Abstract
1726
Objectives: The objective of this study is to enable PET attenuation correction based on MRI and to identify the level of quality of segmentation necessary to produce an acceptable PET image. This can guide the development of automated algorithms to calculate MRI-based attenuation maps. Methods: For MRI-based attenuation maps, one must identify and segment tissue types that differently attenuate the photons and assign each tissue group a corresponding attenuation value to approximate CT-based attenuation. PET, CT, and MRI examinations of the thoracic region were performed. The PET data was first reconstructed both unattenuated and with CT attenuation. The MRI images were then registered to the uncorrected PET data. After registration, the MRI images were segmented into regions of air, lung, and soft tissue, where a value of 0.0 cm-1, 0.030 cm-1, and 0.096 cm-1 were assigned to air, lung, and soft tissue respectively. The segmented MRI attenuation map was then used for PET reconstruction. The unattenuated, CT attenuated, and MRI attenuated PET images are compared using different segmentations. Results: Without attenuation, it is nearly impossible to distinguish between the lungs, mediastinum, and surrounding soft tissue. By segmenting only air and other tissues in MRI, a well-defined contour of the body is present, but different tissue types are difficult to identify. Using segmentation of air, lung, and soft tissue in MRI, the mediastinum, lung, and surrounding tissues are easily identified. However, there are still discrepancies between the MRI attenuated image and the CT attenuated image, such as lack of definition of vessels in the lungs and abnormally low signal in the spine. Conclusions: The MRI-segmented attenuation maps do correspond with CT-based attenuation maps. However, a sub-segmentation of morphologic tissue areas is necessary to achieve the desired details in the anatomical structure of PET images in order to be comparable to those produced by CT-based attenuation. Using only tissue type based segmentation of air, lung, and soft tissue is insufficient and produces only a rudimentary image quality.
- Society of Nuclear Medicine, Inc.