Endocavity ultrasound and photoacoustic system for fetal and maternal imaging: design, implementation, and ex-vivo validation

Abstract.

Purpose: Transvaginal ultrasound (TVUS) is a widely used real-time and non-invasive imaging technique for fetal and maternal care. It can provide structural and functional measurements about the fetal brain, such as blood vessel diameter and blood flow. However, it lacks certain biochemical estimations, such as hemoglobin oxygen saturation (${\mathrm{SO}}_2$), which limits its ability to indicate a fetus at risk of birth asphyxia. Photoacoustic (PA) imaging has been steadily growing in recognition as a complement to ultrasound (US). Studies have shown PA imaging is capable of providing such biochemical estimations as ${\mathrm{SO}}_2$ at relatively high penetration depth (up to 30 mm).

Approach: In this study, we have designed and developed a multi-modal (US, PA, and Doppler) endocavity imaging system (ECUSPA) around a commercialized TVUS probe (Philips ATL C9-5).

Results: The integrated system was evaluated through a set of in-vitro, ex-vivo, and in-vivo studies. Imaging of excised sheep brain tissue demonstrated the system’s utility and penetration depth in transfontanelle imaging conditions. The accuracy of using the spectroscopic PA imaging (sPA) method to estimate ${\mathrm{SO}}_2$ was validated by comparing sPA oximetry results with the gold standard measurements indicated by a blood gas analyzer. The ability of US and Doppler to measure moving blood volume was evaluated in-vivo. Spectral unmixing capabilities were tested using fluorophores within sheep brains.

Conclusion: The developed system is a high resolution (about $200\mu \mathrm{m}$ at 30 mm depth), real-time (at 30 Hz), and quantitative (${\mathrm{SO}}_2$ estimation error $<10\%$) imaging tool with a total diameter less than 30 mm, making it suitable for intrapartum applications such as fetal and maternal diagnostics.

1. Introduction

Transvaginal US (TVUS) plays an important role in maternal and fetal care, especially in the management of labor and delivery because of its notable advantages of high-level safety, compactness, and wide availability.^1,2 In fetal and perinatal care, hypoxia is a significant clinical problem that can lead to serious medical conditions. Each year, $\sim \mathrm{840,000}$ neonatal deaths are related to asphyxia at birth or oxygen deprivation.³ Other severe health problems, such as hypoxic-ischemic encephalopathy, jeopardize the newborns to long-term neurological disorders, such as mental impairment or cerebral palsy.^4,5 Among various methods for evaluating fetal conditions during labor, fetal heart rate monitoring (FHRM) is considered the gold standard for use in clinical settings.^6,7 FHRM aims to identify early signs of fetal hypoxia through monitoring the fetal heart rate. Although FHRM has effectively contributed to reducing fetal mortality, it still has substantial both false-negative ($\sim 50\%$) and false-positive ($\sim 70\%$) diagnoses.^8–12 Therefore, the ability of FHRM to accurately and consistently identify fetuses at risk of metabolic acidosis is still limited. The impact of the limited accuracy of FHRM on maternal care is shown by the continuous increase in the number of C-sections.^13,14 In addition, the low sensitivity of FHRM limits its ability to detect the partial hypoxia status (which can cause neurological problems) in children during delivery.^8,9,15 US efficacy in imaging fetal brain is limited to visualizing blood vessels and measuring blood flow through Doppler imaging. Optical-based imaging modalities, such as near-infrared spectroscopy (NIRS), have been used to monitor fetal health during labor.¹⁶ NIRS (aka fetal pulse oximetry) detects the differences between the optical properties of oxy-hemoglobin (${\mathrm{HbO}}_2$) and deoxyhemoglobin (Hb) and provides a quantitative estimation of fetal blood oxygen saturation (${\mathrm{SO}}_2$).¹⁷ In spite of a good correlation between NIRS and blood oxygenation saturation, practical restrictions, such as displacement of the sensor, fetal movement, or uterine contractions, limit its clinical use.¹⁸ Other imaging modalities such as magnetic resonance imaging and computed tomography (CT) are not a point-of-care for imaging fetuses during labor and delivery. Therefore, there is a need for an alternative (or complement to FHRM) method for intrapartum monitoring of fetal status that can reliably identify early signs of fetal compromise, thus improving clinical decisions and perinatal outcomes.

