Introduction

Radionuclide imaging such as positron emission tomography (PET) and single-photon emission computed tomography (SPECT) can visualize in vivo physiological function even in deep sites of the body with high sensitivity and excellent quantitative capability and are powerful tools for cancer imaging [1]. These techniques can detect tumor masses in vivo, but they cannot visualize interiors of tumor masses because the spatial resolution of conventional radionuclide imaging devices is quite poor (>5 mm for PET and >8 mm for SPECT). However, tumors are never homogeneous. Therefore, a technique with the ability to clearly visualize intratumoral heterogeneity in vivo would be an innovation that improves cancer diagnosis and therapy.

Recently, SPECT scanners dedicated for small animal imaging have been developed and have shown excellent spatial resolution of <1 mm [25]. They appear to hold potential for visualization of intratumoral heterogeneity in vivo. However, most studies on high spatial resolution using these scanners have been conducted with phantoms under ideal conditions [68], and little work has been done with actual small animal imaging. A few studies have reported in vivo SPECT images of small animals with excellent resolution [2, 9]; however, the administered radioactivity in these studies was approximately 200–900 MBq per head using 99mTc- or 123I-labeled probes. This dosage is very high for actual small animal imaging, and hence, is likely to induce very high radiation in the animals [10]. On the other hand, very recently, increasing numbers of studies have reported the use of these scanners for tumor imaging of small animals [4, 1114]. The administered radioactivity used in these studies seems to be reasonable for small animals, but instead, they visualized just tumor masses, not tumor interior, because of insufficient spatial resolution. Although a few studies have reported good resolution [11, 12], they were not necessarily optimized for in vivo high resolution tumor imaging. In this study, we investigated the acceptable and optimal conditions related to high resolution in vivo SPECT imaging for actual small animals, such as acquisition equipment and time, reconstruction parameters, and detection and dosage thresholds for visualization of intratumoral heterogeneity in vivo. We also used this technique to visualize intratumoral heterogeneity in tumor-bearing mice with fine spatial resolution using a small animal SPECT scanner.

Materials and methods

SPECT/CT imaging

A SPECT/CT combined scanner with four detectors dedicated for small animal imaging (NanoSPECT/CT; Bioscan Inc., Washington, DC, USA) was used to acquire sectional images. Each detector was equipped with tungsten-based multiplexing nine-pinhole collimators with diameters of 1.0 or 1.4 mm and a cylindrical field of view (FOV) of 30 × 16 mm (diameter × length). The axial FOV was extended using a step-and-shoot helical SPECT with a user-definable range up to 230 mm, depending on the imaging region. The energy peaks for the camera were set to 171 and 245 keV, and the energy window was set to the peak energy ±10%. CT scans were performed with the following conditions: tube voltage, 45 kV; tube current, 177 μA; and tube current–time product (the product of the tube current and scan time), 32 mAs. CT and SPECT reconstruction was performed using InVivoScope software (Bioscan). CT images were reconstructed with the standard resolution, and SPECT image data were reconstructed by an ordered subset expectation maximization iterative algorithm dedicated for multiplexing multi-pinhole reconstruction. Three operator-dependent parameters were used during the reconstruction of the SPECT image data. They were noise suppression (low, middle, and high), voxel size (fast 0.4 mm; standard 0.3 mm; and fine 0.2 mm), and a reconstruction setting for the iteration/subset number (fast, standard, and fine). The following two sets of reconstruction parameters (conditions A and B) were used in this study. For noise suppression, voxel size, and reconstruction setting, the values were low (25% of voxel size), fine, and fine for condition A, respectively, and middle (35%), standard, and standard for condition B, respectively. The SPECT/CT fusion images were obtained using the automatic fusion software installed on the scanner.

