Apparent diffusion coefficient histogram analysis can evaluate radiation-induced parotid damage and predict late xerostomia degree in nasopharyngeal carcinoma

Abstract

We investigated apparent diffusion coefficient (ADC) histogram analysis to evaluate radiation-induced parotid damage and predict xerostomia degrees in nasopharyngeal carcinoma (NPC) patients receiving radiotherapy. The imaging of bilateral parotid glands in NPC patients was conducted 2 weeks before radiotherapy (time point 1), one month after radiotherapy (time point 2), and four months after radiotherapy (time point 3). From time point 1 to 2, parotid volume, skewness, and kurtosis decreased (<i>P</i> < 0.001, = 0.001, and < 0.001, respectively), but all other ADC histogram parameters increased (all <i>P</i> < 0.001, except <i>P</i> = 0.006 for standard deviation [SD]). From time point 2 to 3, parotid volume continued to decrease (<i>P</i> = 0.022), and SD, 75^th and 90^th percentiles continued to increase (<i>P</i> = 0.024, 0.010, and 0.006, respectively). Early change rates of parotid ADC_mean, ADC_min, kurtosis, and 25^th, 50^th, 75^th, 90^th percentiles (from time point 1 to 2) correlated with late parotid atrophy rate (from time point 1 to 3) (all <i>P</i> < 0.05). Multiple linear regression analysis revealed correlations among parotid volume, time point, and ADC histogram parameters. Early mean change rates for bilateral parotid SD and ADC_max could predict late xerostomia degrees at seven months after radiotherapy (three months after time point 3) with AUC of 0.781 and 0.818 (<i>P</i> = 0.014, 0.005, respectively). ADC histogram parameters were reproducible (intraclass correlation coefficient, 0.830 - 0.999). ADC histogram analysis could be used to evaluate radiation-induced parotid damage noninvasively, and predict late xerostomia degrees of NPC patients treated with radiotherapy.