Effect of rt‐PA on Shear Wave Mechanical Assessment and Quantitative Ultrasound Properties of Blood Clot Kinetics In Vitro

The consequences associated with blood clots are numerous and are responsible for many deaths worldwide. The assessment of treatment efficacy is necessary for patient follow‐up and to detect treatment‐resistant patients. The aim of this study was to characterize the effect of treatment on blood clots in vitro using quantitative ultrasound parameters.

6][7] Anticoagulants are given by injection or orally and are intended to stop the evolution of the clot, prevent it from embolization, and limit possible recurrences.Thrombolytics aim to break down the clot directly by destroying the fibrin mesh that binds it together.][10][11] The duration of administering anticoagulants is influenced by the crucial concern of DVT recurrence risk. 12,13Guidelines exist 5,14,15 but treatment may still be prescribed for several years at the discretion of the treating physician. 13Rapid diagnosis of DVT is important for its management. 13,16,17In addition, quantification of the effectiveness of the treatment is necessary for followup and to identify resistance in efficacy. 18,19There are several methods for treatment management such as the Prothrombin Time (PT) that evaluates blood coagulation in vitro, which is now generalized in the International Normalized Ratio (INR). 20,21However, this test still lacks precision and standardization.The D-dimer test can also be used to prolong treatment but it is generally used to prevent the risk of recurrence. 22A recent study proposed to use clot structure parameters with a gel point analysis as a marker of treatment effect. 16Other in vitro studies used the percentage of mass loss to assert the effect of a treatment 23,24 but this method is not amenable to in vivo situations.
For the above reasons, ultrasound-based methods were investigated as means of characterizing blood clots in vitro and in vivo.Early reports related clot aging to their ability to deform under a manually applied probe pressure. 25The age of blood clots has also been quantitatively related to their rigidity. 24,26- 28An in vivo rabbit model study 27 could successfully separate acute from chronic thrombi with shear wave (SW) elastography.Such results have been less conclusive in humans [29][30][31][32] ; readers may consult a systematic review of clinical studies for details. 33The clot's resistance to thrombolytics has also been related to its stiffness and mass density. 23With clot aging, its stiffness increases and it becomes more resistant to thrombolytics.Recent studies have focused on new quantitative ultrasound parameters to characterize blood clots, such as SW dispersion and SW attenuation, 34 and homodyne K-distribution (HKD) parameters. 35he SW dispersion measures the frequency dependency of the SW speed. 36,379][40][41] Both parameters decreased during coagulation and stabilized after 40 minutes in vitro. 34HKD parameters based on the analysis of ultrasound speckle statistics reflect acoustical and geometric properties of scatterers in the imaged tissue. 42,43Two relevant HKD parameters can be described: the diffuse-to-total signal power ratio 1/(κ + 1), and μ, the intensity of the echo envelope. 43In humans treated with anticoagulants, 35 these parameters were evaluated over a period of 30 days post-hospitalization. Time differences of 1/(κ + 1) and μ were noticed, whereas no effect of anticoagulant treatment was observed for SW elasticity.
To avoid in vivo confounding variables in the interpretation of results, this study aimed to characterize blood clots with different treatment (thrombolytic) conditions in vitro.Properties of clots were assessed using ultrasound SW elasticity, SW dispersion, SW attenuation, and HKD parameters 1/(κ + 1) and μ.

Methods
In vitro experiments were performed using blood from 10 pigs.For each pig, three clots were formed and subjected to different treatment conditions.SW elastography and HKD measurements were made during clots' formation and treatment follow-up.

Blood Clots and Phantom Preparation
Porcine's blood samples (hybrid Yorkshire/Landrace) were obtained from the animal facility of the University of Montreal Hospital Research Center (Montréal, QC, Canada).A quantity of 20 mL was collected in a tube containing 0.9% sodium citrate before any drugs were administered to the pig (to conduct experimental studies independent of the current protocol).The hematocrit was set to 40% using centrifugation, plasma separation, and erythrocyte volume adjustment for all samples.Blood was then split into three tubes and each tube was weighed.

