Abstract
During the treatment of colorectal liver metastases, evaluation of treatment efficacy is of the utmost importance for decision making. The aim of the present study was to explore the ability of preclinical imaging modalities to detect experimental liver metastases. Nine male Wag/Rij rats underwent a laparotomy with intraportal injection of CC531 tumor cells. On days 7, 10, and 14 after tumor induction, sequential positron emission tomography (PET), computed tomography (CT), and magnetic resonance imaging (MRI) scans were acquired of each rat. At each time point, three rats were euthanized and the metastases in the liver were documented histologically. Topographically, the liver was divided into eight segments and the image findings were compared on a segment-by-segment basis with the histopathologic findings. Sixty-four liver segments were analyzed, 20 of which contained tumor deposits. The overall sensitivity of PET, CT, and MRI was 30%, 25%, and 20%, respectively. For the detection of tumors with a histologic diameter exceeding 1 mm (n = 8), the sensitivity of PET, CT, and MRI was 63%, 38%, and 38%, respectively. The overall specificity of PET, CT, and MRI was 98%, 100%, and 93%, respectively. This study showed encouraging detectability and sensitivity for preclinical imaging of small liver tumors and provides valuable information on the imaging techniques for designing future protocols.
DETECTION AND MONITORING of colorectal liver metastases have a major influence on the treatment strategy and prognosis of colorectal cancer (CRC) patients. Detailed information on the presence, localization, size, and number of metastases influences clinical decision making in terms of surgery, adjuvant therapy, and palliative therapy regimens.1,2 Moreover, the use of imaging as an early marker for response is important because it allows a window of opportunity during which treatment regimens can be tailored accordingly, resulting in a decrease in morbidity and undue costs. 3 Also, research on the potential effect of new therapeutic agents, such as antibody-based agents and other biologics, depends highly on response monitoring.4–6 Tumor visualization is traditionally performed using anatomic imaging techniques such as ultrasonography, computed tomography (CT), and magnetic resonance imaging (MRI). 7 In daily practice, 18F-2-deoxy-2-fluoro-D-glucose positron emission tomography (FDG-PET) is mainly performed when CT or MRI is not decisive. Functional imaging by FDG-PET is considered to be of additional value because of high tumor detection rates.
Adequate small-animal imaging of colorectal liver metastases can be of value for several reasons: (1) the efficacy of new therapeutic agents can be studied in preclinical models, in which noninvasive response monitoring limits the number of animals needed; (2) it allows preclinical studies on the role of different imaging modalities in decision making in CRC patients with liver metastases; and (3) it could be used in studies on multimodality imaging, which is considered to enhance diagnostic information. During the past decade, new animal imaging modalities have been developed and improved. Given that PET, CT, and MRI are currently used in the diagnosis and follow-up of patients with colorectal liver metastases, these imaging modalities were chosen as a subject of this study. The aim of the present exploratory study was to determine the image quality, sensitivity, and specificity of these three preclinical imaging modalities to detect small liver metastases in a rat model.
Material and Methods
Animals, Cell Line, and Surgical Procedure
Nine 10- to 12-week-old male Wag/Rij rats, with a mean weight of 254 g, were used (Charles River Laboratories, Sulzfeld, Germany). They were accustomed to laboratory conditions for 1 week before use and housed under nonsterile, standard laboratory conditions (temperature 20–24°C, relative humidity 50–60%, 12-hour light-dark cycle), with free access to animal chow and water. All experiments were conducted in accordance with the principles laid out by the revised Dutch Act on Animal Experimentation (1997) and approved by the institutional Animal Welfare Committee of Radboud University Nijmegen.
The syngeneic rat colonic carcinoma cell line CC531 is derived from colonic tumors of Wag/Rij rats exposed to 1,2-dimethylhydrazine. 8 CC531 cells were cultured and suspended as described previously. 9 All surgical procedures were carried out under clean conditions. Tumor inoculation was performed via a midline laparotomy and portal vein injection as described previously. 10 Rat cages were kept on a warm mattress for the first 24 hours after operation. Before surgery and on the first and second days thereafter, analgesia was given in the form of subcutaneous carprofen injections (5 μg per 100 g body weight).
