Stefano Ursino1, Alessia Giuliano2, Fabio Di Martino3, Paola Cocuzza4, Alessandro Molinari4,5, Antonio Stefanelli6, Patrizia Giusti7, Giacomo Aringhieri7, Riccardo Morganti8, Emanuele Neri7, Claudio Traino3, Fabiola Paiar4. 1. Department of Radiation Oncology, University Hospital S. Chiara, Via Roma 55, 56100, Pisa, Italy. stefano.ursino@med.unipi.it. 2. Department of Physics, S. Luca Hospital, Via Guglielmo Lippi Francesconi 556, 55100, Lucca, Italy. 3. Department of Physics, University Hospital S. Chiara, Via Roma 55, 56100, Pisa, Italy. 4. Department of Radiation Oncology, University Hospital S. Chiara, Via Roma 55, 56100, Pisa, Italy. 5. Department of Radiation Oncology, Ecomedica Institute of Clinical Research, Via Cherubini 2/4, 50053, Empoli, Italy. 6. Department of Radiation Oncology, University Hospital Cona, Via Aldo Moro 8, 44124, Ferrara, Italy. 7. Department of Radiology, University Hospital S. Chiara/Cisanello, Via Roma 55/Via Paradisa 2, 56100, Pisa, Italy. 8. Department of Clinical and Experimental Medicine, Section of Statistics, Via Roma 55, 56100, Pisa, Italy.
Abstract
PURPOSE: To develop a videofluoroscopy-based predictive model of radiation-induced dysphagia (RID) by incorporating DVH parameters of swallowing organs at risk (SWOARs) in a machine learning analysis. METHODS: Videofluoroscopy (VF) was performed to assess the penetration-aspiration score (P/A) at baseline and at 6 and 12 months after RT. An RID predictive model was developed using dose to nine SWOARs and P/A-VF data at 6 and 12 months after treatment. A total of 72 dosimetric features for each patient were extracted from DVH and analyzed with linear support vector machine classification (SVC), logistic regression classification (LRC), and random forest classification (RFC). RESULTS: 38 patients were evaluable. The relevance of SWOARs DVH features emerged both at 6 months (AUC 0.82 with SVC; 0.80 with LRC; and 0.83 with RFC) and at 12 months (AUC 0.85 with SVC; 0.82 with LRC; and 0.94 with RFC). The SWOARs and the corresponding features with the highest relevance at 6 months resulted as the base of tongue (V65 and Dmean), the superior (Dmean) and medium constrictor muscle (V45, V55; V65; Dmp; Dmean; Dmax and Dmin), and the parotid glands (Dmean and Dmp). On the contrary, the features with the highest relevance at 12 months were the medium (V55; Dmin and Dmean) and inferior constrictor muscles (V55, V65 Dmin and Dmax), the glottis (V55 and Dmax), the cricopharyngeal muscle (Dmax), and the cervical esophagus (Dmax). CONCLUSION: We trained and cross-validated an RID predictive model with high discriminative ability at both 6 and 12 months after RT. We expect to improve the predictive power of this model by enlarging the number of training datasets.
PURPOSE: To develop a videofluoroscopy-based predictive model of radiation-induced dysphagia (RID) by incorporating DVH parameters of swallowing organs at risk (SWOARs) in a machine learning analysis. METHODS: Videofluoroscopy (VF) was performed to assess the penetration-aspiration score (P/A) at baseline and at 6 and 12 months after RT. An RID predictive model was developed using dose to nine SWOARs and P/A-VF data at 6 and 12 months after treatment. A total of 72 dosimetric features for each patient were extracted from DVH and analyzed with linear support vector machine classification (SVC), logistic regression classification (LRC), and random forest classification (RFC). RESULTS: 38 patients were evaluable. The relevance of SWOARs DVH features emerged both at 6 months (AUC 0.82 with SVC; 0.80 with LRC; and 0.83 with RFC) and at 12 months (AUC 0.85 with SVC; 0.82 with LRC; and 0.94 with RFC). The SWOARs and the corresponding features with the highest relevance at 6 months resulted as the base of tongue (V65 and Dmean), the superior (Dmean) and medium constrictor muscle (V45, V55; V65; Dmp; Dmean; Dmax and Dmin), and the parotid glands (Dmean and Dmp). On the contrary, the features with the highest relevance at 12 months were the medium (V55; Dmin and Dmean) and inferior constrictor muscles (V55, V65 Dmin and Dmax), the glottis (V55 and Dmax), the cricopharyngeal muscle (Dmax), and the cervical esophagus (Dmax). CONCLUSION: We trained and cross-validated an RID predictive model with high discriminative ability at both 6 and 12 months after RT. We expect to improve the predictive power of this model by enlarging the number of training datasets.
Authors: Miranda E M C Christianen; Irma M Verdonck-de Leeuw; Patricia Doornaert; Olga Chouvalova; Roel J H M Steenbakkers; Phil W Koken; C René Leemans; Sjoukje F Oosting; Jan L N Roodenburg; Bernard F A M van der Laan; Ben J Slotman; Hendrik P Bijl; Johannes A Langendijk Journal: Radiother Oncol Date: 2015-08-25 Impact factor: 6.280