Evaluation of principal component analysis-based data-driven respiratory gating for positron emission tomography

Abstract

Objective: Respiratory motion can degrade PET image quality and lead to inaccurate quantification of lesion uptake. Such motion can be mitigated via respiratory gating. Our objective was to evaluate a data-driven gating (DDG) technique that is being developed commercially for clinical PET/CT. Methods: A data-driven respiratory gating algorithm based on principal component analysis (PCA) was applied to phantom and FDG patient data. An anthropomorphic phantom and a NEMA IEC Body phantom were filled with 18F, placed on a respiratory motion platform, and imaged using a PET/CT scanner. Motion waveforms were measured using an infrared camera [the Real-time Position Management™ system (RPM)] and also extracted from the PET data using the DDG algorithm. The waveforms were compared via calculation of Pearson's correlation coefficients. PET data were reconstructed using quiescent period gating (QPG) and compared via measurement of recovery percentage and background variability. Results: Data-driven gating had similar performance to the external gating system, with correlation coefficients in excess of 0.97. Phantom and patient images were visually clearer with improved contrast when QPG was applied as compared to no motion compensation. Recovery coefficients in the phantoms were not significantly different between DDG- and RPM-based QPG, but were significantly higher than those found for no motion compensation (p < 0.05). Conclusion: A PCA-based DDG algorithm was evaluated and found to provide a reliable respiratory gating signal in anthropomorphic phantom studies and in example patients. Advances in knowledge: The prototype commercial DDG algorithm may enable reliable respiratory gating in routine clinical PET-CT.