Effects of scatter and attenuation correction on quantitative assessment of regional cerebral blood flow with SPECT

Appropriate corrections for scatter and attenuation correction are prerequisites for quantitative SPECT studies. However, in most cerebral SPECT studies, uniform attenuation in the head is assumed, and scatter is usually neglected. This study evaluated the effect of attenuation correction and scatter correction on quantitative values and image contrast.

Studies were performed in six normal volunteers (ages 22-26 yr) following intravenous 123I-IMP administration using a rotating, dual-head gamma camera. A transmission scan was acquired with a 99mTc rod source (74 MBq) placed at the focus of a symmetrical fanbeam collimator. Data were reconstructed using two attenuation coefficient (mu) maps: quantitative mu map from the transmission scan and a uniform mu map generated by edge detection of the reconstructed images. Narrow and broad beam mu values were used with and without scatter correction, respectively. Scatter was corrected with transmission-dependent convolution subtraction and triple-energy window techniques. Quantitative rCBF images were calculated by the previously validated IMP-autoradiographic technique, and they were compared with those obtained by (15)O-water and PET. SPECT and PET images were registered to MRI studies, and rCBF values were compared in 39 ROIs selected on MRI.

Clear differences were observed in rCBF images between the measured and constant mu maps in the lower slices due to the airways and in the higher slices due to increased skull attenuation. However, differences were < 5% in all cerebral tissue regions, thus assumption of uniform mu introduces little bias. The scatter correction was found to increase the image contrast significantly, i.e., rCBF increased by 20%-30% in gray matter and decreased in white matter regions by 10%-20% after scatter correction, increasing gray-to-white ratio to be close to that of PET measurement. The rCBF values from the two scatter correction were not significantly different, but the triple-energy window technique suffered from increased noise. After scatter correction, rCBF values were in good agreement with those measured by PET.

This study shows little loss in accuracy results from assuming uniform mu map. However, scatter correction is required for the quantitative rCBF values and gray-to-white ratios to approach those of PET.