2.1.

Phantom Preparation

Cylindrical phantoms were made out of polymethyl methacrylate (PMMA) tubes ($\text{length}=40\mathrm{mm}$, $\text{inner}Ø=14\mathrm{mm}$), filled with gelatin (300 g Bloom, FormX, Amsterdam, the Netherlands). An acoustic window was realized on both sides of the tube to allow PA and PDUS measurements (Fig. 1).

Fig. 1

Perfusion phantoms: (a) the cylindrical perfusion phantom molds, consisting of a cylindrical PMMA tube with a cut-away to enable acoustic measurements. (b) Needles are used to create channels in the gelatin. The mold is placed within a teflon holder and filled with gelatin.(c) After solidifying, the tube is taken out of the teflon holder and the needles are removed. On the inner wall of the tubes, a screw thread was tapped to prevent leakage and sliding of the gelatin during perfusion.

A 10% mass/volume solution of gelatin was heated to 45°C for 10 min. Under continuous stirring, $3\times {10}^{-3}$ $\text{volume}/\text{volume}$ ($\mathrm{v}/\mathrm{v}$)% black Indian ink (Royal Talens, The Netherlands) and $14\mathrm{v}/\mathrm{v}\%$ Intralipid^® fat emulsion (Fresenius Kabi, Bad Homburg, Germany) were added to the gelatin solution to tune the optical properties to physiological values of muscle tissue (${\mu }_a=0.25{\mathrm{cm}}^{-1}$ and ${\mu }_s^{\prime }=7{\mathrm{cm}}^{-1}$ at $\lambda =800\mathrm{nm}$).^23,24 The optical absorption of the used ink batch was verified with a plate reader (SynergyHT, Biotek, Winooski). Finally, Orgasol (ELF Atochem, Paris, France) was added (1 percent by weight) for acoustic scattering.

The PMMA tubes were placed in teflon holders and filled with gelatin. Needles ($Ø=180\mu \mathrm{m}$, Seirin J-Type, SEIRIN-America, Weymouth) were inserted into the gelatin to create microchannels upon removal, after solidification of the gelatin. The alignment and separation of the needles were ensured by placing them into a teflon disc with drilled grids. After 24 h of cooling at 4°C, the tubes were taken out of the holders and the needles were removed from the gelatin. Four different phantoms were made with an increasing amount of channels, resulting in a total perfused cross-sectional area of 1.3%, 2.7%, 4.0%, and 5.4% respectively, which is in the physiological range of human skeletal muscle.^25,26

2.2.

Probes

PA imaging was performed using the second generation FULLPHASE prototype PA probe, controlled by a MyLab One scanner (ESAOTE Europe, Maastricht, The Netherlands). The integrated diode laser system (QUANTEL, France, OSRAM, Germany), consisting of a stack of highly efficient diode arrays in the probe, operates at $\lambda =805\mathrm{nm}$ with 1-mJ pulses. A customized laser driver (BRIGHTLOOP, France) allows a pulse width of 130 ns at half maximum and a maximum pulse repetition rate of 10 kHz. Using a combination of cylindrical lenses and diffractive optical elements (SILIOS, France), the beam is collimated and homogenized in the axis perpendicular to the diode arrays and reshaped into a rectangular form. The laser beam exits the probe through a small glass window. A linear array transducer (SL3323, $f_c=7.5\mathrm{MHz}$, ESAOTE Europe, The Netherlands), next to the glass window (Fig. 2), constitutes the US transducer of the probe. The spatial resolution of the system is 0.28 mm axially and around 0.5 mm laterally, depending on the imaging depth. Further operation and specification details of the FULLPHASE probe can be found in the paper by Daoudi et al.²⁰

Fig. 2

Schematic of the imaging setup used in the phantom study. The image shows the PA probe, placed orthogonally above the phantom to obtain cross-sectional images.

PDUS imaging was performed using an LA523 probe in combination with a MyLab 70 scanner (ESAOTE Europe, Maastricht, The Netherlands). The element length (5 mm), pitch (0.245 mm), and center frequency (7.5 MHz) of this probe are similar to the US part of the FULLPHASE probe. For US measurements on vasculature, a high-frequency linear array probe was used (LA435, $f_c=12\mathrm{MHz}$, $\text{pitch}=0.2\mathrm{mm}$, ESAOTE Europe, The Netherlands) combined with the MyLab 70 scanner.

2.3.

