Multichannel Resonant Acoustic Rheometry System for Rapid and Efficient Quantification of Human Plasma Coagulation

Resonant Acoustic Rheometry (RAR), a newly developed ultrasound-based technique for non-contact characterization of soft viscoelastic materials, has shown promise for quantitative assessment of plasma coagulation by monitoring the entire dynamic process in real time. Here, we report the development of a multichannel RAR (mRAR) system for simultaneous monitoring of the coagulation of multiple small-volume plasma samples, a capability that is critical to efficiently provide improved assessment of coagulation. The mRAR system was constructed using an array of 4 custom-designed ultrasound transducers at 5.0 MHz and an electronic driving system that controlled the generation of synchronized ultrasound pulses for real time monitoring of multiple samples simultaneously. The mRAR system was tested using Coumadin-treated plasma samples with a range of International Normalized Ratio (INR) values, as well as normal pooled plasma samples. Tracking of dynamic changes in clotting of plasma samples triggered by either kaolin or tissue factor was performed for the entire duration of coagulation. The mRAR system captured distinct changes in the samples and identified parameters including clotting time, clotting speed, and the mechanical properties of the clots that were consistent with Coumadin dose and INR levels Data from this study demonstrate the feasibility of the mRAR system for the rapid, efficient, and accurate characterization of plasma coagulation.


Introduction
Warfarin (brand names Coumadin or Jantoven), a Vitamin K antagonist (VKA), is the most frequently prescribe oral anticoagulant (OAC) for patients at risk for abnormal blood clots 1 . For example, it has been prescribed as a long-term anticoagulant therapy for prevention of stroke in patients with atrial fibrillation (AF) 2,3 , treatment of venous thromboembolism (VTE) 4 , pulmonary embolism 5 , and heart attacks 6 . Coumadin monitoring via regular blood testing of patients is critical to provide guidance on the proper dosage to ensure desired delayed clotting time without the risk of uncontrolled bleeding in the anticoagulation therapy 7 . Conventionally, Coumadin monitoring has relied on the prothrombin time test (PT), which reports the result as the International Normalized Ratio (INR) 8 . However, the pharmacodynamic effect of PT-monitored Coumadin (referred as PT-Coumadin) is highly variable in many patients 9,10 , making the approach less appropriate for accurate Coumadin monitoring and optimal patient care outcome.
PT measures the activity of a single clotting factor (prothrombin), which may not fully reflect the overall clotting characteristics of the blood, In contrast, viscoelasticity-based hemostatic assays (VHA), such as TEG or ROTEM 11,12 , measure the viscoelastic properties of the samples during the entire clotting process. In addition, TEG has the capability of performing multiple assays to examine the effects and interactions of multiple clotting factors with blood components.
Generally, VHAs can provide comprehensive assessment of the true clotting state of the blood 13 .
However, TEG and ROTEM are not widely adopted in the clinic for Coumadin monitoring due to a number of limiting factors. The equipment can be expensive and the procedure and interpretation of results require specialized training. The high cost and complexity of tests with current VHA techniques make them not feasible for point-of-care (POC) or at home tests for frequent and convenient Coumadin monitoring. Multiple assays for investigation of multiple factors require even longer time to results and further increased cost.
Resonant acoustic rheometry (RAR) is a newly developed ultrasound-based technique for non-contact, rapid characterization of viscoelastic soft materials 14 . In a recent study, we demonstrated the feasibility of RAR for quantification of coagulation characteristics of human hemophilia A plasma samples 15 . RAR utilizes a dual-mode ultrasound technique that has unique advantages for longitudinal characterization of temporally evolving soft materials in a non-contact fashion. RAR generates and detects resonant surface waves in a sample housed in a small sample holder, e.g. in a well of a standard 96-well microplate 14  is the surface tension, the mass density, and k the wavenumber of surface wave 16 . In RAR, k is determined by the resonant mode of the surface waves and the radius of the sample surface 14  , which depends on the shear modulus of the bulk material G. Thus by measuring the frequency of the surface waves using RAR, the shear modulus of the material can be obtained in a non-contact fashion and as a function of time.
A single RAR measurement of the resonant surface wave in a sample takes less than 0.5 s, allowing characterization of the dynamic blood coagulation process in near real time by repeated RAR measurements throughout the entire clotting duration, e.g. a few min up to 90 min or longer.
While a single channel RAR system provides rapid measurement of a single sample, it is inefficient to perform multiple samples sequentially to assess the effects of multiple factors on coagulation.
The goal of this study is to develop and validate a multichannel RAR (mRAR) system for viscoelastic characterization of multiple samples simultaneously. To this end, we designed and fabricated an array of single ultrasound transducers that operate in parallel. The array is controlled by a custom-designed electronic driving system that controls the generation of synchronized ultrasound pulses with desired parameters, allowing the measurement of resonant surface waves in multiple temporally changing samples simultaneously. We tested the prototype mRAR system and demonstrated the capability of the mRAR system to quantify the coagulation of normal pooled plasma as well as Coumadin anti-coagulated plasma samples with a range of INR values.

