2.1.1.

Planar imaging systems

The first planar view breast imaging system, the Twente Photoacoustic Mammoscope (PAM), was proposed by Piras et al.³³ In this system, the patient lies in a prone position with the breast pendant through an aperture in a patient bed. The breast is compressed with a glass plate for laser illumination on the cranial side and the US detector on the caudal side. Figure 1(a) illustrates a schematic of the system. A photograph of the breast against the detector is taken and the thickness after compression in the craniocaudal (CC) plane is measured. A flat array detector with 588 elements in an almost circular configuration is used. The size of each element in the array is $2\times 2{\mathrm{mm}}^2$, with an element pitch of 3.175 mm. The central frequency of the transducer is 1 MHz and the field of view is $9\times 8{\mathrm{cm}}^2$. A Q-switched Nd:YAG laser with 10-ns pulse width, 10-Hz pulse repetition rate, and 1064-nm output wavelength is used as the excitation source. The optical fluence on breast surface is $10{\mathrm{mJ}/\mathrm{cm}}^2$ and the light illumination area is $35{\mathrm{cm}}^2$. Using two eight-channel digitizers ($60\mathrm{MS}/\mathrm{s}$, 12 bit), the authors acquired signals from a sequence of 10 elements at a time, and signals from each element was averaged by 100 times. The imaging time is about 10 min for the complete detector to cover an area of $9\times 8{\mathrm{cm}}^2$ on the breast.³⁵ The system has a spatial resolution of 3 to 4 mm in both lateral and axial directions.^33,35–39 As shown in Fig. 1(a), the imaging depth for this system is measured from the cranial side of the breast in the sagittal plane. The maximum imaging depth that can be achieved by this system is 32 mm.³³ Figure 1(b) shows an overlay of lesion identified by Twente PAM on CC x-ray images, obtained from a conventional x-ray mammogram. The photo and compressed thickness of the breast are used for co-registration of PA and x-ray images in order to identify lesion location accurately. The image shows invasive ductal carcinoma (IDC) in a 69-year-old patient. The three-dimensional (3-D) representation of the lesion volume is shown in Fig. 1(c).

Fig. 1

(a) Twente PAM system. The patient lies in a prone position, with the breast pendant. The laser illuminates from the cranial side and the transducer detects on the caudal side of the breast. (b) Maximum intensity projection (MIP) PA image (color) overlaid on top of the left CC view x-ray mammogram (gray scale) of a 69-year-old patient with a 24-mm IDC (grade 2). (c) 3-D volume representation of the lesion region. This work by Heijblom et al.³⁵ is licensed under a Creative Commons Attribution 4.0 International License.

This system could successfully visualize most malignant tumors with high imaging contrast. The authors concluded that PA contrast at 1064 nm is mostly independent of breast density unlike in x-ray mammography. Because the breast is compressed in the CC plane as in an x-ray mammography, it is relatively easy to co-register Twente PAM and x-ray mammogram results. The use of a planar transducer array enables a uniform 3-D spatial resolution. However, similar to other PA systems, optical attenuation in tissue limits the imaging depth of Twente PAM. Since the laser is illuminated from only one side, the imaging depth of the Twente PAM is limited to 32 mm from the illuminated side of the breast [Fig. 1(a)]. Also, due to the use of single wavelength, the system cannot acquire functional information, such as oxygenated hemoglobin (${\mathrm{HbO}}_2$), deoxygenated hemoglobin (Hb), or relative oxygen saturation levels (${\mathrm{sO}}_2$). Furthermore, activating only 10 elements at a time and the need of signal averaging lead to a relatively slow imaging time for the system.