In recent years, a relatively new US-based imaging modality, photoacoustic (PA) imaging, has shown significant potential in providing functional, cellular, and molecular information^19–33 in relatively deep imaging depths for tissue characterization. PA imaging utilizes short (nanoseconds) laser pulses to excite tissue and then detects the acoustic signal resulting from thermal expansion of the chromophores. Previous studies proposed integrating PA imaging into a TVUS probe to provide functional (and molecular) images for measuring cervical and ovarian tissue vascularity and ${\mathrm{SO}}_2$ level in the placenta at clinically relevant depths.^34–38 Those studies indicated the arrangement of a light source around the TVUS probe is the solution for optimizing the PA imaging coverage and penetration depth. However, their designs have certain limitations, such as non-uniform light illumination³⁵ and an open path for laser light.³⁸ Therefore, in this study, we proposed a new design of the light delivery system around the TVUS probe to create a more uniform and focused light delivery. The proposed endocavity US, PA, and Doppler (ECUSPA) imaging system utilizes multi-parametric optical and acoustical information, where US imaging provides structural information, power Doppler monitors the blood flow, and PA imaging provides tissue and vascular oxygenation maps. The proposed system is a non-invasive, low-risk, point-of-care imaging tool with an enclosed light delivery path. The system has the potential capacity to monitor and measure fetal cerebral perfusion and ${\mathrm{SO}}_2$ levels during labor and delivery, which allows for evaluating the cerebral metabolic rate of oxygen (${\mathrm{CMRO}}_2$) of the fetal brain during the delivery operation.

2. Materials and Method

2.1. Design of the ECUSPA Imaging System

The ECUSPA imaging system integrates an end-firing clinical TVUS transducer (Philips, ATL C9-5 ICT, Andover, Massachusetts) surrounded by 18 fibers from a 19 fibers bundle (Thorlabs, multimode, FT1000EMT, Newton, New Jersey), in which the 19^th fiber was used for real-time laser energy monitoring [Fig. 1(a)].³⁹ The proximal end of the bundle is coupled to a high-energy pulsed laser light source (Spectra-Physics, Quanta-Ray PRO 270, 30 Hz, Clara, California) connected with a tunable optical parametric oscillator (OPO, VersaScan730, 400 to 2400 nm, Clara, California). The ECUSPA imaging probe [Fig. 1(b)] has a total diameter size of 29 mm, which satisfies the clinical requirement for the probe diameter ($<30\mathrm{mm}$) to avoid any patient discomfort during the scan.^40,41 The fibers are forced to form a line-like illumination pattern 25 mm away toward the acoustic imaging plane by a compacted custom-designed 3D printed fiber holding sheath [Printing material, VisiJet M3 Crystal, Fig. 1(c)]. Therefore, the PA imaging penetration can be optimized by improving illumination patterns within the limitation of maximum permissible exposure (MPE). By balancing the laser energy and illumination coverage, the PA penetration depth is optimized. Our previous study⁴² demonstrated that forcing the illumination pattern closer to the US imaging plane (the center of the element) can increase PA penetration depth. A 128-channel ultrasound research platform (Verasonics, Vantage 128, Kirkland, Washington) is used to acquire real-time US and PA signals. The acoustic and light source synchronizations are controlled by high-speed field-programmable gate arrays (FPGA, Digilent, BASYS3, clock 200 MHz, Pullman, Washington) for maintaining a constant imaging sequence. The designed imaging sequence was set to 30 frames per second (fps) for PA images, and between each PA frame, three plane wave US images with 21 angle-compounding were acquired.

Fig. 1.

(a) ECUSPA imaging system block diagram. (b) Schematic of the components in the combined probe (diameter 29 mm); red, body structure; yellow, fibers; and blue, locking structure. (c) Photograph of the assembled ECUSPA imaging probe.

2.2. Characterization of ECUSPA Imaging System

A Monte-Carlo light simulation (program “mcxyz.c”, developed by Oregon Medical Laser Center)⁴² was used to study the theoretical PA imaging penetration depth. The simulation introduced an in-silico numerical phantom with three layers (water, human skin, and gray matter) and a blood vessel embodied at 20-mm depth. The optical properties of the chromospheres were adopted from^43–47 for the wavelength of 808 nm. As a result, the ECUSPA naturally can form two illumination patterns, where the probe is directly in contact with the imaged object [Fig. 2(b)] or with a 25-mm distance [Fig. 2(c)]. The laser energy for the probe attached to the tissue surface must be less than 4.5 mJ since the illumination area is $0.1413{\mathrm{cm}}^2$ to comply with the MPE ($32\mathrm{mJ}/{\mathrm{cm}}^2$ at 808 nm), according to ANSI safety limits.⁴⁸ In contrast, the scenario in which the probe is placed at a distance (i.e., detached position) requires higher pulse energy of about 61 mJ (area $2.6{\mathrm{cm}}^2$) to reach the same fluence.

