Detection of Lung Density Variations With Principal Component Analysis in PET

Respiratory motion generates lung volume changes during the breathing cycle. These affect the lung tissue density and therefore influence both the attenuation effect and the radiotracer concentration in PET imaging. To detect and correct for these effects could improve the quantitative accuracy of lung PET imaging. In this work we propose the use of Principal Component Analysis (PCA) to detect respiratory-induced lung density changes in the upper lung, where motion is expected to be minimal. The method is firstly applied to simulation data, specifically generated to simulate density changes only and no motion. Secondly, it is applied on the upper lung bed position of 15 lung cancer patients datasets. The total number of counts in time is also evaluated. The results show that the PCA signal is highly correlated to the respiratory trace obtained from an external device, and also to the variation of total counts in time. As the bed positions taken into account do not include moving organs, the results suggest that PCA is successful in detecting respiratory-induced density changes in the upper lung.