Abstract

Background: Single photon emission computed tomography (SPECT)-alone imaging using the Tc-99m radiopharmaceutical labeled with methylene diphosphonate or similar analogs is usually employed to diagnose metastatic bone and is typically followed by complementary magnetic resonance (MR) imaging for support in clinical decision-making. In this study, two attenuation map generation approaches from MR and SPECT non-attenuation corrected (SPECT-nonAC) images were evaluated in the context of quantitative SPECT imaging. Materials and Methods: The 2class-MR attenuation map was generated via segmenting an MR image into air and soft tissue. Likewise, SPECT-nonAC was segmented into background air and soft tissue to generate a 2class-SPECT attenuation map. The reference attenuation map was generated through manual bone segmentation from an MR image to develop a 3class-bone attenuation map. Standard uptake value (SUV) bias was calculated using the different attenuation maps on 50 vertebrae from normal patients and 16 vertebrae from metastatic patients. Results: The 2class-MR approach resulted in -16 and -8 SUV bias in normal and metastatic groups, respectively, while 2class-SPECT led to 33 and 26 SUV underestimation for the normal and metastatic patient groups, respectively. Conclusion: The 2class-SPECT approach led to a significant underestimation of SUV due to the uncertainty of body contour delineation. However, the 2class-MR approach resulted in less than -9 SUV bias in metastatic patients, demonstrating its potential to support quantitative SPECT imaging.