Preliminary Study Of MR DWI With Different Mathematical Models Of Renal Masses

Objective:To evaluate the value of MR DWI parameters, derived from different mathematical models, in differentiating renal masses. To compare the ability of different mathematical models in representing the signal attenuation pattern of renal masses.Materials and Methods:Twenty-seven patients from Jul2012to Nov2012in Tianjin Medical University were recruited for this study, including sixteen males and eleven females, aged between24and83years, mean60.5years. A total of27lesions in27patients, sized1.3-11.0cm, mean diameter5.4cm. Among of them, nineteen patients were with clear cell renal cell carcinomas(ten cases of male, and nine female patients, aged between47and78years, median age59years), including four cases of nuclear grading Ⅰ, nine cases of nuclear grading Ⅱ, and six cases of nuclear grading Ⅲ. The others included one papillary renal cell carcinoma, one chromophobe renal cell carcinoma, two transitional renal pelvic carcinoma, two metastatic tumors, one leiomyoma, and one angiomyolipoma. All MR examination was performed on a3.0T magnetic system with eight-channel torso coil. Coronal FIESTA, axial fat-suppressed T2WI, axial T1WI and axial fat-suppressed T1WI were acquired on both kidney region. DWI examinations with ten different b value, included(0、20s/mm2)、(0、50s/mm2)、(0、100s/mm2)、(0、200s/mm2)、(0、400s/mm2)、(0、800s/mm2)、(0、1200s/mm2)、(0、2000s/mm2)、(0、3000s/mm2),were obtained, within a single breath-hold time of21s for each scan. To measure the signal intensity of solid portion of the lesions of ten b values, and measuring traditional ADC value of ADC800、ADC1200、ADC2000、 ADC3000.MATLAB (R2012a) was used to perform data fitting, derived relevant parameters of different mathematical models. ADCmono, R_square of monoexponential model. D, f, R_square of biexponential model. DDC, α, R_square of stretched exponential model. Statistical analysis was performed using SPSS software(version17.0). The value of DWI parameters was compared with each other in diagnosising renal masses. Receiver operating characteristic (ROC) analysis was introduced to determine the threshold for differential diagnoses and to compare the abilities of different mathematical models in representing signal attenuation pattern of renal masses. There was a significant difference when P<0.05.Result:There was either no significant difference of all DWI parameters between benign and malignant masses of the kidney, or no significant difference among different nuclear grading. ADC2000, ADC3000, ADCmono, D had significant difference between ccRCC group and non-ccRCC group. All parameters(except a) have significant difference between ccRCC group and malignant non-ccRCC group, ADC800、ADC1200, ADC2000, ADC3000, ADCmono, D, DDC, f values of ccRCC group were higher than malignant non-ccRCC group. For non-ccRCC, there was significant difference of f value between benign and malignant masses (P<0.05). Both AUC of ADCmono and DDC exceed0.9. The specificity of ADC3000, ADCmono, D, DDC, f values in differentiating ccRCC from malignant non-ccRCC were all as high as100%. D value was significantly lower than ADCmono and DDC values. There was significant correlation among D, ADCmono,and DDC. DDC value was higher than ADCmono value, with significant difference. Significant correlation was found between f and ADCmono, DDC, for the pearson correlation coefficient r was0.824and0.863, and weak correlation between f and D(r=0.534). R_square value of monoexponential model was lower than that of biexpoential and stretched exponential models. But there was no significant difference between biexpoential and stretched exponential model.Conclusion:Compared with monoexponential model, biexpoential and stretched exponential models could represent signal attenuation pattern of renal masses better. ADC derived from multiple b values has better diagnostic efficiency. Biexponential mode afforded parameters related to perfusion, and f combined with D had the optimal value in renal masses diagnosis.