[Measurement reproducibility of parameters derived from introvoxel incoherent motion diffusion-weighted MRI imaging...

Objective: To study the measurement reproducibility of parameters derived from introvoxel incoherent motion (IVIM) diffusion-weighted imaging (DWI)-MRI of rectal cancer between- and within- radiologists. Methods: Clinical data of 34 patients with rectal cancer were prospective analyzed. Conventional MRI sequences, IVIM DWI-MRI with sixteen b values and dynamic contrast enhancement (DCE)-MRI sequences of rectum were acquired by GE 3.0-T MRI imager. The IVIM sequence images with b value=1000 sec/mm(2) were selected to measure the maximum axial section of tumor by a radiologist with 15 year-experiences in gastrointestinal cancer imaging.Two radiologists (radiologist 1 and radiologist 2 with 2 and 10 years of experience in gastrointestinal cancer imaging, respectively) independently draw a freehand region of interest (ROI) that contained the largest available tumor area on the selected section. Monoexponential apparent diffusion coefficient (ADC) and biexponential IVIM metrics maps and IVIM parameters were generated automatically by the software. The repeated measurement was performed at an interval of one week. The average values of each measurement were used for statistical analysis. ADC values and IVIM parameters obtained between- and within- radiologists were analyzed by Wilcoxon signed-rank test. Intraclass correlation coefficients (ICC) and Bland-Altaman plots were used to analyze the parameter reproducibility of two measurements between- and within- radiologists. Results: The first and second measured ADC (×10(-3)mm(2)/s), true diffusivity (D, ×10(-3)mm(2)/s), false diffusivity (D(*,) mm(2)/s) and perfusion fraction (f, %) by radiologist 1 were 0.997, 0.692, 0.043, 34.6 and 0.993, 0.691, 0.038, 32.8, respectively. The first and second measured ADC (×10(-3)mm(2)/s), D (×10(-3)mm(2)/s), D(*) (mm(2)/s), f (%) by radiologist 2 were 0.987, 0.651, 0.046, 32.8 and 0.996, 0.689、0.041, 32.7, respectively. No statistically significant differences were observed in ADC and IVIM parameters obtained between- and within- radiologists (P>0.05). The ADC values and the f values of two times were significantly correlated between- and within- radiologists. The D values were significantly correlated within a radiologist, and the correlation of D(*) values within a radiologist was significantly higher than that between radiologists. The 95% limits of agreement (LoA) of ADC values and f values were smaller than those of D values and D(*) values between- and within- radiologists. The 95% LoA of ADC values was the least, while that of D(*) values varied most.The 95% LoA of f values and D values kept steady within a radiologist, and 95% LoA of f values was slightly smaller than that of D values. The 95% LoA of IVIM parameters (ADC, D, f, D(*) values) within radiologists 2 were better than those within radiologist 1. Conclusions: The reproducibilities of ADC and f values are the best, while the reproducibility of D(*) values is relatively poorer in rectal cancer. Measurement reproducibility of parameters derived from IVIM may be improved by increasing radiologists' experiences in drawing ROI.