The dual-scan mammoscope (DSM) proposed by Nyayapathi et al. is a double-sided-view planar imaging system.³⁴ It uses two linear array transducers to scan the breast in the CC plane. The breast is gently compressed between two water tanks, with the patient standing upright. Two 128-element linear transducer arrays are mounted at right angles to the two outputs of a line output fiber bundle. These are held together by 3D-printed holders such that dichroic mirrors (97% transmission at 45-deg illumination) could be placed at a 45-deg angle to both components. This design allows for coplanar light illumination and acoustic detection. A schematic of the system set-up is shown in Fig. 2(a). The central frequency of the transducers is 2.25 MHz, with element pitch of 0.7 mm. A Q-switched Nd:YAG laser with 10-ns pulse width, 10-Hz pulse repetition rate, and 1064-nm output wavelength is used as the excitation source. The laser output is coupled to the circular input of a dual-line output fiber. The optical fluence on the skin surface is measured to be $21{\mathrm{mJ}/\mathrm{cm}}^2$. An imaging region of $8.6\mathrm{cm}\times 10\mathrm{cm}\times 7\mathrm{cm}$ is reported using this system. The latter two dimensions can be extended based on scanning length and breast thickness. The linear array is moved at a speed of $1\mathrm{mm}/\mathrm{s}$, thus scanning over a 6-cm region takes 60 s. For reconstruction, the universal back projection and 3-D focal-line reconstruction⁴⁰ are used. The lateral and elevational resolutions are quantified to be 0.97 and 1.05 mm, respectively. The data acquired from both transducers are combined and aligned to completely visualize the vascular structures in the compressed breast region. This is the first system to report PA imaging of human breast through a field of view of 7-cm deep into the breast,³⁴ as shown in Fig. 2(b). The image is projected in CC view, which is the orientation radiologists are most familiar with. The angiographic features agree with vascular structures from contrast-enhanced MRI. US data are also captured between PA pulses so that PA and US images are naturally co-registered for easy correlation. With double-sided illumination and detection, this planar system has a better depth visualization in comparison to single-sided planar imaging geometries.^{33,35–39,41–43} The imaging depth for this system is measured in the sagittal plane, as shown in Fig. 2(a). The maximum field of view achieved in the sagittal plane is 7 cm from both transducers combined.

Fig. 2

(a) Schematic of DSM. (b) PA image of a volunteer with scattered fibroglandular breast density and thickness after compression of 7 cm. Depth-encoded MIP formed by 3-D focal-line reconstruction algorithm. © 2019 IEEE. Reprinted with permission.³⁴

Similar to the Twente PAM, the DSM system captures data in CC view, enabling easy correlate with x-ray mammograms. Due to double-sided illumination, this system is successful in revealing vascular structures in the middle of the breast, with a combined imaging depth of 7 cm in the sagittal plane [Fig. 2(a)]. The use of a linear transducer array allows for easy implementation of US imaging, which could be useful in locating the suspicious lesion. Because the planar view is achieved by scanning the linear array, the field of view can be customized based on breast size. Limitations of the system include using a single wavelength (1064 nm), which prohibits functional characterization, and the nonuniform spatial resolution, due to the elevation resolution (along the scanning direction) being poorer than the axial or lateral resolutions of the imaging system using a linear array.

2.1.2.

Hemispherical imaging systems

Kruger et al.⁴⁴ proposed the first hemispherical detector array (HDA) for breast imaging. The first-generation HDA consists of 128 elements, 3-mm diameter each, with 5-MHz central frequency. The elements are laid out in a spiral pattern in a hemispherical surface of 100-mm radius of curvature. The rotation to multiple angular positions in a spiral distribution allows for an increase in k-space sampling density while maintaining uniform angular sampling. In the subsequent generations, to increase the field of view, the rotational scanning is replaced by a rectilinear scan of the whole array in a spiral pattern.^45–47 In addition, the number of the array elements is increased to 512, and the central frequency is reduced to 2 MHz. The spiral scan is such that PA data acquired is at equidistant locations.^28,44,48 The advantage of hemispherical detection is that it allows for nearly spatially isotropic resolution in 3D. Figure 3(a) gives a schematic overview of the system geometry. The patient lies in prone position on the patient bed with the breast placed in a breast-holding cup of depth 38 mm placed above the HDA. A small amount of clean water is placed in the breast cup to allow for acoustic coupling between the breast and the breast cup. A Q-switched Alexandrite laser with selectable wavelengths of 755 and 795 nm is used to illuminate conically from the bottom of the cup in a 30-mm radius through a 12-mm diverging lens. The breast is compressed slightly toward the chest wall. Then the HDA scans the breast in a spiral pattern using $XY$ translation stages. Scanning time for a single breast takes about 4 min. The image is reconstructed using the universal back projection algorithm. The imaging depth for this system is measured from the surface of the breast touching the holding cup, radially inward. Toi et al.⁴⁶ quantified the maximum depth by following the continuity of a blood vessel from the subcutaneous layer, deep into the breast. The deepest point from the skin surface is at 27 mm, as marked by the orange arrow in Fig. 3(b).⁴⁶