Fig. 2.

(a) In-silico numerical phantom for Monte-Carlo simulation. Schematic of the illumination pattern when the ECUSPA probe is (b) attached to the imaged object and (c) detached with a 25-mm distance.

We further evaluated the PA imaging penetration depth in a set of ex-vivo experiments. An excised bovine muscle tissue embedded with three thin-wall optically transparent tubes (ZEUS AWG 13, $D=1.8\mathrm{mm}$, Orangeburg, South Carolina) was filled with human blood at various depths from 10 to 30 mm [Fig. 5(a)]. A set of co-registered US and PA images were acquired at a wavelength of 808 nm. The fluence was about $24\mathrm{mJ}/{\mathrm{cm}}^2$. The selected wavelength is the isosbestic point of the oxy- and deoxy-hemoglobin. An average of eighty PA acquisitions was utilized to reduce the PA background noise. The region of interest (ROI) selections for the tubes were based on the US image. The averaged PA signal amplitudes were calculated and plotted along with the imaging depth.

Fig. 5.

(a) Evaluation of PA imaging penetration depth in excised bovine tissue. (b) US and (c) PA ($\lambda =808\mathrm{nm}$) images. (d) Normalized PA signal amplitudes versus depth. Green arrows indicate the boundary of the embedded tubes in the US image.

We then evaluated the system’s lateral and axial resolutions for the US imaging at different imaging depths. Our developed system is an acoustic-resolution PA (AR-PA) imaging system, which shares the same acquisition transducer, receiving circuit, and beamforming (image reconstruction) method as in planar US imaging. Besides, the literature reported that the spatial resolution of the US is comparable for PA imaging.⁴⁹ Therefore, we evaluated the US imaging resolution with six nylon filaments (diameter of $200\mu \mathrm{m}$) that emerged in a water tank [Fig. 6(a)]. The wire phantom was imaged at t distances varying from 10 to 30 mm. The centerline of the resolution phantom was aligned at the centerline of the probe. The full width half maximum (FWHM) was calculated based on the appearance of the wires and used to determine the resolutions of the system. The probe was tilted about 10 deg to avoid the blockage of the acoustic waves. Our previous publication³² selected the minimum variance beamforming algorithm for the US and PA imaging reconstruction. The theoretical resolutions were calculated using the following Eqs. (1)–(3).^43,50

$$\text{Lateral Resolution}=\frac{d\times \lambda }{D},$$ (1)

$$\text{Axial Resolution}=\frac{1}{2}\lambda ,$$ (2)

$$\text{Spatial Pulse length}:\lambda \approx 2\times \frac{c_{\text{water}}}{f},$$ (3)

where $d$ is the axial distance between the object to the transducer surface, $f$ is the bandwidth with a transmit center frequency (7 MHz), $C_{\text{Water}}=1497\mathrm{m}/\mathrm{s}$ is the speed of sound for water at 25°C, and transducer surface width $D=22\mathrm{mm}$.

Fig. 6.

ECUSPA imaging system resolutions evaluated using a calibration wire-phantom ($200\mu \mathrm{m}$) (a); (b) US images at varying imaging depths. (c) Axial and (d) lateral appearance of the wires at different depths.