Phantom studies

System performance evaluation by measurement of the spatial resolution and sensitivity of the SPECT scanner

A Jaszczak phantom with hot rods ranging in diameter from 1.0 to 1.5 mm (Bioscan), which was filled with 100 MBq (50 MBq/mL) of 111InCl3 solution (kind gift by Nihon Medi-Physics, Tokyo, Japan), was imaged according to the procedure described above, using the SPECT scanner equipped with 1.0- and 1.4-mm diameter pinhole apertures. Projection image data were acquired in the step-and-shoot mode from 128 directions over 360° covered by four detectors (32 directions × 4 detectors). The total acquisition count was set at 300 Mcounts. Consequently, the total acquisition time was several hours. The acquired image data were reconstructed using conditions A and B mentioned above. The reconstructed sectional images were evaluated on the basis of count profile curve methods. When the smallest count in the intercolumn space was lower than 50% of the maximum count of the columns, we judged that the columns were separately imaged. Additionally, the smallest diameter at which each column was separately imaged was defined as the spatial resolution. The sensitivity of each aperture for 111In was measured using an 111In point source. Three MBq of 111In was absorbed into a 0.15- to 0.30-mm resin bead, which was placed at the center of the FOV and imaged from 128 (32 × 4) projections with an acquisition time of 30 s per projection. Sensitivity was calculated by dividing the acquired counts per second (cps) by the radioactivity of the point source (Bq). The acquisition was repeated three times.

Measurement of the SPECT image spatial resolution under the conditions simulated conditions of small animal imaging

For simulation of actual in vivo animal imaging, a Jaszczak phantom with rods of 1.0- to 1.5-mm diameter was filled with 10 MBq (5 MBq/mL) of 111InCl3 solution. SPECT image data were acquired in the similar manner to mentioned above, except projection numbers, that is, 24 projections over 360° covered by four detectors (6 × 4). The total acquisition time was set as 5–120 min. The obtained image data were reconstructed in the same manner as described above.

Evaluation of the lower limit for visualization and quantitative linearity of the SPECT images

Six 1-mL plastic syringes with an inner diameter of 4.3 mm were filled with 111InCl3 solution with a concentration between 0 and 5.0 MBq/mL, and the image data were acquired in the same manner as above from 24 (6 × 4) views with 300 s per projection, consequently a total acquisition time was 30 min. The acquired image data were reconstructed using condition B. Regions of interest (ROIs) of the same size were placed on each syringe based on the superimposed CT images, and the radioactivity of each ROI was measured (SPECT value).

Animal studies

Chemicals

Distearoylphosphatidylcholine (DSPC; Coatsome MC-8080) and cholesterol were purchased from NOF Corporation (Tokyo, Japan). Diethylenetriaminepentaacetic acid (DTPA) and 8-hydroxyquinoline were purchased from Wako Pure Chemicals (Osaka, Japan). Isoflurane was purchased from Abbott Japan (Tokyo, Japan). All other chemicals were purchased from Wako Pure Chemicals or Sigma-Aldrich Japan (Tokyo, Japan).

Animals

Male ddY mice were obtained from Japan SLC Inc. (Hamamatsu, Japan). The animals were kept at 12/12-h dark–light cycles and were fed ad libitum. Sarcoma 180 tumor cells (2 × 106 cells/mouse) were subcutaneously transplanted into the left hind leg of 6- to 7-week-old mice. When the tumor weighed between 0.3 and 1.0 g, the mice were used for the experiments. All procedures were performed in accordance with the Guidelines of the Animal Care and Use Committee of the National Cancer Center.

Preparation of 111In-labeled liposomes at high concentration

Liposomes were composed of DSPC and cholesterol at a molar ratio of 2:1 and prepared by a lipid film hydration-extrusion method [15]. In brief, the dried films were dispersed in 10 mM DTPA in 50 mM HEPES/5% mannitol buffer, pH 7.4, at 60°C and then repeatedly extruded through polycarbonate membrane filters (0.2-, 0.1-, and 0.05-μm pore sizes) to adjust their diameters to about 100 nm. The extra chelating ligands were removed by passage through a Sephadex G-50 (GE Healthcare Japan Ltd., Tokyo, Japan) column. The phospholipid concentration was measured by phosphate assay. The liposomes were then labeled with 111In. To achieve a very high level of specific radioactivity, a remote loading method was used [16]; specifically, 111In was loaded inside the liposomes through the liposomal membranes by transchelation between the lipophilic oxine and hydrophilic DTPA ligands.