Fig. 3

(a) Schematic of the hemispherical (HDA) breast imaging system. The breast is placed inside the HDA cup. The whole array is scanned in a spiral pattern along the horizontal plane. (b) PA MIP image of a healthy breast. The orange arrow corresponds to the deepest point from skin surface at 27 mm. This work by Toi et al.⁴⁶ is licensed under a Creative Commons Attribution 4.0 International License.

Since the array is hemispherical, this system eliminates the limited view problem that is prevalent in planar view systems and it provides a spatially isotropic resolution of 0.4 mm. Multiwavelength imaging allows for the quantification of functional characteristics. However, similar to other single-sided illumination systems, this system has a small penetration depth of less than 30 mm. The slow imaging time of 4 min per breast had caused motion artifacts that affected the calculation of ${\mathrm{sO}}_2$ despite applying motion correction processing.^49,50 In addition, the system lacks intrinsic US imaging capability. So far, US imaging could only be achieved using a secondary transducer array, which needs additional image co-registration.⁴⁷

Another hemispherical system implemented by Oraevsky et al.⁵¹ features an arc-shaped transducer that is rotated around the breast. This system, called the laser optoacoustic imaging system assembly (LOUISA-3D), consists of two subsystems—the PA subsystem and the US subsystem. The PA subsystem has a 90-deg arc-shaped array detector with 96 ultrawideband (50 kHz to 6 MHz) transducers. The patient lies on the examination bed with breast placed in a hemispherical plastic cup stabilizer. The complete breast is illuminated with a dual-pulsed Alexandrite laser (50-ns pulses) with toggling wavelengths of 757 and 797 nm, separated by a time delay of 50 or 100 ms. The short time delay is needed in order to accurately co-register the PA images and calculate functional images of total hemoglobin (HbT) and oxygen saturation in blood. The light is delivered using arc-shaped fiber segments in the radial direction toward the focal point of the arc. In order to maintain safe optical fluence of $20{\mathrm{mJ}/\mathrm{cm}}^2$, the system takes up to 10 steps to illuminate an entire breast (surface area $\sim 400{\mathrm{cm}}^2$), with beam area of $40{\mathrm{cm}}^2$. The imaging time per breast is about 10.6 min for a 10-Hz interleaved scan. This accounts for two wavelengths, 10 steps per wavelength for full illumination of the breast, per transducer step for 320 angular positions. A spatial resolution of $\sim 0.3\mathrm{mm}$ can be achieved with this system. The US subsystem consists of another 90-deg arc-shaped transducer array with 192 transducers and 80-mm radius. The central frequency is 7 MHz with a wide bandwidth of $\pm 3.5\mathrm{MHz}$. The 2-D slices of breast anatomy acquired by US subsystem are overlaid with corresponding PA slices from the 3-D images. The maximum imaging depth is 40 mm, measured radially inward [Fig. 4(a).

Fig. 4

(a) LOUISA 3-D schematic and (b) MIP in sagittal view of a 3-D PA image showing small tumorous growth and microvessel-filled nipple. The 3.5-mm tumor is not conclusively ascertained. Therefore, this is used to demonstrate the sensitivity of LOUISA-3D and not its specificity if tumor differentiation. Reproduced with permission.⁵¹

In comparison to the hemispherical system in Fig. 3(a), the LOUISA-3D system achieves better imaging depth because the light illumination is provided at different angles through scanning the arc-shaped fiber. Other benefits of the system include high isotropic spatial resolution, full breast coverage, and dual-wavelength functional imaging. A challenge faced by this system is nonuniform illumination on the breast volume due to arc-shaped light delivery. Park et al.⁵² compensated for this by estimating the nonuniform illumination as a function of polar angle^20,53 and depth-dependent optical attenuation using Beer–Lambert law.⁵⁴ An increase in penetration depth by 67% could be observed in comparison to noncompensated results. However, the acquisition takes over 10 min, which may introduce motion artifacts. Although the system also acquires US data, it uses two different transducer arrays for US and PA imaging, which may further slow the imaging process.

