A High Sensitivity CMUT-Based Passive Cavitation Detector for Monitoring Microbubble Dynamics During Focused Ultrasound Interventions

Abstract

Tracking and controlling microbubble (MB) dynamics in the human brain through acoustic emission monitoring during transcranial focused ultrasound therapy (tFUS) is critical for attaining safe and effective treatments. The low-amplitude MB emissions have harmonic and ultraharmonic components, necessitating a broad bandwidth and low-noise system for monitoring transcranial MB activity. Capacitive Micromachined Ultrasonic Transducers (CMUTs) offer high sensitivity and low noise over a broad bandwidth, especially when they are tightly integrated with electronics, making them a good candidate technology for monitoring the MB activity through human skull.

In this study, we designed a 16-channel Analog Frontend Electronics (AFE) with a low-noise transimpedance amplifier (TIA), a band-gap reference circuit, and an output buffer stage. To assess AFE performance and ability to detect MB acoustic emission, we combined it with a commercial CMUT array. The integrated system has 12.3 – 61.25 mV/Pa receive sensitivity with $0.085-0.23\text{mPa}/\sqrt{\text{Hz}}$ minimum detectable pressure up to 3 MHz for a single element CMUT with 3.78 mm² area. Experiments with free microbubbles in a microfluidic channel demonstrate that our system is able to capture key spectral components of MBs’ harmonics when sonicated at clinically relevant frequencies (0.5 MHz) and pressures (250 kPa). Together our results demonstrate that the proposed CMUT system can support the development of novel passive cavitation detectors to track MB activity for attaining safe and effective FUS treatments.

Graphical Abstract

I. Introduction

Microbubble-enhanced transcranial focused ultrasound (tFUS) has gained significant attention for its potential to improve local drug delivery by alleviating vascular barriers and opening up the blood-brain barrier (BBB) [1], [2], [3], [4].

The tFUS treatments are widely utilized in addressing diseases like brain cancer [5], Alzheimer’s [6], and Parkinson’s [7] diseases. Attaining safe and effective BBB opening requires robust and sensitive methods and technologies to detect and control the MB dynamics. While in some cases up to 40% of patients were under/over-treated with tFUS [8], [9], methods based on real-time monitoring and (close-loop) control of microbubbles (MBs)’ activity are improving the therapeutic index between BBB opening efficacy and treatment-related adverse side effects[10], [11], [12], [13], [14]. These methods monitor the occurrence of different frequency components in the MBs emission spectrum (sub- and ultra-harmonics, harmonics, and in-between signals) to establish a threshold for inertial cavitation and set an upper limit on the applied pressure [4], [15]. Therefore, these algorithms track the harmonic [16], sub-harmonic [17], and ultra-harmonic [18], [16], [19] levels of the MBs acoustic emission (AE) to differentiate between stable and inertial cavitation using passive cavitation detectors (PCDs) [20], [18], [21].

Depending on the subject and requirements, tFUS systems can transmit frequencies ranging from 0.2 to 1 MHz [22]. In addition, pressure waves emitted by MBs have multiple harmonics [23]. Therefore, it is essential to use transducers with high sensitivity and bandwidth to capture the spectral components of the received pressure waves to make more accurate inference on the threshold for inertial cavitation and optimize the therapeutic index. Transducers based on Lead Zirconate Titanate (PZT) do not offer sufficient frequency agility due to their 70–80% fractional bandwidth, which limits their capability for tuning the transmit and receive frequency and capturing the multiple harmonics generated by MBs [24].

As compared to their piezoelectric counterparts, CMUTs offer frequency agility with larger bandwidth and high receive sensitivity, and low noise, especially when used with properly designed electronics [25], [26], [27]. Furthermore, depending on the DC bias level, CMUTs operate either at conventional (low frequency) or collapse (high frequency) regimes with distinct acoustic properties such as higher receive and transmit sensitivity at higher frequencies in collapse mode [28]. While the high flexibility and performance attained by this technology supports its use for tFUS applications [29], it also comes with significant optimization challenges that need to be addressed for each design separately.