Experimental Setup
For each pig, a phantom with three identical cylindrical holes of 12 mm diameter and 60 mm depth was prepared.The phantom was fabricated according to the method of Bhatt et al, 44 using 4% gelatin (product number G-1890, Sigma Chemical, Saint-Louis, MO) and 1.5% agar (product number A-9799, Sigma Chemical).Clot formation was induced by adding CaCl 2 at a concentration of 16.1 mM. 23Liquid blood from the three tubes was poured in each phantom hole just after the addition of CaCl 2 .Ultrasound measurements were then made over a period of 100 minutes.Three recombinant tissue plasminogen activator (rt-PA) treatment conditions, one for each clot, were used: 1) no treatment, 2) rt-PA treatment at 20 minutes, and 3) rt-PA treatment at 60 minutes after blood pouring.The ultrasound data recording corresponding to each rt-PA treatment condition was randomly done for each porcine sample to avoid confounding time effects.The rt-PA dose was 3.15 μg/ mL, as in the in vitro study of Mercado-Shekhar et al. 23 After 100 minutes, clots were carefully removed from the inclusion, dried and reweighed.A summary of the experimental protocol is given in Figure 1.

Data Acquisition
Ultrasound data were acquired using a Verasonics V1 scanner (Verasonics V1, Verasonics Inc., Redmond, WA) equipped with a 128-element linear probe (ATL L7-4, Philips, Bothell, WA) held by a clamp.SWs were generated by three pushes performed one below the other within the clot (in the x direction, as illustrated in Figure 2A) to create a wavefront traveling in the clot longitudinal direction z.Pushes were focused using 64 probe elements centered on the 32nd element.Ultrafast ultrasound images without compounding at 4000 frames per second were acquired to track propagating SWs.Acquisitions were repeated every 4 minutes during blood clotting.

Post-Processing
A pictorial description of post-processing steps is given in Figure 2.An acquisition on a given blood clot included 99 radiofrequency signal frames (4 frames corresponding to the push duration were discarded from data analysis).At each 4-minute time point, 10 successive acquisitions were made for averaging purpose.To measure elastography parameters, three-dimensional particle velocity matrix was generated using Loupas et al's algorithm, 45 incorporating x (width), z (depth), and t (time frames) coordinates (Figure 2C).Directional filtering was applied to minimize reflected waves. 46From the particle velocity matrix, the SW group velocity, SW phase velocity (for dispersion computation), and SW attenuation coefficient were computed using the time-of-flight, 47 shear wave spectroscopy, 48 and revisited frequency-shift (R-FS) 41 methods, respectively.The selected range of frequencies for SW dispersion was 50-500 Hz, 34,36,49 and 0-500 Hz for SW attenuation computation.Examples of SW elasticity, SW dispersion, and SW attenuation maps derived from those measures are given in Figure 2, D-F.The region of interest (ROI) for these maps had a dimension of 150 Â 20 pixels, that is, 0.5 Â 0.5 cm centered within the clot and close to the push, as ROI #3 in Ref [34].
The analysis of the echo envelope statistics was performed on ultrasound compression waves.To avoid SW propagation in processed data, the last 30 frames of each acquisition were used to derive HKD parameters.Echo envelopes were averaged over the 10 acquisitions at each 4-minute time point during clot formation.A new ROI (ROI #2, 150 Â 50 pixels or 0.5 Â 1.2 cm) centered within the clot and with its central axis corresponding to the pushes' location was selected (Figure 2, G and H).As in Destrempes et al, 43 before the estimation of HKD parameters, pixels in the ROI #2 were classified with an unsupervised method into three labels.The estimation of the HK distribution of each label assumes that the sample is drawn from a single distribution.However, the selected ROI may contain different HK distributions corresponding to different tissue characteristics.To obtain a global value of 1/(κ + 1) and μ over the three labels in the ROI, weighted sums of parameters were computed according to eqs. 3 and 4 of Destrempes et al. 43

Statistical Analysis
A one-way analysis of variance (ANOVA) with repeated measures and multiple post hoc pairwise comparisons (Tukey test) was performed to analyze the effect of treatment (ie, no treatment and treatment at 20 or 60 minutes) on blood clot mass loss.Linear mixed models (LMM) (independent variables: time and treatment condition) were used to confirm the impact of independent variables on SW elasticity, SW dispersion, SW attenuation, and HKD parameters.To account for repeated data acquisitions over the pigs, a random intercept was included in the LMM.Linear regression coefficients were assessed to quantify the effect (ie, if it is an increase, positive value, or decrease, negative value).The statistical significance level was fixed at P < .05.

