Evaluation of morphological and structural variations in parotid glands during radiotherapy in the head and neck district

The treatment of head and neck cancer (HNC) patients is one of the widest applications of radiation treatment. One of the main organs involved in dose delivery and playing a relevant role in the quality of life of patients treated for HNC is the parotid gland, the major salivary gland, whose damage causes a pathologic condition called xerostomia. During the treatment, HNC patients may undergo significant anatomical changes; it was generally reported that parotids shrink during RT with a shift of the center of mass (COM) toward the midline. These modifications cause a translation of the glands into the high dose region, thus the effective delivered dose on parotids is significantly higher than that planned. Therefore, the study of local changes in parotid glands due to irradiation is still an open challenge, and the exploitation of the information content in medical images acquired during the radiation treatment is of practical interest for the non-invasiveness of these techniques. The aim of this work was thus the evaluation of morphological and structural modifications induced by RT on parotid glands using image-based indices extracted from CT images acquired during the treatment. In particular, in this PhD project it was proposed to estimate morphological changes by the deformation of the glands, evaluated by applying a non-rigid registration method expressly optimized and validated on CT images of the HN district. Secondly, a method of image analysis based on textural feature extraction was proposed to characterize structural variations of parotid tissue. All these methods were evaluated and optimized for the clinical problem and all these parameters were studied in relation to dosimetric and clinical indices and to the clinical outcome, in order to better assess their contribution in this research field. Parotid deformation was here estimated through an image registration and contour propagation method, based on Free Form Deformation and Mutual Information. A deep validation of this method was performed to assess the accuracy of contour propagation method, to ensure a reliable quantification of parotid deformation. The quantification of parotid deformation was carried out starting from the estimated deformation field, by calculating the Jacobian index, while the structural characterization of parotid tissue was performed by the extraction of features related to the intensity patterns of the image, quantified by texture analysis. We evidenced that parotid shrinkage is highly related to low dose DVH values; at the same time, density variation measured after RT is significantly correlated with volume decrease, suggesting that the loss of acinar cells and the consequent increase in the percentage of the fatty component is related to the decrease in functionality. We proved that the major reduction in density is concentrated in the first half of the treatment, while the decrease in volume and in tissue complexity is quite constant during the whole treatment. Finally, we found that the most accurate prediction of parotid shrinkage is achieved by the combination of geometrical and structural variations measured in the first two weeks and that the early prediction of xerostomia score was preliminary achieved by early density and volume variation.