Most of the previous work on CMUTs focused on ultrasound imaging, which exploits the broad bandwidth of CMUTs [30], [26], [31]. While IVUS, harmonic imaging, Doppler flow, and 3D imaging are some of the applications that CMUT-based systems were successfully implemented [32], [33], [34], [35], [26], [31]; CMUTs are only recently being tested for monitoring MB activity during tFUS [36]. A CMUT-based PCD was recently introduced with a center frequency of 3 MHz [36]. However, due to the CMUT operating frequency and lack of tightly integrated electronics, the performance of the PCD was limited. As receivers, CMUTs generate current signals proportional to the input pressure. Therefore, a transimpedance amplifier (TIA) is frequently used as an analog front-end (AFE), especially when the impedance of the transducer is large [37], [38], [39], [30], [40].

This study presents a CMUT-based PCD for monitoring MB-enhanced tFUS using a 16-channel custom-designed application-specific integrated circuit (ASIC) front-end receiver electronics operating in the DC - 3 MHz frequency range (Figure 1). The details of the electronic design have been briefly introduced at a conference proceeding [41]. Here, the capability of this custom-designed ASIC combined with a commercial CMUT array to detect the key spectral components of MB generated pressure signals when sonicated at clinically relevant frequencies and pressures is investigated.

Fig. 1:

Overlayed micrograph (left) and layout schematic (right) of the fabricated ASIC with16 receive (Rx) channels. Each Rx channel consists of a TIA and a buffer. Required internal voltage levels are generated using on chip BGR and Biasing network (zoomed in section on the right).

II. Materials and Methods

A. CMUT Electrical Characterization

For this study, a commercially available CMUT array (CM5, Philips Engineering Solutions, Eindhoven, The Netherlands) was utilized as a transducer. The array features 64 elements, each with 0.315 × 12 mm in size (Figure 2a). The impedance and admittance of both a single water-loaded CMUT element and 15 combined CMUT elements were characterized using a network analyzer (8753ES S-parameter Network Analyzer, Keysight, Santa Rosa, CA. USA) at 120 V DC bias, so that the CMUTs are operated in the collapsed mode (collapse-mode).

Fig. 2:

(a) The 64 element CMUT array has 12 mm by 21 mm aperture with pitch size between element of 315 μm. The measured CMUT AC response showing at conventional (blue) and collapse-mode (red) operating regime (b). Upon Collapse, the CMUTs bandwidth and sensitivity increases. (c)The acoustic setup for detecting MB dynamics.

Based on the measurements, an equivalent current noise density of the CMUT elements [27] was calculated using (1)

$$I_N=\sqrt{4kT\times real(Y_{CMUT})}$$ (1)

where $k$ is Boltzmann constant, and $T$ is the temperature and real$({\text{Y}}_{\text{CMUT}})$ is the real part of the CMUT admittance. In the calculations, the temperature was assumed to be 300° kelvin.

B. TIA characterization

The fabricated ASIC with 16 receive (Rx) channels and its schematic are shown in Figure 1. Each Rx channel consists of a resistive feedback TIA with a feedback resistor of 800 kΩ. A Band Gap Reference (BGR) circuitry and a biasing network were implemented on the chip. The TIA output signal is then buffered using a compact class AB amplifier to match the output impedance.

To improve the system’s signal-to-noise ratio to a level that it could detect microbubble acoustic emissions through human skull (≪ 1 Pa), a TIA with a lower input-referred current noise level than transducer noise is desired. Equation 1 calculates the transducer noise level based on the measurements. According to these measurements, a TIA with 118 dBΩ gain was designed to achieve this criterion. In general, and assuming ideal electronics, a larger array surface area provides higher receive sensitivity and enhances the directivity as compared to a single element of the CMUT array. The TIA design was optimized for a single CMUT element, but its performance was also examined by connecting 15 CMUT elements in parallel (a larger array area) and feeding the output to a single TIA.

For gain characterization, a 200 mVpp 10-cycle sine wave in 100 kHz - 3 MHz range was applied to a 1 MΩ resistor generating a 200 nA current. The current was then fed into the TIA input, and the output voltage of the receive channel was recorded with an oscilloscope.