In vivo SPECT/CT imaging with high spatial resolution

The 111In-labeled liposomes (8.5–15 MBq/0.2 mL/mouse) were injected intravenously into sarcoma 180-bearing mice, and CT and SPECT images were acquired 24 or 48 h later. Mice were anesthetized by inhalation of 1.5% isoflurane, and their body temperatures were kept at about 37°C during the scans. Acquisition and reconstruction of the CT and SPECT image data were performed under the conditions optimized by the phantom studies simulating animal imaging. SPECT projection data were acquired in a step-and-shoot mode from 24 (6 × 4) views covering the whole body or the tumor site. The acquisition time was set to 150–300 s per projection, and image data were acquired from 24 projections taking a total of 1–2 h. The acquired projection data were reconstructed using InVivoScope software with condition B.

Measurement of accumulated radioactivity and autoradiography (ARG) of the excised tumors

After the in vivo imaging, the mice were killed, and the tumors were carefully excised. The excised tumors were immediately weighed, and their radioactivity was measured by means of a dose calibrator (IGC-7; Aloka Co., Ltd., Tokyo, Japan) or an Auto Well Gamma System (ARC-380CL; Aloka). Then, the tumors were embedded in OCT compound (Sakura Finetek Japan, Tokyo, Japan), frozen and cut into 6-μm-thick sections for hematoxylin–eosin staining and 20-μm-thick sections for ARG using a cryomicrotome (CM 3050S; Leica Instruments, Germany). The sections were prepared in the same direction of SPECT slice. The section corresponding to the SPECT image was selected according to the slice thickness and used for ARG. For ARG, the samples were placed in contact with an imaging plate (MS2040; Fujifilm Co., Tokyo, Japan), and the exposed imaging plates were analyzed with an FLA-7000 reader (Fujifilm Co.) and Multi Gauge v3.0 software (Fujifilm Co.).

Results

Initially, we evaluated the performance of the scanner for spatial resolution and sensitivity under ideal conditions. Figure 1 shows the SPECT images and their count profile curves of a Jaszczak phantom filled with 111InCl3 solution as high as 50 MBq/mL. Under the conditions of high radioactivity, long acquisition time, and reconstruction parameters for fine image quality (condition A), the resolution obtained was excellent. Hot rods as thin as 1.0 mm on the SPECT images were separately visualized using both 1.0- and 1.4-mm pinhole apertures (Fig. 1a, b). The resolution was slightly better using pinhole apertures of 1.0-mm diameter than of 1.4-mm diameter by count profile curve analysis, but there was little difference on visual evaluation. On the other hand, the sensitivity of the 1.4-mm pinhole apertures was 2.0 kcps/MBq for 111In, which was two times higher than that of the 1.0-mm pinhole apertures (0.97 kcps/MBq).

Fig. 1
figure 1

SPECT images of an 111In Jaszczak phantom and their count profile curves under the best performance condition. The Jaszczak phantom was filled with 111InCl3 of 100 MBq (50 MBq/mL), and images were obtained on a SPECT scanner equipped with pinhole apertures with diameters of 1.0 mm (a) and 1.4 mm (b, c). Projection data were acquired with 128 directions over 360° covered by four detectors (32 directions × 4 detectors). Total acquisition counts were set at 300 Mcounts. The acquired projection data were reconstructed using InVivoScope software based on ordered subset expectation maximization (OSEM) with two distinct conditions, A (a, b) and B (c)

The reconstruction parameters (noise suppression, voxel size, and iteration/subset number) had considerable influence on the images. Figure 1b, c shows images that were obtained from the same projection data but were reconstructed under different conditions. Figure 1b was reconstructed under condition A and Fig. 1c under condition B. The condition A image showed higher resolution than the condition B image when sufficient counts were acquired, as in the system performance evaluation experiments. However, ideal acquisition conditions were not possible in the animal imaging experiments, because a large radioactive dose cannot be administered to small animals and the animal position cannot be kept stable for such a long time. Therefore, phantom experiments simulating in vivo small animal imaging were performed.

