Importance of bone attenuation in brain SPECT quantification

The purpose of this study was to determine the effects of nonuniform attenuation on relative quantification in brain SPECT and to compare the ability of the Chang and Sorenson uniform attenuation corrections (UACs) to achieve volumetric relative quantification. Methods: Three head phantoms (dry human skull, Rando and Radiology Support Devices (RSD) phantoms) were compared with a human head using a gamma camera transmission CT (γTCT) SPECT system and x-ray CT. Subsequently, the RSD phantom's brain reservoir was filled with a uniform water solution of ^99mTc, and SPECT and γTCT data were acquired using fanbeam collimation. The attenuating effects of bone, scalp and head-holder in individual projections were determined by an analytical projection technique using the SPECT and γTCT reconstructions. The Chang UAC used brain and head contours that were segmented from the γTCT reconstruction to demarcate its attenuation map, whereas the Sorenson UAC fit slice-specific ellipses to the SPECT projection data. For each UAC, volumetric relative quantification was measured with varying attenuation coefficients (μs) of the attenuation map. Results: Gamma camera transmission CT and x-ray CT scans showed that the dry skull and Rando phantoms suffered from a dried trabecular pone compartment. The RSD phantom most closely reproduced the attenuation coefficients of the human γTCT and x-ray CT scans. The analytical projections showed that the attenuating effects of bone, scalp and head-holder were nonuniform across the projections and accounted for 18%-37% of the total count loss. Volumetric relative quantification was best achieved with the Chang (zero iterations) attenuation correction using the head contour and μ = 0.075 cm^-1; however, cortical activity was found to be 10% higher than cerebellar activity. For all UACs, the optimal choices of μ were experimentally found to be lower than the recommended 0.12 cm^-1 for brain tissue. This result is theoretically supported here. Conclusion: The magnitude of errors resulting from uniform attenuation corrections can be greater than the magnitudes of regional cerebral blood flow deficits in patients with dementia, as compared with normal controls. This suggests that nonuniform attenuation correction in brain SPECT imaging must be applied to accurately estimate regional cerebral blood flow.