An Accurate Attenuation and Scatter Correction Method for PET/MR Scanners with Lutetium-based Scintillators

A novel method for attenuation and scatter correction of PET data in PET/MR is investigated in this work. Given the absence of an accurate attenuation map in PET/MR, the use of an activity and attenuation joint reconstruction algorithm such as TOF-MLAA is studied. To address issues in TOF-MLAA, a number of, improvements were investigated: 1) the use of background radiation from lutetium-based scintillators as a weak transmission source, 2) denoising attenuation maps with artificial intelligence, and 3) the use of an energy-based scatter method that does not require any prior anatomical knowledge. The proposed method is evaluated on PET data from a Siemens Biograph Vision Quadra scanner using phantom and patient data. Images of NEMA NU-2 image quality showed that the proposed method retains the contrast recovery and background variability of the PET images within 5% of the reference CT-based reconstructed PET images. Similarly, organ uptake in PET images using the proposed method showed an average of 5% error. In conclusions, the proposed reconstruction framework showed promising results towards quantitative CT-less PET suitable for PET/MR.