Imaging Setup for Phantom Measurements

The phantoms were connected to a syringe pump (STC-521, Terumo Europe, Leuven, Belgium) using silicone tubing and perfused with an ink solution (Fig. 2). The optical absorption of the ink solution (0.013% $\mathrm{v}/\mathrm{v}$) resembled that of diluted blood²⁷ at 805 nm ($\text{hematocrit}=5\%$, ${\mu }_a=1.0{\mathrm{cm}}^{-1}$). The pump flow rate was adjusted to establish an average flow velocity of $20\mathrm{mm}/\mathrm{s}$ in all phantoms. A customized probe holder was used to place the PA and PDUS probe orthogonally above the phantoms. PA and PDUS imaging were performed during perfusion while the tubing, phantom, and probe tips were submerged in water at room temperature. Additionally, a motorized translation stage (M-403.2DG, Physik Instrumente, Germany) was used to move the PA probe horizontally over the phantom at a constant velocity during scanning, to allow for 3-D PA visualization, using the consecutive cross-sectional images at different locations to reconstruct a volume data set.

2.4.

Scanning Protocol for Phantom Measurements

Cross-sectional images of each phantom at the center of the acoustic window were acquired, first with PDUS and then with PA. The distance between the US detector surface and the phantom surface was kept at 5 mm. In PDUS, the US frequency was set to 6.3 MHz and the pulse repetition frequency of the PDUS mode to 1.5 kHz, with the lowest possible wall filter setting. Data were acquired for 10 s at a frame rate of 33 Hz. In PA imaging, 100 frames were acquired at a frame rate of 40 Hz. Additionally, a PA scan was made over the whole length of the acoustic window (20 mm) using the motorized stage, at a constant velocity of $1.2\mathrm{mm}/\mathrm{s}$ and a frame rate of 40 Hz.

2.5.

Data Processing and Analysis

2.5.1.

Photoacoustic data

Raw radio frequency (RF) data, stored at a sample frequency of 50 MHz, were reconstructed offline using MATLAB^® (MATLAB^® 2014b, MathWorks, Natick, Massachusetts). Multiple filtering steps were performed, i.e., DC blocking, moving average filtering ($N=19$;), and bandpass filtering (0.8 to 5.5 MHz). The filter cutoff frequencies are based on a spectral analysis of the raw PA signal from superficial tissue and vasculature, obtained with this system, which typically showed strong signal contribution around 3.5 MHz. Next, the beam-formed RF data were reconstructed with the delay-and-sum method. To improve image quality, the individual frames in the beam-formed data were deconvolved with the Richardson–Lucy algorithm using the system’s point spread function (PSF), measured on a cylindrical black object ($d=100\mu \mathrm{m}$) at a depth of 10 mm. Forty consecutive frames were averaged, and the intensity was thresholded at $-23\mathrm{dB}$. This threshold is based on the SNR of the signal from the channel closest to the probe (in the upper part of the image), compared to the mean intensity value of the noise in the lower part of the image, where no channel signal is visible. An overview of the signal processing sequence is given in Fig. 3. In the resulting image, the visible cross-sectional area of the phantom was considered as the region of interest (ROI). The perfused cross-sectional area was calculated by dividing the amount of pixels in the ROI with an intensity level above the threshold by the total number of pixels in the masked area. Because of the limited penetration depth of the PA imaging ($<8\mathrm{mm}$), only the upper half of the phantoms was taken into account. A visualization of the reconstructed 3-D volume based on the PA scanning was realized using MATLAB^® after stacking all acquired 2-D frames and thresholding the data at noise level.

Fig. 3

Block diagram of the signal processing, for both PA and PDUS.

2.6.

Scanning Setup and Protocol for In Vivo Photoacoustic Measurements

To test the feasibility of visualizing skin and superficial blood vessel geometries using the FULLPHASE PA probe, a small area on the ball of the hand of a volunteer was imaged. Also, in four different volunteers, the skin on the inside of the wrist was imaged. The volunteer study was approved by the research ethics committee of the Radboud University Nijmegen Medical Centre (Nijmegen, The Netherlands), and requirements for informed consent were waived.

During scanning, the forearm of the volunteer and the tip of the probe were submerged in a water tank. To avoid vasoconstriction, the water was kept at a temperature of approximately 30°C. The probe was mounted in a custom-made probe holder and attached to the aforementioned translation stage [Fig. 4(a)]. After positioning the probe at a distance of 5 mm from the skin, areas of approximately $25\times 15\mathrm{mm}$ (image width) were imaged at a frame rate of 40 Hz, while again moving the probe at a constant velocity of $1.2\mathrm{mm}/\mathrm{s}$ using the translation stage. The scanning method including a water tank and a translation stage [Fig. 4(b)] is related to the one described in the work by Fronheiser et al.²⁸ The scan area at the wrist was demarcated by water resistant band-aids, provided with ink lining [Fig. 4(c)]. This functioned as a marker in both PA and US imaging.

Fig. 4

PA setup for skin imaging. (a) The PA probe is mounted into a customized probe holder (I) and attached to the translation stage (II). (b) The water tank for in vivo scanning, showing the fixed handgrip for support. (c) The skin areas at the wrist are demarcated with band-aid and ink.

