Recently, a statistical model based on the gamma distribution (gamma model) has been proposed to characterize the non-Gaussian diffusion behavior in prostate cancer^{1, 2}. The gamma model proved suitable for describing diffusion signal decay curves in prostate cancer and may provide better correlation between diffusion signal decay and histological information in prostate cancer using the acquired parameters, that is, the area fraction for D<1.0×10^-3mm²/s [Frac<1.0] and D>3.0×10^-3mm²/s [Frac>3.0]). In this study, we aimed to investigate whether the parameters obtained from the gamma model could distinguish a Gleason 6 from a Gleason 7 or higher disease, and help to improve the prediction of insignificant prostate cancer in active surveillance (AS) candidates.

Gamma model

Gamma model is a type of the statistical model that presumes a continuous distribution of diffusion coefficients within the imaging voxel. Thus, a continuous distribution of diffusion coefficients can be obtained even from a single voxel. Recent studies have demonstrated that a statistical model based on the gamma distribution was more suitable than that based on the Gaussian distribution for prostate cancer^{1, 2}.

Diffusion-weighted imaging (DWI)

All examinations were performed on a 3T or 1.5T MRI scanner (Achieva 3T and Ingenia 1.5T, Philips Healthcare, Eindhoven, the Netherlands) using 6–32 channel coils. Sixty-three foci of prostate cancer from 59 patients (mean age, 66.3±5.0 , mean PSA, 11.4±7.9 ng/ml) who received MRI prior to undergoing radical prostatectomy were included in this study. DWI parameters were as follows: TR/TE=4277–6499/40–69ms, 3 orthogonal diffusion gradients applied, b=(0, 10, 20, 30, 50, 80, 100, 200, 400, 1000) s/mm² (n=13), b=(0, 200, 400, 1000, 1500) s/mm² (n=15), b=(0, 500, 1000, 1500, 2000) s/mm² (n=31).

Data analysis

1) Ability to Differentiate Gleason score 6 from Gleason score≥7

Regions of interest (ROIs) were placed in cancer on DWI, and signal intensities were measured for each b-value with a copy-paste operation. The measured signal intensities versus b-value curves were fitted to the gamma model. The parameters of the gamma model (Frac<1.0, 0.8, 0.5, Frac>3.0) were compared between Gleason score 6 and Gleason score≥7 using unpaired Student's t tests.

2) Ability of the parameters to predict insignificant prostate cancer among AS candidates

We selected 14 AS candidates who met the criteria for AS (a Gleason score 6 on biopsy, PSA<10ng/ml, clinical stage T1c or T2 disease, and positive biopsy cores≦25%). Correlation between the parameters (including ADC) and presence of a Gleason score≥7 or extraprostatic extension at final pathology (adverse pathologic outcome) were assessed using unpaired Student t tests and receiver operating characteristic (ROC) analyses.

Frac<0.8 and Frac<0.5 were significantly higher in Gleason ≥ 7 prostate cancer than in Gleason 6 prostate cancer. Frac<1.0 and Frac>3.0 were not significantly different between Gleason ≥ 7 and Gleason 6 prostate cancer (Fig.1). With regard to AS candidates, there were six organ-confined Gleason 6 cancers and eight Gleason≥7 cancers. Frac<1.0, 0.8, 0.5 were significantly higher, and Frac>3.0 and ADC were significantly lower in patients with adverse pathologic outcomes (Fig.2). ROC analyses for predicting adverse pathologic outcomes in AS candidates showed that the area under the curves (AUC) of the parameters were from 0.81 to 0.88, and were not significantly different among them (Fig.3).AS has emerged as an alternative to radical treatment in men with low-grade prostate cancer. The most important issue for successful AS is selection of the patients eligible for AS. ADC seems to be a promising metric for patient selection on AS, however, the utility of ADC in predicting clinical outcomes in AS candidates still remains controversial^{3, 4}. We have found that the parameters of the gamma model as well as ADC to be useful for predicting insignificant prostate cancer. There are some limitations in the current study. First, our study was a retrospective analysis with a limited number of patients. Larger patient populations with a broad range of tumor grades are necessary for further investigation. Second, we have analyzed DWI data sets with uneven maximum b-values. The different maximum b-values may impact on our results.The parameters obtained from gamma model were useful to distinguish a Gleason 6 from a Gleason 7 or higher disease. DWI using the gamma model might help to improve the prediction of insignificant prostate cancer in AS candidates.