A Patient-Specific Correspondence Model to Track Tumor Location in Thorax during Radiation Therapy

University of Minnesota Ph.D. dissertation. 2021. Major: Biophysical Sciences and Medical Physics. Advisor: Parham Alaei. 1 computer file (PDF); 105 pages.In radiation therapy of tumors in the thorax and abdomen, respiratory-induced motion creates challenges in different steps of the treatment, such as simulation, treatment planning, and the daily patient positioning. Different methods are clinically applied to reduce the impact of respiration in radiation therapy. However, most techniques involve invasive procedures or expose the patient to extra radiation. In this thesis, first, the application of combining stereotactic body frames with a surface imaging system for patient positioning in stereotactic body radiotherapy (SBRT) of lung cancer was evaluated. Stereotactic body frames are safe and non-invasive devices that reduce the impact of respiratory-induced motions by decreasing the range of the tumor motion. Surface imaging systems are also safe and dose-free technologies for patient positioning that have the ability to monitor the patient’s motion during radiation delivery with high precision. Combining the use of a surface imaging system with SBRT positioning frames adds another level of accuracy to patient setup and provides monitoring of the movements during treatment. The results of this evaluation indicated difficulty in using surface imaging with SBRT frames primarily due to the presence of the compression plate.Next, a patient-specific correspondence model was developed to track tumors in the thorax during radiation therapy treatments using surface displacement as the surrogate signal. The proposed model is made prior to the treatment for each patient. Two types of data are used for model construction: Four-dimensional computed tomography (4DCT) images of the patient and the displacement of two points on the patient’s skin on the thoracic area (surrogate signals). The two types of data are acquired simultaneously. The 4DCT images are sorted by the amplitude-binning algorithm to account for hysteresis and breathing irregularities. A deformable image registration algorithm is applied to the 4DCT images to calculate the deformation vector fields as the knowledge of the patient’s internal motion. Principal component analysis is used to fit the correspondence model. The model incorporates recorded surrogate signals during radiation delivery as an input and delivers the 3D trajectory of the tumor or other anatomy of the interest. This model accounts for hysteresis and irregular breathing. The accuracy of the proposed model was evaluated on a respiratory phantom and five lung cancer patients. The results showed a primary validation for localizing the tumors in the lung. Testing the model on a larger population of patients will examine the accuracy of the localization more precisely