Details

Autor(en) / Beteiligte

• Dumouchel, T
• de Kemp, R A

Titel

Recovery of partial volume losses in cardiac mouse PET imaging using a combined 1D/2D and a combined 1D/3D model

Ist Teil von

• IEEE Nuclear Science Symposuim & Medical Imaging Conference, 2010, p.2303-2307

Ort / Verlag

IEEE

Erscheinungsjahr

2010

Quelle

IEEE Electronic Library Online

Beschreibungen/Notizen

• Micro-PET image resolution is on the order of the left ventricle (LV) wall thickness in a mouse heart. Mouse LV images are thus subject to partial volume (PV) losses, impeding the ability to quantify tracer activity in cardiac muscle. In this study, 2D and 3D PV correction (PVC) models are proposed for mouse imaging. ECG gated PET images are acquired and a 1D model is used to extract the LV wall contours and adjacent activity. This information is used to build either 2D or 3D images to derive the regional recovery coefficients in 2D or 3D. A 2D mouse heart phantom was created physically and digitally to test the 2D algorithm. The phantom was designed with 8 cardiac gates and variable wall thicknesses. The physical phantom was imaged with the Inveon small animal PET scanner and an effective Gaussian resolution of 1.3 mm FWHM was derived from the image. A 3D simulation was created based on the MOBY phantom assuming isotropic resolution. The 3D PVC model was applied to the simulation. Finally, ECG gated FDG mouse images were obtained with the Inveon and the 2D PVC algorithm was applied in the basal slices of the heart, with resolution also estimated directly from the image. The 2D PVC algorithm was found to reduce bias in the 2D measured and simulated phantom activity from 40% to 5% while also restoring image uniformity. The 3D PVC performed on the MOBY simulation reduced bias from 40% to 15% while increasing image homogeneity across all planes throughout the heart. A similar pattern was observed in the FDG mouse hearts using an estimated transverse resolution of 1.35 mm. With current technology, PVC is mandatory to restore quantitative accuracy in small animal cardiac PET imaging. This study indicates that the proposed methodology has the ability to partially restore the expected activity distribution. While mouse imaging was the focus of the present study, this algorithm could be used for LV imaging in other species where LV thickness is on the order of the system resolution.