A Jaszczak phantom was filled with 111InCl3 solution of 10 MBq (5 MBq/mL), and images were acquired for 5–120 min using 1.0- and 1.4-mm pinhole apertures. The acquired data were reconstructed with two different conditions, conditions A and B described above. Under these experimental conditions, images acquired with the 1.4-mm pinhole apertures were always better than those acquired with the 1.0-mm apertures (data not shown). The images acquired with the 1.4-mm pinhole apertures are shown in Fig. 2. The SPECT images reconstructed with condition B (upper row in Fig. 2) always showed higher resolution than those reconstructed with condition A (lower row), unlike the images shown in Fig. 1b, c. Under condition B, the spatial resolution greatly depended on acquisition time, specifically, on total acquisition counts. An acquisition time of 30 min provided images with acceptable spatial resolution of 1.3–1.4 mm under the experimental conditions.

Fig. 2
figure 2

SPECT images of an 111In Jaszczak phantom under conditions simulating in vivo small animal imaging. The Jaszczak phantom was filled with 111InCl3 of 10 MBq (5 MBq/mL), and the images were acquired for 5–120 min with 1.4-mm pinhole apertures and 24 (6 × 4) directions. The acquired data were reconstructed with conditions A (lower row) and B (upper row) using InVivoScope software

We also evaluated the lower limit of radioactivity for visualization. Syringe phantoms were filled with 111InCl3 solution with concentrations of 0–5.0 MBq/mL for simulated animal imaging. Image data were acquired for 30 min and reconstructed using condition B. Figure 3a shows axial images of the syringes filled with 0.1–5 MBq/mL, and Fig. 3b shows images of the syringes filled with 0–0.5 MBq/mL. Under these conditions, the syringes with radioactivity concentrations lower than 0.1 MBq/mL could not be visualized on the reconstructed SPECT images irrespective of the scale range. Therefore, the threshold for visualization of 111In was estimated to be 0.2–0.5 MBq/mL (MBq/g) under these test conditions.

Fig. 3
figure 3

The minimum radioactivity threshold for visualization and quantitative linearity of SPECT images. Six 1-mL plastic syringes were filled with 111InCl3 of 0.1–5.0 MBq/mL (a, c) or 0–0.5 MBq/mL (b, d), and their SPECT images were acquired. Regions of interest of the same size were placed on each syringe based on the superimposed CT images, and the radioactivity was measured (SPECT value). a, b Axial SPECT images of the phantoms. c, d The correlation between 111InCl3 concentration and the SPECT value

The quantitative linearity of the SPECT images was also examined. ROIs of the same size were placed on each syringe based on the superimposed CT images, and their radioactivity was measured (SPECT value). Figure 3c, d shows the correlation between the radioactivity concentration of 111InCl3 and the SPECT value. For both concentration ranges described above, including the points under 0.1 MBq/mL, excellent linearity was shown (R 2 > 0.99).

After optimization of the conditions needed to obtain high resolution SPECT images in vivo, we then attempted to visualize intratumoral heterogeneity. The 111In-labeled liposomes were intravenously injected into sarcoma 180 tumor-bearing mice (8–15 MBq/mouse), and SPECT/CT fusion images were obtained at 24 h after injection (Fig. 4). The images of case #1 are a whole body maximum intensity projection (MIP) image and representative slice images containing the tumor. Images of case #2 are sectional images of the tumor site of another mouse. In both cases, the 111In-labeled liposomes were strongly accumulated in the tumor, and the heterogeneous intratumoral localization in vivo was successfully visualized using SPECT with good spatial resolution. The liposomes were strongly located in the marginal regions of the tumors. We also found that the 111In-labeled liposomes were strongly accumulated in the liver and spleen as shown in the case #1 images. These results corresponded well with our previous results for biodistribution of radioactive liposomes [15, 1719].

