Data Driven Respiratory Signal Detection in PET Taking Advantage of Time-of-Flight Data

Respiratory gating is a powerful tool for tackling motion-related issues in chest PET imaging. On current scanners the respiratory signal is obtained from external devices, whereas with Data-Driven methods it can be extracted directly from the data. The aim of this work is to show the increased potential of the application of Principal Component Analysis (PCA) on TOF data. We propose a methodology that retains the TOF information and compare it to the non-TOF method. We tested the method on 16 FDG oncology patients, monitored with an RPM camera. To further investigate the benefit of TOF, PCA was selectively applied to sets of TOF bins equidistant from the center. The correlation with the RPM, the level of noise and the respiratory-likeness were analysed for all the obtained respiratory signals. The results of our analysis showed that retaining the TOF information into the sinograms considerably increased the quality of the extracted respiratory signals.