The TIA noise levels with and without the CMUT elements were measured using an Agilent 4395A spectrum analyzer (Keysight, Santa Rosa, CA. USA). When the CMUT array was connected to the TIAs, the noise analysis was performed on the TIA with water-loaded CMUT elements.

C. Acoustic Characterization

The AC response of the CMUT was characterized at 40 V DC and 120 V DC using a hydrophone (HGL-Onda, Sunnyvale, CA, USA). Figure 2b shows the CMUT AC responses that are normalized to the maximum of the responses which happens at 40 V DC (conventional mode). In collapse mode, the CMUTs become more sensitive above 0.75 MHz and over a broader frequency range which is important for the detection of MB generated harmonic signals for 0.5 MHz excitation frequency.

The receive sensitivity of the CMUT + TIA system was analyzed in a pitch-catch experiment. First, a single element, PZT-based transducer (V303-SU, Olympus) was characterized to obtain the transmit transfer function (output pressure/input voltage amplitude in Pa/V) at seven different frequencies in the 0.4–2 MHz range (0.4, 0.5, 0.8, 1, 1.2, 1.5, and 2 MHz) using the hydrophone at a distance of 15 cm. Second, the characterized V303- SU was used in transmit mode and the CMUT array was placed at the same location as the hydrophone. A 120 V DC bias was applied to the CMUT array (collapse-mode operation), and the received signals were recorded. Consequently, the receive transfer function of the CMUT + TIA system was characterized at these relevant frequencies including 0.5 MHz. To evaluate the system sensitivity, the minimum detectable pressure or noise equivalent pressure of the system was calculated from the sensitivity and output noise measurements.

D. MB acoustic emission

A standard piezo-based PCD previously utilized in preclinical studies [13] was used to confirm the ultrasound arrival and the MBs’ activity by detecting the spectral content of MBs’ AE. The piezo-PCD has a 6 mm in diameter and 3.5 MHz center frequency (Imasonic, Voray-sur-l’Ognon, France). Its output signals were high-pass filtered with a cut-off frequency of 1 kHz, amplified by 40 dB (Krohn-hite, model 3944), and simultaneously digitized using a multi-channel oscilloscope (PicoScope 5242B, 16-bit, 62.5 MS/s).

Figure 2c shows the setup consisting of a 500 kHz, single-element transducer (A305, Olympus), a previously characterized microfluidic channel (Ibidi μ-slide, 200 μm channel height, 5 mm channel width) [42]. This microfluidic channel was selected due to its simplicity and because its in-situ ultrasound pressure and attenuation are similar with skull attenuation. The ultrasound attenuation levels of the ibidi μ-slide at 2^nd (1 MHz) and 4^th (2 MHz) harmonics are 5.5 dB and 7.6 dB respectively, which is similar with the attenuation in human skull (6.9 dB/MHz/cm) [43], [44]. The transmit transfer function and −6 dB beam-width of the A305 transducer were characterized using the hydrophone at 65.5 mm. The setup also includes the standard piezo-PCD, and the CMUT array connected to the AFE, forming the CMUT-PCD.

The ibidi μ-slide was positioned in 65.5 mm of the A305 transducer with its thick-walled facing the ultrasound field. The piezo-PCD and CMUT-PCD were geometrically aligned with respect to the incident pressure field and positioned at their natural focal distance from the ibidi. To prevent signal saturation, they were both positioned on the side of the water tank such that the main ultrasound path was orthogonal to the piezo-PCD and CMUT-PCD.

The ibidi slide was incubated with bovine serum albumin (BSA) at room temperature for 1 hour in order to prevent nonspecific binding of MBs and the slide [42]. Following the incubation, the ibidi slide was rinsed three times.

Definity-type MBs with an initial concentration of 1 − 12 × 10⁹ per mL after activation were used to monitor the MB activity. The initial concentration was diluted 2400 times to approximate the MB concentration found in the brain during MB-FUS experiments. This amount of concentration is also comparable with a clinically recommended dosage of 10 μL/kg for imaging in the human body [45] (See Appendix).

