By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Cardiovascular pathologies are the major cause of morbidity and mortality in industrialized countries (1). These pathologies, including atherosclerosis, have a progressive chronic inflammatory evolution characterized by a series of specific cellular and molecular responses, from the initial inflammatory lesion to the advanced lesion stages at risk of rupture and thrombosis (2).
To monitor the progression of atherosclerotic plaques and assess molecules targeted by new, specific MRI markers, it is important to be able to image the aortic root and the carotid origin, where atherosclerosis predominantly develops. Mouse models of atherosclerosis, where high spatial resolution is essential, are of limited use without efficient cardiac and respiratory gating (3).
The impact of motion on the maximum achievable intrinsic spatial resolution with in vivo MRI has been well described (4–9). Various gating strategies have been used to limit this impact, including ventilation synchronization (10, 11), cardiac gating (12), cardiac and respiratory autogating (13), and radial encoding (14).
Simultaneous acquisition of cardiac and respiratory signals during MR exams enables the MR sequence to be synchronized to both the cardiac and respiratory cycles. MR acquisition can then be performed during the expiratory phase, which corresponds to minimal breathing motion. However, the fundamental resolution limit in any study employing synchronous and/or gated acquisition is determined by the reliability of the gating in capturing the moving anatomy at the same phase of the physiological cycle (4). Mouse cardiovascular MRI is particularly affected by cardiac and respiratory motion artifacts, exacerbated by the high heart and respiratory rates and the small cardiac anatomy of mice and by the high magnetic field strengths (>4.7 T) of the research systems (12, 15). These artifacts appear as motional blurring of the image in the readout direction, and by image ghosting in the phase-encoding direction (7, 12, 16).
In previous studies (4–9), respiratory gating was achieved using a pressure sensor. The present study developed a new approach, referred to as “single-sensor gating,” based on a breathing-modulated ECG signal, generating a demodulated respiratory signal and consequently a cardiorespiratory gating window. The method derived from a digital real-time cardiac gating system recently published (17).
The first aim of the present study was to assess the feasibility and performance of high-resolution MRI in combination with two different gating strategies: 1) double-sensor cardiac and respiratory gating, using both ECG and respiratory sensor signals; and 2) single-sensor cardiorespiratory gating using demodulated ECG.
The second aim was to apply the better of these two gating strategies to dynamic contrast-enhanced (CE) vessel-wall imaging of the aortic arch and carotid bifurcation, a mouse model of atherosclerotic plaque.
MATERIALS AND METHODS
As previously described, ECG detection used three neonatal electrodes (Red Dot; 3M, Saint Paul, MN, USA), a nonmagnetic electrical-to-optical converter and amplifier, and fiber-optic transmission to an optical-to-electrical conversion receiver (18). A solid-state ultra-low pressure sensor (SURSENSE; Honeywell, IL, USA), connected to an air-filled latex belt surrounding the abdomen of the mouse, was used to generate a respiratory signal. Using the real-time toolbox of Simulink (The MathWorks Inc., Natick, MA, USA), in-house written applications were developed to acquire and monitor the physiological signals and parameters (ECG R-waves and respiration-expiratory phases) and to control image acquisition (17).
The analog respiratory and ECG signal were sampled at 1 kHz using a data acquisition board (Daq PCI-6024E; National Instruments, TX, USA). Figure 1 shows a simplified block diagram of the gating system. The ECG signal was processed using a digital filter composed of low-pass infinite impulse response (IIR) filtering, followed by signal amplification. The filter order of 3 was selected for better MR interference elimination with minimal delay, and the cutoff frequency was set to 12 Hz to cope with the QRS component bandwidth (17). For both denoised ECG and respiratory digital signals, threshold detection with conversion of the physiological signals into transistor-to-transistor logic (TTL) window acquisition signals was applied. The TTL signals were transferred to the data acquisition board outputs, and then fed as analog data into the MR scan management controller.
