Quality measurement on 3D ultrasound volumes reconstructed from 2D slices and its application in radiation therapy

Abstract

Radiation therapy relies on exact patient positioning to spare organs at risk and thus be able to deliver higher and more effective radiation doses with fewer side effects to surrounding tissue. Thereby it is also possible to apply better dose distribution on the tumor. There are several methods available for patient positioning, which are discussed, but most of them imply additional radiation dose for the patient. There are also systems on the market which rely on ultrasound as imaging modality. As the system should be used to align patients suffering from soft tissue cancer for example in the region of the prostate, the advantage is the good contrast of soft tissue compared to systems using X-ray. This short overview suggests a new 3D ultrasound reconstructed from 2D ultrasound image slices. To use such a method several tools are of need which are presented and discussed. A tracker device tracks the position of the ultrasound probe during image acquisition. Thereby, by ultrasound calibration, the transformation between the probe sensor and the recorded image can be calculated. Therefore a software is used which performs calibration, recording of the images as well as reconstruction of the 3D volume out of the recorded 2D images. After reconstruction it is compared which of the US systems is better in quality and accuracy, the reconstructed 3D ultrasound or the classical 3D ultrasound. This is of importance for patient positioning as then the better method can be used. There are several ways to perform such a comparison. One of the most promising for the underlying application is 3D/3D ultrasound image registration. This is done via comparison of two separate recorded and reconstructed image volumes. Registration tries to align two images that way that both images, when possible, exactly overly and are congruent. Image registration then gives a transformation matrix which contains the translation and rotation of one image compared to the other. Out of this it is concluded which of the two systems is better in quality and thus more accurate. For comparison two different datasets where used. Results show an overall lower quality of reconstructed image volumes compared to general 3D ultrasound devices when using patient data. Using a figure phantom show better results but quality in terms of accuracy still remains behind general 3D ultrasound machines. Thus applicability in image-guided radiation therapy cannot be recommended as when using it patient positioning would be too inaccurate.