Before measuring MB AE, the ibidi μ-slide (ibidi) was filled with degassed water, and a volume of 8.5 mm × 5mm × 0.2 mm was insonified with a single 20-cycle, 500 kHz, 250 kPa peak-to-peak (MI = 0.17) sine wave to acquire the background signals. Next, the MBs were injected into the channel, and the received signals including acoustic emissions of the MBs were recorded. The MBs were stationary during the ultrasound insonification. The presence of MBs in the insonification field and the sufficient level of transmit pressure were verified by confirming the harmonic response of MBs received by the piezo-PCD. The MBs’ AE was simultaneously recorded with piezo-PCD and the 15 CMUT elements. After confirming the MBs’ AE, another experiment was carried out during which the digitization channel between the piezo-PCD and a single CMUT element was switched to receive the MBs’ AE using 1 and 15 CMUT elements. It’s important to note that the same region of MBs was insonified only once and using fresh MBs.

III. Results

A. CMUT + TIA Electrical Characterization

Figures 3a and 3b depict the electrical impedance of the water-loaded, one, and 15 CMUT elements, respectively. We biased the CMUTs with 120 V DC and operated them in collapsed mode. As expected, around 0.5 MHz and 1 MHz, the impedance shows variations corresponding to the dips in the frequency response. Figure 3c depicts the theoretical input-referred current noise density of the CMUTs that we calculated it using Equation 1. Note that for 15 CMUT elements, the current noise density is higher and increases in frequency with increasing admittance (Eq. 1), approaching a theoretical value of $40pA/\sqrt{Hz}$ for a large transducer at higher frequencies.

Fig. 3:

The real (a) and imaginary (b) part of the measured electrical impedance for 1 (in blue) and 15 (in red) CMUT elements with water loading and in collapse-mode, (c) is the calculated input current noise density using Eq.1

Figure 4a shows the measured gain of the TIA as a function of frequency, which is 117.5 dBΩ, in agreement with the design target, and remains flat up to 3 MHz. The circuit can start in all process corners, and the frequency response of the TIA is adequate for capturing up to the 6^th harmonic of the MB response for the excitation frequency of 500 kHz.

Fig. 4:

(a), The measured TIA AC response shows a 117.5 dBΩ gain which is flat up to 3 MHz. (b) the output voltage noise density of water-loaded, one and 15 CMUT elements operating in the collapse-mode, when the elements are connected to the TIA. The open-input, output-referred voltage noise density of the TIA is in good agreement with the Post-layout simulation.

Figure 4b shows the measured output-referred voltage noise density of the TIA when connected to a single CMUT element and a 15 CMUT element array. We also compared this measured noise with the noise of a single receive channel without a connected CMUT. This comparison shows a good agreement with the post-layout simulation (green and gray graphs in Figure 4b). Connecting 1 or 15 CMUT elements to a single receive channel alters the output noise level. Based on the results (red and blue graphs in Figure 4b), the system’s output noise level with 15 CMUT elements is higher than that with a single CMUT element, following the same trend observed in the calculated current noise spectrum (Figure 3c).

The RMS output-referred voltage noise of the collapse-mode, water-loaded, single, and 15 CMUT elements receiving channel are 137.83 μV and 559.06 μV, respectively over the measured frequency of 100 kHz − 3 MHz (Figure 4b). Therefore, for an output voltage of 1.2 Vpp, the system achieves the dynamic ranges of 69.7 dB and 57.6 dB, respectively.

The open-circuit input TIA noise measurement (green graph in Figure 4b) and the measured TIA gain (Figure 4a) determine the input-referred current noise density of the TIA at 2 MHz as follows:

$$\frac{568nV/\sqrt{Hz}}{117.75dB}=735fA/\sqrt{Hz}.$$ (2)

The calculated $735fA/\sqrt{Hz}$ input-referred current noise density based on the measured noise and gain is slightly higher than the $312fA/\sqrt{Hz}$ obtained from post-layout simulation at 2 MHz. This discrepancy is due to the parasitic capacitors of the PCB at the input of the TIA that degrades the noise performance. However, the measured $735fA/\sqrt{Hz}$ inputreferred current noise density is still lower than a single CMUT element theoretical noise ($6.3pA/\sqrt{Hz}$ at 2 MHz, Figure 3c), as desired.

