Hand-Held Optoacoustic System for the Localization of Mid-Depth Blood Vessels

Abstract

The ability to rapidly locate blood vessels in patients is important in many clinical applications, e.g., in catheterization procedures. Optical techniques, including visual inspection, generally suffer from a reduced performance at depths below 1 mm, while ultrasound and optoacoustic tomography are better suited to a typical depth on the scale of 1 cm and require an additional spacer between the tissue and transducer in order to image the superficial structures at the focus plane. For this work, we developed a hand-held optoacoustic probe, designed for localizing blood vessels from the contact point down to a depth of 1 cm, without the use of a spacer. The probe employs a flat lens-free ultrasound array, enabling a largely depth-independent response down to a depth of 1 cm, at the expense of low elevational resolution. Specifically, while in lens-based probes, the acoustic signals from outside the focal region suffer from distortion, in our probe, only the amplitude of the signal varies with depth, thus leading to an imaging quality that is largely depth-independent in the imaged region. To facilitate miniaturization, dark-field illumination is used, whereby light scattering from the tissue is exploited to homogenize the sensitivity field.

1. Introduction

The non-invasive visualization and localization of blood vessels in the clinic are important for various applications, including guiding transradial access [1,2], intravenous cannulation [3], and the measurement of blood flow [4]. While optical methods can image superficial blood vessels, their contrast and resolution rapidly deteriorate at depths exceeding 1 mm, due to light scattering by the tissues [5,6,7,8]. Additionally, optical images generally lack depth information, which may be useful in applications such as catheterization procedures. Although some depth information may be recovered optically, using diffuse optical tomography (DOT) [5], this comes at the price of slow imaging rates and bulky apparatus, limiting the clinical applications.

Pulse-echo ultrasound offers considerably higher penetration than optical methods, which is conventionally several centimeters, but at the cost of significantly lower contrast between the blood vessel and surrounding tissue [1,6,7]. While the contrast may be improved using Doppler ultrasound [8], which images the flow, Doppler ultrasound is compatible only with relatively large blood vessels and fast flow [4].

In the last decade, there has been increasing interest in the use of optoacoustic tomography (OAT) for imaging blood vessels [6,9]. OAT is a hybrid imaging technique that uses pulse excitation and ultrasound detection to create depth-resolved maps of optical absorption in the tissue, with acoustic resolution and typical penetration depths of 2–3 cm [6,10]. Similar to conventional optical techniques, when the illumination is in the first near-infrared (NIR) imaging window [9,11] at 650 to 950 nm, the main contrast mechanism of OAT in the subcutaneous tissue is blood, making blood vessels the most visually apparent objects on the OAT image.

Numerous configurations of hand-held OAT probes have been demonstrated, which differ in how the illumination is delivered into the tissue and how the resulting acoustic signals are detected [6,12,13]. To achieve optimal image quality in OAT, a full tomographic view is desired, which may be achieved by curved transducers fitted with detection surfaces that surround the imaged region [14,15,16]. In some systems, a large or full tomographic view is achieved by immersing the imaged organ in water [14,15,17], necessitating a stationary imaging apparatus, whereas, in other systems, the water cavity is integrated into the curved ultrasound transducer to enable a handheld device [18,19,20].

Because of the complexity of curved arrays and the integration of the coupling medium, such as water cavities, most hand-held OAT probes rely on linear ultrasound transducers, leading to a lighter design at the expense of a loss of tangential resolution. Similar to pulse-echo ultrasound, the linear arrays in hand-held OAT probes employ an acoustic lens for focusing the probe in the elevational direction [13]. Conventionally, when the probe is in contact with the tissue, the acoustic focusing is achieved at depths from several millimeters to centimeters below the skin, limiting the system’s ability to image more superficial blood vessels (Figure 1a). Numerous configurations have been developed to optimize the imaging range of linear OAT probes. Some of these configurations include optically transparent and acoustically matching spacers to offset the transducer from the tissue, to enable light to reach the area in front of the transducer [6,12,18]. Other designs have used a hybrid beam combiner [12] that enabled the co-axial propagation of the optical illumination and the returning acoustic waves. Although the focus of this work is on localizing isolated blood vessels, it is important to note that linear OAT probes may be used for numerous applications, e.g., the Imagio^® Breast-Imaging System from Seno Medical, which has recently received premarket approval from the Food and Drug Administration [21].

