Histological grades of rectal cancer: whole-volume histogram analysis of apparent diffusion coefficient based on reduced field-of-view diffusion-weighted imaging.

Abstract

To explore the role of whole-lesion histogram analysis of apparent diffusion coefficient (ADC) derived from reduced field-of-view (rFOV) diffusion-weighted imaging (DWI) technique in discriminating histological grades of rectal carcinoma. Altogether, 49 patients with rectal cancer were enrolled in this retrospective study. All patients received preoperative 3.0 T MR scan. Histogram parameters from rFOV DWI were calculated and correlated with histological differentiation of rectal cancer. The parameters were compared between different histological grades of rectal cancer by independent Student's t-test or Man-Whitney U-test. The Spearman correlation test analyzed correlations between histological grade and histogram parameters. The diagnostic performance of individual parameters for distinguishing poorly from well-/moderately differentiated tumors was assessed by receiver operating characteristic curve (ROC) analysis. There were significant differences for ADCmean, 25th, 50th, 75th, 90th, 95th percentiles, skewness, and kurtosis of rFOV DWI sequence between well-, moderately, and poorly differentiated rectal cancers (P<0.05). Significant correlations were noted between histological grades and the above histogram parameters (r=0.679, 0.540, 0.701, 0.730, 0.669, 0.574, -0.730, and -0.760 respectively, P<0.001). Among the individual histogram parameter, kurtosis achieved the highest AUC of 0.882 with an optimal cutoff value of 1.934 in distinguishing poorly from well-/moderately differentiated rectal cancers. The combination of ADCmean, 75th percentile, and kurtosis yielded the highest AUC of 0.927 with a sensitivity of 88.00% and a sensitivity of 91.7% using logistic regression. Quantitative whole-lesion ADC histogram analysis based on the rFOV DWI technique could help differentiate histological grades of rectal cancer. The combination of ADCmean, 75th percentile, and kurtosis may be the best choice.