2.1.3.

Cylindrical imaging systems

Li et al.³² proposed a ring-shaped transducer array system named the functional photoacoustic tomography (fPAT). In fPAT, the patient lies in a prone position with the breast pendant through an aperture in the patient bed. The breast is mildly compressed into a cylindrical shape with diameter about 10 cm and thickness $<6\mathrm{cm}$. To achieve this, the ring array is adjusted to fit the breast. The transducer elements are moved radially inward and the polyethylene terephthalate glycol plastic plate is used to compress the breast toward the chest wall. The ring array consists of 64 detectors, arranged in 2 rows of 32 elements (element size $2.3\times 30\mathrm{mm}$), based on a polyvinylidene fluoride film. The $-6\text{-}\mathrm{dB}$ bandwidth of each transducer is from 380 kHz to 1.48 MHz with a maximum frequency response up to 2 MHz. In order to cover the complete breast, the light is delivered from the bottom of the table through a concave lens and ground glass. A large illumination area is achieved by scanning the whole breast in 2-D using light delivery system. The transducers are connected to a 16-channel preamplifier, and a 50-MHz 16-channel data acquisition board via mechanical switching. The diameter of the aperture is adjustable and can be fit to the size of the breast [Fig. 5(a)]. This system consists of a tunable pulsed Ti: sapphire laser, which is optically pumped with a Q-switched Nd:YAG laser. The pulse repetition rate is 10 Hz with a pulse width of 8 to 25 ns. A $13{\mathrm{mJ}/\mathrm{cm}}^2$ energy density is generated by the light beam. The multispectral image reconstruction was performed using a total variation minimization-based finite-element method to achieve deeper penetration depth.⁵⁵ The resolution achieved by this system is 0.5 mm.⁵⁶ The functional characteristics used for fPAT diagnosis are calculated at the tumor sites as shown in Fig. 5(b). The increased hemoglobin concentration is observed for a lesion area of $1.9\times 2.7{\mathrm{cm}}^2$ [Fig. 5(b)]. According to MRI, this region is located at 5 cm in the anterior–posterior dimension and 2.6 cm in CC dimension, and 5.6 cm posterior to the nipple. The lesion dimensions, measured by the full-width at half-maximum of HbT profiles, agree with the MRI report. The functional characteristics for the healthy (right) breast were shown in Fig. 5(c), where the distribution of HbT concentration and oxygen saturation are relatively uniform.

Fig. 5

(a) The top view of fPAT breast interface. Red solid arrows indicate that the diameter of the breast interface is radially adjustable; red dashed arrows indicate the path of the laser beam. (b) An invasive mammary carcinoma with high-grade ductal carcinoma in situ (DCIS) in the left breast. Coronal HbT concentration and ${\mathrm{sO}}_2$ (oxygen saturation) maps of a breast with invasive mammary carcinoma with high-grade DCIS. The arrows indicate the suspicious lesion area. (c) Coronal HbT and ${\mathrm{sO}}_2$ maps for the contralateral healthy breast. © 2015 Am. Assoc. Phys. Med. Reproduced with permission.³²

Compared to hemispherical imaging systems, the ring-array system can capture a cross-sectional image without moving the transducer array with a high spatial resolution of 0.5 mm. The transducer array used in the fPAT system can be adjusted to fit different breast sizes. The use of multiple wavelengths further enables functional quantification of breast tissue. Also the complete breast is scanned in 2D with light delivery system for efficient illumination. However, this system does not capture US data, which may be useful in locating the tumor region. In addition, the transducer array has a 30-mm slice thickness, which essentially limits the sectioning capability of the imaging system. Therefore, the exact location of the tumor in 3D is difficult to retrieve.

