Evaluation Methodology for Respiratory Signal Extraction from Clinical Cone-Beam CT (CBCT) using Data-Driven Methods

The absence of a ground truth for internal motion in clinical studies has always been challenge to evaluate developed methods extract respiratory especially during 60-second cone-beam CT (CBCT) scan Image-Guided Radiotherapy Treatment (IGRT). unavailability gold standard led this study present methodology manually track on clinically acquired CBCT projection data set over 360° view angle. tracked signal is then used as reference assess the performance four data-driven extraction, namely: Amsterdam Shroud (AS), Local Principal Component Analysis (LPCA), Intensity (IA), and Fourier Transform (FT)-based that do not require additional equipment nor protocol existing treatment delivery. assessment using includes both quantitative qualitative analysis. It found out quantitatively all managed signals are highly correlated with signal, LPCA method displaying highest correlation coefficient value at 0.9108. Furthermore, normalized root-mean-squared amplitude error detected peaks troughs within from also lowest 1.6529 % compared other methods. This result further supported by analysis via visual inspection each extracted plotted same axes.