Simulation of lung motions using an artificial neural network

• Published in: Cancer radiothérapie 2011-04, Vol.15 (2), p.123-129
• Main Authors: Laurent, R, Henriet, J, Salomon, M, Sauget, M, Nguyen, F, Gschwind, R, Makovicka, L
• Format: Article
• Language: fre
• Subjects: Feasibility Studies; Four-Dimensional Computed Tomography - methods; Humans; Learning Curve; Lung - diagnostic imaging; Lung Neoplasms - diagnostic imaging; Lung Neoplasms - pathology; Lung Neoplasms - radiotherapy; Medical Physics; Movement; Neural Networks (Computer); Physics; Radiotherapy Dosage; Radiotherapy Planning, Computer-Assisted - methods; Respiration; Tomography, X-Ray Computed; Tumor Burden
• ISSN: 1278-3218; 1769-6658
• DOI: 10.1016/j.canrad.2010.07.636

A way to improve the accuracy of lung radiotherapy for a patient is to get a better understanding of its lung motion. Indeed, thanks to this knowledge it becomes possible to follow the displacements of the clinical target volume (CTV) induced by the lung breathing. This paper presents a feasibility study of an original method to simulate the positions of points in patient's lung at all breathing phases. This method, based on an artificial neural network, allowed learning the lung motion on real cases and then to simulate it for new patients for which only the beginning and the end breathing data are known. The neural network learning set is made up of more than 600 points. These points, shared out on three patients and gathered on a specific lung area, were plotted by a MD. The first results are promising: an average accuracy of 1mm is obtained for a spatial resolution of 1 × 1 × 2.5mm(3). We have demonstrated that it is possible to simulate lung motion with accuracy using an artificial neural network. As future work we plan to improve the accuracy of our method with the addition of new patient data and a coverage of the whole lungs.