SU-E-T-245: Survival Fractions for Head-And-Neck Cancer Derived from Tumor-Volume Variation Curves Using a 2-Level...

Volumetric tumor response to radiotherapy is an integrated process which includes several radiobiological mechanisms, such as cell killing, cell proliferation, dead-cell removal and tumor reoxygenation. Our goal is to reconstruct the information about these underlying radiobiological processes and specifically the cell survival fractions by fitting a 2-level cell-population tumor-volume model to imaging-derived tumor-volume variation curves obtained during radiotherapy for head-and-neck cancer.

Modeling tumor-volume during radiotherapy is a challenging problem because it is described by a sum of exponentials; therefore, the problem of accurately fitting a model to measured data is ill-posed. As an initial point of this research, we utilize a simplest 2-level cell-population tumor-volume model which separates the entire tumor-cell population into oxygenated viable cells and oxygenated lethally damaged cells. The 2-level cell population tumor model has the advantage of being conditionally well-posed. We integrated this parameterized radiobiological model with a least squares objective function and a simulated annealing optimization algorithm to characterize individual patients' time-dependent tumor-volume regression rates. The measured tumor-volume variation curves were taken from a clinical study on tumor-volume variation during radiotherapy for 14 head-and-neck cancer patients in which an integrated CT/linac system was used for tumor-volume measurements.

The 2-level tumor volume modeling is able to predict tumor behavior throughout an entire treatment for 8 of 14 patients. The average survival fraction 0.44 agrees very well with the published survival fraction of 0.45 for the head-and-neck squamous cell carcinoma. However, the 2-level model cannot describe the variation of the cell disintegration rate which is observed at the end of treatment for some of the head-and-neck cancer patients.

The 2-level cell population model is an acceptable approximation for the tumor-volume for some clinical cases, but it cannot describe all tumor-volume regression cases. This may be explained by omitting hypoxia in the 2-level model.