Design, fabrication, and calibration of ultrasound transducers for mRAR
To develop an mRAR system for measuring multiple samples housed in a standard 96-well microplate, computational simulation was performed before fabrication and testing. COMSOL simulation determined that the 3-dB width of a focused ultrasound beam should be ~ 0.5 mm at the focus in order to generate resonant surface waves in a sample within a well of radius of 3.25 mm in a standard 96-well microplate, as previously described 14 (Fig. 1A). Simulation was also performed using the MATLAB K-WAVE package that determined the focal distance of a 5.0 MHz transducer (with an active surface diameter 5.0 mm) to be 12.3 mm for the desired mRAR system.
Rapid prototyping was employed to fabricate single element transducers 17  To demonstrate the general principle of mRAR, a prototype mRAR transducer assembly was constructed including 4 single transducers with the same design parameters. These single transducers were mounted on a 3D-printed supporting scaffold that holds the transducers (Fig. 1B) with a well-to-well distance from each other that physically aligned them to 4 adjacent wells in a 96-well microplate (9 mm center-to-center spacing) (Greiner Bio-One), allowing for testing of 4 samples simultaneously (Fig. 1C).
Calibration of the transducers was performed using a fiber optic hydrophone (Onda HFO 690) by measuring the acoustic pressure field of the transducers in de-gassed, room temperature water in a free field condition. The hydrophone was operated at a sensitivity of 7.00 mV/MPa or higher and was spatially scanned at a step size of 0.1 mm to measure 3D field maps of acoustic pressure distribution.

Electrical circuit for mRAR operation
An electric driving circuit was designed and constructed to operate the array of transducers to function in both transmit-only and transmit-receive (or pulse-echo) mode in a synchronized fashion for mRAR measurement. A class-D driver circuit architecture used for early histotripsy research 18 was modified to use Silicon Carbide transistors for operation at 5.0 MHz. This design uses a pair of transistors operating in a "push-pull" mode to generate a high voltage square wave burst which is then filtered by a series inductor to produce a higher voltage, nearly sinusoidal output for driving the transducers (Fig. S1). Varying the main power supply voltage to the transistors varies the amplitude of the square wave and ultimately the transmit pulse amplitude.
For these experiments, the maximum square wave was 75 V, producing a driving sinusoid of about 190 V peak-peak. To receive acoustic echoes from the surface of the sample, a 50 Ohm current sensing resistor was placed in series in the ground return path from each transducer. The voltage on this resistor was digitized by a model 5443 Picoscope (Pico Technology, St Neots, UK). On transmit, this resistor is bypassed by a pair of transistors to maximize power output and prevent saturating the digitizer (Fig. S1).
A custom microcontroller implemented on a field programmable gate array (FPGA) interfaced the circuit board to a host computer with software control through MATLAB. The electrical circuitry was designed to generate ultrasound pulses with controllable acoustic pressure amplitude, number of cycles (pulse duration), and pulse repetition frequency of the ultrasound pulses such that the same transducer served for both pushing and detection in RAR, as described previously where two separate transducers and driving systems were used for these different tasks 14 .
For RAR measurement, the transducers were driven to generate an ultrasound tone burst in transmit mode (excitation pulse) to induce surface perturbation leading to resonant surface waves in the sample. After generation of the tone burst, the transducers were immediately switched electronically to pulse-echo mode, first generating a short pulse (detection pulse) and then receiving its backscattered signal (via the Picoscope) from the sample surface at a high pulse repetition frequency (PRF) (e.g. 10 kHz). Thus the resonant surface wave in the samples was Signal processing to determine the surface displacement from RAR measurements As described previously 14 , a custom-developed MATLAB script was used to analyze the measured backscattered signals during pulse-echo detection during RAR to determine the surface displacement in a sample after application of each "push" ultrasound pulse. The temporal change was generated (vertical axis: fast time ; horizontal axis: observation time). Correspondingly, a spectrogram was generated by displaying the power spectra of ( , ) during coagulation. RAR measurement data was collected on each sample for a total of 60 minutes.