B. Acoustic Characterization

Figure 5a shows the receive sensitivity of both the single CMUT element and 15 CMUT elements as a function of frequency. The receive sensitivity spans from 12.3 − 19.8 mV/Pa for the single CMUT element to 24.3 − 61.25 mV/Pa for the 15 CMUT elements over the frequency of interest. As expected, the 15 CMUT elements exhibit higher receive sensitivity than the single CMUT element. For instance, it is 2.9 times higher at 2 MHz. However, this increase is significantly less than expected for this measurement. This is partly due to additional capacitance introduced by the large array which impacts the gain of the TIA reducing the apparent sensitivity. Indeed, an analysis that considers the noise model and gain of a TIA with a CMUT input ([26]) confirms that connecting 15 CMUT elements to a TIA designed for a single element reduces the gain by approximately 9 dB at 2 MHz and limits the system bandwidth to about 1 MHz. More importantly, the non-uniform amplitude and phase of the incident pressure field due to misalignment and other imperfections have a more significant impact on the larger area CMUT. Consequently, this results in a less significant sensitivity increase compared to noise at the output, impacting the minimum detectable pressure for this particular measurement. As will be seen later in the MB emission experiments (Figure 7), the sensitivity difference is 3 dB higher between the 15 element CMUT and single CMUT element as compared to Figure 5a due to the more uniform pressure field on the receivers in those experiments.

Fig. 5:

(a), The receive sensitivity of the collapse-mode, 1 and 15 CMUT elements. (b), the minimum detectable pressure of the system.

Fig. 7:

The MB’s acoustic emission comparison between the single CMUT element (green) and 15 CMUT elements (blue)

Figure 5b shows the minimum detectable pressure (MDP) or noise equivalent pressure of the system. According to this result, the MDP ranges from $0.085-0.15mPa/\sqrt{Hz}$ for the single CMUT element and $0.12-0.23mPa/\sqrt{Hz}$ for the 15 CMUT elements, with outliers at 0.5 and 1 MHz due to the discussed decrease in receive sensitivity.

C. MB acoustic emission

The A305 transmitter used in the experiments and at a distance of 65.5 mm produces a 10 mm in diameter, −6 dB beam width at 0.5 MHz. Thus, we can consider the pressure field on the MBs and inside the 5 mm channel width of the ibidi μ-slide quasi-uniform for the 45-degree incidence angle.

The ultrasound gets attenuated more through the thick-walled surface of the ibidi μ-slide, with the highest attenuation occurring for the 200 μm channel height [42]. For example, at a 45° incident angle, researchers have reported attenuation values of 5.5 dB and 7.6 dB for excitations at 1 and 2 MHz, respectively.

Figure 6a shows two sets of signals that the 15 CMUT elements received under this incident ultrasound excitation condition. The background signal, shown in gray, was obtained without MBs in the ibidi channel, whereas the blue signal was acquired with MBs present in the channel. The signals exhibit a DC offset of 0.9 V due to the bias at the AFE output.

Fig. 6:

(a) The time signal of the background and the acoustic emission response of the 15 CMUT elements. The comparison between the acoustic emission response of the piezo-based PCD (red) with its background signal (gray) and the 15 CMUT elements (blue) with its background (gray) in (b). The 15 CMUT elements detect harmonics and ultra-harmonics.

The frequency spectrum of the time signal depicted at Figure 6b indicates that the AE response of the MBs, acquired with the 15 CMUT elements, contains harmonics and ultraharmonics signatures, while the background (no MB) signal shows no higher harmonics. Figure 6b also presents spectra captured by the piezo-based PCD (in red), confirming the presence of harmonics and verifying that the MBs are excited in the nonlinear regime. These results demonstrate the 15 CMUT elements, together with the AFE ASIC, have sufficient sensitivity and bandwidth to capture harmonic and ultraharmonic responses in a broad frequency range up to the 6th harmonic for 0.5 MHz excitation.

