Oxygen-15 labeled CO₂, O₂, and CO PET in small animals: evaluation using a 3D-mode microPET scanner and impact of reconstruction algorithms

¹⁵O-labeled gases in positron emission tomography (PET) examinations have been used for the clinical evaluation of cerebral blood flow (CBF) and oxygen metabolism in patients with steno-occlusive disease of the cerebral artery. Despite the recent increase in the use of the arterial spin labeling method for magnetic resonance imaging (MRI) or perfusion imaging using computed tomography (CT) for the evaluation of CBF, PET using ¹⁵O-labeled gases has been the reference standard for quantitative evaluations of CBF, the oxygen extraction fraction (OEF), the cerebral metabolic rate of oxygen (CMRO₂), and the cerebral blood volume (CBV). Among these quantitative values, oxygen utility information, such as the OEF, can only be obtained using PET with ¹⁵O-labeled gases, and these parameters are regarded as the best predictors of stroke recurrence [1].

In the preclinical field, PET studies using ¹⁵O-labeled gases are not widely performed mainly because of the technical difficulties associated with handling labeled gases with a short half-life in small animals. We have established a methodology for the quantitative evaluation of ¹⁵O-labeled gases using PET in small animals and have reported the CBF, OEF, CMRO₂, and CBV of normal anesthetized rats examined using the 2D-mode of a clinical PET scanner [2]. However, recently developed microPET scanners are dedicated to 3D-mode acquisition similar to clinical PET scanners, and the scatter fraction and random coincidence can be problems for quantitative PET examinations, especially in studies involving the continuous inhalation of ¹⁵O-labeled gases [3, 4].

Meanwhile, iterative reconstruction algorithms, such as an ordered-subset expectation maximization (OSEM), are often used in small animal studies because they enable better image quality and spatial resolution, compared with the filtered back projection (FBP) algorithm [5, 6]. In previous studies, the spatial resolutions measured using iterative reconstruction algorithms were better than those measured using FBP, although the results depended on the parameter used for the iterative reconstruction algorithms [7‐9]. OSEM2D and OSEM3D were major reconstruction methods for iterative reconstruction. In OSEM2D, 3D data is generally converted into 2D data using a rebinning step such as Fourier rebinning method to reduce the reconstruction time. In the clinical field, OSEM3D reportedly improves the image quality, compared with OSEM2D [10]. In addition, OSEM3D with point spread function correction, such as OSEM3D followed by maximum a posteriori (MAP), is used in the preclinical field and has been reported to exhibit a higher uniformity, higher recovery coefficient, and lower spill over ratio than either OSEM2D or FBP [11].

Regardless of the clinical situation and preclinical examinations, FBP is primarily selected as the reconstruction algorithm for quantitative PET measurement [2‐4, 12‐14]. Although FBP is a standard reconstruction method for quantitative PET examinations, the influence of iterative reconstruction algorithms on ¹⁵O-labeled gases studies in small animals has not been previously evaluated.

In the present study, we evaluated quantitative values using ¹⁵O-labeled gases and 3D-mode microPET, focusing on the scatter fraction from the lung, and examined the impact of different reconstruction algorithms on the quantitative values of CBF, OEF, CMRO₂, and CBV.

The relationship between the radioactivity of flowing ¹⁵O-labeled gases and the radioactivity concentration is shown in Fig. 1 (Additional file 1: Figure S3). The radioactivity count as measured using PET was significantly lower in the lung phantom than that measured in the brain (P < 0.01). In all the reconstruction methods, the PET radioactivity count in the brain showed a small and minor increase compared to the count without the supply of ¹⁵O-labeled gases.

The mean pH, PaCO₂, PaO₂, SaO₂ hematocrit, hemoglobin concentration, heart rate, SBP, and DBP are shown in Table 1. These parameters were within physiological ranges. The stability of radioactivity was confirmed by the time activity curve in the brain during each PET measurement (Fig. 2). A comparison of the PET radioactivity counts in the whole brain is shown in Fig. 3. In both the ¹⁵O-CO₂ and ¹⁵O-O₂ studies, the radioactivity counts as determined using OSEM2D and OSEM3D-MAP were significantly higher than those determined using FBP (P < 0.01 and P < 0.01, respectively). In the ¹⁵O-CO studies, the radioactivity count value as determined using OSEM2D was significantly lower than that determined using FBP. Representative PET images during the steady state inhalation of ¹⁵O-CO₂ gas are shown in Fig. 4.