2.3. Spectroscopic PA Oximetry

The developed ECUSPA imaging system can acquire spectroscopic PA (sPA) images and use a wavelength unmixing method to calculate the blood oxygen saturation (${\mathrm{SO}}_2$) changes. A set of ex-vivo experiments with controlled blood ${\mathrm{SO}}_2$ levels were performed to evaluate the system’s capability in sPA oximetry. The control of blood ${\mathrm{SO}}_2$ levels was achieved by exposing the blood to ${\mathrm{O}}_2$ or ${\mathrm{CO}}_2$ combined with ${\mathrm{N}}_2$ in a three-neck flask and stirring at body temperature (37°C). An air-tight syringe drew the blood into a shelled glass tube (inner diameter 3 mm, outer diameter 4 mm) when the blood reached the desired ${\mathrm{SO}}_2$ level. The range of ${\mathrm{SO}}_2$ was controlled from 100% down to about 40%, which mimics a healthy range for the fetus during labor.⁴⁴ The ${\mathrm{SO}}_2$ levels were determined by a gold standard clinical blood gas analyzer (BGA, OPTI CCA-TS2, Optimedical, Roswell, Georgia). In addition, a real-time in-situ oxygen monitoring probe (NEOFOX-Kit-Probe, Ocean Optics, Largo, Florida) was used to monitor the ${\mathrm{SO}}_2$ level changes simultaneously with the BGA and provide measurements of blood ${\mathrm{SO}}_2$ when it went below BGA limitations. The following wavelength unmixing method [Eqs. (4) and (5)] adopted from Refs. 45 and 46 were used to calculate ${\mathrm{SO}}_2$ (multi-wavelengths unmixing).

$$PA({\lambda }_i)=(\epsilon Hb({\lambda }_i)[Hb]+\epsilon Hb{O}_2({\lambda }_i)[Hb{O}_2])F({\lambda }_i),$$ (4)

$$S{O}_{2_{(x,y,z)}}=\frac{{[Hb{O}_2]}_{(x,y,z)}}{{[Hb{O}_2]}_{(x,y,z)}+{[Hb]}_{(x,y,z)}}=\frac{F_2·{\epsilon }_{H{b}_{\lambda 2}}·P{A}_{{\lambda }_1(x,y,z)}-F_1·{\epsilon }_{H{b}_{\lambda 1}}·P{A}_{{\lambda }_2(x,y,z)}}{F_2·P{A}_{{\lambda }_1(x,y,z)}({\epsilon }_{H{b}_{{\lambda }_2}}-{\epsilon }_{Hb{O}_{\lambda 2}})-F_1·P{A}_{{\lambda }_2(x,y,z)}({\epsilon }_{H{b}_{{\lambda }_1}}-{\epsilon }_{Hb{O}_{\lambda 1}})},$$ (5)

where the $\epsilon \mathrm{Hb}(\lambda \mathrm{i})$ and $\epsilon {\mathrm{HbO}}_2(\lambda \mathrm{i})$ are the known molar extinction coefficients (${\mathrm{cm}}^{-1}/\mathrm{M}$) of Hb and ${\mathrm{HbO}}_2$ at wavelength $\lambda \mathrm{i}$, respectively, and $[Hb]$ and $[Hb{O}_2]$ are the concentrations of deoxy- and oxy-hemoglobin. The $\mathrm{F}(\lambda \mathrm{i})$ is the fluence at each wavelength deposited to the absorbers, and the spatial coordinate of a voxel in the image was indicated as $x$, $y$, and $z$. Although, in theory, two wavelengths are sufficient to solve the linear equation of ${\mathrm{SO}}_2$ level, we used three wavelengths (750, 780, and 800 nm) with a constant fluence ($24\mathrm{mJ}/{\mathrm{cm}}^2$) to reduce the error of measurement using the least-squares fitting method. For ${\mathrm{SO}}_2$ measurements, wavelengths of 750, 780, and 800 nm were chosen to minimize water absorption. Besides, the water has a low scatterer within these wavelengths. Studies have shown that fluence compensation can increase the accuracy of PA oximetry,^47,51 and our measurements were performed in a small diameter tube located inside a homogenous water tank with no addition of scattering agents and the absorption of water at wavelengths we used for ${\mathrm{SO}}_2$ measurements are negligible. Therefore, no fluence compensation was required. However, fluence composition is required in future in-vivo imaging applications.

2.4. Transfontanelle US and sPA Imaging

One major obstacle in ultrasonic imaging of adult human brains is the acoustic distortions and limitations of light penetration due to the presence of a thick skull.⁵² However, in fetuses and neonates, this issue is less severe due to the significantly lower thickness of the skull and the existence of fontanelle openings.⁵³ The system was first tested in an adult sheep head with an artificial fontanelle. The artificial fontanelle had a diameter of 40 mm, which mimicked the actual size of anterior fetal fontanelles.⁵⁴ The dura maters were kept intact on the brain. The probe of the ECUSPA imaging system has a total diameter of 29 mm, which makes the system suitable for imaging through the anterior fontanelle openings. Several injections were made to generate visible inclusions: (1) 8% gelatin mixed with 0.1% cellulose for US images; (2) 8% gelatin mixed with sheep blood for PA images. The injections were located at depths varying from 1 to 2.5 cm. US and PA images were acquired at 740 nm with a fluence of $23\mathrm{mJ}/{\mathrm{cm}}^2$ before and after injection.