Determination of coagulation parameters and statistical analysis
The frequency of the resonant surface wave was identified as the peak in the power spectrum of the surface displacement measured in RAR. A spectrogram of the frequency is generated to show the changing power spectrum as a function of coagulation time, and was used to extract a set of parameters for the quantification of plasma coagulation 15 . The initial frequency, , was calculated as the averaged frequency measured during the first minute of RAR measurements of a sample while the final frequency, , as the averaged frequency measured during the last 5 minutes of RAR measurements. The clotting start time, , was defined as the time at which the resonant frequency began to increase by 5% from determined from the spectrogram. The clotting end time, , was defined as the time at which the frequency reached 95% of , indicating the end of the active clotting process. The clotting duration was the difference between these times, − , A visual representation of these parameters is shown in Fig. S2.
The results obtained from coagulation of pooled normal plasma as well as plasma samples with low, medium, and high INR triggered by either TF or Kaolin were compared. Linear regression with interaction effects was performed. Differences among the groups were determined to be statistically significant if p < 0.05.

Streamlined mRAR system operation based on dual-mode operation of array of single transducers
As shown in Fig. 1, in the array driven by the control circuit, each of the transducers produced short ultrasound pulses (Fig. 2a) for pulse-echo mode operation with a pulse duration of 0.6 µs, attaining an acoustic pressure amplitude of 3.4 ± 0.8 MPa (n = 4) at a 75.0 V input setting.
The transducers generated symmetric focused ultrasound fields, as shown by the spatial distribution of the acoustic pressure in the focal plane (Fig. 2b). The full width half maximum (FWHM) of the focal zone was measured to be 1.02 ± 0.07 mm and 1.02 ± 0.06 mm in the in the X and Y direction respectively (Fig. 2b), satisfactorily meeting the design requirement for the mRAR system. The FWHM in the Z (axial) direction was 12.41 ± 4.11 mm, showing an elongated focal depth (Fig. 2c), providing flexibility for axial placement of the focus relative to the sample surface during RAR measurements. The transducers were also able to generate tone bursts with an adjustable number of cycles to serve as the excitation or pushing pulse to induce surface waves in the samples, as shown by the examples of tone burst with 20 cycles and 100 cycles respectively (Fig. 2d&e). As expected, tone bursts with longer duration achieved slightly higher acoustic pressure amplitudes than the 2-cycle pulses for pulse-echo operation (imaging mode). The spatial-peak temporal-peak intensity (Isptp) for a 2-cycle pulse (measured across the four transducers) was 484.3 ± 187.5 W/cm 2 . The Isptp was 548.6 ± 192.9 W/cm 2 and 640.0 ± 311.2 W/cm 2 for a 20-cycle and 50-cycle pulse respectively.
The mRAR system successfully produced a set of synchronized pulses for both excitation and detection of surface waves in 4 samples simultaneously for RAR measurement. As illustrated in Fig. 3a, for each RAR measurement at a given time, 20 imaging/detection pulses (PRF from 5 to 10 KHz) were applied to obtain the base-line signals from the surface of the samples at equilibrium before application of a single tone burst or a 'push' pulse (duration 20 µs) to induce surface resonant wave in the samples. The surface movement was measured using 500 detection pulses immediately after the "push" pulse to detect the temporal shift of the echo signals from the sample surface relative to the equilibrium. The example in Fig. 3b shows the A-line signal measured by one of the transducers operated in pulse-echo mode, showing the reflective signals from different interfaces including the sample surface.