For this work, we developed a new type of hand-held OAT probe for imaging mid-depth (<1 cm) blood vessels, without the need for an acoustic spacer. In contrast to previous designs [12,13], our OAT probe employs an unfocused linear ultrasound transducer to enable imaging from the skin surface without the use of a spacer. In addition, a fiber bundle with two output arms is used, to illuminate the tissue in a dark-field configuration (Figure 1b), thus avoiding the need for a hybrid beam combiner [12,13]. While dark-field illumination is conventionally used in optoacoustic microscopy to reduce undesired signals received from the tissue surface, in this work, this was also used to enable a spacer-free contact probe, relying on light-scattering to provide sufficient illumination to the superficial tissue layers. The use of a direct-contact transducer, in contrast to spacer-based designs, yields three advantages: a more compact probe, wider tomographic coverage, and a stronger acoustic signal, due to a reduction in the propagation distance. In addition, the use of unfocused detection elements did not introduce spatially dependent signal deformation and, therefore, was compatible with the numerically efficient back-projection algorithm [19], which is often used in hand-held OAT probes.

From a clinical perspective, the main advantage of our OAT probe is its small size and monolithic design, which simplifies its handling in manual scanning over the region of interest. This design may also be beneficial for detecting blood vessels during surgical applications in which the miniaturization of the imaging probe is crucial [22]. The main disadvantage of our design is the loss of elevational resolution, which could potentially reduce the image quality. However, since the peripheral blood vessels are relatively parallel to the tissue surface, they lack significant spatial information in the elevational direction and may be imaged without elevational focusing. Effectively, our detection geometry produces a biased image in which tubular structures parallel to the surface are amplified, making it optimal for the detection of blood vessels.

The performance of the OAT is characterized using several types of measurements. First, the optical and acoustic properties of the probe are tested in a scattering-free imaging scenario. Second, tissue-mimicking phantoms are used to evaluate the dependence of signal strength on the depth of tissue. Third, the OAT probe is tested on healthy volunteers to showcase its ability to image peripheral blood vessels at typical depths ranging from 1 mm to 1 cm.

2. Methods

2.1. System Components

The OAT probe (Figure 2) consisted of a custom-made piezoelectric transducer (IMASONIC SAS, 70190 Voray-sur-l’Ognon, France), a two-arm fiber bundle (CeramOptec GmbH, Brühler Straße 30, Bonn, Germany), and a custom-made holder to hold the components together. The transducer had a flat surface and consisted of 64 detection elements with a center frequency of 4 MHz and a bandwidth of 75% at −6 dB. The elements had a pitch of 0.2 mm and a length of 5 mm, yielding a total sensing area of 12.5 mm × 5 mm (Figure 2b), whereas the total transducer surface with the casing was 22 mm × 12 mm. The fiber bundle had two arms, each with 310 fibers, with a numerical aperture of 0.37, which were combined into a single arm at the back end of the system. The fibers had an inclination of 10° with respect to the axial direction of the transducer and their distance from the edge of the sensing surface was 4.7 mm.

The back end of the system consisted of a data acquisition (DAQ) module (PhotoSound Technologies, Inc., 9511 Town Park Drive, Houston, TX, USA) and an OPO laser capable of producing optical pulses with a typical energy of 30 mJ in the wavelength band of 680–980 nm (InnoLas Laser GmbH, Krailling, Germany). The DAQ has 128 channels, with a 40 MHz sampling rate and 12-bit resolution, which included a 40 dB preamplifier and an additional tunable amplifier, with a gain of 6 to 51 dB.

2.2. Image Reconstruction

Image reconstruction was performed using the universal back-projection (BP) algorithm [19], which was implemented on a graphical processing unit (GPU) for real-time reconstruction. The reconstruction was performed over a square grid with a side length of 12 mm, which is approximately the width of the sensing surface (Figure 2b), with a pixel size of 0.0375 mm. Prior to using the BP algorithm, the raw signals were filtered with a bandpass filter of 1–7 MHz to eliminate noise at frequencies outside the transducer’s bandwidth.