2.5. Evaluating the Accuracy of sPA in Measurement of Chromophores’ Concentrations

A set of ex-vivo sPA imaging experiments utilizing excised sheep brain with four tubes (ZEUS AWG 13, Orangeburg, South Carolina) filled with PA contrast agents was performed to quantify the sPA estimations accuracy of relative concentration using the same wavelength unmixing method. The contrast agents were Indocyanine Green (ICG, S2265, Few Chemicals GmbH, Bitterfeld-Wolfen, Germany) and Cyanine5.5 (Cy5.5, 1628790-37-3, Lumiprobe, Maryland). The four tubes were filled with four different combinations of the dyes at different ratios: (1) Pure ICG dye ($31\mathrm{nM}/\mathrm{ml}$), (2) Pure Cy5.5 dye ($20\mathrm{nM}/\mathrm{ml}$), (3) 30% ICG dye mixed with 70% by volume Cy5.5 dye, and (4) 70% ICG dye mixed with 30% Cy5.5 dye (in volume). The tubes were embedded about 10-mm deep inside the excised sheep brain measured absorptions of each combination are shown (Fig. 3). In addition, the chromophores’ optical density (OD) was measured by a microplate reader (SpectraMax ABS Plus, VWR, Radnor, Pennsylvania). Several sPA images were acquired at the wavelengths 680, 730, and 780 nm, covering the localized peaks of ICG and Cy5.5 dyes and the isosbestic point. Given the smaller size of the excised sheep brain compared to the fetus,⁵⁵ the US images were acquired with a linear-array US probe (L11-4v) at 9 MHz to visualize the structures of the brain. The wavelength unmixing method [Eqs. (4) and (5)] was used for the relative concentration calculation by replacing extinction coefficients of ICG as ${\mathrm{O}}_2$ and Cy5.5 as ${\mathrm{HbO}}_2$.

Fig. 3.

OD of different chromophores: pure ICG ($31\mathrm{nM}/\mathrm{ml}$), pure Cy5.5 ($20\mathrm{nM}/\mathrm{ml}$), 30% ICG dye mixed with 70% Cy5.5 dye (in volume), and 70% ICG dye mixed with 30% Cy5.5 dye (in volume).

2.6. In-vivo Fetal Brain Blood Flow and Volume Estimation

ECUSPA imaging system has the potential to monitor the fetal cerebral metabolic rate of oxygen consumption (${\mathrm{CMRO}}_2$) during labor and delivery by combining the PA measurements (blood ${\mathrm{SO}}_2$) and US measurements such as blood flow and perfusion within the arterial and venous system. The US measurements of blood flow and perfusion have already been established and utilized in the clinic. However, it is worth mentioning that the same TVUS transducer used to develop the PA probe can acquire all required US measurements for calculating ${\mathrm{CMRO}}_2$. Literature reported that regional blood flow perfusion could be estimated using power Doppler (PDU) and the fractional moving blood volume (FMBV) method.⁵⁶ Briefly, FMBV is a mathematical algorithm designed to reduce acoustic attenuation in-depth tissue imaging and tissue interphases on the PDU signals. The algorithm can provide a more accurate estimation of blood movement,⁵⁷ in which lower and higher points of ROI are described. The Doppler signals of blood movements from the center of the vessels were located by defining a normalization value (NV). The FMBV algorithm created a cumulative distribution of all PDU intensity values to define the NV and applied a two-tangent line technique.⁵⁸ Then the normalized signals were converted into fractions based on the NV and were expressed as the FMBV. Hence, the FMBV represents the percentage of blood flow movements in a defined ROI.⁵⁹ All PDU intensity values above the NV were assigned a value of 1 (they were considered similar to the NV), and all Doppler values below the NV were converted to fractions of the NV. The final FMBV estimation was calculated by averaging all the normalized Doppler intensity values per number of pixels. This estimate ranges from 0 to 1 and, when converted into a percentage, expresses the fraction of the ROI occupied by moving blood. By implementing this FMBV method, the feasibility of ECUPSA measurements was evaluated in a clinical setting during the stages of labor and associated uterine contractions. In-vivo performances of FMBV measurements were evaluated in a pilot study with a limited number ($n=10$) of patients at about 30 weeks of gestation during patient prenatal care, conducted at Detroit Medical Center/Hutzel Women’s Hospital Perinatology Research Branch and under an approved protocol by Institutional Review Board at Wayne State University (Protocol: NIH IRB 18-CH-N063).

