Characterization of thrombus composition with multimodality CT based imaging: an in-vitro study

Abstract

Background:

Computed tomography (CT) is most common for acute ischemic stroke evaluation. There is growing interest to use pre-operative imaging to characterize clot composition in stroke. We performed an in-vitro study examining the ability of various CT techniques in differentiation between different clot types.

Methods:

Five clot types with varying fibrin and red blood cells (RBCs) densities (5% RBC and 95% fibrin; 25% RBC and 75% fibrin; 50% RBC and 50% fibrin; 75% RBC and 25% fibrin; 95% RBC and 5% fibrin) were prepared and scanned using various CT scanning protocols (single-energy, dual-energy, photon-counting detector CT, mixed images, and virtual monoenergetic images). Martius Scarlett Blue trichrome (MSB) staining was performed to confirm the composition of each clot. Mean CT value of each type of clot under different scanning protocols was calculated and compared.

Results:

Mean CT values of the CT numbers in the five clot specimens for 5%, 25%, and 50% RBC clot were similar across modalities, and increased significantly for 75% and 95% RBC clots (p < 0.0001). Mean CT values are highest in the Mono+ 50 keV images in each type of clot, and they were also significant larger than any other imaging modules (p < 0.001). Dual energy CT with Mono+ 50 keV images showed the greatest difference between attenuation in each type of clots.

Conclusion:

Mono+ 50 keV dual-energy CT scan may be helpful for differentiating RBC-rich and fibrin-rich thrombi seen in large vessel occlusion patients.

Introduction

There is growing interest in characterization of thrombus composition in acute ischemic stroke (AIS) patients presented with large vessel occlusion as thrombus composition has been shown to be correlated with stroke etiology and even prognosis and success with intravenous thrombolysis and mechanical thrombectomy (1–5). In general, it is thought that RBC-rich clots are associated with improved recanalization rates with intravenous thrombolysis and mechanical thrombectomy techniques. In-vitro studies also show that clot composition could impact the performances of stent retriever devices and aspiration catheters (6, 7) (8).

Since computed tomography (CT) is the most common imaging modality for screening AIS patients, developing CT-based techniques for improving clot characterization could potentially be useful. We tried to find out which CT scanning method would provide reliable imaging information for identify thrombus composition in AIS patients using in-vitro clot model.

Materials and Methods

Clot Analogues Preparation

Following Institutional Review Board (IRB) approval from our institute, whole blood was obtained from volunteer from the Blood Transfusion Service after quality control. Blood was separated into plasma, buffy-coat, and erythrocyte-rich layers after centrifugation at 1,500g for 12 minutes. The plasma and erythrocytes were harvested independently and then recombined in controlled ratios to form five different clot analogues with 2.06% Calcium Chloride and 2% thrombin added (5 analogues for each composition): Group A, fibrin-rich (95% plasma:5% RBCs; Group B, fibrin-rich (75% plasma:25% RBCs); Group C, intermediate (50% plasma:50% RBCs); Group D, RBC-rich (25% plasma:75% RBCs,) and Group E, RBC-rich (5% plasma:95% RBCs). Clotting was initiated by adding a 2.06% calcium chloride (CaCl₂) solution in a 1:9 ratio (CaCl₂ solution:blood mixture) followed by 1% thrombin (Sigma-Aldrich, St Louis, MO, USA) in a 1:1000 ratio (thrombin:blood mixture). The blood clot mixtures were quickly loaded into 3 cc plastic Luer-lock syringes, which spun at 20 rpm in a hybridization incubator for 1 hour at 37˚C to prevent the natural sedimentation of red blood cells at static conditions (9).

Phantom and CT Scans

To emulate the attenuation of the brain and skull, the clots were placed in a stadium shaped water tank measured 20 cm laterally. For each clot type, five specimens were placed in the water tank. The phantom was then scanned using a commercial dual-source CT system (Somatom Definition Flash, Siemens Healthineers, Forchheim, Germany) with both a single- (SECT) and dual-energy (DECT) clinical head protocols, and also with a research photon-counting detector CT (PCD-CT) system (Somatom CounT, Siemens Healthineers, Forchheim, Germany) (10, 11). Dual-energy scans with virtual monoenergetic images (VMI) at 50 keV (the Mono+ technique) were also used for imaging acquisition (12). Online supplementary file 1 lists the acquisition parameters for each CT scan.