3. Results

3.1. Characterization

3.1.1. The Spatially Dependent Impulse Response of the Transducer

To measure the spatial dependency of the transducer’s impulse response, we created an acoustic point source on the tip of a cleaved optical fiber with a core diameter of 50 µm, which was coated with black ink and polydimethylsiloxane. Optical pulses from the OPO laser, with λ = 730 nm, were coupled to the fiber and absorbed at the tip, creating the acoustic point source via the optoacoustic effect. The fiber tip was positioned facing the transducer and was scanned in three directions while recording the acoustic signals measured by the transducer. Figure 3 shows the acoustic response of one of the transducer elements for the different scan positions in the elevational (Figure 3a,b), axial (Figure 3c,e), and lateral (Figure 3e,f) directions. In the elevational direction, the response did not significantly vary within the 5 mm length of the elements, indicating that the transducer effectively performs an integration over the sources in that region. In the axial direction, the signals’ delays varied in their delay, as dictated by the time-of-flight principle, but the shape of the pulses did not vary significantly over the scan region. In the axial direction, the decay was mostly as a result of the $1/r$ decay of acoustic sources, where $r$ denotes the distance from the source, whereas in the lateral dimension, the decay was mostly due to the limited angular sensitivity of the piezoelectric elements. Normalizing the signal to account for the distance resulted in an angular sensitivity with a full width at the half-maximum of $\pm 9$°.

3.1.2. Resolution

Using the same data, axial and lateral resolution were measured as a function of distance from the source (Figure 4). As expected, the axial resolution remains constant because it depends only on the transducer bandwidth and the speed of sound. The lateral resolution, on the other hand, depends on the tomographic view of the object and the number of elements that can see it. Up to a depth of 10 mm, the lateral resolution remains relatively constant and is limited by the ±9° acceptance angle of the individual element. At greater depths, the total angular coverage of the transducer becomes the limiting factor, leading to a reconstruction width that is proportional to the depth. The elevational resolution is approximately equal to the length of the transducer single element, i.e., 5 mm, as can be seen in Figure 3b.

3.1.3. Illumination

The illumination profile as a function of distance from the probe was characterized in air by coupling a white-light source to the back end of the fiber bundle and placing a sheet of paper at a specific set of distances from the front end. The back of the paper was imaged using a CCD camera to reveal the illumination profile (Figure 5). As the propagation distance increases, the two beams are near each other due to the inclination of the fiber-bundle arms and broaden due to the high NA of the fibers.

3.1.4. OAT Probe Imaging of a Wire

The depth-dependent sensitivity of the OAT probe was tested using a thin metal wire to mimic the geometry of a blood vessel. The wire, the diameter of which was 0.5 mm, was surrounded by a black plastic sleeve with a diameter of 1.4 mm. The wire was imaged in two types of media: a transparent, non-scattering medium, water, and a tissue-mimicking solution with a reduced scattering coefficient of $20{ \mathrm{cm}}^{-1}$ and an absorption coefficient of $0.1 {\mathrm{cm}}^{-1}$, produced following the authors of [20]. The measured peak-to-peak signal versus depth for water and the tissue-mimicking solution are shown in Figure 6. In water, the response for distances below 10 mm was low, due to the poor overlap between the beams and the detection region of the ultrasound transducer (Figure 5). The response reached its peak at a distance of 20 mm and then declined, due to the 1/r decay of the acoustic source and the broadening of the beam. In the tissue-mimicking solution, the response was already significant at zero distance and reached its maximum at a distance of 5 mm. At larger distances, a rapid decrease in signal was observed, in comparison to the propagation in water, due to the optical attenuation of light in the scattering medium.

In order to assess the effect of the probe orientation on the appearance of tubular structures in the reconstruction, a 200-micrometer copper wire was positioned in a plane parallel to the transducer’s surface and was imaged for different in-plane orientations. The results, shown in Figure 7, show that when the wire was parallel to the direction of the detection elements, a localized reconstruction was obtained, representing the wire’s cross-section, which is typical for OAT systems with elevational focusing. Rotating the probe in-plane led to the broadening of the reconstruction, with a full 90-degree range leading to a longitudinal image of the wire along its length. The figure clearly shows that the contrast of the wire against the background is much stronger in the cross-sectional image, which is to be expected from the coherent summation of all the acoustic signals in a small region. In the longitudinal image, the same acoustic signals generated a much larger structure, inevitably leading to lower image magnitude and contrast.

3.2. In Vivo Imaging of Peripheral Blood Vessels

The OAT probe was used to image the peripheral blood vessels of healthy volunteers in different regions of the arm. The illumination had a wavelength of 808 nm, a repetition rate of 50 Hz, and a pulse energy of 1.3 mJ, leading to a fluence of $5.8\frac{\mathrm{mJ}}{{\mathrm{cm}}^2}$ and an intensity of 289 $\frac{\mathrm{mW}}{{\mathrm{cm}}^2}$ at the output of the fiber bundle, fulfilling the ANSI safety limit of 32 $\frac{\mathrm{mJ}}{{\mathrm{cm}}^2}$ per pulse and an intensity of 328 $\frac{\mathrm{mW}}{{\mathrm{cm}}^2}$ [23], detailed calculation can be found in Appendix A. The OAT images were reconstructed in real time and were presented on a monitor to provide feedback to the operator. The OAT probe was in direct contact with the skin, where only a minimal amount of ultrasound gel (Aquasonic clear, Parker Laboratories) was used to facilitate the efficient transfer of the acoustic waves to the transducer.