RAR detected dynamic changes during plasma coagulation in real time
To test the performance of the mRAR system, we conducted experiments to measure the changes of plasma samples during coagulation. The results in Fig. 4 show a representative example of a series of RAR measurements (measurement interval of 6.0 s for a duration of 60.0) of coagulation of normal plasma sample triggered by TF. The heat map of the surface displacement ( Fig. 4a) displays the surface displacement at the center of the sample surface as a function of coagulation (elapsed) time, revealing significant changes in the duration and amplitude of the surface waves in the sample during the coagulation process (Fig. 4a&b). Clearly, the resonant surface waves or surface movement generated in the samples exhibited the characteristics of a damped harmonic oscillator (Fig. 4b).
Overall, the RAR measurements reveal several distinct stages during plasma coagulation. During the initial phase of the coagulation, i.e. T < 90.0 s, the amplitude and period of the surface displacement remained relatively constant (Fig. 4a),  . 4a&b).
The corresponding spectrogram (Fig. 4c), which is the normalized power spectrum of the surface displacement/wave vs. coagulation time, also provided a global view of the changes occurring in the samples during coagulation. It revealed several stages throughout the coagulation process (Fig. 4c&d). During the lag phase, the frequency of the resonant surface waves remained constant (Fig. 4c), suggesting a liquid state of the samples. After the lag phase, the frequency of the resonant surface wave started to undergo a period of rapid increase (Fig. 4), suggesting the emergence of elasticity in the sample as it transitioned from liquid to solid, with the surface wave transitioning from CWs to Rayleigh behavior 14,16 . The frequency peak widened concurrently with the rapid increase of frequency (Fig. 4c&d), suggesting increased viscosity of the sample due to clotting 16 . Lastly, when T > 150.0 s, the surface wave frequency stabilized (Fig. 4c), reflecting completion of the clotting process and formation of a stable clot with stabilized shear modulus G.
Taken together, these results demonstrate the capability of our mRAR system for capturing the dynamic progression of plasma coagulation. The drastic decreases in resonant wave amplitude and duration, as well as the increase in the frequency of the resonant surface waves, provide a clear and sensitive indicator of clot formation dynamics as the sample transitioned from fluid to solid.

Characterization of Coumadin effects on plasma coagulation using mRAR
Experiments were performed using the mRAR system to characterize the effects of Coumadin on the coagulation of human plasma samples induced by Tissue factor (TF) or Kaolin, respectively, as described in Ma terials and Methods. As shown by the example in Fig. 5, heatmaps of surface displacement and corresponding spectrograms revealed clear differences between groups in terms of the dynamic progression of the samples during TF-triggered (Fig. 5) or Kaolin-triggered (Fig. 6)  We performed quantitative assessments of the coagulation dynamics using parameters derived from the measured frequency of the resonant surface waves (Fig. S2), and the results are shown in Fig. 7. Overall, RAR detected statistically significant differences in coagulation between groups triggered by TF or Kaolin (Tables S1-S3  (Fig. 7a), potentially capturing the effects of different coagulation pathways. In terms of clotting duration, the medium and high INR samples took longer to coagulate than the normal and low INR samples triggered by TF (Fig. 7b). However, the high INR samples exhibited significantly increased clotting duration triggered by Kaolin compared to samples in other groups, suggesting that clotting duration is especially sensitive to high INR values when triggered with Kaolin (Fig. 7b). Interestingly, the final frequency of the resonant surface waves in normal plasma samples measured at the end of coagulation was significantly lower than those measured in plasmas samples from patients treated with Coumadin (Fig. 7c) (Fig. 7c). The frequency results of Kaolintriggered coagulation exhibited a trend opposite to the clotting duration vs. INR values (Fig. 7b), indicating that higher INR values delayed coagulation and formed softer clots, although still much stiffer than normal clots.