Lin et al.³¹ used a 512-element ring-shaped transducer array to achieve cylindrical imaging of the breast. In their single-breath-hold photoacoustic computed tomography (SBH-PACT) system, the patient lies prone on a patient bed with breast suspended into a water tank through a hole in the table. The breast is slightly compressed toward the chest wall and scanned perpendicular to it as shown in Fig. 6(a). The detector is a 512-element 2.25-MHz ring transducer array, with 5-mm element elevation height, 1.35-mm element pitch, and 0.7-mm interelement spacing. Four sets of 128-channel data acquisition systems are used to capture data from all 512 elements. A 1064-nm laser (10-Hz pulse repetition rate and 8- to 12-ns pulse width) light is incident in a donut-shaped beam in order to achieve uniform illumination inside the breast. The entire scan lasts for an SBH ($\sim 10$ to 15 s), which eliminates breathing-induced motion artifacts. The image is reconstructed using the universal back projection algorithm. After 3-D image reconstruction, the elevational resolution is 5.6 mm and the in-plane resolution is 0.258 mm. The maximum imaging depth achieved is 40 mm, measured in the CC plane from the nipple toward the chest wall as shown in Fig. 6(a).

Fig. 6

(a) Schematic of the SBH-PACT system. The breast is scanned using a 512-element ring array transducer. The laser illuminates in a donut-shaped beam from the bottom. (b) Depth-encoded angiogram of a 44-year-old patient with fibroadenoma in the right breast acquired by SBH-PACT. The white dashed circle represents the tumor region. (c) PA MAP image of vessels deeper than the nipple of a 44-year-old female patient with an IDC in the left breast. Vessel density map (red) is overlaid on the MAP image but does not clearly demarcate the tumor region. (d) The PA elastographic image for the relative area change during breathing in blue dashed region in (c) is shown. Tumors are less compliant in comparison to normal breast tissue; therefore, they have smaller detectable relative area change when PA elastography is performed. The tumor could not be identified by SBH-PACT alone, however, elastographic study revealed the tumor marked by the red circle. This work by Lin et al.³¹ is licensed under a Creative Commons Attribution 4.0 International License.

Figure 6(b) shows the depth-encoded angiogram acquired by SBH-PACT. The dashed circle represents the tumor region. Figure 6(c) shows the PA maximum amplitude projection (MAP) image (gray) and the corresponding vessel density map (red) of another breast. In this image, the tumor region is not easily identifiable. In order to improve the sensitivity of the system, the group also performed PA elastography. Since breast tumors are less compliant than normal breast tissue, when compression is applied, the tumor area would deform less than the normal breast tissue. In SBH-PACT, the compression is achieved through patient breathing, which compresses the breast against an agar pillow. After acquiring temporal image frames, the breast deformation is quantified by analyzing the change of area between blood vessels. With PA elastography, the system successfully identified a tumor [Fig. 6(d)] that can hardly be seen in the vessel density map [Fig. 6(c)]. The total time needed by SBH-PACT to simultaneously observe blood vessel density and breast tissue compliance is within $\sim 30\mathrm{s}$.

The SBH-PACT significantly reduces motion artifacts as the imaging itself occurs in a single breath-hold. With the addition of PA elastography, the system sensitivity is significantly improved with a negligible increase in imaging time. Also using a donut-shaped illumination allows for less energy deposition in the nipple and areola regions, which have higher concentrations of pigment. However, the illumination and detection schemes have limited the SBH-PACT’s access to tumors close to the chest wall. Other limitations include the lack of US imaging capability and the use of a single wavelength.

2.2.1.