Figure 8 shows the typical images of arteries and veins obtained with the OAT probe. In Figure 8a, the radial artery and four veins are visible. The artery could easily be identified since its pulsation was visible in the real-time reconstructions (see Supplementary Video S1). In all the images, collapsed veins are visible, which are the result of the pressure applied by the OAT probe. In Figure 8c, the radial artery was captured at a depth of 6.5 mm.

In Figure 9, one of the wrist veins was imaged in several orientations of the probe. As in the case of the wire-phantom images of Figure 7, the zero-degree orientation led to a cross-sectional image, while rotation of the transducer led to a broadening of the reconstruction. In the longitudinal image (at a 90-degree orientation), the visualization of the blood vessel was along its length, revealing a structure that is approximately parallel to the transducer surface. Similar to the wire phantom, the cross-sectional image achieved a higher contrast, making it easier to identify the vein in that orientation.

The images in Figure 8 and Figure 9 share several visual features that are common in handheld OAT probes. First, none of the blood vessels had a fully round shape in the cross-sectional images, but instead showed an elongated eye-like shape, which is theoretically to be expected in the case of limited-view tomography [24]. Second, only the surface of the blood vessels is visible, whereas their interiors are hollow. This result may be explained by the high optical attenuation in the blood, with a low detection sensitivity for low acoustic frequencies (<1 MHz), leading to an effective high-pass of the image [24], and the inaccuracy of the back-projection formula, which is known to accentuate borders in the cross-sectional images [25]. Third, the top boundaries of the blood vessels are stronger than the lower ones, which may be explained by the high attenuation of light in biological tissue.

4. Discussion and Conclusions

In this work, we demonstrate a new geometry for a handheld OAT probe that is specifically designed for the visualization and localization of blood vessels. Our geometry is based on an unfocused ultrasound transducer and dark-field illumination, which lead to a compact probe that can be operated in contact with the tissue, i.e., without the use of a spacer. Although the use of unfocused detectors leads to a relatively low elevational resolution of approximately 5 mm, this resolution is shown to be sufficient to image the peripheral blood vessels, which are relatively parallel to the tissue surface. When the transducer is oriented with its elevational axis held in the longitudinal direction of the blood vessel, the cross-section of the blood vessel can be visualized at a quality comparable to systems in which focused transducers are used [13]. Additionally, since, in this orientation, the blood vessels generally lack variations in the elevational direction, neighboring blood vessels may be individually identified (Figure 8a).

The main limitation of our design is the possible smearing in the cross-section of blood vessels that are not parallel to the tissue surface. Nonetheless, when the inclination of the blood vessels is not too high, it may be possible to orient the transducer to be parallel to the blood vessels, either by applying pressure to the tissue or by using larger amounts of ultrasound gel to maintain contact between the tissue and the maneuvered transducer. In the case of a blood vessel with a high inclination, visualization may be still possible along its length, using the orientation applied in Figure 8d and Figure 9c.

The images in this work were reconstructed using the universal back-projection algorithm, which is commonly used in cross-sectional OAT based on focused detectors. Although this algorithm is exact for only a specific set of 3D OAT geometries, it is generally compatible with any detection geometry that obeys the time-of-flight principle. Indeed, as shown in Figure 3e,f, the signal delays detected by the transducer elements are governed by the time-of-flight for in-plane translations. In terms of the elevational direction, the signal summation is performed with the same delay being applied at all elevational values, preventing distortion as long as the imaged object does not change in that direction.

In our work, the optoacoustic excitation was performed using an OPO laser, which is capable of typical pulse energies of 30 mJ. However, the experimental demonstration was performed with an energy of merely 1.3 mJ, and no averaging was used to form the images. Thus, the probe is sufficiently sensitive to operate with cost-effective laser diodes, which have previously been demonstrated with focused ultrasound transducers [26]. We note that the background noise in the in vivo images is not due to the electric noise from the transducer but is rather the result of acoustic signals that did not obey the time-flight principle used in the image formation, e.g., an out-of-plane signal or back-scattering of acoustic waves generated at the tissue surface. Indeed, blood vessels were still visible with energy levels taken down to 0.19 mJ, without applying averaging.