Date of Award

January 2016

Document Type

Thesis

Degree Name

Medical Doctor (MD)

Department

Medicine

First Advisor

Li Guang

Second Advisor

Roy Decker

Abstract

A NOVEL RESPIRATORY MOTION PERTURBATION MODEL ADAPTABLE TO PATIENT BREATHING IRREGULARITIES. Amy Yuan a, Carl P. Gaebler a, Jie Wei b, Hailiang Huang a, Devin Olek a, Guang Li a. Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York a; Department of Computer Science, City College of New York, New York b. (Sponsored by Jun Deng, Department of Therapeutic Radiology, Yale University School of Medicine).

Respiratory motion is a major source of uncertainty in lung radiation therapy. Existing methods of respiratory control and prediction encounter fundamental limitations in the presence of breathing irregularities. We have developed a physics-law-based motion perturbation model to estimate irregular tumor motion by accounting for variations in tidal volume and breathing pattern from simulation to treatment.

A novel respiratory motion perturbation (RMP) model was developed to estimate tumor motion variation based on surrogates of breathing irregularities: variations in dynamic tidal volume and breathing pattern. The RMP model contained two terms: a base motion trajectory (BMT) measured from 4-dimensional computed tomography (4DCT) at simulation and a motion perturbation (∆MT) term calculated from the breathing variations of ∆TV and ∆BP at treatment, with reference to simulation condition. The motion perturbation was derived from patient-specific anatomy, tumor-specific location, and time-dependent breathing variations. Ten patients were studied and two 4DCT images for each patient were acquired: one at simulation and the other one two weeks later during treatment. Amplitude binning was applied in 4DCT reconstruction. The motion trajectories of 40 corresponding bifurcation points in both 4DCT images of each patient were obtained using deformable image registration. To facilitate this study, an in-house 4D toolkit in MATLAB was developed to automatically calculate the breathing parameters of TV and BP associated with each breathing phase. The motion prediction was performed from 4DCTsim to 4DCTtxt, and vice versa, resulting in a total of 800 predictions. As comparisons, an established 5D model and a standard of care Correlation model with external sensor were trained using one set of 4DCT to predict motion in another 4DCT.

Interfractional tidal volume difference ranged from 10-248ml and breathing pattern varied from 0-19% across the 10 patients. The mean range of motion in the SI direction was 9.4±4.4mm, and in the AP direction was 2.7±1.4mm. Prior to motion prediction, the interfractional mean SI displacement was 2.0 ± 2.8mm. After applying the RMP model, the mean displacement difference reduces significantly to 1.2±1.8 mm (p = 0.0018). In comparison, the 5D model produced an accuracy of 1.2±1.8 mm, similar to the RMP model (p = 0.72). In contrast, the Correlation model performed significantly worse than the uncorrected data, at 3.1± 4.0mm (p = 0.0005).

A new physical motion perturbation model has been developed and evaluated with two sets of 4DCT images in each of ten patients. The prediction accuracy of 1.2±1.8 mm on average is similar to an established lung motion model. This physical perturbation model is analytically derived and has the capacity to adapt to breathing irregularities. Further improvement of this RMP model is neCessary and under investigation.

Comments

This thesis is restricted to Yale network users only. This thesis is permanently embargoed from public release.

Share

COinS