Knowledge-based segmentation of attenuation-relevant regions of the head in T1-weighted MR images for attenuation correction in MR/PET systems

Attenuation correction is an important prerequisite for quantitative brain PET. In dedicated PET, attenuation correction is based on a transmission scan which is not available in MR/PET systems. In contrast to CT images used in PET-CT systems, MR images from MR/PET systems do not provide unam-biguous values which can be directly transformed into PET attenuation coefficients. The new method segments the attenuation-differing regions of the patient's head based on its routinely acquired T1-weighted MR data set. Since tissue types of differing PET attenuation coefficients show the same relaxation- and proton density-dependent intensity values in T1-weighted MR images, these regions cannot be separated purely on the intensity values. Thus, additional anatomical knowledge about the regions' position to each other and its rough shape is utilized in the segmentation process. So far, the new knowledge-based segmentation method was applied to 47 MPRAGE data sets of the head (11 M & 10 F tumor patients, 15 M & 11 F healthy volunteers). 17 data sets were acquired at a Siemens 1.5T Avanto MR system, 16 data sets at a Siemens 3T MR/PET TimTrio system without PET insert and 14 data sets with PET insert. Even though the variability of the image quality and of anatomical and tumor regions was large in this study, the method yields excellent results in separating brain tissue, extracerebral soft tissue, cra-niofacial cavities, and the mastoid process, and good to excellent results for the neurocranial and craniofacial bone segmentation. The application of the resulting segmented data sets for PET attenuation correction shows very promising results compared to measured and CT-based attenuation correction.