SU-E-J-143: Characteristics of Tumor-Motion Surrogate Signals Analyzed Using Empirical Mode Decomposition and...

We introduce a novel technique for analyzing tumor-motion surrogate signals using Empirical Mode Decomposition (EMD) and Hilbert-Huang Transformation (HHT).

The tumor-motion surrogate signals were acquired (with RPM/Varian), from 20 lung-cancer patients in free-breathing method and its data were decomposed into Intrinsic Mode Functions (IMFs) using EMD. HHT was then applied to each IMF to obtain instantaneous frequency as a function of time. The Result of the frequency information was compared to Fast Fourier Transformation (FFT) and manual calculation of frequency. Correlation of each IMF with the surrogate signal was used to determine the adequate IMF as a faithful tumor-motion surrogate.

The surrogate RPM signals were decomposed to 10 ± 1 IMFs on average. The decomposed IMFs showed three categories of frequencies: (1) high frequencies (1 - 10 Hz) such as a noise-like signal, (2) medium frequencies (0.1 - 0.3 Hz), which is potentially a true breathing signal, and (3) low frequencies (0.003 - 0.09 Hz), which behave a baseline drift. The marginal frequency, which is a measure of total amplitude contribution from each frequency, showed an average difference of -0.03 ± 0.07 from the FFT and -0.02 ± 0.05 with the manual calculations. Each surrogate signal showed a high correlation with one IMF (0.747 on average) and, a low correlation with the rest of the IMFs (0.139 on average). The IMF with a high correlation alone represented the surrogate signal well in terms of breathing frequency and amplitude.

The EMD and HHT were used to analyze the cyclic components of nonlinear and non-stationary surrogate signals in the time domain. Since the EMD decomposes the signal into physically-meaningful modes, it was possible to determine IMFs that represent the tumor motion faithfully after removing noise-like signals. Further investigation on physical meanings of the IMFs is the next step of the study.