3. Results

3.1. Characterization of ECUSPA Imaging System

The Monte-Carlo simulation results for the propagation of light photons are presented in two scenarios: (a) when the ECUSPA system is directly attached to the surface of the object, and (b) 25-mm away (detached) from the object were shown in (Fig. 4). Figures 4(a) and 4(c) show the transverse plane at a depth of 6 mm within the tissue, and Figs. 4(b) and 4(d) show the fluence distribution of the median plane (along the centerline), respectively. Thus, both illumination strategies could deposit more than $0.5\mathrm{mJ}/{\mathrm{cm}}^2$ into the object at an imaging depth of 20 mm. Simultaneously, the probe’s detached placement led to three times more fluence ($>2\mathrm{mJ}/{\mathrm{cm}}^2$). Therefore, the detached illumination strategy may achieve a higher PA penetration depth.

Fig. 4.

Monte-Carlo simulation results to evaluate the PA imaging penetration depth of ECUSPA. (a) Transverse at 6-mm depth and (b) frontal plane at the middle of the fluence distribution when the ECUSPA probe is attached to the surface of the object. (c) Transverse at 6-mm depth and (d) frontal plane at the middle median view of the fluence distribution for detached illumination. (e) Fluence comparison at 20-mm depth for two illumination strategies.

The US and PA images of excised bovine tissue with inserted tubes are shown in Figs. 5(b) and 5(c), respectively. The US images visualize the inserted tubes. The PA signal amplitudes arising from the inserted tubes were calculated and plotted in Fig. 5(d). The US images measured the inserted tubes’ diameters to be 2.2 mm (real diameter was 1.8 mm). The PA images demonstrate that the PA signals from the inserted tube at 30 mm (deepest depth) were detected by ECUSPA, indicating a penetration depth of 30 mm in-vivo.

The evaluation results of the system’s spatial resolution are shown in (Fig. 6). The US images of the wire phantom at different imaging depths were shown in Fig. 6(b), demonstrating a lower lateral resolution at greater depths. We quantified the axial resolutions at different imaging depths for the ECUSPA system: $200\mu \mathrm{m}$ at 10 mm, $200\mu \mathrm{m}$ at 20 mm, and $202\mu \mathrm{m}$ at 30 mm [Fig. 6(c)]. The lateral resolutions were also quantified at the same depths: $286\mu \mathrm{m}$ at 10 mm, $381\mu \mathrm{m}$ at 20 mm, and $525\mu \mathrm{m}$ at 30 mm [Fig. 6(c)].

3.2. sPA Oximetry

The experimental results of the sPA oximeter compare with the gold standard BGA results are shown in Fig. 7. In this experiment, the blood ${\mathrm{SO}}_2$ has a range from 46% to 100%. The color-coded sPA oximetry results at each ${\mathrm{SO}}_2$ level are shown in Figs. 7(a)–7(f), representing the ${\mathrm{SO}}_2$ maps of the cross-section of the blood tube. The non-uniformity of ${\mathrm{SO}}_2$ maps could be due to the bias induced by light diffusion, which can affect spectral unmixing and sPA analysis, non-uniform blood-gas exchange within our custom-built chamber,^60,61 and the error of the linear inversion algorithm.⁶² However, the sPA oximetry reveals a relatively high accuracy in measuring ${\mathrm{SO}}_2$ ($R^2=0.8727$) over a range of values relevant to clinical conditions.

Fig. 7.

Comparison between ex-vivo sPA oximetry estimation of ${\mathrm{SO}}_2$ and BGA results. (a)–(f) sPA oximetry measurements of ${\mathrm{SO}}_2$ in an artificial blood vessel (sPA measurement/BGA reading). (g) sPA oximetry estimations correlated with BGA ${\mathrm{SO}}_2$ measurements.

3.3. Ex-vivo Transfontanelle US and sPA Imaging in Sheep Brain

The ex-vivo transfontanelle PA images before and after the injection of contrast agents in a sheep brain are shown in (Fig. 8). While the PA images are capable of measuring at a depth up to 30 mm, the excised sheep brain model lacks blood flow, which has associated signal attenuation. A more accurate model would include consideration for the effect of hemodynamics on signal propagation and acquisition.

