This work deals with the improvement of motion monitoring for cancer treatment with external radiotherapy. Errors in imaging-detected tumor volume localization strongly degrade the treatment quality as it limits dose delivery accuracy. Clinical management consists of the addition of fixed margins around the targeted volume in order to ensure a complete coverage, although it also leads to undesired irradiation of nearby organs-at-risk and potentially chance of secondary cancer development. An efficient monitoring of motion would allow for a reduction of these margins thereby improving overall dosimetry to both target and organs-at-risk. The first motion-related degrading factor for tumor localization concerns patient positioning at each fraction of the treatment. The second is related to lungs and breasts displacement with respiratory motion. In order to address these two issues in clinical practice, 3D scanners based on infrared light demonstrated promising efficacy for contactless, harmless and real-time 3D observation a surface: on the one hand, the prediction of internal breathing through real-time estimation of external motion allows respiratory motion compensation. On the other hand, positioning can be performed by exploiting the associated surfaces. This work investigated the Time-of-Flight camera validation for these applications. The understanding of the impact of the observation conditions, developments for measurement noise reduction, as well as the creation of a system comprising two calibrated cameras allowed to significantly improve the accuracy and robustness of motion estimations. This work led to the development of validated new methods allowing fast and exhaustive extraction of motion information from 3D points clouds. Finally, this work also demonstrated that the deformation of the patient surface during breathing in the anterior-posterior direction can be estimated at the level of a point, in real time, with an accuracy of about a millimeter. Furthermore, fusion of 2D and 3D information provided by the Time-of-Flight camera allows estimating elastic deformation with an error of up to two millimeters, several times per second. Finally, a clinical study about patient positioning for radiotherapy treatment showed the efficiency of a double-cameras observation system for the detection and quantification of patient¿s translation, although it also indicated that positioning using external surfaces does not allow for the detection of internal misalignment.