Linear-view imaging system

Oraevsky et al.⁵⁷ proposed a linear-array-based handheld PA system (Imagio) for functional imaging of breast. The system integrates diffused laser light through fiber bundles placed on either side and an US linear phased array transducer into a handheld probe [Fig. 7(a)]. Functional PA imaging of ${\mathrm{HbO}}_2$ and Hb is achieved using a dual-wavelength laser system made by Seno Medical Instruments. It comprises of a 15-ns pulsed Nd:YAG laser emitting 1064 nm wavelength and a 50-ns pulsed Alexandrite laser emitting 757 nm wavelength. In order to minimize the effect of tissue motion, the functional imaging is co-registered such that there is a 5-ms delay between pulses emitted at two wavelengths. The laser pulses are delivered in cycling pairs with a repetition rate of 5 Hz. The transducer is a 128-element linear array with element width of 0.25 mm and a wide bandwidth 0.1 to 12 MHz. The reconstruction method is a modified filtered back-projection algorithm⁵⁸ that takes account of frequency response, angular response, and spatial impulse response of each transducer in the array. The axial resolution achieved is 0.47 mm for the Alexandrite laser (long-pulse width) and 0.42 mm for the Nd:YAG laser (short pulse width). Lateral resolution is measured to be the highest at the center of field and was 0.81 mm for the Alexandrite and 0.73 mm for the Nd:YAG laser. For objects located at 3 cm or farther from the probe and in close proximity to the probe, the lateral resolution decreased to $\sim 1\mathrm{mm}$ for Alexandrite laser and $\sim 0.92\mathrm{mm}$ for Nd:YAG laser.⁵⁷ At a video frame rate of 10 PA/US images per second, co-registered functional PA and B-mode US images are displayed simultaneously in real time. An example of a B-mode US image from the system is shown in Fig. 7(b), where a hypoechoic solid mass is shown at the center. Figure 7(c) quantifies the relative HbT concentration within and around the tumor region. Dense areas of tumor angiogenesis-related microvasculature and major vasculature are represented in yellow. The ${\mathrm{sO}}_2$ map of the same tumor is shown in Fig. 7(d), where the green color represents the typical ${\mathrm{sO}}_2$ level in normally oxygenated tissues, and the red color represents below normal ${\mathrm{sO}}_2$ level, which is a typical indication of malignancy [Fig. 7(d)]. The authors conclude that, using functional parameters such as HbT concentration and ${\mathrm{sO}}_2$, the specificity of diagnostic US can be enhanced.

Fig. 7

(a) A schematic of the handheld Imagio system. (b) Gray-scale US image of a 2.6-cm irregular intensely hypoechoic solid mass with indistinct margins. The mass is suspicious for malignancy. (c) Overlaid image of US and PA HbT map shows increased internal HbT concentration in more than half of the tumor region and a major vessel above the tumor. (d) Overlaid image of US and PA ${\mathrm{sO}}_2$ map shows diffused internal blood deoxygenation, with numerous radiating parasitized feeding arteries (green) in tissue that surround the mass. The entire mass is identified as predominantly hypoxic. The mass was biopsied later and the histology revealed grade III IDC. This work by Oraevsky et al.⁵⁷ is licensed under a Creative Commons Attribution 4.0 International License.

The Imagio system has a high translational potential, because the system operation is similar to that of a breast US. The use of two wavelengths enables functional characterization of breast tissue, which improves the system’s specificity. However, compared to tomographic imaging systems, the handheld system is operator-dependent and it cannot provide 3-D information of the tumor.

2.2.2.

Spherical array-based imaging system

Dean-Ben et al.²⁹ proposed a spherical-array volumetric handheld scanner for real-time PA imaging in dense breast. A custom-made 256-element 2-D array of detectors is arranged on a spherical surface with radius 40 mm and covering a solid angle of 90 deg. The central frequency of the transducer is 4 MHz with a size of $3\mathrm{mm}\times 3\mathrm{mm}$. A wavelength-tunable optical parametric oscillator laser ranging from 690 to 900 nm emits $<10\mathrm{ns}$ pulses with a pulse repetition rate of 10 Hz. Three wavelengths of 730, 760, and 850 nm are used. The light is delivered to the center of the detection array through a fused silica fiber bundle for coaxial illumination as shown in Fig. 8(a). A transparent polyethylene membrane is used to enclose the active detection surface. To ensure acoustic coupling, the volume in between is filled with water. For data acquisition, 256 parallel analog to digital converters are arranged in 16 acquisition cards with 16 channels each, and a sampling rate of $40\mathrm{MS}/\mathrm{s}$ is employed. The spatial resolution is quantified to be in the range of $200\mu \mathrm{m}$.⁵⁹ The reconstruction is performed using a 3-D model-based procedure for spectral unmixing.⁶⁰ Distribution maps for oxygenated and deoxygenated hemoglobin and melanin are obtained by least-square fitting the reconstructed optical absorption in each image voxel to the known molar extinction coefficient of the corresponding tissue chromophores [Fig. 8(b)]. The angiographic features of the breast are visualized up to a depth of 22 mm.²⁹

