Effects of Attenuation Correction and 3D-Reconstruction Algorithms on Brain Lesions Detectability

We compared lesion detectability in the brain images reconstructed with the three-dimensional (3-D) reprojection algorithm, with and without attenuation correction, and with the 3-D ordered subsets expectation maximization (OSEM) algorithm with and without post-filtering. For this purpose, three small lesion phantoms with diameters of 3, 5 and 8.6 mm were embedded into the Hoffman brain phantom. The approximate ratio of the activity concentration in the lesions to the surrounding brain gray matter was ranging from 1.5 to 10. We used the high-resolution MDAPET, a prototype 3D-PET camera, to scan the phantoms. This camera is a multi-ring scanner with an intrinsic transaxial resolution of 2.8 mm. For practicality, to minimize the number of imaging sessions to be taken, data for the hot lesion phantoms and the normal warm Hoffman brain phantom were taken separately. Then, before performing image reconstruction, the two sets of the sinograms data were selectively combined to generate the sinograms data for the desired SUVs. When the lesions data were taken, the lesions were placed inside the Hoffman brain phantom filled only with water to provide the attenuation and scatter effects. We observed a slight improvement in lesion detectability for the 5 mm lesion located near the center of image for attenuation corrected over non-attenuation corrected sinograms when images were reconstructed with the 3D-reprojection algorithm and also images visually looked better. Images reconstructed with 3-D OSEM followed by post-filtering show slight improvement in lesion detection over 3DRP.