Application of 3d digital surface imaging for patient positioning

Abstract

Purpose/Objective: Accurate patient positioning is an important aspect of treatment especially for conformal treatments such as IMRT, which has high gradients and smaller margins. Studies have identified target positional errors of up to 5mm for even well immobilized patients using standard techniques. These errors may reduce the dose delivered to the target volume in the range of 7 to 15%, which can significantly reduce the probability of tumour control. The purpose of the study is to prove that improved accuracy and efficiency in patient positioning can be achieved by monitoring the external 3D patient surface and correcting the patient set-up using the 3D surface displacement information. This will provide improvements over iterative conventional set-up methodologies. The superior resolution of the system should enable more accurate set-up than the existing techniques. Materials/Methods: This technique uses a photogrammetry based 3D Digital Surface Imaging system to improve accuracy and efficiency in patient set-up. The Digital Surface Imaging system consists of two stereoscopic video cameras, a LCD projector, and image acquisition and registration software. The cameras and the projector are mounted on a stand, which allows rotation and translation with 5 degrees of freedom. Well-defined dot patterns are projected onto a surface and the cameras simultaneously acquire sequences of 2D digital images. Using triangulation between the 2 stereoscopic images and the projected dots a 3D digital surface image is constructed. An initial reference 3D surface-image or a CT surface-image is fused with subsequent surface images and the displacement between the two image sets is measured in terms of translation and rotation. The displacement information can be used to re-position a patient in the planned reference position. In the Phantom study the phantom is deliberately translated and rotated to a known position and the Digital Surface Imaging system is used to predict the translated and rotated positions. The results are compared to evaluate the system accuracy. Similar measurements were performed on volunteers. The system was characterized in terms of its resolution, effect of ambient light and the distance of the object from the camera. Results: In the 3D Digital Surface Imaging system the resolution is characterized by using a step-wedge with steps of 2mm height. The height of the steps is reproduced to within 0.3mm for a camera-to-object distance of 1m. To test the system’s registration accuracy known translations and rotations of a phantom head are compared to the displacements measured by the system. The deviation from the known x- and z-translations is 0.2mm at 3σ in the interval -3.5 to 3.5cm, and for all rotational directions a 0.2° deviation at 3σ is observed for a known rotation in the interval -5 to 5°. The clinical use of the system is currently tested on volunteers wearing a non-invasive head frame and positioned on a GE CT-sim couch. A set of markers on the face of the volunteer is used to measure the repositioning accuracy by reading the laser co-ordinates at the laser-marker intersection. The resulting displacement due to repositioning as well as deliberate displacement is compared to the registered displacement from the 3D Digital Surface Imaging system. Preliminary results agree well within 1mm. Conclusions: We have demonstrated that a 3D Digital Surface Imaging system can accurately calculate the rotation and translation to within 1mm and 1° accuracy. The data obtained with volunteers indicate that this system is capable of reproducing the patient set-up to within 1mm.