Bone SPECT of the spine: a comparison of attenuation correction techniques

Image artifacts from variable self-attenuation are recognized as major sources of diagnostic uncertainty in SPECT. For myocardial perfusion studies, an attenuation map is often obtained from a separate transmission study. However, for many applications such as bone SPECT, it has been believed to be unnecessary to obtain a transmission study to correct for the effects of attenuation. We have had significant success in clinical management of lower spine pain using bone SPECT. This success has led us to consider SPECT for the management of cervical spine pain. Cervical spine reconstructions without attenuation correction are difficult to interpret, because the high attenuation in the mandible and skull tends to decrease estimates of activity of the upper cervical spine, and the lower cervical/upper thoracic vertebrae are obscured by the shoulders. We present a technique that uses downscatter to provide attenuation correction for these acquisitions and compare it with other recognized attenuation correction techniques.

An emission study is acquired using two windows: one for obtaining the photopeak data and another for obtaining the downscattered photons. A body outline is estimated from these datasets using a projection data thresholding method. From this outline, a uniform attenuation map is created using attenuation coefficients appropriate for 99mTc in water (0.154 cm(-1)). These maps are used in SPECT reconstruction using ordered-subset expectation maximization (OSEM). This method is compared with (a) no attenuation correction (NC), (b) conventional Chang attenuation correction based on the interactive determination of the body outline from the 99mTc emission photopeak data (ChangAC) and (c) OSEM correction using attenuation maps estimated with a line source and fanbeam collimators (transAC).

Patient studies using scatterAC demonstrated a significant improvement in the uniformity of estimated cervical spine uptake in normal patients, compared with either NC or ChangAC. Results using scatterAC were similar to those of transAC. We also observed significant improvement in uniformity using scatterAC in SPECT of the lower back in obese patients, as well as the relative limitations of scatterAC versus nonuniform, transmission-based attenuation correction.

Comparisons with reconstructions using transmission data for estimating attenuation demonstrate that reasonable quantitative accuracy can be obtained in SPECT of the cervical spine using this simple attenuation estimate. Both scatterAC and transAC appear to provide consistent and expected uniform spine uptake in the cervical spines of normal patients.