The first gating method was double-sensor cardiac and respiratory gating, using two physiological signals. Image acquisition was initiated by specific triggers from biological references, such as ECG QRS and/or respiration phase. QRS complexes were detected by applying a Schmitt trigger to the denoised signal. For efficient, correct R-wave detection, the threshold was updated by automatic adjustment; when automatic adjustment was not sufficient, a manual adjustment mode was selected (17). The expiratory phases were selected as time segments with sensor voltage below a constant threshold (1.95 V in our particular case).
The second gating method (single-sensor) was based on a single ECG acquisition combined with breathing modulation detection. The respiratory signal was obtained by numerical demodulation of the breathing-modulated ECG signal, using an interpolation of the amplitudes of the detected peaks of QRS complexes, after low-pass filtering (0.8 Hz cutoff frequency, adapted to mouse breathing rates). The expiratory phases were detected when the demodulated respiratory signal was below its mean value.
For both gating methods, the acquisition window period was triggered by the generation of a TTL signal derived from QRS during the expiration phase. The period of triggered gated MR acquisition was fixed as one-third of the RR cycle in systolic and two-thirds in diastolic imaging.
A user-friendly interface was developed, allowing the gating technique to be selected by analog switches, with display of gating parameters such as R-R interval, heart rate (HR), and gating delay between TTL rising edge and ECG R-wave.
Healthy male C57BL/6 mice (Charles River Laboratories, L'Arbresle, France) and mice deficient in apolipoprotein E (male ApoE−/−, 22 weeks) (Taconic, Denmark) were used.
The mice were anesthetized by intraperitoneal injection of sodium pentobarbital (60 mg/kg).
The following procedure was used to obtain the ECG and to minimize the noise level:
1Paws were epilated and a contact gel (Ten 20; Weaver and Company, Aurora, CO, USA) was used;
2The limb electrodes were positioned close together;
3The axis between the limb electrodes and the leg electrode was held parallel to the magnetic flux lines;
4Limb and leg electrodes were positioned as close together as possible and positioned in the center of the MR imager;
5The respiratory sensor was placed in the abdominal area.
All experiments were performed on a 2-Tesla horizontal MR system (Oxford magnet and MRRS console) equipped with a 180 mT/m gradient set. The mice were placed supine in a homemade Alderman-Grant coil (19) (30-mm long, 30-mm diameter). Body temperature was maintained with a water-circulating heating blanket.
After manual shim adjustment, axial and coronal scout images of the imaged area (heart, aortic root, and the carotid origin) were acquired using a two-dimensional (2D) gradient echo (GE) sequence.
To evaluate gating robustness, three types of high-resolution MRI sequence were tested: GE, spin echo (SE), and fast SE (FSE).
Acquisition parameters were TR/TE = 385/10 ms, 450/18 ms, and 2300/50 ms, for GE, T1-weighted SE, and T2-weighted FSE, respectively. A total of six slices of 1-mm thickness were acquired with an in-plane resolution of 100 μm. The delay between R-wave triggering and TTL gating signal rising edge was set at 20 ms.
For the application to vessel-wall imaging, pre- and postcontrast images were acquired. Axial T1-weighted images of the carotid origin were obtained using a 2D multislice SE sequence (5 slices): TR = R-R = 290 ms, TE = 18 ms, matrix size = 256 × 256, slice thickness = 1 mm, pixel size = 90 μm, and signal averages = 4. MR acquisition was performed before and over 1 h after tail-vein injection of 0.016 mmol/kg Vistarem (P792; Guerbet, France), a gadolinium blood-pool contrast agent with a predominantly vascular distribution, a molecular weight of 6.47 kDa, high relaxivity (r1 = 39 s−1mM−1 at 20 MHz and 37°C) and fast renal elimination (20, 21). Image intensity was calibrated using the signal from an external reference: a 1-mm diameter catheter filled with 98 μmol/liter of P717 diluted in water (22).