Results
The weight loss between blood before pouring it in the gel inclusion (before t = 0 min) and the final clot formation considered in this study (after t = 100 min) is plotted in Figure 3.The repeated ANOVA showed a general effect of treatment (P < .001)with a significant increase between untreated and treated pig blood clots at 20 (P < .005)and 60 minutes (P < .003).
Mechanical properties characterized by the Young's modulus, SW dispersion, and SW attenuation are plotted over 100 minutes of coagulation and three different treatment conditions in Figure 4.The variability between pigs is displayed by the standard deviation.The LMM confirmed a significant effect of time on the Young's modulus (P < .001)with a linear regression coefficient of 0.0016 (ie, an increase over time, Figure 4A).Treatment conditions effected the Young's modulus (P < .001)when comparing no treatment and both treated groups (with linear regression coefficients of À0.056 and À0.042 for clots treated at 20 and 60 minutes, respectively; ie, decreases over time).
No statistically significant effect of time was observed for SW dispersion (Figure 4B).A significant treatment effect was found (P < .001)for rt-PA added at 60 minutes vs untreated samples (linear regression coefficient of À8.2e À5 ; ie, a decrease with treatment).Significant reductions over time were noticed for SW attenuation (P < .001, Figure 4C), and the linear regression coefficient was À0.045 (ie, a decrease over time).A significant effect was found between untreated clots and clots treated at 20 minutes (P < .031),and the linear regression coefficient was 0.57 (ie, an increase with treatment).
As seen in Figure 5A, no time or treatment effect was found for HKD parameter 1/(κ + 1).Parameter μ varied over time (P < .032, Figure 5B), with a linear regression coefficient of À1.2e 3 (ie, a decrease with time).Significant differences were found for μ between untreated samples and clots treated at 20 minutes (P < .001,with a linear regression coefficient of À1.5e 5 ), and 60 minutes (P < .02,with a linear regression coefficient of À0.99e 5 ).Negative values of regression coefficients signify reductions of μ with treatments at both time points.