Figure 5 and Table 2 show the quantitative values in each brain region. The CBF in the whole brain was 46.6 ± 12.5 mL/100 mL/min when reconstructed using FBP, 47.9 ± 13.2 mL/100 mL/min when reconstructed using OSEM2D, and 56.0 ± 16.3 mL/100 mL/min when reconstructed using OSEM3D-MAP. The OEF in the whole brain was 63.7 ± 7.2% when reconstructed using FBP, 64.1 ± 7.3% when reconstructed using OSEM2D, and 60.5 ± 6.0% when reconstructed using OSEM3D-MAP. The CMRO₂ in the whole brain was 5.72 ± 0.34 mL/100 mL/min when reconstructed using FBP, 5.94 ± 0.36 mL/100 mL/min when reconstructed using OSEM2D, and 6.56 ± 0.56 mL/100 mL/min when reconstructed using OSEM3D-MAP. The CBV in the whole brain was 5.66 ± 0.34 mL/100 mL when reconstructed using FBP, 5.58 ± 0.33 mL/100 mL when reconstructed using OSEM2D, and 5.62 ± 0.37 mL/100 mL when reconstructed using OSEM3D-MAP. The CBF and CMRO₂ values were significantly higher when the OSEM2D and OSEM3D-MAP reconstruction methods were used, compared with the results obtained using FBP reconstruction, whereas the OEF values were significantly lower when reconstructed using OSEM3D-MAP.

Representative images of CBF, OEF, CMRO₂, and CBV are shown in Fig. 6. Table 2 shows that a significantly higher OEF was present in the cortex region and significantly higher CBF, OEF, and CMRO₂ values were present in the striatum, the thalamus, the hippocampus, and the midbrain, compared with the cerebellum.

In the present study, the primary purpose was to evaluate quantitative values using ¹⁵O-labeled gases and 3D-mode microPET. In 3D-mode acquisition, the scatter fractions from both inside and outside the FOV can be problematic for quantitative evaluations. The high radioactivity of ¹⁵O-labeled gases in the face mask, which is included in the FOV, reportedly affects the PET image in clinical situations [3, 4]. Using a clinical PET scanner, Ibaraki et al. reported the influence of random events and scatter coincidences from outside the FOV when using a 3D-mode PET scanner as well as the utility of a neck shield, which improved the noise equivalent count ratio by reducing the scatter effects [4]. Hori et al. evaluated the influence of scatter from inside and outside the FOV on PET images and demonstrated the validity of using a 3D PET scanner for ¹⁵O-labeled gas-inhalation PET studies [3]. When using a microPET scanner with a long axial-FOV, such as Inveon for small animals, the main scatter source is located inside the FOV. Consequently, the increase in scattered and random events because of the high radioactivity in the lung can potentially affect the quantitative values. We evaluated the influence of scattered and random events in the lung on the radioactivity in the brain and found that it was relatively small when using a 3D-mode microPET scanner with a long axial-FOV. In the lung phantom study, the increase in the radioactivity count for the whole brain was only 0.7–1.1% in the ¹⁵O-CO₂ study and 3.1–3.4% in the ¹⁵O-O₂ study (Additional file 1: Figure S4). Using clinical PET, quantitative studies can be performed using a 3D PET scanner even though the scatter coincidence is larger in 3D-mode than in 2D-mode [12]. The scatter fractions of clinical PET scanners have been reported for both 2D-mode and 3D-mode (13.1–21.4% and 31.8–45%, respectively) [22‐24]. The scatter fraction of the small animal PET scanner that was used in the present study was 17.2% for a rat-sized phantom, which was equivalent to that of a clinical PET scanner used in the 2D-mode [25]. In another report, Konik et al. investigated the relationship between the scatter fraction and an object’s size using a Monte-Carlo-based simulation and various sizes of digital phantoms [26]. The scatter fraction ranged from 12 to 25% for rat-sized phantoms (phantom diameters of 3–9 cm) and from 26 to 46% for human-sized phantoms (phantom diameters of 10–56 cm). For the small animal PET scanner, the scatter fraction increased linearly with the phantom diameter. The scatter fraction for small animal PET studies is less than that for human studies since the subject size is very small. In quantitative studies using a small animal 3D PET scanner, the scatter fraction is roughly equivalent to that of the 2D-mode using a clinical PET scanner. In the present study, we focused on the influence of the radioactivity in the lung which was considered to be problematic in a previous clinical study. As a result, the radioactivity in the lung and the tracheal tube was found to be relatively low, as shown in Fig. 4, and uptake in the myocardium or liver was comparable to the distribution in other PET studies, such as those for FDG PET. However, the influence of high radioactivity levels in other organs should be evaluated in a future study. Figure 4 showed a higher radioactivity in the brain using OSEM3D-MAP, compared to that using FBP and OSEM2D. However, the radioactivity in the body as determined using OSEM3D-MAP was lower than the levels determined using FBP and OSEM2D. Mannheim et al. reported that differences in object size and the reconstruction method influenced the quantitative accuracy [27]. The influence arising from the difference in object size between FBP and OSEM2D was minimal, but OSEM3D and MAP showed a different tendency than that of FBP and OSEM2D. Therefore, the radioactivity in the brain and the body trunk as evaluated using OSEM3D-MAP might exhibit a different tendency from the radioactivities determined using FBP and OSEM2D.

