Completion of input images of the elastic registration process for IMRT applied to breast cancer using spline-based interpolation

Abstract

Cancer is one of the leading causes of death worldwide. Radiotherapy plays a fundamental role in its treatment, but the collateral damage of the process can affect patients' quality of life. Intensity-modulated radiation therapy (IMRT) is an advanced technique that offers promising benefits, but input imaging for IMRT incurs high costs. To mitigate collateral damage and increase treatment efficacy, we propose a methodology to expand the input set for IMRT. In this study, we focused on reducing the uncertainty of the data by preprocessing the input images. By employing bootstrapping, a non-parametric statistical sampling technique, we reduce the input images to regions of interest. Interpolating this information using polynomial splines and B-splines generates intermediate images. Our findings show that both interpolation methods, specifically the degree 1 polynomial spline, effectively reduce the uncertainty of the data. The methods are tested using Pearson correlation tests and bootstrap hypothesis tests, finding them accurate. By expanding the input data set and minimizing uncertainty, our approach promises to improve treatment planning and enhance patient outcomes in radiotherapy.