Respiratory gated PET based on time activity curve analysis

Abstract

1741

Objectives: Respiratory motion in PET degrades images and limits detectability and characterization of small or low-contrast lesions. Although image quality can in principle be improved using respiratory-gating hardware, this adds to the complexity and expense of acquiring PET data. We aimed to develop a data-driven method, based on individual voxel signal fluctuations, for accomplishing electronic respiratory gating of PET data acquired in a clinically practical manner, requiring no additional hardware or end-user input. Methods: PET scan simulations were generated based on the 4D NCAT torso phantom. Images were convolved with a Gaussian blurring kernel (9.4 mm FWHM) then degraded by a Poisson noise function corresponding to a 10 mCi FDG injection. They were then altered along a defined axis of respiratory motion falling predominantly along the z (sup-inf) axis, to produce representations of thoracic PET images acquired at a rate of 4 per sec with a mean 5-sec respiratory cycle. Each individual scan was forward-projected into sinogram space, then time activity curves (TACs) of selected voxels were filtered for respiratory frequencies and evaluated for fluctuations in their signal. Parameters derived from this process were applied to thousands of voxels, with each one individually contributing an estimated gate trigger to a global temporal histogram used to identify triggers for defining the start of each respiratory cycle. Finally, image data were divided into 10 equal time bins per cycle. Results: Our methods correctly identified the start frame of each respiratory cycle defined for the phantom within ± half a bin. Resultant gated images demonstrated improved effective resolution: FWHM [mm] in (x, y, z) planes of centrally and peripherally located tumors before gating were (2.2, 2.2, 3.9) and (2.2, 2.2, 4.0) respectively, and after gating were (1.9, 1.9, 1.8) and (2.0, 2.0, 1.8). SUV measurements from 6 tumors (located in central, peripheral, basal, apical, anterior and posterior lung locations) had a mean of 1.8 ± 0.2 (mean ± SD) in the ungated images, and 3.6 ± 0.3 in the gated images (measured as mean within 70% max contour). Conclusions: Data-based respiratory gating of PET images may be achieved without the need for gating hardware or additional user input, in a manner capable of improving effective resolution and increasing lesion detectability.