Likelihood-Fuzzy Analysis of Parotid Gland Shrinkage in Radiotherapy Patients Marco POTA a,1 , Elisa SCALCO b , Giuseppe SANGUINETI c , Maria Luisa BELLI d , Giovanni Mauro CATTANEO d , Massimo ESPOSITO a and Giovanna RIZZO b a Institute for High Performance Computing and Networking (ICAR-CNR), Napoli, Italy b Institute of Molecular Bioimaging and Physiology (IBFM-CNR), Segrate MI, Italy c Radiotherapy, Istituto Nazionale Tumori Regina Elena, Roma, Italy d Medical Physics Department, San Raffaele Scientific Institute, Milano, Italy Abstract. In head-and-neck radiotherapy, an early detection of patients who will undergo parotid glands shrinkage during the treatment is of primary importance, since this condition has been found to be associated with acute toxicity. In this work, a recently proposed approach, here named Likelihood-Fuzzy Analysis, based on both statistical learning and Fuzzy Logic, is proposed to support the identification of early predictors of parotid shrinkage from Computed Tomography images acquired during radiotherapy. For this purpose, a set of textural image parameters was extracted and considered as candidate of parotid shrinkage prediction; for all these parameters and combinations of maximum three of them, a fuzzy rule base was extracted, gaining very good results in terms of accuracy, sensitivity and specificity. The performance of classification was also compared to a classical Fisher’s Linear Discriminant Analysis and found to provide better results. Moreover, the use of Fuzzy Logic allowed obtaining an interpretable description of the relations between textural features and the shrinkage process. Keywords. Fuzzy Logic, Statistics, Parotid Gland Shrinkage, Radiotherapy. Introduction It is widely evidenced that in head-and-neck Radiotherapy (RT), parotid glands, the major salivary glands, progressively shrink during the treatment [1]. This alteration, which is known to be related to an increased dose delivered to the glands [1], causing morphological and functional variations, can lead to a higher risk of acute toxicity [2]. In this context, the early identification of those subjects who will be most probably interested by a loss of salivary functionality after RT is of primary importance, since supportive treatment cares or adaptive RT (ART) can be provided for these patients [3]. Recently, the availability of Computed Tomography (CT) images acquired during the treatment has led to the possibility of monitoring parotid modifications in terms of anatomical and structural variations, thus enabling the early selection of patients interested by high parotid shrinkage induced by RT [4-8]. In particular, a very recent work [8] dealt with this problem by extracting a set of textural features from CT images acquired in the first two weeks of RT, with the aim of 1 Corresponding Author: Marco Pota, Institute for High Performance Computing and Networking (ICAR-CNR), via Pietro Castellino 111, 80131 Napoli, Italy; E-mail: marco.pota@na.icar.cnr.it.