Density variation of parotid glands during IMRT for head-neck cancer: Correlation with treatment and anatomical parameters

PURPOSE: Measuring parotid density changes in patients treated with IMRT for head-neck cancer (HNC) and assessing correlation with treatment-related parameters. PATIENTS AND MATERIALS: Data of 84 patients treated with IMRT for different HNC were pooled from three institutions. Parotid deformation and average Hounsfield number changes (?HU) were evaluated through MVCT (with Helical Tomotherapy) or diagnostic kVCT images taken at the treatment start/end. Parotids were delineated in the first image and propagated to the last using a previously validated algorithm based on elastic registration. The correlation between ?HU and several treatment-related parameters was tested; then, logistic uni- and multi-variate analyses taking "large" ?HU as end-point were carried out. Due to the better image quality, analyses were repeated considering only kVCT data. RESULTS: ?HU was negative in 116/168 parotids (69%; for kVCT patients: 72/92, 78%). The average ?HU was significantly different from zero (-7.3, 0.20-0.25HU/fraction, p<0.001). Individual ?HU was highly correlated with parotid deformation both in terms of volume change and mean value of the Jacobian of the deformation field (Jac_mean), and with neck thickness variation; these correlations were much stronger for kVCT data. Logistic analyses considering ?HU<-11 (quartile value) as the end-point showed a two-variable model including large deformation (Jac_mean<0.68) and initial neck thickness to be the most predictive variables (p<0.0005, AUC=0.683; AUC=0.776 for kVCT); the odd ratio of large vs moderate/small parotid deformation was 3.8 and 8.0 for the whole and the kVCT population respectively. CONCLUSIONS: Parotid density reduced in most patients during IMRT and this phenomenon was highly correlated with parotid deformation. The individual assessment of density changes was highly reliable just with diagnostic KvCT. Density changes should be considered as an additional objective measurement of early parotid radiation-induced modifications; further research is warranted.