All image sets were reconstructed using a medium smooth kernel (Q40) with 1 mm slice thickness at an increment of 0.8 mm. The images from the PCD-CT system were reconstructed using photons with energy thresholds of 25 and 140 keV (Threshold low, Tlow) and 25 and 75 keV only (Bin 1). Images from Tlow and Bin 1 were compared regarding to image contrast and signal-to-noise ratio (SNR).

Image Analysis

The images from the DE acquisition were processed in commercial software (Syngo VIA, Monoenergetic Plus, Siemens) to produce virtual monoenergetic images (VMI) at a photon energy of 50 keV (mono+ 50 keV). The dual-energy data were also used to produce a linear blended (LB) image set with a blending ratio of 0.4/0.6 between the low and high kV images. The mean CT number in each clot specimen was measured using a region growing tool (Syngo VIA, Siemens) with lower and upper thresholds of 0 and 100 HU in order to reject the low attenuating syringe and high attenuating rubber stopper.

Histologic analysis

Following CT scanning, each clot analogue was fixed in 10% phosphate-buffered formalin for 48 hours and then clot analogues went through standard tissue processing and were embedded in paraffin. Formalin fix paraffin embedded analogues were cut into 3μm slices and then slides were stained with Martius Scarlett Blue trichrome (MSB). Histological quantification was performed using Orbit Image Analysis software that uses machine-learning method for analysis (13). Relative density of each component including RBC, Fibrin/PLT was determined.

Statistical Analysis

Mean value and standard deviation of the CT numbers of five specimens of each clot type from different scanning modules were calculated. Clustered chart was used to compare those values. Trend of change in mean CT value with all five compositions under the mono+ 50 keV protocol was shown using marked line graph. Differences of the CT value between different composition and scanning modules were performed using the Student’s t test. P < 0.05 was considered as significant.

Results

Mean values of the CT numbers measured in the five clot specimens for each of the five clot types are listed in Table 1. There was little difference between the measurements in the low density clots, but as the RBC concentration of the clots exceeded 50%, there was an increase of approximately 10–20 HU with each subsequent clot type. Mean CT values of clots with 75% and 95% RBC content was significant than larger than that of 5%, 25%, and 50% RBC content (p < 0.0001). Mean CT values are highest in the Mono+ 50 keV images in each type of clot (Figure 1), and these values under Mono + 50keV protocol was significant larger than any other imaging modules (p < 0.001) (Figure 1, 2). Among all the four types of CT scanning, the ability to distinguish between different clot types was highest for the DE Mono + 50keV (Figure 3).

Table 1.

Mean CT Values of Clots with Different Composition

		Clot RBC Content (%)
	5%	25%	50%	75%	95%
Mono+	30.2±0.8	30.0±1.0	33.4±2.5	54.0±5.7	64.6±5.2
PCD Tlow	24.3±0.5	23.6±3.2	27.6±2.1	46.4±7.0	59.8±5.7
DE Mixed	24.3±0.4	24.4±0.5	26.8±2.2	47.8±8.1	56.8±5.4
SE 120 kV	24.0±0.4	24.2±0.8	27.0±2.2	48.8±6.6	61.2±7.3

Figure 1.

Showing mean CT value of the five clot types with four CT scanning methods. Clots from 5% to 50% RBC content indicate similar CT value, while 75% and 95% RBC clots showed big increase by comparing to the lower lever of RBC content. Highest CT value is shown in each type of clot with DE Mono+ 50 keV CT scanning.

Figure 2.

Representative Mono+ 50 keV images of different type of clot. A, 5% RBC rich; B, 25% RBC rich; C, 50% RBC rich; D, 75% RBC rich; E, 95% RBC rich. Clot density is increased from A to E (white arrows from each type). Representative histological images (F to J) of 5 types of clots (MSB staining, magnification 20X), from left to right, the percentages of RBC range from 5% to 95%. The color of clot area changes from light red to dark yellow, which indicates less fibrin and more RBC as the percentage of RBC increases.