Fig. 8

(a) Schematic of the volumetric 3-D handheld scanner. (b) Results from functional PA breast angiography using multiwavelength data acquisition as it simultaneously resolves major tissue chromophores in the breast of a 40-year-old volunteer in 3D, in real time. The breast surface is located in the upper part of the cubes. © 2015 by Wiley-VCH. Reproduced with permission.²⁹

The spherical transducer array allows acquisition of 3-D volumetric images in real time with a single laser shot and it minimizes the limited-view problem. The fast wavelength-switching laser enables multispectral imaging at a high speed. However, the system is also operator dependent and the imaging depth is only about 22 mm, which may pose a problem in locating deep-seated lesions.

2.2.3.

Curved array-based imaging system

Diot et al.³⁰ proposed a curved-array handheld multispectral optoacoustic tomography (MSOT) system with 28 wavelengths. The 256-element transducer array has a diameter 120 mm and central frequency of 5 MHz. The transducer elements are arranged in an arc spanning 174 deg.⁶¹ The transducer is enclosed in a transparent low-density polyethylene membrane, with the cavity between the arc and membrane filled with heavy water (${\mathrm{D}}_2\mathrm{O}$) for acoustic coupling as shown in Fig. 9(a). ${\mathrm{D}}_2\mathrm{O}$ is used because it absorbs less light than water in the used wavelength range.⁶² A tunable pulsed laser with a working range of 680 to 980 nm, pulse length of 8 ns, and repetition rate of 50 Hz is coupled with a custom made line output ($40\times 1{\mathrm{mm}}^2$) fiber. For data acquisition, a custom-built analog-to-digital converter is used at a sampling rate of $40\mathrm{MS}/\mathrm{s}$ and 12-bit digital resolution. A delay and sum reconstruction algorithm is used for real-time viewing. The patient is scanned in a supine position. The tumor is first located using conventional US, and then the area is scanned by handheld MSOT. During the scan, slight pressure is applied to reduce tumor depth. For each cross-sectional slice, 28 frames are collected at wavelengths ranging from 700 to 970 nm in 10 nm steps. Each multiwavelength slice is acquired in $\sim 0.56\mathrm{s}$, with a total examination time of 2 to 4 min. MSOT can achieve a high spatial resolution of $\sim 250\mu \mathrm{m}$,⁶³ but it is limited to a penetration depth of 2 to 3 cm due to light attenuation.⁶² After spectral unmixing is applied to all 28 wavelengths, four images are produced representing the absorption of ${\mathrm{HbO}}_2$, Hb, lipid, and water. Figure 9(b) shows the composite image of all four absorbers in breast tissue. The high spectral resolution of MSOT due to multiple wavelengths allowed calculating the total blood volume (TBV) gradient, which measures the spatial heterogeneity from periphery to the center of the tumor. Using TBV gradient as the metric, authors observed fivefold variability in between tumors. The authors suggested that TBV gradient could be used to differentiate various tumor subtypes.

Fig. 9

(a) A schematic of the handheld MSOT probe. (b) Results show a composite image of all four absorbers (Hb, ${\mathrm{HbO}}_2$, lipid, and water), revealing the layered structure of the breast. (Yellow, skin; pink, lipid; light blue, mammary tissue; green, connective tissue known as Cooper ligament). © 2017 AACR. Reprinted with permission.³⁰

The handheld MSOT provides information about physiological features of cancer. With the use of 28 wavelengths and generation of four functional maps, this system has a unique capability to analyze blood oxygenation levels, blood volume, lipid, and water behavior in the tumor region with improved quantitative accuracy. The authors state that TBV gradient characteristics could be used to differentiate tumor subtypes. However, being a handheld device, the MSOT is operator-dependent and provides only 2-D tumor information. In addition, prior to imaging with handheld MSOT, the tumor region has to be located using a conventional US system.