All-reflective ring illumination system for photoacoustic tomography

Abstract. Given that breast cancer is the second leading cause of cancer-related deaths among women in the United States, it is necessary to continue improving the sensitivity and specificity of breast imaging systems that diagnose breast lesions. Photoacoustic (PA) imaging can provide functional information during in vivo studies and can augment the structural information provided by ultrasound (US) imaging. A full-ring, all-reflective, illumination system for photoacoustic tomography (PAT) coupled to a full-ring US receiver is developed and tested. The US/PA tomography system utilizes a cone mirror and conical reflectors to optimize light delivery for PAT imaging and has the potential to image objects that are placed within the ring US transducer. The conical reflector used in this system distributes the laser energy over a circular cross-sectional area, thereby reducing the overall fluence. This, in turn, allows the operator to increase the laser energy achieving better cross-sectional penetration depth. A proof-of-concept design utilizing a single cone mirror and a parabolic reflector is used for imaging cylindrical phantoms with light-absorbing objects. For the given phantoms, it has been shown that there was no restriction in imaging a given targeted cross-sectional area irrespective of vertical depth, demonstrating the potential of mirror-based, ring-illuminated PAT system. In addition, the all-reflective ring illumination method shows a uniform PA signal across the scanned cross-sectional area.


Introduction
The combined ultrasound (US) and photoacoustic (PA) tomographic imaging system described in this paper has broad imaging applicability. With its full-ring US receiver and illumination source, one potential application could be breast imaging. Breast cancer is a common cancer type among women and is a major health concern affecting many lives worldwide. In 2018, it was estimated that the number of newly diagnosed breast cancer cases around the world will be above 2 million. 1 Mammography, magnetic resonance imaging (MRI), and B-mode US are three of the most common imaging modalities used for breast cancer screening, 2,3 and each has its unique shortcomings. Mammography has low sensitivity in detecting breast lesions in women with high-density breast tissue, which is critical since this population is considered to be at a higher risk of developing breast cancer. 3,4 MRI can be used in conjunction with mammography to detect breast tumors in dense breasts. 5,6 However, the availability and cost of MRI imaging restrict the accessibility of this modality. Conventional B-mode US is one of the most widely used medical imaging techniques for screening various types of human tissues, and it is a highsensitivity, non-ionizing, and low-cost tool that can produce images in real time. 7,8 Yet its low specificity in breast screening can lead to unnecessary biopsies. 9,10 Therefore, it is essential to develop new or complementary breast cancer imaging modalities that minimize or eliminate the existing limitations.
Ultrasound tomography (UST), employing a ring-shaped US transducer, has shown promise for breast cancer screening. [11][12][13][14][15] Moreover, PA imaging has demonstrated potential in detecting carcinomas 16,17 and angiogenesis due to tumor growth. 18,19 The addition of photoacoustic tomography (PAT) to the UST imaging can enhance a physician's diagnostic capability by providing functional information about the tissue of interest. Moreover, PA imaging can be easily integrated with UST since the two modalities share the same acquisition hardware. Previous PA/ PAT visualization tools 20-24 either suffer from the use of inefficient illumination methodologies or distort the breast tissue, reducing tissue circulation which may affect the PA results.
A significant challenge for PAT breast imaging is providing sufficient fluence for the desired cross section, while avoiding the maximum permissible exposure (MPE) limit for the tissue. For a full-ring illumination system, expanding the illumination area will help deliver a higher laser energy per pulse to the targeted cross-sectional slice, while keeping fluences below the American National Standards Institute limits. 25 For example, a 10-mm diameter laser beam, with 200-mJ∕pulse energy will result in fluence of 253 mJ∕cm 2 . However, using the coneshaped reflector and the parabolic reflector to create the omnidirectional ring illumination pattern, with a 5-mm thickness and a 10-cm diameter cross-sectional area, will result in a fluence of 12.7 mJ∕cm 2 . In this example, the fluence of the full-ring illumination system is about 20 times lower than that of direct *Address all correspondence to Mohammad Mehrmohammadi, E-mail: mehr@ wayne.edu illumination, and it is below the MPE limit which is 20 mJ∕cm 2 for 532 nm.
The full-ring illumination system has been used in other applications by deploying either a cone mirror or conical lens combined with an acoustically penetrable optical reflector (APOR). [26][27][28] However, the APOR's low power threshold and acoustical transparency make it an impractical solution for full-ring illuminated systems. At the same time, the transmission coefficient of the APOR is affected by the incidence angle of the acoustical signal, 26 so a 45-deg APOR, which should be used in this prototype, will have about 65% US transmittance. Conversely, any design using lenses to create a ring beam 29,30 reduces the wavelengths available for spectroscopic imaging due to possible chromatic aberrations. To overcome earlier limitations, the full-ring US/PA tomography system using omnidirectional optical reflectors is proposed, which can deliver the needed energy for imaging and is vertically translatable to image the entire breast. [31][32][33] The all-reflective, ring illumination PAT system proposed in this paper provides a practical, easily scalable, low fluence imaging system, capable of imaging tissue-mimicking phantoms with significant depth. The three-mirror system presented in this paper is capable of dispersing the light energy over a greater area to reduce the energy fluence, and in the process illuminates more of the tissue which overcomes the light penetration limitations for PAT imaging. In the first two experiments, a ring illumination source was used with a linear US transducer to study the efficiency of full-ring illumination in tissue-mimicking phantoms. In the final experiment, a proof-of-concept PAT imaging prototype consisting of a cone-shaped reflector, a large parabolic reflector, a tunable laser source, and a UST engine equipped with a ring US transducer, was used to image a tissue-mimicking phantom. The data generated using this system are presented with an eye toward laying the groundwork for a future three-mirror system with near-normal tissue illumination and even greater imaging depth.