Fig. 4
figure 4

In vivo SPECT/CT images of sarcoma 180-bearing mice 24 h after administration of 111In-labeled liposomes. Case #1: whole body images containing maximum intensity projection (MIP) and representative sagittal, coronal, and axial slice images containing the tumor. Case #2: representative sagittal, coronal, and axial slice images of the tumor site in another mouse. The SPECT projection data were acquired under the optimized conditions for in vivo imaging. The acquisition time was 150–300 s per projection with 24 (6 × 4) directions, total = 1–2 h. The acquired projection data were reconstructed using InVivoScope software with condition B

Figure 5a shows serial axial slices of SPECT tumor images with an interval of 0.2 mm. Each image shows good spatial resolution. We also demonstrated that the images were different from each other, even though all were part of the same tumor mass. This indicated that the distribution of the radioactive liposomes in the tumor appeared to be heterogeneous and was successfully visualized in vivo. Figure 5b shows ex vivo ARGs of sections of an excised tumor and the corresponding in vivo SPECT images. The ARGs showed excellent resolution and confirmed that the distribution of the 111In-labeled liposomes in the tumor was indeed heterogeneous, and in vivo SPECT images accurately reflected the heterogeneity.

Fig. 5
figure 5

Visualization of the intratumor distribution of 111In-labeled liposomes. a In vivo serial axial SPECT images of 0.2-mm thickness from the same tumor. b Ex vivo autoradiograms of excised tumor sections and the corresponding in vivo SPECT images

Figure 6 shows in vivo axial SPECT images of various-sized tumors obtained from four individual tumor-bearing mice that were administered 111In-labeled liposomes. For every case, the intratumoral distribution could be clearly visualized. After the acquisition, the tumors were removed, and their weight and radioactivity were measured (Table 1). The tumor sizes were between 0.51 and 1.61 g, and the accumulated radioactivity in the isolated tumors was between 0.26 and 1.2 MBq/g of tumor (5–10% of the injected dose/g).

Fig. 6
figure 6

Representative axial slice images of tumors obtained from four individual tumor-bearing mice administered 111In-labeled liposomes. Each image was acquired 24 or 48 h after administration of the 111In-labeled liposomes. The greatest dimension of the serial axial SPECT images for each mouse is shown. After imaging, the tumors were removed, and their weight and uptake of radioactivity were measured (Table 1)

Table 1 Radioactivity accumulation in the tumors that were imaged in vivo with high spatial resolution

Discussion

The aim of this study was to investigate the optimal conditions required for SPECT with spatial resolution high enough to visualize intratumoral heterogeneity using in vivo small animal imaging. These approaches have not often been attempted because scanners with high spatial resolution were needed for this purpose. Conventional radionuclide imaging devices have shown quite poor spatial resolution. However, recently developed SPECT scanners dedicated for small animals excel at spatial resolution and appear to hold potential for intratumoral heterogeneity imaging applications. Our SPECT scanner provided excellent spatial resolution of <1 mm in the phantom study under ideal acquisition conditions (Fig. 1a). However, it included conditions such as too high radioactivity and too long acquisition time, which are not practical in actual animal experiments. For practical purpose, most imaging studies using small animal scanners employ a more reasonable radioactive dosage and acquisition time; these conditions, however, should be insufficient for visualization of tumor heterogeneity using high resolution SPECT imaging. Therefore, in the first part of our study, we investigated the acceptable and optimal conditions required for acquisition and reconstruction of images in simulating actual animal experiments with limited radioactivity and acquisition time (Fig. 2).

As a result, we determined that adjustment of three parameters was required for optimal small animal imaging. First, we found that a pinhole aperture of 1.4 mm diameter was better for our purpose than an aperture of 1.0 mm diameter. Images from the 1.0-mm aperture were slightly higher in resolution than those from the 1.4-mm aperture as shown in Fig. 1, but the differences were not significant by visual inspection. Evaluation of the obtained results showed that the number of acquired counts strongly affected the spatial resolution in the simulation of small animal imaging. We judged that the 1.4-mm aperture had an advantage because it gave sensitivity that was twice as high as that obtained using the 1.0-mm aperture, as shown by the total acquired counts that were two times higher. Second, the acquisition time should be more than 30 min under these experimental conditions for the same reason as mentioned above (to acquire a sufficient number of counts). The time required is relatively long for in vivo animal imaging. Therefore, anesthesia and the stability of the animal position are very important. Third, the reconstruction parameters, such as noise suppression, voxel size, and iteration/subset number, should be optimized for imaging with limited acquisition counts and will be different from the parameter values required for the best system performance under ideal conditions. It is very important to choose appropriate reconstruction parameters according to the experimental objectives because the image quality varies greatly depending on the chosen values, even though the images are reconstructed from the same projection data [6, 20]. The best spatial resolution obtained under these conditions was approx. 1.3 mm. It would be acceptable for our purpose, though it did not quite match the best possible performance of the scanner (1.0 mm). The deterioration in resolution was due to the limited acquisition counts. Therefore, the SPECT requirements for high resolution animal images in vivo are of necessity distinct from those that can provide the best scanner performance.