Study Design
On day 0, all rats underwent the surgical procedure as described above. On days 7, 10, and 14 after the operation, all rats underwent PET/CT and MRI under general anesthesia (isoflurane, oxygen, and nitrous oxide). At each time point, three rats were euthanized using oxygen/carbon dioxide asphyxiation and immediately dissected. The liver was fixed in formalin, embedded in paraffin in the anatomic position, and cut in the coronal plane for routine histopathologic hematoxylin-eosin staining (5 μm thick slices at each 1 mm of liver). To determine the localization of each lesion, the liver was topographically divided into eight segments (right/left, ventral/dorsal, cranial/caudal). Of each imaging modality, the metastasis detection rate in each segment was compared to the histopathologic findings (12.5× magnification).
MicroPET Imaging
PET images were acquired with an Inveon small-animal PET/CT scanner (Siemens Preclinical Solutions, Knoxville, TN) with an intrinsic spatial resolution of 1.5 mm. 11 All rats were fasted 6 hours before an intravenous injection of 10 MBq 18F-FDG (Cyclotron BV, Amsterdam, the Netherlands). One hour after injection, the animals were placed in a prone position in the scanner and body temperature was maintained at 37°C by placing the rats on a warmed mattress. PET emission scans were acquired for 15 minutes. Scans were reconstructed using Inveon Acquisition Workplace software version 1.2 (Siemens Preclinical Solutions). PET reconstruction was performed using an ordered subset expectation maximization–three-dimensional/maximum a posteriori (OSEM3D/MAP) algorithm with the following parameters: matrix size 256 × 256 × 159, pixel size 0.43 × 0.43 × 0.8 mm3, and a uniform variance MAP smoothing parameter of β = 0.05.
MicroCT Imaging
CT was performed in the same imaging session as PET, avoiding the need to reposition the animals. CT images were acquired with the same Inveon animal PET/CT scanner. A two-bed-positions scan was acquired with 20% overlap (80 kV, 500 μA, exposure time 300 ms per step, 360° rotation in 180 steps, low magnification, bin by 4, 0.5 mm aluminium filter). Two hours prior to scanning, rats received 1.5 mL of a reticuloendothelial system–specific contrast via the tail vein (glyceryl 2-oleoyl 1, 3-bis [7-(3-amino-2, 4, 6-triiodophenyl)] alkanoate, Fenestra LC, Advanced Research Technologies, Inc, Montreal, QC). Twenty minutes prior to scanning, rats were given an intraperitoneal injection of 1.5 mL ioversol (Optiray 320, Covidien, Petten, The Netherlands; 1:2 dilution). Scans were reconstructed using COBRA software version 6 (Exxim Computing Corporation, Pleasanton, CA) using a proprietary algorithm (SAMARA) with a Shepp-Logan reconstruction filter, a polynomial-based soft tissue beam-hardening correction, and a medium noise-ring artifact reduction. The reconstructed scans had the following parameters: matrix size 768 × 768 × 922, pixel size 0.11 × 0.11 × 0.11 mm3.