Fig. 8.

Ex-vivo transfontanelle US and PA images (a) before and (b) after the injection of contrast agents by the ECUPSA system. The yellow arrows indicate the injection of contrast agents for PA, and red arrows for US, respectively.

3.4. Accuracy of sPA Measurement of Absorbers’ Relative Concentration in Excised Brain Tissue

The results for estimating the relative concentration of two spectroscopic separated chromophores with the ECUSPA system are shown in (Fig. 9). Figure 9(a) shows the processed high-resolution US image of the sheep brain where the four inserted tubes of different combination ICG and Cy5.5 dyes are visualized. More anatomical features were detected and marked in Fig. 9(e). The detected PA images [Figs. 9(b)–9(d)] demonstrate that the PA signals arising from the dyes have different amplitudes, which followed the reported absorption spectrum of the ICG and Cy5.5. The pure ICG showed the strongest signal at the wavelength $\lambda =780\mathrm{nm}$, and Cy5.5 exhibited a strong signal at $\lambda =680\mathrm{nm}$, as was anticipated from the optical properties of the dyes, shown in Fig. 3. The sPA color map, which estimated the relative concentration of ICG and Cy5.5, is shown in Fig. 3(e). The average relative dye concentrations were calculated as the average concentration in the selected ROIs and are listed in Table 1. sPA accuracy of chromophores’ concentration within the brain tissue was calculated to have an error of 6.66% based on comparison to actual concentration. While the sensitivity and accuracy of in-vivo PA oximetry is a subject of a future study, we anticipate a comparable accuracy for the same sPA imaging and spectral unmixing method since hemoglobin is a stronger absorber compared to the dyes we used in our studies.

Fig. 9.

sPA estimation of the relative concentration of ICG and Cy5.5. (a) US background. (b)–(d) PA images at different wavelengths overlapped with the US background. (e) Spectral unmixed sPA measurements for the relative concentration of ICG and Cy5.5. Color coded maps indicating the relative concentration of ICG and Cy5.5 and showing the ability of sPA to distinguish and measure the concentration of two spectrally different absorbers. In order to increase the visibility of the sheep brain anatomical structures, a higher frequency linear US probe (L11-4v) was used to acquire the US background.

Table 1.

Relative concentration of the ICG and Cy5.5 mixed solution based on OD and sPA measurements.

Solution Name	Optical reader measurements		sPA wavelength unmixing
	ICG (%)	Cy5.5 (%)	ICG (%)	Cy5.5 (%)
Pure ICG	100.00	0.00	98.55	1.45
70% ICG + 30% Cy5.5	63.48	36.52	70.14	29.87
30% ICG + 70% Cy5.5	23.20	76.80	24.67	75.33
Pure Cy5.5	0.00	100.00	0.61	99.39

3.5. In-vivo Fetal Brain Blood Volume Estimation Using FMBV

The FMBV measurement of a fetal brain acquired with the TVUS transducer used in the ECUSPA system is shown in Fig. 10. The Doppler image [Fig. 10(a)] shows the accessibility of fetal cerebral vasculature for the potential ECUSPA system overlapping on top of US images. The cumulative distribution of Doppler signals and threshold selection for normalization steps are shown in Fig. 10(b). Our FMBV measurements indicate that the parietal cortex has a mean perfusion value between 45% and 60%, which equates to a final FMBV estimation of 31% within the field of view [Fig. 10(c)].

Fig. 10.

(a) Power-Doppler ultrasound of cortical arterial and venous blood flow in fetus brain (in-vivo). (b) Cumulative PDU intensity distribution (histogram, blue); selection of intersection of global best linear fit (red line); tangent lines (light green lines) and their intersect to define the normalization level (dashed dark green), (c) Normalized FMBV image (green mask) overlaid on top of acquired Doppler image, indicating the blood volume is 31%.

