Air breath control radiotherapy in severe insufficiency respiratory patients with NSCL: application for deformable registration method in thoracic radiotherapy

摘要

PURPOSE: Using deformable registration methods from a phase two clinical study of air breath control during radiotherapy in patients suffering from severe respiratory insufficiency and non-small cell lung carcinoma. PATIENTS AND METHODS: Between April 2002 and November 2005, 22 patients with severe respiratory insufficiency were treated with curative intent by conformal therapy combined with active breathing control. RESULTS: After a mean of follow-up of 22 months, the local control rate is 28% and the method is feasible despite the severe respiratory insufficiency. However the overall survival is still poor due to metastatic widespread. For the second part of the study, the clinical protocol was also used for two studies using deformable registration methods. In the first study, a deformable registration method has been developed in order to register several breath-hold 3D CT of the same patient acquired at several days of interval. It allowed quantifying the interfraction breath-hold reproducibility by analysing the resulting displacement field. For 6 patients, the breath-hold was effective, while for 2 patients, motion greater than 10 mm were detected. The second study aimed to simulate 4D images from 3D breath-hold images. Developing an ad-hoc methodology based on the interpolation of 3D dense deformation fields performed it. The approach has been validated with expert selected landmarks, with accuracy lower than 3 mm. CONCLUSION: ABC is feasible, even in case of severe insufficiency respiratory syndrome but metastatic widespread disease is still a major challenge even with an acceptable local control rate without serious side effects: regarding the deformable registration method. Such artificial 4D images could allow decreasing the dose need to acquire a full 4D image, to simulate irregular breathing pattern and to be used for 4D dosimetry planning.