MicroMRI
All MRI experiments were performed on a 7 T small-animal ClinScan 70/30 USR (Bruker BioSpin, Ettlingen, Germany). The scanner has a B-GA 20S gradient insert with a clear bore size of 200 mm in diameter, maximum gradient strength of 290 mT/m, and a slew rate of 1160 T/m/s. The ClinScan is interfaced with the Siemens Magnetom Avanto Syngo user interface (Siemens Medical Solutions, Erlangen, Germany). A dedicated 50 × 70 mm homebuilt transmit/receive microstrip radiofrequency surface coil was used with a customized plastic cradle to accommodate the rat in a stable prone position to reduce motion artifacts, as described in detail by Gambarota and colleagues. 12 One day prior to scanning, the rats were given an intravenous injection of 0.1 mmol/kg ultrasmall superparamagnetic iron oxide (Sinerem, Guerbet Group, Villepinte, France) to increase liver to tumor contrast. During all measurements, the body temperature and breathing frequency of the rats were monitored using the magnetic resonance–compatible model 1025 monitoring and gating system (SA Instruments, Stony Brook, NY). The body temperature was kept constant at 37 ± 1°C with a small-rodent heater system (SA Instruments), and the breathing cycle was kept constant at 45 ± 10 times/min for stable respiratory triggering to further reduce motion artifacts. After initial imaging of the liver with a fast gradient echo localizer, 54 contiguous coronal turbo spin echo (TSE) images (ie, parallel to the radiofrequency coil) and 64 contiguous TSE sagittal images were acquired followed by a FLASH 3D VIBE. The TSE images were used for anatomic localization of the liver and detection and measurement of the tumors. The FLASH 3D VIBE was used for additional anatomic localization of the tumors, especially in relation to the vessels in the liver. Imaging parameters were as follows: for the TSE images: field of view (FOV) = 120 × 120 mm, matrix size = 256 × 256, slice thickness = 0.7 mm, repetition time (TR) = 1,350 ± 150 ms, echo time (TE) = 34 ms, turbo factor = 7, total acquisition time (TA) ≈ 6 minutes. To reduce motion artifacts, all TSE acquisitions were triggered in real time by the breathing cycle of the rat so that, for each image, each set of k lines was acquired in the same phase of the breathing cycle. This caused the small variations in TR. For the FLASH 3D VIBE: FOV = 120 × 120 × 120 mm, matrix size = 384 × 384 × 384, TR = 40 ms, TE = 1.2 ms, TA = 19:43 minutes.
Image Analysis
The images were arranged randomly, and blinded interpretation was done by independent experts. MRI and CT images were analyzed by two radiologists (H.M.D. and R.D.M.M., respectively). PET images were reviewed by a nuclear medicine physician (M.G.). Experts were asked to record the presence of intrahepatic lesions in each liver segment. When lesions were present, they were asked to describe the localization (eg, segment) and estimated size of the six largest lesions in the sagittal plane. FDG uptake was scored semiquantitatively on a 1 to 4 scale as well as by drawing regions of interests (ROI) centered over the tumors and the intestines without correction for partial volume effects. The mean amount of activity, expressed as the standardized uptake value (SUV, defined as the mean uptake in the lesion, scaled by the administered activity and the body weight), was calculated. The imaging findings were then compared to the histopathologic findings on a segment-by-segment basis.
Data Analysis
The sensitivity and specificity of each imaging modality were determined. Regression analysis of the detection rate of each modality as a function of the size of the lesions was performed. Retrospectively, the chronological detection of growing metastases was analyzed. In this exploratory study, the number of observations was limited and the data were not independent, so no statistical level of significance was determined.
Results
Surgical Procedure
One rat died shortly after the operation, probably owing to hypothermia. All other surgical procedures went uneventfully.
Image Quality
A typical example of a 1.8 mm lesion in the liver at dissection, the corresponding microscopic image, and the corresponding PET, CT, and MRI images are displayed in Figure 1. The image quality of the PET images was judged as reasonable. Interpretation of a few images was impeded owing to difficult alignment of the liver. Moreover, in 20 of 44 cases, the expert noted enhanced uptake in the laparotomy wound and physiologic uptake of FDG in the intestines, complicating correct interpretation of the PET images (Figure 2, A and B). The mean SUV of the tumors was 2.7 (range 0.7–8.4), whereas the mean SUV of the intestines was 1.9 (range 1.2–4.8). The mean ratio of the SUVmean (tumor) to SUVmean (intestines) was 1.5. The animal CT scanner is a conventional scanner, with an average scan time of 15 minutes per rat, inevitably causing artifacts by breathing motion. Breathing artifacts appeared as blurring of the structural contours and often as doubling or tripling of these contours. They can seriously deteriorate image quality and therefore hinder correct image interpretation (Figure 2C). Moreover, beam hardening influenced the image quality of the CT scans. (Figure 2D). The image quality for MRI was judged as good. At the postcontrast TSE images, the liver was dark gray and the metastatic lesions were hyperdense (white), facilitating detection of the tumors. However, the quality of the dorsal planes was lower than the quality of the ventral planes because the coil was localized at the anterior side of the animal. Some motion artifacts, mainly owing to respiration and intestinal motility, were present. However, because image acquisition was triggered by movement in the transverse plane, the artifacts were not present in the coronal plane and did not disturb the interpretation of the liver region (see Figure 2E).