2.7.

Visualization of In Vivo Photoacoustic Measurements

Data were filtered and reconstructed offline as described in the phantom study. The 2-D frames were stacked and intensities were thresholded at noise level ($-23\mathrm{dB}$) before 3-D renderings were reconstructed for visualization. For visualization purposes, the epidermis and underlying blood vessels were represented with a different color. Segmentation in each 2-D frame was performed automatically using the pronounced curved signal of the epidermis in the images.

2.8.

Validation of In Vivo Photoacoustic Measurements with Ultrasound

To validate the PA blood vessel visualization in the wrist, the vessel geometry of the same area was also obtained with high-resolution US. In the same way as with PA, cross-sectional B-mode images were acquired at a frame rate of 40 Hz while moving the US probe over the skin. After transferring the DICOM data to an external PC, the skin contour and individual vessels were manually segmented in selected frames, using MATLAB^® software. By interpolating the manual segmentations over the whole data set, the contours of the vessels were reconstructed and visualized. The 3-D vascular geometries obtained in both PA and US were compared.

3.1.

Phantom Measurements

Figure 5 shows the 3-D rendered PA images of the four different phantoms. The signal originating from the upper edge of the phantom is clearly visible on top of the signals from the channels. Due to a limited depth of light penetration, only part of the channels is visible. The increase in the amount of channels can be appreciated from these images.

Fig. 5

3-D visualization of the PA scans of the phantoms. The four different phantoms are shown, with increasing perfused area from left to right. The yellow plane indicates the location of the cross-sectional images on which the analysis is performed.

Figure 6 shows raw and processed PA and PDUS cross-sectional images of the four phantoms. In the raw images, the PDUS signal is represented by the red values. In the most perfused phantom (5.4%), the PDUS signal of several channels seems to overlap due to the limited resolution, creating an overshoot of signal. In PDUS, the whole phantom is clearly visible, showing the difference in imaging depth between PDUS and PA.

Fig. 6

Cross-sectional PA and PDUS images of the phantoms with increasing perfused area (I to IV). For each modality, the raw total images and the filtered signal in the perfused area are shown. In the raw PDUS images, the dashed squares indicate the PDUS region of interest.

The measured amount of signal in each phantom is shown in Fig. 7, for both PA and PDUS. In PA, the relative amount of pixels that exceeded the intensity threshold seems to be proportional to the real perfused area in the phantoms. The results are validated with PDUS, where the same correlation is found. However, the visual overshoot of PDUS signal that was encountered in the cross-sectional images of the most perfused phantom is also found in the quantitative analysis.

Fig. 7

The relative amount of signal in (a) PA imaging and (b) PD, proportional to the relative amount of perfused cross-sectional area in the phantoms. The signal in PDUS shows an overshoot in the most perfused phantom. Error bars indicate one standard deviation, and the dashed line represents the identity line ($y=x$).

3.2.

In Vivo Measurements

An example of PA scans in the human hand is shown in Fig. 8. A typical cross-sectional image of these areas reveals a relatively strong, curved signal from melanin in the epidermis on top of signals from (sub) millimeter size blood vessels underneath. Another, low intensity, curved signal at the location of the deep vascular plexus is observed approximately 1 to 1.5 mm below the epidermis. The achieved imaging depth is around 5 mm. 3-D visualization shows the geometry of the vessels while segmentation offers the possibility of discriminating between epidermis and blood vessels. Total acquisition time for these areas of $3{\mathrm{cm}}^2$ was 17 s.

Fig. 8

PA scanning of a small skin area on the hand. (a) Measurement area on the thumb ($1.5\times 2.0\mathrm{cm}$) and (b) a typical cross-sectional image that is obtained during this scan. The image shows the clear signal from the epidermis and underlying blood vessels. (c) 3-D volumes of the scanned area are obtained by combining and segmenting all cross-sectional 2-D images.

Figure 9 shows the validation of in vivo PA vasculature imaging with high-frequency US imaging in four volunteers. The image width of US was 37 and 15 mm in PA. In the third volunteer, two PA scans on adjacent skin areas were combined. In the second volunteer, only one half of the scanned area could be used because of technical issues. In the acquired 2-D US images, the skin layer contour and the (sub) millimeter vasculature in the hypodermis could easily be observed up to a depth of 1 cm. In the 3-D renderings, white striped areas indicate the marking band-aid that was clearly visible in both modalities. The visualization of skin and vasculature shows the resemblance of the imaged vasculature geometries in both PA and US in terms of number of vessels and vessel size. The images also show that a higher resolution is obtained with US in this setup.

Fig. 9

3-D visualization of PA and US acquisition of subdermal vasculature in the wrist of four volunteers. Point of view is inside the arm. Skin markers are represented white striped while the dotted lines in the US images indicate the corresponding area in the PA scan