Objective:

To explore the influence of the region of interest(ROI) range changes on the histogram features that based on the apparent diffusion coefficient(ADC) maps of the different risk prostate carcinoma (PCa).

Methods:

Thirty-six patients admitted in Fudan University Shanghai Cancer Center with prostatectomy were collected in this study. At the largest slice of every lesion

we drew the ROIs of different sizes

and then calculated the histogram features of different risk PCa. The ROIs Were defined as follows. ROI1: We described the ROI on ADC maps built on the confine of the largest lesion slice on T2WI; ROI2: It was based on the border of the largest lesion slice on ADC maps; ROI3 and ROI4: the shape was as same as ROI1

but the area was 3/4 and 1/2 of ROI1

respectively; ROI5: We depicted the areas on all of the whose-lesion slices of ADC maps according to axial T2WI. We mainly performed students t test

and then screen out those parameters that have statistical difference between low- and high-risk PCa. Furthermore

we analyzed the differences of different parameters in different ROIs. In order

to achieve the best parameter

we performed multiparameter receiver operating characteristic (ROC) curveanalysis to ADC values of different ROIs in every percentile.

Results:

The 48 lesions consisted of 17 low-grade (LG) PCa lesions and 31 high-grade (HG) lesions. All of skewnesses and kurtosises had no statistical difference between LG and HG PCa

but in every kind of ROIs

all of mean ADCs and percentiles from LG PCa were higher than those from HG PCa. Some of standard deviation (SD) values in ROI1 and ROI2 between LG and HG PCa had significant difference. Additionally

with the decrease of ROI

all of mean ADC

percentiles

and SD values were reducing except the 10th ADC values

but the respondent gaps of these parameters between LG and HG PCa were decr

easing. Mean ADC

percentiles

and SD values between ROI5 and ROI1 were similar. To ADC values in 10th percentiles

the one of ROI1 had the largest area under curve (AUC)

hadno significant differences comparing with ones of ROI2 (0.891

vs

0.860)

and had significant differences comparing with ones of ROI 3 and ROI 4 (0.891

vs

0.825

0.891

vs

0.817). All of statistical computes were based on

P

＜0.05.

Conclusion:

The area of ROI has an influence on ADC values and percentiles

and the impact extents are different to LG and HG PCa. Histogram analysis can compute additional valuable parameters. The 10th percentile of ADC value

which is based on the largest area of lesions on T2WI

may have nice efficiency to distinguish HG and LG PCa.