A typical example of a 1.8 mm lesion (green arrows) in the left, ventral, cranial segment of the liver at dissection (A), the corresponding microscopic view (B, ×12.5 original magnification), the 18F-FDG PET image (C), the CT image (D), and the MRI (E).
Typical findings and artifacts (arrows) on PET, CT, and MRI. Misinterpretation on PET by enhanced FDG uptake in the laparotomy wound (A) and by physiologic uptake of FDG in the intestines (B). Breathing artifcats on CT at the level of the diaphragm (C). Beam-hardening artifacts (D). Motion artifacts in the coronal plane on MRI (E).
Per-Segment Analysis
Of the 64 segments (eight rats with 8 segments in the liver), 20 segments contained tumors at the time of dissection. These tumors seemed evenly well distributed over the segments, with two to four tumors in each segment. Only one segment did not contain any tumor in any of the livers (left dorsal cranial segment). The largest tumors in these segments had diameters of < 0.50 mm, 0.50 to 1.00 mm, and > 1.0 mm in 3, 9, and 8 segments, respectively. Forty-four of the 64 segments did not contain any microscopically detectable tumors. The sensitivity and specificity of each imaging modality as determined for different tumor sizes are summarized in Table 1. The overall sensitivity of all modalities was poor (20–30%). However, the overall specificity was high, especially for PET and CT (98% and 100%, respectively), whereas the specificity of MRI was 93%. The sensitivity of the three modalities was similar for tumors exceeding 0.50 mm in diameter. The sensitivity for detecting tumors with a diameter larger than 1.00 mm was 63%, 38%, and 38% for PET, CT, and MRI, respectively, without major changes in the specificity of the three modalities. This positive correlation between the tumor diameter and the sensitivity of each modality was confirmed by regression analysis. The correlation coefficient (R) between PET, CT, and MRI and tumor diameter was .42, .30, and .43, respectively, indicating a moderate correlation. 13
Sensitivity and Specificity of Preclinical 18F-FDG-PET, CT, and MRI for the Detection of Small Liver Metastases in Rats
CT = computed tomography; 18F-FDG-PET = 18F-2-deoxy-2-fluoro-D-glucose; MRI = magnetic resonance imaging; PET = positron emission tomography.
In retrospect, all scans of the rat with the largest tumor at day 14 were analyzed, and the results are summarized in Figure 3. On day 10 after injection, both PET and MRI detected a tumor in the same liver segment in this rat. On day 7 after injection, MRI also detected a 1 mm lesion in this segment. At that time point, PET only revealed a lesion with low uptake in another segment (right, dorsal, caudal side of the liver), most likely owing to physiologic uptake in the intestines (not shown). On the CT scans of the same animal, no tumors were detected on day 7 or 10 after injection.
Image analysis of longitudinal scans of rat with a large confluent (13 mm diameter) tumor on the right, ventral, cranial side and multiple small tumors at other locations on day 14 after tumor inoculation, including estimated size or uptake of the lesion. *Comment by expert: only a little uptake at the dorsal side (not shown), possibly intestinal uptake. **Comment by expert: multiple tumors.
