A system for visual guidance in patient set-up for external-beam radiotherapy procedures was developed using augmented reality. The system uses video cameras to obtain views of the linear accelerator, and the live images are displayed on a monitor in the treatment room. A 3D model of the patient's external surface, obtained from planning CT data, is superimposed onto the treatment couch in the camera images. The augmented monitor can then be viewed, and alignment performed against the virtual contour. The system provides an intuitive method for set-up guidance, and allows non-rigid deformations to patient pose to be visualised. It also allows changes to patient geometry between treatment fractions to become observable, and can remain in operation throughout the treatment procedure, so that patient motion becomes apparent. Coordinate registration between the camera view and the linac is performed using a cube which is aligned with the linac isocentre using room lasers or cone-beam CT. The AR tracking software detects planar fiducial tracking markers attached to the cube faces, and determines their positions in order to perform pose estimation of the 3D model on-screen. Experimental results with an anthropomorphic phantom in a clinical environment have shown that the system can be used to position a rigid-body with a translational error of 3 mm, and a rotational error of 0.19 degrees, 0.06 degrees and 0.27 degrees, corresponding to pitch, roll and yaw respectively. With further developments to optimise the system accuracy and its interface, it could be made into a valuable tool for radiotherapy clinics. The outcome of the project has been encouraging, and has shown that augmented reality for patient set-up guidance has great potential.

A system for visual guidance in patient set-up for external-beam radiotherapy procedures was developed using augmented reality. The system uses video cameras to obtain views of the linear accelerator, and the live images are displayed on a monitor in the treatment room. A 3D model of the patient's external surface, obtained from planning CT data, is superimposed onto the treatment couch in the camera images. The augmented monitor can then be viewed, and alignment performed against the virtual contour. The system provides an intuitive method for set-up guidance, and allows non-rigid deformations to patient pose to be visualised. It also allows changes to patient geometry between treatment fractions to become observable, and can remain in operation throughout the treatment procedure, so that patient motion becomes apparent. Coordinate registration between the camera view and the linac is performed using a cube which is aligned with the linac isocentre using room lasers or cone-beam CT. The AR tracking software detects planar fiducial tracking markers attached to the cube faces, and determines their positions in order to perform pose estimation of the 3D model on-screen. Experimental results with an anthropomorphic phantom in a clinical environment have shown that the system can be used to position a rigid-body with a translational error of 3 mm, and a rotational error of 0.19 degrees, 0.06 degrees and 0.27 degrees, corresponding to pitch, roll and yaw respectively. With further developments to optimise the system accuracy and its interface, it could be made into a valuable tool for radiotherapy clinics. The outcome of the project has been encouraging, and has shown that augmented reality for patient set-up guidance has great potential.