Discussion
The age of a clot, reflected by its Young's modulus, is decisive in the choice of treatment. 50Internal bleeding is one of the side effects of treatment, so it is important to be able to quickly attest on its effectiveness.The aim of this study was to investigate mechanical and acoustic properties of treated and untreated blood clots to characterize an effect of treatment non-invasively.Ultrasound parameters used for the characterization were the Young's elasticity modulus, the SW dispersion, the SW attenuation, and two HKD parameters, which depend on acoustical and geometric properties of clot scatterers.However, HKD parameters are also affected by ultrasound system's settings, which were fixed in this work for all acquisitions.The rt-PA treatment was administrated at two different time points to see an effect according to the blood clot maturation duration.
Clot weights were measured before pouring them in inclusion phantoms, and after 100 minutes of coagulation for untreated samples and two rt-PA treatment conditions.The action of rt-PA is to activate the transformation of plasminogen, which is present in the plasma, into plasmin.The plasmin degrades the fibrin strands, breaking the matrix that holds the red blood cells and platelets in place. 51][54] This effect was shown in Figure 3, where treated clots had lost significantly more mass over time than untreated ones.One would expect that clots treated at 20 minutes would have lost more weight than those treated at 60 minutes.However, in our study, the time at which rt-PA was injected had no significant effect on the weight loss.In our design, no additional plasma was added to the clot, so the only plasminogen available for activation was that present in the sample to form the clot.This plasminogen limitation could explain the lack of difference in mass loss between clots treated at 20 and 60 minutes.
Mercado-Shekhar et al showed that highly retracted clots had higher Young's modulus and were losing less mass with rt-PA treatment. 23We expected that by breaking fibrin strands, the clot would become less dense and its' Young's modulus would decrease.In two previous studies, 35,55 the Young's modulus was evaluated for different treatment conditions without showing any statistical effect.The first study consisted of blood clot elastograms measured at 3 time points (day 0, day 7, and day 30) in vivo.This condition gave to the clot more fibrin to continue its evolution.In addition, the treatment (heparin and oral anticoagulant) used in our prior work was an anticoagulant that does not lyse the thrombus directly.The second study performed elastography ex vivo on excised thrombi.Extracted clots were cut and different pieces were immersed in plasma at 37 C or in plasma and rt-PA at the same temperature (the bath duration was not specified).Clots were embedded in phantoms and elastography was performed.The nonstatistically significant effect of the treatment on the Young's modulus in their study could be explained by the age of clots (more than 10 days in general), because clots do not differ significantly in composition in vitro Figure 4. Shear wave elastography parameters.The Young's elasticity modulus is displayed in panel A, the shear wave dispersion in panel B, and the shear wave attenuation in panel C, over time and for different treatment conditions.The linear mixed model showed a statistically significant effect of time on the Young's modulus (P < .001).Differences were also noticed between untreated and both treatment conditions (P < .001).A statistically significant effect of treatment was observed for shear wave dispersion between untreated clots and clots treated at 60 min (P < .001).Significant reductions over time were noticed for shear wave attenuation (P < .001),and the treatment at 20 minutes impacted this metric (P < .031).A total of 10 porcine blood samples split into three clots each were used, so that for each curve n = 10.The mean AE standard deviation over the different samples are plotted in gray.
after 10 days. 56,57In the case of the present study, significant effects of the treatment on elasticity were found between the two treatment conditions and untreated clots.The significant increase in the Young's modulus with clotting time is known in the literature, 24,[26][27][28] and was confirmed in our study.
As the clot forms, it becomes denser and its matrix is reinforced with fibrin strands.Our hypothesis was that by destroying these fibrin strands, the treatment would increase the clot viscosity assessed either using the SW dispersion or SW attenuation.Lower SW dispersion values for clots treated at 60 minutes does not follow our hypothesis.However, the SW attenuation increasing with treatment when given at 20 minutes is emphasizing the importance of an early detection and treatment of blood clots.The SW dispersion and SW attenuation have already been reported in a previous in vitro study, 34 and significant decreases were noticed over time for both parameters.Similar results were observed here for SW attenuation.Notice that the ROI selection for SW attenuation computation was made to minimize spectral resonance effects, which was recommended in our prior report. 34D parameters are usually computed on compression wave echo envelopes.Here, SWs were generated and compressional waves were used to track SW displacements.To avoid SW motion during HKD parameter assessments, we observed SW displacement maps at different time points, as in Figure 6.After 10 minutes of coagulation, SWs were no longer present in ROI #2 over the last 30 frames (Figure 6B).Consequently, results in Figure 5 ignored the first 10 minutes of coagulation to avoid SW motion.
The main ultrasound scatterers in blood clots are red blood cells and fibrin strands. 58,59The HKD diffuse-to-total signal power ratio 1/(κ + 1) of a medium containing loosely structured scatterers should have a high value.In a previous report, 35 an increase in 1/(κ + 1) between clots treated with anticoagulant for 7 and 30 days was found in vivo.Here, measurements were taken for 100 minutes, not days, as clots in vitro would degrade after a few days.The treatment had no significant effect on the structure of scatterers measured with 1/(κ + 1).
On the other hand, the average intensity of the echo envelope μ is a measure of echogenicity.Studies have shown that the intensity of echoes increased  with clotting, 60 and that dense fibrin clots were more echogenic than loose fibrin ones. 61We expected that μ would increase with time as fibrin strands are forming but we observed the opposite (likely because of the effect of treatments since the behavior of μ for untreated clots seems to indicate an increase).The decrease of μ with treatment as fibrin strands are broken by rt-PA was anticipated.Indeed, we observed lower values of μ for clots treated at 20 or 60 minutes vs untreated ones.

