Detection of thrombus size and protein content by ex vivo magnetization transfer and diffusion weighted MRI

Atherosclerosis was induced in adult male (n = 17) New Zealand White rabbits (Charles River Laboratories, Wilmington, MA) as previously described [31, 38]. Briefly, rabbits were fed a 1% cholesterol diet (PharmaServe, Framingham, MA) for 2 weeks before and 6 weeks after balloon injury of the abdominal aorta, followed by 4 weeks of normal diet. Plaque disruption and thrombosis was induced with Russell’s viper venom (0.15 mg/kg, IP, Enzyme Research, South Bend, IN) followed by histamine 30 min later (0.02 mg/kg, IV, Sigma-Aldrich, St. Louis, MO). This procedure was performed twice within 48 h. Finally, 24 h after the last pharmacological triggering, the animals were sacrificed with sodium pentobarbital (100 mg/kg, IV). Heparin was infused to inhibit postmortem coagulation. The aortas, from the renal branches to the iliac bifurcation, were harvested, sectioned and catalogued. The segments were fixed in 10% formalin overnight and then transferred in PBS for ex vivo MRI followed by histology. All experiments were performed under the approval of the Institutional Animal Care and Use Committee on Animal Investigations.

The specimens used for this ex vivo study were derived from our previous work that focused on in vivo imaging of atherosclerosis [31]. Ex vivo MRI was performed in a vertical-bore Bruker Avance spectrometer (11.7-Tesla) fitted with gradient coils (bore size = 89 mm; maximum gradient strength = 906.6 mT/m). Aortic segments were imaged in phosphate-buffer saline (PBS) using a 10 mm birdcage transmitter/receiver coil. During data acquisition the samples were maintained at 37°C, using a thermocouple-heating element. 2D T2W spin-echo images were acquired with: TR = 3 s, TE = 30 ms, averages = 128, slice thickness = 1.0 mm, matrix = 128 × 128, FOV = 6.5 mm (resolution = 50x50μm), scan time = 7 min. 2D T1W spoiled-gradient-echo images with and without MT were acquired with: TR = 330 ms, TE = 4 ms, flip angle = 30°, averages = 192, slice thickness = 0.5 mm, matrix = 256 × 256, FOV = 6.5 mm (resolution = 25 × 25μm), scan time = 1 h. A Gaussian saturation MT pre-pulse was applied 10000 Hz off-resonance with duration of 12 ms and power of 20μT once every TR. These conditions were previously optimized in our lab using phantoms and carotid endarterectomy specimens [37]. MT employs an off-resonance pulse that selectively saturates the water molecules bound to macromolecules compared to those in the free water pool, resulting in images with reduced signal intensity that is proportional to the amount of macromolecules [39‐41]. Finally, 2D DW images were acquired with: TR = 1 s, TE = 25 ms, Δ = 12.6 ms, δ = 5 ms, averages = 32, slice thickness = 1 mm, matrix = 128 × 128, and FOV = 6.5 mm (spatial resolution = 50 × 50 μm), scan time = 6.5 h. The apparent diffusion coefficient (ADC) was calculated from seven b-values = 0, 196, 442, 637, 867, 1770 and 2409 s/mm². For DW imaging a b-value = 442 s/mm² created the best contrast between the thrombus and the vessel wall and was used to calculate the thrombus area. T2 relaxation was determined using a Carr-Purcell-Meiboom-Gill sequence with TR = 3 s and ten different echo times with echo spacing of 6.3 ms, scan time = 3.3 h. T1 relaxation was determined using a spin–echo sequence with 10 different TR values ranging from 100–8000 ms, scan time = 7 h. The rest of the acquisition parameters were similar to those listed for the T2W and T1W images above.

Aortic segments were embedded in tissue freezing medium and stored at -80°C. Serial 10 μm thick cross-sections were used for Masson’s trichrome staining to detect collagen and immunohistochemistry to detect fibrin. Immunohistochemistry was performed by the avidin-biotin-peroxidase method (Vector Laboratories, Burlingame, CA). An anti-rabbit mouse monoclonal antibody for fibrin (American Diagnostica Inc, Stamford, CT, No. 350) was used. Disrupted rabbit aortic plaques contained an overlying thrombus similarly to what has been observed for human coronary plaques [42].