For each mouse strain, ECG and respiratory signals were recorded during various MR sequences. To assess gating strategy performance, 10-s time windows were selected from the recorded sample signals for each mouse/sequence combination. Each sample lasted for at least six respiratory cycles, thus containing several cardiorespiratory acquisition windows. These samples were representative of the various events in the sequence, such as RF pulsing, read, phase, and slice encoding.
The QRS number was not constant because of heart rate variations between strains and individual mice.
Thus, the study was performed on physiological signals obtained using three different sequences on seven mice.
Gating efficiency for R-wave detection was assessed by measuring sensitivity (S) and positive predictive value (P), given respectively by:
where TP is the number of true positives (TTL signals corresponding to an R-wave), FP the number of false positives (TTL signals not corresponding to an R wave), and FN the number of false negatives (R-waves not giving a TTL signal) (23).
To evaluate the stability of detection of the cardiorespiratory gating window, cardiorespiratory accuracy (CRA) and cardiorespiratory regularity (CRR) percentages were measured, given, respectively, by:
where B is the number of expiratory phases (low phases) per sample and F is the number of cardiorespiratory gating windows (Fig. 2) per sample. D is the number of cardiorespiratory gating windows having a number of TTL gating signals equal to the mean number of TTL signals in gating windows per sample (Fig. 2). This mean value is obtained from the ratio C/F, where C is the total number of TTL signals during sample signal time.
Signal enhancement was measured from pre- and postcontrast T1-weighted images using dedicated homemade arterial wall analysis software written in C++ (Creatools) (24). Arterial wall contours were determined manually by an internal and an external contour. For each slice, the lumen (internal) contour was obtained from the postcontrast image giving optimal wall determination, and propagated to the pre- and postcontrast images. The arterial wall signal was measured in the region between the two contours. Image intensity was calibrated using the signal from the external reference catheter. The mean signal intensity from a region of interest (ROI) in the reference tube was used to normalize pre- and postcontrast signal intensity. An ROI was defined outside the animal to measure the SD of the noise. The signal-to-noise ratios of the arterial wall (SNRwall) and of the reference (SNRref) were calculated by dividing the signal by the SD of the noise. For each pre- and postcontrast image, (SNRwall)pre and (SNRwall)post were measured. For each slice, the percentage enhancement (%ENH) of the aortic wall was determined as follows:
The signal enhancement from muscle tissue was also calculated by the same equation.
A total of seven mice were investigated: C57BL/6 (N = 4, mean weight = 25 ± 2 g, mean heart rate = 240 ± 50 bpm) and ApoE−/− (N = 3, mean weight = 32 ± 3 g, mean heart rate = 330 ± 100 bpm). A total of 20 min was necessary to prepare the animal and to place it in the magnet to start the MRI protocol.
Gating efficiency parameters are summarized in Table 1 (double-sensor cardiac and respiratory gating) and Table 2 (single-sensor gating). Sensitivity (S) and positive predictive value (P) were 100% in both strategies.
Table 1. Efficiency of Double-Sensor Cardiac and Respiratory Gating Using Respiratory Sensor and Automatic Trigger Adjustment for ECG for the Various High-Resolution MR Sequences
NA = nonapplicable values.
82 ± 8
94 ± 6
96 ± 4
62.5 ± 18
59.5 ± 3.5
93 ± 5.7
67.6 ± 6.8
Table 2. Efficiency of Single-Sensor Cardio-Respiratory Demodulation Gating Using Automatic Trigger Adjustment for the Various High-Resolution MR Sequences
NA = nonapplicable values.
86 ± 14
62.3 ± 6.5
76.3 ± 12.8
76.3 ± 12.8
87.5 ± 12.5
45 ± 31
65.5 ± 29
73.5 ± 9.2
85.5 ± 14.5
51 ± 18.5
Typical physiological signals derived from a C57BL/6 mouse positioned in the magnet during an SE sequence are shown, after signal processing, in Fig. 2.
Cardiac images obtained without gating, with cardiac gating only and with cardiorespiratory gating are shown for the two gating strategies in Figs. 3 and 4, respectively. With both cardiorespiratory gating strategies, the ghosting artifact due to thoracic motion in the respiratory cycle disappeared, SNR improved, and smaller anatomic details appeared at the level of the aortic root and the carotid origin.