Discussion
The results of this study demonstrate the potential of small-animal imaging to be used to monitor tumor growth in rats. The sensitivity of all modalities correlated with the diameter of the tumor. Small tumors with a diameter of approximately 1 mm can be detected by FDG-PET with a sensitivity > 60%, whereas the sensitivity of CT and MRI for detecting these lesions was 38%. The specificity of all three modalities was high, allowing monitoring of therapy response in future studies. Because of the limited number of observations, these results should be considered an indication of the in vivo visualization potential of these three imaging modalities. Moreover, a possible effect of the postoperative effects of surgery and the various agents administered to optimize tumor visualization cannot be excluded. Visualization of tumors, or lesion detectability, for all three imaging modalities is basically determined by three parameters: (1) the average or maximum signal level in the lesion versus the average background level (tumor to background ratio), (2) the lesion size compared to the spatial resolution of the imaging modality, and (3) the noise level in the image. The interplay between these parameters determines how well lesions can be detected.
Although it was beyond the scope of this study to assess all parameters that affect lesion detectability for the three modalities, multiple measures have been taken to optimize the settings of all three imaging modalities for this particular experiment. For microCT, we used two contrast agents to improve the alignment of the liver and to enhance the intrahepatic tumor to background contrast. The image quality of microPET was optimized by minimizing uptake of FDG in other organs by anesthesia, fasting, and warming of the animals. 14 For microMRI, the breathing cycle was kept constant, allowing respiratory triggering to reduce motion artifacts. However, all three small-animal imaging modalities suffered from artifacts, so image quality could potentially be improved further.
The FDG-PET instrumentation could be further improved. The spatial resolution could possibly be increased by decreasing the size of the detection crystals in future scanner designs. The signal to noise ratio can be enhanced by increasing the aspect ratio of the PET scanner, which increases its sensitivity because more photon pairs are being detected. 11 Moreover, the false-positive rate for FDG-PET, owing to uptake in inflammatory tissue, could be a potential aspect of improvement. If suture material is still present in the scar, macrophage accumulation may cause high FDG uptake, resembling tumor uptake. This may explain the difficulties localizing the site of FDG uptake to the liver or the abdominal scar.
Simultaneous CT imaging would help better localize the site of uptake. Also, in the case of nonspecific bowel uptake, simultaneous PET/CT scanning will help in exactly localizing a potential lesion. Furthermore, gated imaging may help avoid breathing artifacts so that on combined PET/CT images, the anatomic correlation would be optimized in the area of the liver, where breathing artifacts are more pronounced owing to the close proximity to the diaphragm. The disadvantage of gating would be the considerable increase in acquisition time, so there will always be a tradeoff between anatomic matching, imaging quality, and duration of the scan. Improving image quality of CT could be achieved by reducing beam hardening. Beam hardening refers to the process of selective removal of soft x-rays that are easily attenuated from the x-ray beam. As these x-rays are removed, the remaining beam becomes more penetrating (hard x-rays). Beam hardening is the result of the combination of the initial x-ray spectrum and the composition and thickness of the tissue traversed. This results in beam-hardening artifacts in CT images. Unfortunately, beam hardening in CT scans of bony areas is hard to prevent. As for MRI, respiratory gating was already conducted to reduce motion artifacts. By means of this technique the MRIs in the coronal and sagittal planes were of acceptable quality. Because of the rapid breathing rate of a rat, further optimization of MRIs will be difficult to achieve.
This study on preclinical imaging modalities for detection of liver metastases shows encouraging sensitivity and specificity scores for preclinical imaging of small liver tumors. Moreover, this study provides information on image artifacts and practical considerations for designing future protocols. Each of these modalities has its particular advantages and disadvantages and could therefore suit specific research questions. Multimodality imaging, which is considered to enhance diagnostic information, is currently studied in several (pre)clinical studies and faces many challenges.15–17 The quality of (preclinical) multimodality imaging is likely to benefit from improved performance of each modality and does not require the development of multimodality probes. 18 The results in this study support further studies on detection of liver metastases in an experimental setting and can easily be implemented in studies on multi-modality imaging.
Footnotes
Acknowledgments
We wish to thank Andor Veltien and Arend Heerschap (Department of Radiology), Roger M.L.M. Lomme (Department of Surgery), Bianca A.M.G. Lemmers-van de Weem, and Kitty J.H. Lemmens-Hermans (The Central Animal Laboratory), Radboud University Nijmegen Medical Centre, for excellent technical assistance.
Financial disclosure of authors and reviewers: None reported.