Discussion
A previous RAR system 14 used two co-linearly aligned transducers with different center frequencies and characteristics for generating either a tone burst excitation pulse or short imaging pulses for pulse-echo detection. While such a dual transducer system may optimize excitation (at lower frequency of 1.5 MHz) and detection (at higher frequency of 7.0 MHz), it requires two separate electronic systems with different functionalities that need to synchronize and spatially align the ultrasound pulses, and is therefore not well-suited for the construction of multichannel systems due to increased cost and complexity. While the throughput of such a system can hypothetically be increased by multiplexing and automatically translating the transducer assembly to different samples, these functionalities are prohibitive to constructing a multi-channel RAR system for simultaneous RAR measurements of multiple samples in systems with fast-changing properties such as hypercoagulation. In contrast, the mRAR system described here employed the same ultrasound transducer for both excitation and detection with an innovative design for a streamlined electronic driving system that achieved both transmit and transmit-receive modes with controllable pulse duration and pulse repetition frequency.
The multi-channel RAR system has the general advantages of rapid testing of plasma coagulation objectively, rapidly, and efficiently. The system is cost-efficient and can be easily constructed. It allows for testing of sample replicates under identical conditions, or for testing sample from the same patients using multiple assays simultaneously. Importantly, mRAR is a VHA-based technique that provides quantitative information on sample properties over the entire clotting process in a cost-efficient fashion. Therefore, mRAR offers a promising approach as a point-of-care (POC) device for more frequent and convenient testing for patients at home or in the clinic for personalized health care 19,20 . The technique overcomes the limitations of both conventional PT and VHA techniques such as TEG or ROTEM 11,12 . A criticism of clinicallyavailable VHA methodologies is that they considerably deform the clot sample, as they use either a rotating cup and pin system (e.g., TEG 5000, ROTEM delta, and ROTEM sigma), vertical cup and pin system (e.g., Sonoclot), or vibration detection by light emitting diode (e.g., TEG 6s) [21][22][23] .
New and emerging technologies of point-of-care (POC) hemostatic assays have used iterations of ultrasonic deformation methodology 12 . For example, sonic estimation of elasticity via resonance (SEER) is a rheological method that addresses this drawback of contact-based techniques by utilizing ultrasound technology. Quantra® is a fully-automated, ultrasound-based test that uses dry reagents, thus simplifying quality assurance and improving reproducibility 12,24,25 . However, these VHAs require expensive systems and reagents that are not readily available at every institution that could benefit from these devices.
RAR technology is based on an entirely different operating principle from other ultrasound-based approaches 26 . It specifically measures the surface waves and extracts the viscoelasticity based on the dispersion relation of the surface waves in different mechanical regimes. It therefore provides a unique way to measure viscoelastic properties including surface tension, viscosity, and shear modulus all at once 14 . While a 4-channel mRAR system was developed in this study, the mRAR platform is easily expandable to more channels. Since RAR measurement requires less than 0.5 s and clotting is much slower, an alternative strategy would be to use a motion controller to scan the array of 4-transducers to measure additional sets of samples in a multiplexed fashion to improve throughput.
The clotting times measured in this study using mRAR are in the same range as those found in a previous study using ROTEM ®13

Conclusion
The results of this study show that a prototype mRAR platform provided streamlined operation for real-time monitoring of the coagulation process in multiple plasma samples