Figure 7 illustrates the received signal, including the MB AE response, simultaneously recorded by both a single and 15 CMUT elements acquired in another experiment. The signal of a single CMUT element appears in green, and the signal of 15 CMUT elements is in blue, with their respective background signals shown in gray. A comparison of the background signals indicates that the noise level of the single CMUT element is, on average, 10 dB lower than that of the 15 CMUT elements.

IV. Discussion

This study presents low noise and highly sensitive analog front-end electronics for a CMUT-based PCD array that is developed for monitoring MB dynamics. It shows the developed system can detect MB-generated AE spectral components with low MB concentration in clinically relevant frequency (0.5 MHz) and pressure (125 kPa peak-negative pressure) and can be used as a PCD. The CMUT-PCD system has a minimum detectable pressure of $0.085-0.23mPa/\sqrt{Hz}$ which can detect harmonics and ultra-harmonics from MBs.

There are several trends and results that require further clarification. The receive sensitivity of the collapse-mode CMUT depicted in Figure 5a follows a trend except at the two distinct frequencies of 0.5 and 1 MHz. This behavior results from the CMUT array frequency response. There are two dips in the frequency response of the collapse-mode (120V DC) CMUT array (Figure 2b). The dips in the frequency response of the CMUT array were associated with peaks in radiation reactance of the array which are attributed to the array design [46], [47]. It means the CMUT array doesn’t efficiently radiate or receive acoustic pressure at those frequencies as the CMUT elements vibrate out of phase and lower the total output current. This behavior is specific to this particular CMUT array and can be avoided by design changes such as using a smaller CMUT membrane and a higher fill factor to move them out of the frequency band [46], [47].

A second harmonic signal is observed in the background signals for both piezo PCD and CMUT-PCD (Figure 6b). Hydrophone measurements indicate that this is present in the output of the piezo transmitter (A305) used in the study. Therefore, in this particular case, the difference between the background and MB cases for frequencies above the second harmonic is a better indicator of MB nonlinear behavior.

In the presence of the MBs, the amplitude of AE response at higher harmonics, from 3^rd to 6^th, is detected with at least 10dB SNR by the CMUT-PCD, taking advantage of the broad bandwidth of the device (Figure 6b).

MBs are highly nonlinear oscillators that produce strong harmonics, sub- and ultra-harmonics [48]. Furthermore, the CMUT-PCD spectrum in Figure 6b shows regular peaks at with their own background signal (no MBs) in gray ultra-harmonic frequencies such as 1.25, 1.75, 2.25, and 2.75 MHz. Since those peaks are not present in the background signals (no MBs), and the only difference is the introduction of the MBs into the ibidi channel, the data strongly suggests that the peaks at ultra-harmonic frequencies originated from the MBs activity.

Based on [48], the ultra-harmonics level is relatively low. For instance, at 2 MHz excitation frequency, a difference of 45–50 dB between fundamental and 25 −30 dB between 2^nd and 3/2 ultra-harmonic were reported. Given that the ultraharmonic signal levels are very low as compared to other harmonics, these results show the low-noise performance of the CMUT-PCD.

The minimum detectable pressure of a single CMUT element is between $0.085-0.23mPa/\sqrt{Hz}$. We demonstrated that with current MDP levels, our system is good enough to detect MB acoustic emission signals through an attenuating media of the ibidi.

The ultrasound attenuation in the human skull is 6.9 dB/MHz/cm [43], [44]. The thickness of the skull varies between 3.1 mm to 14.0 mm depending on region and gender, with an average of 6.5 mm and a standard deviation of 1.7 mm [49]. Therefore, by assuming an average skull thickness of 6.5 mm for humans, the ultrasound attenuation levels at 2^nd (1 MHz) and 4^th (2 MHz) harmonics are 4.48 dB and 8.9 dB respectively. These values are comparable with the ultrasound attenuation in the ibidi. Therefore, these data suggest that our system will be able to detect MB acoustic emission through the human skull.