For the CE-MRI application, an optimal processing technique was selected (double-sensor gating strategy) based on the results summarized in Tables 1 and 2. These criteria may depend on features of the experimental setup such as stable expiratory phase, diastole window, R-R period, and delay.
MR follow-up of contrast enhancement after injection of P792 was performed successfully in both groups of mice: i.e., C57BL/6 (N = 3) and ApoE−/− mice (N = 5). For each mouse, signal intensity was measured at the carotid origin before and 80 min following injection. Signal analysis was performed in eight mice on a total of 50 MR images.
Figure 5 shows a GE coronal image from an ApoE−/− mouse. The carotids and the aortic arch appear with a pixel size of 89 μm. The axial T1-weighted images from the same mouse show increased signal intensity in the carotid wall at 28 min postcontrast injection.
On postcontrast images, the SNR of the carotid wall increased in ApoE−/− mice, peaking at 20 to 35 min postcontrast (postcontrast: 5.1 ± 1.2 vs. precontrast: 4.1 ± 1.2) and then decreasing.
In ApoE−/− mice, the mean percentage arterial wall signal enhancement (%ENH; Fig. 6) increased and peaked (at 32%) 27 min postinjection. Signal enhancement in muscle tissue was limited and decreased rapidly (Fig. 6).
In C57BL/6 mice, signal enhancement did not change from pre- to postinjection and carotid wall thickness was below pixel size.
In the present study, we obtained high-resolution MR images of atherosclerotic plaque at the aortic root and the carotid origin in ApoE−/− mice using digital cardiac and respiratory gating.
For both cardiorespiratory gating strategies, low-pass filtering with automatic adjustment gave 100% correct gating without false triggers or FN for all sequences and all mouse models. In our experimental setting (i.e., imaging at 2 Tesla, pentobarbital anesthesia, C57BL/6 genetic background), sensor detection (double-sensor gating) performed better than single-sensor demodulated cardiorespiratory gating. Moreover, with demodulation-based gating, cardiorespiratory gating accuracy decreased markedly in case of low heart rate (data not shown), whereas the double-sensor performances were independent of cardiac rhythm. Double-sensor gating using sensor signals showed better expiration (low) phase detection, with percentage accuracy (CRA) between 74 and 100, whereas the accuracy obtained with the single-sensor demodulation strategy was affected by heart rate variation.
CRR sheds light on the stability of the number of TTL gating signals per acquisition window during MRI. CRR was satisfactory with the double-sensor strategy (Table 1), but only moderate (mean = 53%) with the single-sensor demodulation-based strategy (Table 2). This may be explained by the fact that expiratory phase duration detection was disturbed by heart rate changes. For higher heart rates (isoflurane anesthesia and a higher magnetic field) the demodulation strategy was more efficient (data not shown). Thus, both strategies of cardiac and respiratory gating provide high-quality and high-resolution images, but the single-sensor gating technique (using numerical demodulation of the respiration-modulated ECG signal) has advantages compared to the second technique: 1) no respiratory sensor is necessary; and 2) less animal preparation time is required. Detecting the demodulated respiratory signal requires a stable heart rate, which can be obtained by using anesthetic gas such as isoflurane, and a higher magnetic field than in the single-sensor approach (15).
In this MRI study at 2 Tesla, we obtained sufficient spatial resolution for cardiovascular imaging in mice (pixel size 78 μm to 156 μm) (22, 25). Our digital synchronization system proved well suited to the fast and highly variable heart rate of mice and to overcoming ECG disturbance induced by high-resolution MR sequences. Visualization of the heart area, aortic root, and carotid origin was previously hindered in mice due to cardiac and respiratory motion artifacts.
