Liver Motion Tracking and Correlation to Surrogate Signals using {3D} Ultrasound

Type of publication:  beliebiger Eintrag
Jahr: 2009
Monat: August
Howpublished: BSc Thesis, University of Lübeck
Notiz: Motion Compensation in Radiosurgery
Abriss: In this work two different types of correlation models are described and tested with modeled and real breathing data with regard to their use in the stereotactic radiosurgery. The CyberKnife system is one device in the stereotactic radiosurgery to treat a tumor without the fixation of the patient. Correlation models are needed in this system to compute the position of the tumor in real time, so that the radiation beam can be focused on the tumor during the whole treatment. Depending on the accuracy of the correlation model the tumor can be irradiated more exactly and the healthy tissue around it can be saved. The correlation model used in the CyberKnife system is the polynomial model. A new model based on "-SVR is compared to this model in this work. Therefore the two models are described first. Afterwards the models are compared with modeled and real breathing data. The focus is on the observation of liver movement, because the liver moves with breathing. Therefore the treatment of a tumor in the liver is more difficult. The liver movement were determined with 3D ultrasound. The movement of LEDs on the chest of the test person were used as external markers for the correlation models. The comparison of the two correlation models shows that the new model based on e-SVR represents the correlation between internal target and external marker better than the polynomial actually model used in the CyberKnife. With all test data the error between the real and the calculated correlation of the polynomial model is at least twice as high as the error of the SVR model. The same result is observed with the maximum error ofthe models. As conclusion of this work can be stated in the examples analysed in this work an irradiation of a tumor using the SVR model would be more exactly than using the polynomial model.
Nutzerfelder: file={s_09.pdf:s_09.pdf:PDF}
Autoren: Senger, Lisa
  • s_09.pdf