This performance should be improved by increasing the aperture area of the CMUT-based PCD array. In the present implementation, this was not the case as the AFE was optimized for a single CMUT element. However, since the AFE has 16 channels it is possible to connect 16 elements to separate TIAs and coherently add the output signals. In this case the SNR, hence the MDP should be improved by 4 times, or 12 dB as compared to a single CMUT element case, enabling even lower levels of MB emissions detection.

V. Conclusion

In this study, we have successfully developed and tested a CMUT-based passive cavitation detector for tFUS with a low-noise TIA front-end circuit system. The system has 12.3 − 61.25 mV/Pa receive sensitivity with $0.085-0.23mPa/\sqrt{Hz}$ minimum detectable pressure, and a flat bandwidth up to 3 MHz. We demonstrated that the CMUT-PCD system is sensitive enough to detect MB-generated harmonics and ultra-harmonics with 500 kHz excitation frequency when MB signals are attenuated at a level similar to that of a human skull. Given that performance improvements are possible with straightforward changes to the system, our findings support the use of this technology for monitoring and controlling the MB activity during MB-enhanced FUS interventions in the brain.

Highlights.

• A CMUT-Based PCD with an analog front-end receiver was developed for monitoring the BBB Opening in tFUS, showing the ability in capturing MBs AE spectral components when sonicated at clinically relevant frequencies (0.5 MHz) and pressures (250 kPa pp).

• The developed CMUT-PCD system has a 12.3 − 61.25 mV/Pa receive sensitivity with $0.085-0.23\text{mPa}/\sqrt{\text{Hz}}$ minimum detectable pressure up to 3 MHz. It detects harmonic and ultra-harmonic signals of MBs.

• The observed responses support the use of CMUT-PCD for monitoring and controlling the MBs dynamics during FUS-mediated BBB disruption.

APPENDIX I. Calculation of the dilution factor

Typically, a clinical dosage of 10 μL/kg is recommended for DEFINITY type MB with initial concentration of 12 × 10⁹ [45]. In bolus injection of the MBs into the bloodstream, the MBs concentration gets diluted in the whole blood. To mimic the bolus injection and MB concentration in the human brain, we used diluted MBs and injected them directly into the ibidi μ-slide. Our calculations to justify the dilution factor are:

$$\begin{array}{cc}\text{An average adult human blood:}& 4.5-5\text{Liter of blood}\\ \text{An average American male weight:}& 70-90\text{kg}\\ \text{DEFINITY concentration:}& 1.2\times {10}^{10}\\ \text{Recommended Clinical Dosage use:}& 10\text{μL}/\text{kg}\end{array}$$

Therefore, for an average adult, the amount of MBs would be:

$$\begin{gathered} 10\text{μL}/\text{kg}\times 90\text{kg}=900\text{μL} \\ 900\text{μL}/1000\text{μL}\times 1.2\times {10}^{10}=1.08\times {10}^{10}\text{MBs} \end{gathered}$$

Therefore, the concentration of the MBs in an average human blood assuming an even distribution would be:

$$1.08\times {10}^{10}\text{MBs}/5000\text{mL}=2.16\times {10}^6\text{MB}/\text{mL}$$

The average human brain blood volume is approximately 750 milliliters, which constitutes about 15–20% of the total cardiac output [50], [51]. Therefore, the number of MBs in adult human brain assuming 1–10% of MBs reaches to the brain:

$$\begin{gathered} 2.16\times {10}^{6}MBs/mL\times 750mL\times (0.01-0.1)= \\ 16.2\times {10}^6-162\times {10}^{6}MBs \end{gathered}$$

The ratio of the average human brain volume to the average volume of all the vessels, including capillaries, veins, and arteries, approximately is 4:1.

Assuming all the MBs are distributed in all the cerebral vessels, the MBs concentration would be:

$$\begin{gathered} 16.2\times {10}^6-162\times {10}^{6}MBs/(750/4)mL= \\ 4.05\times {10}^6-40.5\times {10}^{6}MBs/mL \end{gathered}$$

The concentration after diluting 2400 times:

$$1.2\times {10}^{10}\text{MBs}/2400=5\times {10}^6\text{MB}/\text{mL}$$

Consequently, our dilution is at the lower end of the spectrum of MBs in an average adult human brain.