Applicability of emission-based attenuation map for rapid CBF, OEF, and CMRO2 measurements using gaseous 15O-labeled compounds

In the present study, we generated CBF, OEF, and CMRO₂ images applying an emission-based attenuation map rather than transmission scan data.
The differences between the emission- and transmission-based methods in the CBF, OEF,
and CMRO₂ values were around or smaller than 10%, though some of them differed significantly
such as in the parietal, occipital, and cerebellum regions. The regression analysis
showed close correlations with r?=?0.89 to 0.99 between the methods, meaning that the regional contrast in the images
was similar. These findings suggest that the present approach eliminating the transmission
scan is applicable for clinical examination, particularly for patients with acute
stroke.

The present application attainable here is as follows. First, several studies have
already demonstrated emission-based attenuation correction like in the present method;
however, they performed the scan by administering the tracer via only injection but
not gas inhalation. It was not specified how the inhaled gas affected, particularly
in the nasal cavity, the generated attenuation map by the emission-based method. The
present study clarified the applicability for the gas inhalation method. Second, because
of the considerable shortening of the total examination duration by eliminating the
transmission scan, the patient’s burden, in particular, the need to stay still, would
be reduced. Third, the applicability can be extended to the acute stroke patient,
for whom the most rapid possible intervention is needed 15], and furthermore to assessments of hemodynamic change in a subject receiving a time-limited
balloon occlusion test 16]. Bai et al. suggested that images without attenuation correction can give misleading,
namely, a sphere in a thorax phantom was not visible without attenuation correction
17]. Careful attention is warranted to apply the present method when uniform structure
cannot be assumed.

Some studies have focused on the elimination of transmission scans aside from PET/CT
or PET/MRI. Weinzapfel and Hutchins in a CBF study during activation with and without
transmission-based attenuation map found no significant difference between the methods
18]. Montandon and Zaidi demonstrated a method of template-based attenuation 19]. Kaneko et al. conducted a FDG study without transmission scan and found less adequate
qualitative measurement in the uppermost and lowermost parts 20]. In generating the attenuation map by the emission-based method in the present study,
we used the sinogram from 12 min to the end to define the edge of a brain contour
and confirmed that the detected edge in the sinogram follows exactly the brain edge
contour on a reconstructed image. When a sinogram from the total duration was used,
the edge was blurred due to spillover of inhaled and exhaled ¹⁵O-labeled gas, resulting in an inappropriate attenuation map. Thus, extracting a phase
not affected by the labeled gas is crucial for quantification of CBF, OEF, and CMRO₂ in studies in which ¹⁵O-labeled gas compounds are administered.

The tissue coefficient value applied was 0.1 cm^?1, which was obtained from the mean of measured attenuation maps for ten subjects randomly
chosen from the present data set. For the ten values, mean and SD were 0.0996?±?0.0013 cm^?1, suggesting quite similar values across subjects. In fact, quite a similar value
and variation of 0.099?±?0.002 cm^?1 was also demonstrated in a previous CBF study 18]. Some factors are potential sources of inaccuracy, such as sex, ethnic group, and
age, and thickness of skull. When we applied the software for estimating attenuation
map, the thickness of skull was not involved for the estimation, because that is not
same across subjects and level of head. However, we do not anticipate significant
errors attributable to these factors.

The CBF and CMRO₂ values were obtained from uptake rate in water and oxygen phase and thus directly
affected by the pixel value, thus degree of difference of these two parametric values
between the methods were similar. The variations in OEF were less because that is
computed as a rate of uptake rates in oxygen and water, meaning that the bias in the
estimated pixel values was canceled.

