Histogram Analysis Of Diffusion Kurtosis Imaging For Differentiating Malignant From Benign Masses In Head And Neck Region

Purpose: To evaluate the diagnostic performance of histogram parameters derived from diffusion kurtosis imaging(DKI)for differentiating malignant from benign masses in head and neck region.Materials and methods: Our study retrospectively enrolled 80 patients with head and neck masses who had undergone DKI scan for pre-treatment evaluation.Patients were grouped into malignant group [n = 64,including squamous cell carcinoma(n = 56),lymphoma(n = 5)and adenocarcinoma(n = 3)] and benign group [n = 16,including schwannoma(n = 6),pleomorphic adenoma(n = 6),adenoid tissue(n = 2),hamartoma(n = 1)and ectopic meningioma(n = 1)].Apparent diffusion for Gaussian distribution(Dapp)and apparent kurtosis coefficient(Kapp)were generated.Histogram parameters,including mean,median,10 th,90th percentiles,skewness and kurtosis of Dapp and Kapp were calculated.Independent-sample t test and Mann-Whitney U test were used to compare the differences of DKI derived parameters between malignant and benign groups.Differentiating performances of significant parameters were determined by using receiver operating characteristic(ROC)curve analyses.Intra-class correlation coefficient(ICC)was used to evaluate the inter-reader agreement of quantitative measurements.Results: Malignant group showed significantly lower Dapp-mean,Dapp-median,Dapp-10 th and Dapp-90 th,while higher Kapp-mean,Kapp-median,Kapp-10 th and Kapp-90 th than benign group(All Ps < 0.001).Dapp-10 th demonstrated significantly higher differentiating performance than Dapp-mean [area under curve(AUC),0.963 vs 0.917;P = 0.039],while Kapp-90 th also demonstrated significantly higher differentiating performance than Kapp-mean(AUC,0.894 vs 0.849;P = 0.008).Setting Dapp-10 th ≤ 0.989×10-3 mm2/s as the threshold value,the sensitivity for differentiating malignant from benign head and neck masses was 81.3% and the specificity was 100.0%.Setting Kapp-90 th > 1.116 as the threshold value,the sensitivity for differentiating malignant from benign head and neck masses was 89.1% and the specificity was 81.3%.Moderate to excellent agreements were obtained for the measurements of DKI-derived histogram parameters(ICC,ranged from 0.702 to 0.958).Conclusion: Histogram analysis of DKI holds promising for exhibiting the difference of tumor heterogeneity between malignant and benign head and neck masses.