We also evaluated the lower limit of radioactivity needed for visualization in vivo. We found that radioactivity <0.1 MBq/mL could not be visualized on the scanner images under the acquisition conditions described above (Fig. 3). The threshold of radioactivity for visualization was estimated to be 0.2–0.5 MBq/mL (or g) under the examined conditions. There are few reports on the threshold for visualization with small animal dedicated SPECT scanners. Pissarek et al. [21] reported a threshold level of 0.4–2 MBq/g of tissue for visualization of mouse brain and heart using [99mTc] hexamethylpropyleneamine oxime, [123I] iodophenylpentadecanoic acid, and [99mTc] sestamibi using a Triad 88/Trionix system with three 6-pinhole collimators. Their values were consistent with our results, and the reported image resolution was also similar to ours. Therefore, our threshold values seem to be much to the point. We should note that very good correlation was observed between the SPECT values based on the ROI and the actual radioactivity even for levels <0.1 MBq/mL, under the visualization threshold. These results indicate the excellent quantitative performance of the scanner.

It is no easy task to accumulate radioactivity counts in the tumor higher than the threshold level mentioned above. To achieve the necessary level, administration of a substantial amount of radiopharmaceuticals, such as 10–74 MBq/mouse, would be needed. Actually, this and higher amounts of radiopharmaceuticals have been administered in studies on small animal SPECT imaging [2, 11, 13, 21, 22]. The level is almost the same as the dose per human in conventional clinical SPECT tests. On the other hand, the body weight ratio of a mouse to a human is about 1:2000–3000, and the volume that can be intravenously injected into a mouse is 0.2–0.3 mL at most. Therefore, radiopharmaceuticals with high specific activity and high concentration are needed [23]. To meet these requirements, we used 111In-labeled liposomes. We have previously reported on the preparation of 111In-labeled liposomes with high radioactivity and excellent tumor accumulation of the liposomes [15, 16]. Currently, liposomes are drawing attention as multifunctional nanoparticles for imaging, delivery, and targeting in cancer therapy [12, 24, 25].

The 111In-labeled liposomes of 8–15 MBq were administered to the tumor-bearing mice, and their SPECT/CT images were obtained. As shown in Figs. 4, 5 and 6, the intratumoral distribution of the liposomes was successfully visualized in vivo with good spatial resolution. As shown in the serial images of Fig. 5a, the interiors of the tumors were visualized in detail in vivo and were apparently heterogeneous. Comparison of in vivo SPECT images and ARGs (Fig. 5b) confirmed the heterogeneity. The accumulated radioactivity in the tumor that could be imaged was 0.26–1.22 MBq/mL (Fig. 6), which corresponded well with the threshold for visualization that we established from the phantom experiments (Fig. 3). Previous studies have shown intratumoral heterogeneity by ex vivo ARGs [2628], but earlier techniques have not enabled in vivo visualization of tumor interiors in detail. The results of the present study are innovative. We have already imaged hypoxia-inducible factor-1-active regions in tumors using our SPECT/CT scanner under the optimal conditions defined here [29].

In conclusion, the optimal conditions required for SPECT imaging with high spatial resolution that enabled visualization of intratumoral heterogeneity in in vivo small animal imaging were investigated, and detailed visualization of tumor interiors using a SPECT scanner was achieved.