MR images acquired at 11.7 T were processed with the Paravision 3.02 software (Bruker, Ettlingen, Germany). Images acquired at 3 T were processed with Osirix (OsiriX Foundation, Geneva, Switzerland). Regions of interest (ROI) encompassing the thrombi and plaques were manually segmented on all sequences, and the mean signal intensity (SI) of each ROI was recorded. The signal-to-noise ratio (SNR) of thrombus and plaque was calculated as SNR = (SI_component-SI_noise)/SD_noise Noise was determined within an ROI drawn outside of the specimen in the buffer. CNR between thrombus and plaque was calculated as (SI_thrombus-SI_plaque)/SD_noise. All values were expressed as absolute numbers. Furthermore, images acquired with and without magnetization transfer were used to calculate the percent magnetization transfer rate (% MTR) based on the formula: % MTR = [(SI _{without MT} - SI _{with MT})/SI _{without MT}]* 100. The same ROIs were used to calculate the apparent diffusion coefficient (ADC) based on the equation ADC = ln(S₀-S₁)/b₁-b₀ (mm²/s), and the T1 and T2 relaxation times using the DW, T1 and T2 images, respectively.

Thrombus cross-sectional area (mm²) in rabbit specimens was measured on trichrome stained sections and on ex vivo MR images by computerized planimetry (ImageJ, NIH). In the case of MT, the thrombus area was calculated using the magnetization contrast images (MTC) generated by subtracting the image with MT from that without MT. Computer-assisted color image analysis was used to quantify the collagen area on trichrome stained sections and the positive areas for fibrin on immunohistological sections, which were expressed as percentages of the total thrombus area.

The Statistical Package for the Social Sciences 18.0 (SPSS Inc.) was used and data are presented as mean ± SD. Probability values of P < 0.05 were considered significant. For 2-group comparisons, continuous variables were compared with a Student’s t-test. The correlations between continuous variables were assessed using Pearson’s correlation. Multiple group comparisons were performed by one-way analysis of variance (Anova). The receiver operating characteristic curve was used to compare the accuracy of different MRI sequences for detecting the presence or absence of thrombi. Bland-Altman plots were used to assess the agreement between different MRI sequences and histology in measuring thrombus area. Two independent observers (A.P and N.G) analyzed the MR images and were blinded to all histological data. The inter-observer variability was assessed using the intra-class correlation coefficient (ICC).

Histology revealed that all animals had aortic atherosclerotic plaques and that 9 out of 17 rabbits (53%) had luminal thrombi. Of a total of 59 plaques studied, 34 had superimposed thrombosis whereas 25 did not have any evidence of thrombus.

Multi-contrast images of a disrupted plaque with a superimposed thrombus are illustrated in Figure 1. The T1W (Figure 1A) and T2W (Figure 1B) images revealed thickening of the vessel wall but did not discriminate the thrombus from the underlying plaque. By comparison, DW images (Figure 1C; ADC_thrombus = 0.58 × 10^-3 mm²/s versus ADC_plaque = 0.25 × 10^-3 mm²/s) and MTC images (Figure 1D; MTR_thrombus = 42.6% versus MTR_plaque = 27.4%) increased the conspicuity between the thrombus and the underlying plaque/vessel wall. MRI assignments were corroborated by histopathology (Figure 1E and 1F), which illustrated a platelet- and fibrin-rich thrombus overlying the plaque.

MTC and DWI at 11.7 T provide superior contrast between the thrombus and the underlying plaque and a more accurate assessment of thrombus size compared to T1W and T2W images.

Quantitative assessment of the SNR for the thrombus and the plaque showed that all image sequences provided sufficient SNR for both components (Figure 2A). However, the CNR between the thrombus and the underlying plaque was significantly higher on MTC (CNR = 8.5 ± 1.1) and DW (CNR = 6.0 ± 0.8) images compared to T1W (CNR = 1.8 ± 0.5) and T2W (CNR = 3.0 ± 1.0) images (Figure 2B).

