Aim/ Introduction:
In this work, we revisit the generation of attenuation maps for Positron Emission Tomography (PET) and its application in neuroimaging for neuroscience. These more accurate attenuation maps will be applied to the in-house developed BrainPET insert for a human 7T MR scanner, with the aim to significantly improve PET quantitation accuracy for typical neuroscientific applications as activation studies and neuroreceptor imaging.
Methods:
As a first step, an accurate transformation from Hounsfield units (HU) to linear attenuation coefficients for 511 keV gamma photos is needed to provide accurate attenuation correction maps for quantitative PET. Spectral CT systems allow of reconstruct Virtual monochromatic images (VMI) from the inherently multiple polyenergetic data sets of CT. VMIs can be used to obtain Virtual non-contrast (VNC) image and bone mineral density measurements. Currently, CT-derived PET attenuation maps are based on conversion schemes from conventional CT scanners with polyenergetic X-ray spectra and it is known that the conversion from polyenergetic spectra to the monoenergetic energy of 511 keV unavoidably introduces systematic errors in the attenuation coefficients for PET. Inaccuracies in the transformation will also propagate into attenuation maps for hybrid MR-PET devices, if the attenuation correction is based on CT images, e.g., using template or atlas methods. In our approach, we plan to use angiographic spectral CT scans with administration of iodinated contrast agents and to compute VNC images from the obtained dataset. For improving the accuracy of the measured bone mineralization, which is of high relevance for gamma attenuation, we combine the VNC images with bone mineral density measurements. For this, we segmented all bone values from the bone mineralization image and replaced the corresponding image voxel values in the VNC image.
Results:
The resulting combined monochromatic CT image (at 67 keV) is converted to linear attenuation coefficients at 511 keV. The conversion scheme from monochromatic CT images to linear attenuation coefficients at 511 keV was obtained by acquiring spectral CT data for different tissue types using the electron density phantom developed for spectral CTs and a photon counting CT scanner. We will present conversion schemes from HU to attenuation coefficients for different tube voltages (120 kV and 140 kV) virtual energies (64 keV, 67 keV, 76 keV, 80keV, and 100 keV). Further we will evaluate the accuracy and precision of the conversion scheme using previously acquired angiographic CTs.
Conclusion:
Spectral CT allows to minimize the systematic error when converting HU values to linear attenuation coefficients for 511 keV gamma rays using virtual monochromatic images. Further, VNC images allow to compute these attenuation coefficients from CT angiographic images after administration of iodinated contrast agents.