In the present study, the CBF, CMRO₂, and OEF images were calculated using the steady-state method, which was established by Frackowiak et al. [28]. Although the relationship was not linear, a positive correlation was observed between CBF and the ¹⁵O-CO₂ PET count as well as CMRO₂ and the ¹⁵O-O₂ PET count. When either the ¹⁵O-CO₂ PET count or the ¹⁵O-O₂ PET count increases, the CBF or CMRO₂ also increases. However, changes in the OEF depend on the relationship between the CBF and CMRO₂, as the OEF is the ratio of the CMRO₂ to CBF. For CBF as reconstructed using OSME3D-MAP, the percent differences from the values calculated using FBP were 19.6% for the whole brain, 36.7% for the frontal cortex, 28.9% for the somatosensory cortex, and 27.4% for the striatum. The differences in CBF in the deep regions of the brain (thalamus, pons, and midbrain) were relatively small, compared with those in the cortex regions. The degree of difference was smaller for CMRO₂ than for CBF when reconstructed using OSEM3D-MAP, resulting in lower OEF values for the OSEM3D-MAP reconstructions for most regions of the brain, compared with the values calculated using FBP (percent changes − 8.42–0.31%).

In whole-body PET imaging in the field of oncology, OSEM reconstruction improves both the signal-to-noise ratio and tumor detectability [29]. Kim et al. compared reconstruction methods in a small animal cardiac PET study and concluded that OSEM3D provides better image quality and reduced image noise, compared with FBP [30]. On the other hand, Bahri et al. reported that the PET count values determined using OSEM3D were overestimated by 5%, compared with the true radioactivity values, although the PET count values determined using FBP and OSEM2D were within 99% of the true radioactivity value [31]. Weber et al. demonstrated the effect of different reconstruction algorithms on the results of a dynamic PET study [15]. They reported that the time-activity curve for a tumor as obtained using OSEM2D was consistent with that obtained using FBP; however, the curve obtained using OSEM3D-MAP was 10% higher than that obtained using FBP. The results of these reports regarding OSEM3D-MAP are similar to our results. Although OSEM3D improves tumor detectability in oncology and provides better image quality in cardiology, it may be suitable to use FBP or OSEM2D for quantitative ¹⁵O-gas PET studies.

In our previous study, the CBF was underestimated as 32.3 ± 4.5 mL/100 mL/min in normal anesthetized rats because of the partial volume effect arising from the use of a clinical PET scanner. Kobayashi et al. used the H₂ ¹⁵O steady state method with a bolus and a slowly increasing injection using a multiprogrammable syringe pump and reported that the CBF was 49.2 ± 5.4 mL/100 mL/min [14]. In our study, the CBF in the whole brain was 46.6 ± 12.5 mL/100 mL/min when reconstructed using FBP, which was in close agreement with the above-mentioned report. In this study, the OEF was calculated as 63.7 ± 7.2% in the whole brain, which was consistent with our previous study (64.6 ± 9.1%). This OEF value in rats was higher than that in normal humans (44%) [32]. In our previous study, a low PaO₂ was observed because of the low oxygen content in ¹⁵O-labeled gases and the effect of anesthesia [2]. In the present study, we introduced the use of a gas mixture device to supply the O₂, and we were able to perform the PET experiments successfully while maintaining the arterial gas parameters of rats, such as PaO₂ and PaCO₂, within a physiological range.

The disadvantage of our quantitative PET technique using ¹⁵O-labeled gases was that we needed to perform an arterial blood sampling for each ¹⁵O-labeled gas measurement. The next challenge will be to derive input from the PET images, since the heart or large vessels were included in the PET FOV in the repetitive PET experiments. A limitation of the present study was that only a single parameter was selected for the OSEM reconstruction. Since these parameters may affect quantitative images, they should be evaluated in a future study.

This study examined the use of 3D-mode micro-PET and the differences in quantitative values among various reconstruction algorithms. The iterative reconstruction method resulted in significantly higher quantitative values for CBF and CMRO₂, compared with the FBP reconstruction method.