Accurate PET/MR Quantification Using Time of Flight MLAA Image Reconstruction

Quantification of positron emission tomography/magnetic resonance imaging (PET/MRI) studies is hampered by inaccurate MR-based attenuation correction (MR-AC). To date, most studies on MR-AC have been performed using PET/MR systems without time of flight (TOF). Maximum likelihood reconstruction of attenuation and activity (MLAA), however, has the potential to improve MR-AC by exploiting TOF. The purpose of this study is to assess the impact of MR-AC on PET image quantification for TOF-PET/MR systems and to evaluate PET accuracy when using TOF in combination with MLAA (TOF-MLAA). Simulations were designed to evaluate (1) the impact of MR-AC on PET quantification for different TOF windows (667, 500, 333 and 167 ps) and (2) use of TOF-MLAA for improving PET quantification. TOF-ordered subset expectation maximisation (OSEM) and TOF-MLAA reconstructions using MR-AC were compared with those obtained using TOF-OSEM with computed tomography-based AC (CT-AC). OSEM reconstructions without TOF showed a negative MR-AC-induced bias of -50 % in the bone. TOF-OSEM was able to reduce this bias down to -15 %, with more accurate results for better TOF. TOF-MLAA was able to reduce the bias to within 5 % but at the cost of a similar to 40 % increase in image variance. TOF-MLAA can improve quantitative PET accuracy of PET/MR studies. Further improvements are anticipated with improving TOF performance.