion effects, which affects the overall signal predominantly at low b values, but the application of this method is not yet widespread included in clinical practice. A recent paper stated that ADC maps provide better diagnostic performance than IVIM maps for prostate carcinomas detection² and suggested that the reduction of ADC observed in prostate cancer can derive not only from changes in cellularity but also from perfusion effects.This study was aimed at measuring the water diffusion in tissue separating and estimating the perfusion component by adopting a simple protocol easy to be implemented in clinical practice and evaluate its diagnostic performance.

Patients: Thirty-four patient with suspected PCa underwent mp-MRI of the prostate performed at 3.0 T (GE MR750). Among them, 26 patients were histologically-proven PCa.

MRI: T2w images were acquired using a T2w, radial multi-shot PROPELLER FRFSE sequence for minimize motion artifacts (FOV=22cm, TR/TE= 7000ms/110ms, slice thickness 3mm, matrix 320x320, echo train length 16, NEX=3). DWI images were acquired using a single shot echo planar imaging (FOV=22cm, TR/TE 4000ms/80ms, slice thickness 3mm, matrix 96x128, NEX=2 (for b=0,200,500), 4 (b=1000), 8 (b=3000)) and two sets of b values : 1) b-values 0,500,1000,3000s/mm², a protocol widely adopted for measure ADC, which include the fast diffusing component (perfusive) component, or 2) b=200,500,1000,3000s/mm² to exclude the perfusive contribution (we call it ADC-P parameter).

Tumor delineation: Two radiologists blinded for histology delineated probable tumor based on mp-MRI, and on T2w, diffusion-weighted imaging (DWI) (b3000), and apparent diffusion coefficient maps separately. DWI signal for ADC analyses were taken in pathologic (ROI1), contralateral (ROI2) and central gland (benign prostate hyperplasia, BPH) regions (ROI3).

Data analysis: The ADC parameters were derived from protocol 1 and 2 as monoexponential fit (ADC and ADC-P). Spearman’s rank correlation test was performed to determine the relationship between ADC, ADC-P, PIRADS with Gleason score and Spearman’s Rho was calculated. ROC analysis and Partial Least Squares Discriminant Analysis (PLS-DA) were applied to compare the diagnostic performance of ADC, ADC-P and PIRADS in distinguish between low grade (LG) and intermediate/high grade (HG) PCa when compared with Gleason score (LG, Gleason score ≤6; HG, Gleason score >6) . Moreover, we used a two-tailed independent samples t-test to investigate differences in ADC and ADC-P parameters. Statistics was performed using SPSS 22 and p=0.05 was considered significant.

In agreement with the findings of several other groups³, the ADC determined in this study was significantly decreased in PCa compared to contralateral and BPH. Differences in the ADC values measured by using the two different protocols were detected for all the PIRADS and Gleason scores in all ROIs (Figure 1). A significant negative correlation between ADC and Gleason score (Rho=-0.443, p=0.024), between ADC-P and Gleason score (Rho=-0.464, p=0.017), the differences in the two ADC (the perfusion component) and Gleason score >6 (Rho=-0.453, p=0.023) was observed. A positive correlation with PIRADS and Gleason score > 6 (Rho=0.666, p<0.0001) was found, as expected. The area under the curve (AUC) for predicting pathologic high risk patients with ADC, ADC-P and PIRADS were 0.76, 0.78 and 0.88, respectively. PLS-DA found that both ADC and ADC-P have a better diagnostic performance of PIRADS in distinguish between LG and HG PCa. (Figure 2). Interestingly, the differences between ADC and ADC-P in HG PCa (with Gleason score > 6) were significant reduced with respect to LG PCa (Gleason score ≤6), being 0.6±0.5 and 1.1±0.5 mm²/s in HG and LG, respectively (p=0.04).

Our data showed the presence of a perfusive contribution in normally acquired ADC of PCa, which can be easily excluded by using b values larger than 150 s/mm² in DWI acquisition. This contribution is significantly reduced in histologically-proven HG PCa thus indicating different microscopic tissue structure features, and can help in the differential diagnosis from significant tumor to indolent disease.