The Bland-Altman analysis was applied to test the agreement in thrombus area calculated using different MRI sequences and histology (Figure 3). Although the area of the thrombus measured on different MRI sequences (Figure 3A-3D) was within the 95% limit of agreement, the level of agreement was higher for MTC (Figure 3A) and DW (Figure 3B) images, as indicated by the tighter clustering of the data points around the median value. Conversely, larger differences were observed in the area of the thrombus calculated using the T1W (Figure 3C) and T2W images (Figure 3D) as compared to histology (scattered data points). Furthermore, there was a trend for MTC and DWI to slightly underestimate the area of the thrombus compared to histology (6% and 17.6%, respectively). Conversely, the thrombus area was overestimated on T1W images (53.7%) and T2W images (46.4%).

A summary of the % MTR, ADC, T2 and T1 relaxation times calculated for all the thrombi and plaques included in this study is presented in Figure 4. The %MTR and ADC are significantly different between the thrombus and the plaque and thus can be used to generate endogenous contrast between these two components. Conversely, the T2 and T1 relaxation times are similar between the two components, which results in a lower CNR.

Image characteristics of coagulated erythrocytes and thrombus associated with plaque disruption within the same rabbit vessel is illustrated in Figure 5. Coagulated erythrocytes, located in the center of the lumen, had a bright periphery and a darker center on the T1W image (Figure 5A) and a uniform hypointensity on the T2W image (Figure 5B). On the MTC image (Figure 5C), coagulated blood appeared light grey, indicating the presence of low protein content, and showed very low signal on the DW image (Figure 5D). Thrombus associated with plaque disruption was well differentiated from the coagulated blood, appearing hyperintense on T1W (Figure 5A) and isointense on T2W images (Figure 5B), but was not well delineated from the plaque. Conversely, thrombus was hypointense on the MTC image (Figure 5C) and hyperintense on the DW image (Figure 5D) and well discriminated from the underlying plaque. Histology verified the presence of coagulated erythrocytes interspersed in a loose fibrin network (Figure 5E and 5F) overlying a thrombus associated with plaque disruption enriched in collagen (Figure 5G) and fibrin (Figure 5H).

To investigate the relationship between the % MTR and the protein content of the thrombus, the amount of fibrin and collagen were quantified using immunohistochemistry and trichrome staining, respectively (Figure 5I). The % MTR correlated strongly with the % fibrin (r = 0.73, P = 0.003) and the % collagen (r = 0.9, P = 0.004) content of thrombi. The correlation was more significant when the % MTR was plotted versus the total protein content (fibrin and collagen) of the thrombus (r = 0.89, P < 0.001).

Another example illustrating the correlation between the thrombus area as measured on the subtracted MT images and the corresponding histological detection of collagen and fibrin is shown in Figure 6. The subtracted MT image (Figure 6A), the Masson’s trichrome staining (Figure 6B), and the immunohistochemical staining for fibrin (Figure 6C) show the vessel wall and the luminal thrombus. The corresponding segmented images show that the hypointense region seen on the MTC image (Figure 6D) corresponds to histological regions rich in collagen (Figure 6E) and intermixed with fibrin (Figure 6F).

Because the rabbit aortas were fixed in formalin, collagen residues were modified with new cross-links. Although little is known about the effects of formalin-fixation on the MR properties of tissues, in cartilage it results in a significant decrease of T1, T2 relaxation times, and ADC, and an increase of the % MTR [56]. However, other studies of multiple sclerosis specimens have shown a reduction of MTR [57] while the ADC of gray matter remained more or less constant [58] after formalin fixation. Although the experiments with the rabbit plaques were performed at high field, where the MT effect is greater because of the longer relaxation time of water, and in the absence of physiological motion, previous studies have shown the feasibility to apply these pulse sequences in vivo at clinically relevant fields. MT has been used for imaging coronary veins [45, 46], arteries [47] and myocardial perfusion [48] whereas DWI has been applied for imaging human carotid atherosclerosis [34, 35].