There were significant differences in the CBF, OEF, and CMRO₂ values between the emission- and transmission-based methods, such as in the parietal,
frontal, cerebellum, and occipital regions. Such significant differences are likely
attributable to some factors not taken into account in estimating the attenuation
map, such as head-rest and skull for parietal region, and eyeball, muscle, ventricles,
and bone, which could have different attenuation coefficients from that of the brain
tissue, for occipital and cerebellar regions, and also any ¹⁵O-labeled gas retained in the nasal cavity and trachea. These factors appeared to
bias the attenuation map in the emission-based method. Also, there is a methodological
difference, namely, the transmission-based method is always affected by noise, while
the other assumes uniform distribution of tissue coefficients, and thus is not affected
by noise. This factor appeared as regional variation in attenuation maps from the
transmission-based method. Such factors could result in regional dependency of over-
or underestimation in the reconstructed pixel value and thus in estimated CBF, OEF,
and CMRO₂ values. In fact, the obtained CBF and CMRO₂ values by the present method appeared lower than those by the transmission method
in the parietal and occipital regions and cerebellum, due to the assumption that tissue
coefficient values are uniform not only for brain but also for other tissues whereas
they are in fact subject to bias from bone and soft tissue regions. In contrast, if
the emission-based attenuation map was generated from a sinogram including a gas inhaling
phase, the nasal cavity filled with ¹⁵O gas and its surrounding region might be considered brain tissue, resulting in overestimation
of the pixel value and thus higher CBF and CMRO₂ values. The CBF and CMRO₂ values were obtained from the uptake rate in water and oxygen phase and thus directly
affected by the pixel value. Thus, the degree of the difference in these two parametric
values between the methods was similar. The variations in OEF were less marked because
it is computed as a rate of uptake rates in oxygen and water, meaning that the bias
in the estimated pixel values was canceled. The differences, however, were around
or less than 10% in CBF and CMRO₂ in the parietal, occipital, and cerebellum regions for the present emission-based
method and were smaller than the SDs. As a whole, any apparent difference would not
adversely affect any clinical determination of optimal treatments.

Recently, most PET systems are integrated PET/CT or PET/MRI scanners, and the 2D acquisition
mode, which the present study applied, is not available, while only the 3D mode is.
It would be important to refer the applicability of the present method to the 3D mode,
but it would be important to directly test its validity. In theory, the present method
might be extended to the 3D mode, because edge contour determination on the sinogram
would be possible in the 3D mode like in the 2D mode. Then we could proceed with the
same procedure. It should be noted that quantitative estimation of CBF, OEF, and CMRO₂ has been achieved with 2D mode, while scatter coincidence events disturb the quantitative
nature for 3D mode 20]. To overcome this, a hybrid dual-energy window method (HDW) 21],22] was applied and the validity of CBF, OEF, and CMRO₂ images was demonstrated 23]. When scatter coincidence events interfere with the edge detection on sinograms,
it would be possible to detect them on reconstructed images without attenuation correction,
implementing the above HDW and proceeding with the following procedure. Thus, implementation
of a hybrid scatter correction method would be essential for applying the 3D mode.

In the present study, the duration of transmission was 5 min, and the total true count
was more than 50 M counts depending on the Ge/Ga rod source. In a previous study that
estimated noise levels in CBF, OEF, and CMRO₂ enhanced from transmission data, an N-index was introduced and the noise level was found not to be enhanced when the transmission
true count was more than 40 M counts 6]. A simple comparison between the present and previous counts may not be feasible
because of the different protocols as well as different PET scanners used, but, the
count level in our study is higher, and thus, the transmission duration would be sufficient
for the image quality. We also measured the N-index in our data set for those parametric images and found no significant deterioration
in quality on parametric images.

We separated the present subjects into three groups with (two grades) and without
cerebrovascular disorders and tested the validity of the present method, particularly
in regions with elevated CBV and reduced CBF. The obtained parametric images showed
that the CBF reduced and OEF elevated regions in the transmission-based method can
also be identified with the emission-based method. The regional ROI values did not
differ between the methods in any of the groups. These findings suggest that the present
method with gas study is applicable to patients with cerebrovascular disorders, particularly
acute stroke.

We set the time interval between the sequential administration of two tracers to 11 min,
which is much longer than the 3 min the conventional methodology allows 4],5], in the present examination. The reason for this longer interval was due to limitations
of the synthesizer system. The limitation is due to the need for a sequential supply
of two radioactive compounds, namely, ¹⁵O₂ and C¹⁵O₂, within a short interval, including radio-synthesis, quality control, and purity
examination before administration. An automated synthesis system able to operate under
the same operation system as for the ¹⁵O-dedicated cyclotron would help to improve the logistics necessary in the procedures
24].