Whole-Volume Apparent Diffusion Coefficient Histogram Analysis for Prediction of Lymph Node Metastasis in Periampullary Carcinomas

Abstract

Background: To evaluate whether whole-volume apparent diffusion coefficient (ADC) histogram parameters of the primary tumor were useful to predict regional lymph node metastasis (LNM) in periampullary carcinomas.     

Methods: Thirty-eight patients with periampullary carcinoma who underwent pancreaticoduodenectomy between January 2016 to April 2019 were retrospectively enrolled. Whole-volume ADC histogram analysis of the primary tumor was performed by two radiologists independently. Clinical factors, pathological results and histogram parameters were recorded and evaluated. Intraclass correlation coefficient (ICC) was used to assess agreement between observers. Receiver operating characteristic (ROC) analysis was performed to evaluate the performance of parameters in differentiating LNM-positive group and LNM-negative group.

Results: Interobserver agreements were good to excellent for histogram analysis between two radiologists, with ICCs ranging from 0.766 to 0.967. Tumor size, MR-reported LN status and most ADC histogram parameters (including mean, minimum ADC value, 10th, 25th, 50th, 75th, and 90th percentile, and kurtosis) were significantly different between LNM-positive group and LNM-negative group (p < 0.05), and revealed significant correlations with LNM (p < 0.05). At ROC analysis, tumor size and minimum ADC value generated highest area under the curve (AUC) (AUC = 0.764, 95% confidence interval [CI]: 0.599, 0.886). When diagnostic predictive values were calculated with the combined model incorporating tumor size, MR-reported LN status and 75th percentile, the best diagnosis performance was obtained, with AUC of 0.879 (95% CI: 0.771, 0.986), sensitivity of 100.0%, and specificity of 75.0%. 

Conclusion: Whole-volume ADC histogram parameters of the primary tumor held great potential in differentiating regional LNM in periampullary carcinomas.