UCLA

Breathing motion in radiotherapy is commonly managed with four-dimensional computed tomography (4DCT). 4DCT datasets consist of multiple breathing-gated images that display motion of the subject’s anatomy over a breathing period. Commercial 4DCT protocols are

susceptible to image artifacts when the subject breathes irregularly. Unlike commercial protocols, model-based 4DCT techniques describe a correspondence between tissue motion and an external signal used as a surrogate for respiratory phase. One such technique, termed ’5DCT,’ uses free-breathing fast helical acquisition and characterizes lung tissue motion as function of five degrees of freedom: x,y, and z position in a reference geometry, breathing amplitude, and breathing rate. The overarching goal of this dissertation is to develop the 5DCT technique for clinical use.

A validation study was conducted involving comparing 5DCT images to commercial 4DCT using an animal model. Reproducible and periodic breathing patterns were achieved through mechanical ventilation and image similarity was quantified using landmark displacement.

Differences in measured tumor motion between 5DCT and commercial 4DCT were examined in a cohort of 20 lung cancer patients. Solutions for the unique challenges of using model model-based techniques clinically, such as appropriate amplitude interval selection and presentation of a quantitative error map, were developed.

A quality management program was developed to ensure the safety of the protocol using risk analysis methods such as process mapping and failure modes and effects analysis. In addition, safety of the in-house software required to implement the technique was managed through use cases and quantitative, testable safety requirements.

The physiological significance of the 5D model parameters was investigated, particularly their dependence on breathing rate during acquisition. It was shown that the model parameter relating tissue motion to breathing amplitude was largely invariant with breathing rate during acquisition.

A prospective scanning method was developed to reduce the number of fast helical scans, and associated imaging dose, necessary to perform 5DCT while maintaining motion modeling accuracy. A simulation study was conducted using patient breathing traces, and the results demonstrated that adequate sampling of the respiratory cycle could be achieved using six scans, compared to the previously published protocol that employs twenty-five.