Differentiation between nasopharyngeal carcinoma and lymphoma at the primary site using whole-tumor histogram analysis of apparent diffusion coefficient maps

• DOI: 10.1007/s11547-020-01152-8
• Publication Date: 2020-02-18T14:03:04Z
• Authors: Shanshan Lian; Cheng Zhang; Jun Chi; Yingyi Huang; Feng Shi; Chuanmiao Xie
• Keywords: Adult; Aged, 80 and over; Male; Nasopharyngeal Carcinoma; Lymphoma; Nasopharyngeal Neoplasms; Middle Aged; Sensitivity and Specificity; Diagnosis, Differential; 03 medical and health sciences; Diffusion Magnetic Resonance Imaging; 0302 clinical medicine; Humans; Female; Aged; Retrospective Studies

To determine the value of whole-tumor histogram analysis of apparent diffusion coefficient (ADC) maps in differentiating nasopharyngeal carcinoma (NPC) from lymphoma (NPL) at the primary site METHOD AND MATERIALS: One hundred forty-seven patients with nasopharyngeal tumors (89 NPCs and 38 NPLs) who had undergone magnetic resonance imaging (MRI) and diffusion-weighted imaging were retrospectively analyzed. ADC histogram-derived parameters were compared between the NPC and NPL groups by using the Mann-Whitney U test. Receiver operating characteristic (ROC) curves of the histogram parameters were plotted for diagnostic accuracy. Sensitivity and specificity were calculated for each histogram parameter.In whole-tumor histogram analysis, the mean, median, and 10th and 25th percentiles of ADC were all significantly higher in NPC than NPL (P = 0.045, P = 0.035, P = 0.005, and P = 0.016, respectively). Uniformity was significantly higher in NPC than NPL (P = 0.001). Skewness was significantly lower in NPC than NPL (P = 0.039). For the conventional ROI-based method, ADCmean values were significantly higher in NPC than in NPL (P = 0.009). The ROC curve analysis showed that uniformity yielded the largest area under the curve (AUC = 0.768) for differentiating NPC from NPL among all ADC metrics, followed by 10th percentiles of ADC (AUC = 0.725); sensitivity and specificity were 76.5% and 71.4%, respectively.Whole-tumor histogram analysis of ADC maps could be helpful for differentiating NPC from NPL.