Limitations
Part of the variability in results is due to physiological inter-subject clotting differences between pigs.Measuring blood coagulation conditions with rotational thromboelastometry (ROTEM) tests 62 could have enabled us to reduce or put into perspective this variability.Moreover, in our phantom experiments, the amount of plasminogen available (used by rt-PA to transform it into plasmin for clot lysis) was limited by the volume of blood in each sample.Enhanced treatment efficacy could have been achieved by adding plasma or plasminogen during the experiment.
The elastography sequence for data acquisition was implemented on the Verasonics V1 research scanner.If the sequence had been modified to include radiofrequency data acquisition before radiation pressure pushes, it would have been possible to obtain the evolution of HKD parameters at all acquisition times.Indeed, we had to discard the first 10 minutes to avoid SW motion within the selected ROI.
The diameter of the in vitro blood clots in this study was 12 mm, the size of a clot typically obstructing completely the vena cava. 63However, in vivo blood clots do not always completely occlude the vein or are present in smaller veins.In this study, the position and size of the ROI were selected to limit the impact of wave reflections and mechanical resonances on computed SW parameters. 34In the latter report using gel and blood clot inclusions embedded within cylindrical phantoms, we noticed more resonances for smaller inclusions.Additional efforts would be required to better understand SW propagation in partially or totally occluding cylindrical vessels; considering guided waves 64 could be an avenue to explore.
Finally in this work, clots were considered homogeneous and isotropic.To address the issues of heterogeneity and anisotropy, a solution could have been to image the clots at different locations, as in Lui et al, 57 or to change the orientation of the probe to be sensitive to the anisotropy of the clots. 65,66

Conclusion
The monitoring of clot aging reflected by changes in mechanical or structural properties has the potential to be determined by ultrasound means.In this in vitro study, the evolution of blood clots over 100 minutes and the effect of rt-PA treatment could be demonstrated, which may be helpful to detect lytic resistance.The Young's modulus of blood clots decreased significantly with treatment, making it an interesting parameter for patient monitoring.The SW dispersion was sensitive to late treatment whereas SW attenuation was modified by the early treatment and evolved with time.The analysis of the speckle statistics with the HDK parameter μ also decreased with treatment; it has thus the potential of being used as a complement to SW viscoelastography measures.

Figure 1 .
Figure 1.Experimental setup.In A, a phantom with three inclusions was made.In B, the blood was split into three tubes and weighed.The phantom's holes were filled with blood (panel C); three different treatment conditions (panels D-F) were applied to the blood clots one after another; measurements were made over 100 minutes on each clot (G).Clots were weighed again at the end of experiments (H).

Figure 2 .
Figure 2. Post-processing steps with the second clot of pig #9 as an example.A schematic version of the acquisitions is presented in panel A. The ROI is shown for display only; the ROIs used are ROI #1 in panels C to F and ROI #2 in panels B and G. Panel B represents echo envelope maps (B-mode display) and panel C is showing particle velocity maps; images are displayed at frame number 10 (ie, 2.5 ms), where the propagation of shear waves is visible.Ultrasound radiation pressure pushes are represented by blue crosses in panels A-C, ROIs by blue rectangles, and the clot position by red rectangles.Panels D-F represent SW parameters overlaid on the B-mode background.Panel G corresponds to the labeled map, on which HKD parameters were computed (H) overlaid on the B-mode image for display.Units are in pixels and time frames, the x-axis goes from 0 to 1000 pixels (ie, 0-3.62 cm), the z-axis ranges from 0 to 128 (ie, 3.56 cm), and frames are displayed from 5 to 99 (ie, 1.3 to 24.8 ms).

Figure 3 .
Figure 3. Mass loss for different treatment conditions; there are n = 10 clots per each condition.ANOVA with repeated measures showed a significant difference of the treatment between untreated clots and treated ones (P < .001).Symbol asterisks indicates a significant difference computed with pairwise comparisons with P < .005.

Figure 5 .
Figure 5. Homodyne K-distribution parameters over time and for different treatment conditions.Panel A represents the diffuse-to-total signal power ratio 1/(κ + 1), and panel B corresponds to the intensity of the echo envelope μ.The linear mixed model confirmed a time effect for μ (P < .032).This parameter also differed between untreated clots and clots treated at 20 and 60 minutes (P < .001and P < .02).A total of 10 porcine blood samples split into three clots each were used, so that for each curve n = 10.The mean AE standard deviation over the different samples are plotted in gray color.

Figure 6 .
Figure 6.Velocity maps at different times for pig #9 as an example; color bars represent the normalized shear wave speed amplitude in m/s.Panel A corresponds to velocity maps just after pouring the blood into the gel phantom inclusion; shear waves (characterized by high and low speed amplitudes) do not propagate far from the pushes' locations (z = 32 pixels).Blue arrows indicate the shear wave's propagation direction and time increased with frames going from 10 to 99.After 10 minutes of coagulation (panel B), shear waves propagated faster and disappeared from the image at frame 99.Red rectangles correspond to the position of the clot.Directional filtering was applied.