4. Discussion

A new type of integrated US, PA, and Power Doppler imaging system around a clinical TVUS transducer (C9-5) was developed. It is a minimally invasive and point-of-care quantitative imaging system. A set of ex-vivo and in-vivo studies evaluated the ECUSPA imaging system quality inside the brain at a depth up to 30 mm. It has high accuracy in quantifying the relative concentration of separated chromophores. In addition, it can accurately quantify the relative concentration of blood ${\mathrm{SO}}_2$ as demonstrated by a high correlation to the standard BGA measurements. The developed system uses a novel fiber-based light delivery strategy that provides more uniform, focused, and efficient light illumination for in-vivo fetal brain imaging during labor and delivery. The proposed ECUSPA imaging system is able to operate with a full light enclosure (unlike other common systems that use an air beam), complies with safety requirements, provides real-time laser energy monitoring, and acquires a combination of US and PA biomarkers. These advantages make the system suitable for fetal brain monitoring in clinical settings.

The current study has certain limitations, which need to be addressed prior to translating the technology to in-vivo studies. The demonstrated PA imaging depth might be reduced due to the presence of blood in the scalp. Moreover, the presence of the fetal skull and scalp will induce acoustic aberration and limit the penetration depth and accuracy of sPA imaging.⁶³ However, given the intended use of placing the probe close to the head of the fetus (during active stages of labor) and the relatively shallow penetration depth required to access venous blood in the superior sagittal sinus (SSS) and arterial blood flow within cortical branches of the anterior cerebral artery (ACA), we anticipate that the system is capable of in-vivo imaging of the fetal brain. Although studies have indicated that the fetus’s anterior fontanelle is covered with a thin and soft bone layer that causes acoustic aberration, this has insignificant effects on US images.⁶⁴ In our transfontanelle imaging of the sheep brain, the animal head’s skull was completely removed; thus, the possible skull aberration effects of PA imaging were not evaluated.

Another limitation of the developed ECUSPA imaging system is that it is only suitable to be used in certain fetal positions: occiput anterior (OA), occiput posterior (OP), and occiput transverse (OT). However, delivery positions are the most common ($\sim 94\%$ of total deliveries).⁶⁵

US imaging is one of the safest imaging modalities for fetal imaging and is established as a part of labor and delivery management. However, the use of laser for sPA imaging requires careful considerations of safety. All studies performed in this work used laser fluences below the ANSI safety limits. However, keeping laser fluence levels within the safety limitations will likely prevent this device from accessing tissue significantly deeper than 30 mm in the future. Considering the relative shallow penetration depth required to access the SSS and cortical branches of ACA, we anticipate the ECSUPA system is capable of maintaining its performance while operating at a lower fluence for PA imaging.

The ECUSPA imaging system is a multi-parametric imaging system that aims to visualize blood flow and functional information of brain activities, such as ${\mathrm{SO}}_2$ in the SSS and ACA. Estimating ${\mathrm{SO}}_2$ in those vessels allows for estimating the cerebral metabolic rate of oxygen (${\mathrm{CMRO}}_2$) during the delivery operation. The blood flow measured by PDU has a much higher penetration depth and performs measurements directly in the SSS and ACA. Hence for PA imaging, the SSS is relatively accessible to the system and not for ACA. The fluence may be reduced during in-vivo studies and kept the possibility to perform sPA oximetry in cortical branches of ACA, which are closer to the fetal head and have similar ${\mathrm{SO}}_2$ levels. Therefore, by measuring the ${\mathrm{SO}}_2$ in SSS and cortical branches of ACA, combined with the FMBV, it is feasible to measure the ${\mathrm{CMRO}}_2$.

For future work, the accuracy of the ECUSPA system for measuring blood ${\mathrm{SO}}_2$ and blood volume needs to be quantified. Furthermore, more data is needed to determine the full range of imaging depth, specifically while including the attenuation dynamics of blood flow on the PA signal.

5. Conclusion

A novel, minimally invasive, low-risk, and point-of-care system was developed combining endocavity ultrasound, photoacoustic, and Doppler imaging into a clinical transvaginal ultrasound probe. The performance of the ECUSPA system was evaluated with a series of ex-vivo and in-vivo human experiments. The system has the capabilities to image deep-lying tissues ($\sim 30\mathrm{mm}$) within excised sheep brains (in the absence of the skull). In addition, the proposed imaging system can acquire real-time functional information such as estimating blood oxygen saturation (${\mathrm{SO}}_2$) and blood flow, which can be used to measure the metabolic rate of oxygen consumption of the fetal brain during labor and delivery. Thus, the development and implementation of the ECUSPA imaging system may provide a tool for better fetal and maternal health care and will potentially lower the risk of brain injury during delivery.

Biography

Biographies of the authors are not available.

Disclosures

No conflicts of interest, financial, or otherwise, are declared by the authors.