Intravoxel Incoherent Motion Model in Differentiating the Pathological Grades of Esophageal Carcinoma: Comparison of Mono-Exponential and Bi-Exponential Fit Model

PURPOSE: To compare the diagnostic efficiency of the mono-exponential model and bi-exponential model deriving from intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) in differentiating the pathological grade of esophageal squamous cell carcinoma (ESCC). METHODS: Fifty-four patients with ESCC were divided into three groups of poorly-differentiated (PD), moderately-differentiated (MD), and well-differentiated (WD), and underwent the IVIM-DWI scan. Mono-exponential (D(mono), D*(mono), and f(mono)) and bi-exponential fit parameters (D(bi), D*(bi), and f(bi)) were calculated using the IVIM data for the tumors. Mean parameter values of three groups were compared using a one-way ANOVA followed by post hoc tests. The receiver operating characteristic curve was drawn for differentiating pathological grade of ESCC. Correlations between pathological grades and IVIM parameters were analyzed. RESULTS: There were significant differences in f(mono) and f(bi) among the PD, MD and WD ESCC groups (all p<0.05). The f(mono) were 0.32 ± 0.07, 0.23 ± 0.08, and 0.16 ± 0.05, respectively, and the f(bi) were 0.35 ± 0.08, 0.26 ± 0.10, and 0.18 ± 0.07, respectively. There was a significant difference in the D(mono) between the WD and the PD group (1.48 ± 0.51* 10(-3) mm(2)/s versus 1.05 ± 0.44*10(-3) mm(2)/s, p<0.05), but there was no significant difference between the WD and MD groups, MD and PD groups (all p>0.05). The D*(mono), D(bi), and D*(bi) showed no significant difference among the three groups (all p>0.05). The area under the curve (AUC) of D(mono), f(mono) and f(bi) in differentiating WD from PD ESCC were 0.764, 0.961 and 0.932, and the sensitivity and specificity were 92.9% and 60%, 92.9% and 90%, 85.7% and 100%, respectively. The AUC of f(mono) and f(bi) in differentiating MD from PD ESCC were 0.839 and 0.757, and the sensitivity and specificity were 78.6% and 80%, 85.7% and 70%, respectively. The AUC of f(mono) and f(bi) in differentiating MD from WD ESCC were 0.746 and 0.740, and the sensitivity and specificity were 65% and 85%, 80% and 60%, respectively. The pathologically differentiated grade was correlated with all IVIM parameters (all p<0.05). CONCLUSIONS: The mono-exponential IVIM model is superior to the bi-exponential IVIM model in differentiating pathological grades of ESCC, which may be a promising imaging method to predict pathological grades of ESCC.