Carotid wall MRI without gating is feasible in the neck area (26), but not in the thorax, where cardiorespiratory synchronization is essential both for visualizing the ascending aorta at heart level and for high-resolution imaging of the carotids (27). With cardiorespiratory gating, a much higher spatial resolution (89 × 89 μm2) is achieved, compared to that obtained with cardiac synchronization only (105 × 105 μm2) (17). This spatial resolution is even more of a challenge with a clinical field of 2 Tesla (25), where signal accumulation is required.
For MRI characterization of atherosclerosis lesions in vivo, multicontrast acquisition (i.e., T1-weighted, T2-weighted, and proton density-weighted [PDW]) is required (28). In order to obtain all these multicontrast acquisitions with cardiorespiratory gating, Cartesian sampling was preferred to other sampling methods, considering its robustness and the requirement of high spatial resolution (29).
It has recently been shown that MR contrast agents can improve atherosclerotic plaque characterization by providing an index of plaque activity. Dynamic contrast enhancement in carotid plaques using nonspecific contrast agent correlated to neovascularization and increased permeability (30). However, the fast and nonspecific kinetics of standard gadolinium compounds precludes further kinetic modeling to differentiate perfusion and permeability processes. Using a blood-pool agent in tumors or in myocardial ischemia has proven useful in measuring both tissue perfusion and permeability. In the mouse abdominal aorta, differential contrast kinetics with a blood agent enabled discrimination between early lesions and more advanced plaques (22). In the present study, P792, a fast blood-pool agent, was injected and dynamic contrast enhancement of the carotids was measured. Image analysis showed the efficacy of P792: fast elimination from the blood pool enables the vessel-wall SNR to be measured without signal contamination from the lumen. Fast blood kinetics is illustrated by the muscle tissue SNR curve, showing only modest enhancement (6%) at 10 min postinjection and a rapid decrease (Fig. 6). The macromolecular nature of P792 and its high relaxivity enables tissue permeability measurement from contrast kinetics. In our example, the vessel-wall signal in ApoE−/− mice increased directly after injection, peaked at 30 min (>30% enhancement) and was heterogeneous 50 min later (Fig. 6).
The present study led us to investigate the optimal repetition time for each acquisition (i.e., the TR for T1-weighted gated images) to minimize scan time while performing dynamic measurements under equivalent conditions. In practice, TR is initially chosen so as to maximize the T1 effect of the contrast agent. With ECG gating, TR is a function of the R-R interval. With double gating strategies, the effective TR is determined by cardiac and respiratory rates, which both depend on many factors and physiological conditions (mouse strain, animal temperature, and type of anesthesia).
Some anesthetics cause cardiac and/or respiratory depression: cardiac and respiratory rates decrease (9). Consequently, the TR increases and signal intensity is modified as the time to acquire the center of the Fourier plane changes during acquisition. Dynamic contrast agent study using cardiorespiratory gating, therefore, required an external reference signal to correct for SNR variability due to TR changes. In the present study, TR changes during the hour-long dynamic study were limited and the mean acquisition time per T1-weighted sequence was about 12 min (Fig. 6).
This study has a few limitations. 1) Images were obtained at 2 Tesla with a long acquisition time, as the number of accumulations had to be increased to obtain a good SNR. This time could be reduced using higher field MRI. 2) The study was intended to be simply a feasibility study of contrast enhancement of atherosclerotic plaque: the number of animals was too small for statistical analysis.
In conclusion, this study demonstrated the efficiency of both of the cardiorespiratory gating strategies employed. Digital processing ensures considerable time savings (17) compared to available cardiorespiratory gating systems. With the development of molecular imaging (3), it is important to differentiate the delivery mechanisms of the specific agent to the target, depending on the stage of the lesion. The method allows analysis of dynamic vessel-wall enhancement in mice at the aortic root and the carotid origin, where atherosclerosis predominantly develops initially. It will be applied to molecular imaging of atherosclerosis lesions in mice.
This work was supported by the IPA Program, Rhône-Alpes Region grants to H.A. We thank Valbex (Centre de Bioexpérimentation, Université Lyon 1) for mouse housing and care.