Figure 3.

Representative images of four different image types of 95% RBC clot. A. SE 120 kV (white arrow); B. PCDCT Bin 1 (25–75 keV) (white arrow); C. DE 0.6 linear blend (white arrow); and D. Mono+ 50 keV (white arrow). Thrombus in D is shown better than other three image types.

Bin 1 images for PCD-CT were noisy by comparing with the threshold low images in all the clots. The threshold low images were better as they contained additional information from the photons at energies between 75 and 140 keV, which helped to decrease the noise magnitude and increase the SNR (online supplementary file 2). The SNR was increased from 2.3 (in Bin 1 group) to 3.8 (in Tlow group).

Increase of mean CT value was less than 4 HU from 5% to 50% RBC clots from DE Mono+ 50 keV group. That value jumped 20 HU for 75% RBC clot. There was additional 10 HU increase for the clots with 95% RBC (online supplementary file 3).

Discussion

Our in vitro study examining the ability of a variety of CT imaging techniques to characterize stroke emboli composition demonstrated a number of interesting findings. First, we found that across all modalities, there was no increase in thrombus attenuation until the RBC density reached 50%. These findings are important as they suggest that CT would likely not be able to differentiate between mixed emboli and emboli which are very rich in fibrin and platelets (i.e. RBC poor). Another interesting finding from our study was that the use of the Mono +50keV dual energy CT technique offered the best means to distinguish between higher RBC density emboli. These findings are important as they suggest that this imaging technique may be useful for clot characterization.

The dual-energy data was used to produce a linear blended (LB) image set with a blending ratio of 0.4/0.6, which was previously shown to improve clot characterization when compared with SECT (14). For PCD-CT, we compared Bin 1 images (with energy thresholds of 25 and 75 keV) and threshold low images (with energy thresholds of 25 and 140 keV and found Bin 1 images were noisy by comparing with the threshold low images. The reason for that is the threshold low images include information seen in the Bin 1 images, but also include information from the photons at energies between 75 and 140 keV, which decreases the noise magnitude (10, 11). Also, the mean CT value increased sharply as the RBC content increased from 50% to 95% compared to that of clots with 5% to 50% RBC density, which indicated that relationship between RBC% and CT number is not linear. The exact reason is unknown.

Association between clot histology and imaging characteristics on CT and magnetic resonance imaging for patients with acute ischemic stroke has been studied by several researchers (1, 5, 15–19). Hyperdense versus non-hyperdense clots have consistently been correlated with RBC content. One systematic review and meta-analysis of clot characterization indicated that hyperdense thrombi had a mean RBC % of 45.2% compared to 23.3% for non-RBC-rich clots (20). Also, it should be noticed that the definition of an RBC clot varied from study to study as do the various CT protocols that were used.

Characterization of clot characteristics on imaging is important for both diagnostic and interventional purposes. This study indicates RBC-rich clot is shown as hyperdense on NCCT by comparing with fibrin-rich and platelet-rich clot, which are shown as low or isodense (21). In addition, NCCT is helpful to overcome the limitations of MRI for differentiation of clot composition in some challenging situations (22). Clot composition analysis is helpful in predicting the prognosis after intravenous thrombolysis as studies have found that clots which are RBC-rich or hyperdense are more responsive to tissue plasminogen activator (23). Clots which are rich in RBCs are softer than those which are rich in organized fibrin. Some researchers have found that softer clots are equally well engaged using both unsheathing and pushing techniques when deploying a stent retriever, while stiffer clots are better engaged using a pushing technique (6). Studies have found that and better outcomes can be achieved with catheter aspiration in RBC-rich clots as well (8), and that hyperdense and RBC-rich thrombi are associated with improved recanalization outcomes with endovascular intervention (24, 25).

This study has limitations. It is an in-vitro study, thus the findings may not directly correlate to clinical findings. There was no bone or other highly dense structure surrounding the samples. Addition of dense bone, such as that seen in the skull, could alter the attenuation characteristics of the clots.

Conclusion

Dual energy CT using Mono+ 50 keV dual-energy techniques may improve the ability to characterize RBC rich from RBC poor thrombi. Further studies are needed to validate this technique in vivo.