Development and Validation of Ring-Illumination
Optical System with Linear Array Acquisition The performance of the ring illumination mode was evaluated with regard to the PA imaging depth using two tissue-mimicking phantoms made of 10% porcine gelatin (G2500, SIGMA-ALDRICH, Missouri) and 0.2% cellulose (S5504, SIGMA-ALDRICH, Missouri). For the first experiment, a 700-μm diameter graphite absorber was inserted obliquely into a gelatin phantom, and the PA signal was measured with respect to the horizontal depth [ Fig. 1(a)]. The illumination source for the experiment was a pulsed 532 nm, Nd:YAG, 8-ns laser (Quanta-Ray Pro 270, Spectra-physics, California) with an output of 10 mJ per pulse. A 100-mm diameter parabolic mirror (45-944, Edmund Optics, New Jersey) and cone-shaped reflector (68-791, Edmund Optics, New Jersey) were used to create the ring beam by adjusting the distance between the cone mirror and the parabolic ring reflector [ Fig. 1(b)]. For the second experiment, a 70-mm diameter gelatin phantom containing two graphite absorbers with 500-μm diameters were placed in two planes [ Fig. 1(c)] and imaged to determine the maximum imaging depth and the plane selectivity of the full-ring illumination method. Large scale parabolic reflector (P19-0300, Optiforms Inc., California) with 243-mm diameter was utilized with a cone-shaped reflector to create a large ring-shaped beam with 4-mm thickness. A 532-nm laser source (PhocusCore, Optotek, California) was used, resulting in 4.77 mJ∕cm 2 fluence in the targeted cross-sectional area. In this experiment, it was not possible to adjust the position of the ring beam, so the beam was 17 mm below the targeted cross-sectional area [ Fig. 1

Photoacoustic Tomography Using Full-Ring Illumination and Ring-Shaped Transducer
The full-ring US transducer and illumination system, shown in Fig. 2(a), also used a gelatin phantom for system characterization. In this experiment, an 89 mJ∕pulse laser source (PhocusCore, Optotek, California, USA) was expanded to a ring shape, resulting in 4.7 mJ∕cm 2 of fluence at 532 nm. The ring US transducer has a 200-mm inner diameter and comprised 256 elements, with an element pitch of 2.45 mm and a height of 9 mm. In PAT mode, the scattered signals were recorded by all 256 elements at a sampling rate of 8.33 MHz. The 243-mm diameter parabolic reflector and the cone-shaped reflector were used to create the ring-shaped beam with 4-mm thickness on the phantom surface. The phantom used for this experiment was 7.5 cm in diameter and made of 12% porcine skin gelatin (G2500, SIGMA-ALDRICH, Missouri) mixed with 0.4% cellulose (S5504, SIGMA-ALDRICH, Missouri). Three graphite rods, with 2-mm thickness, were placed horizontally inside the phantom in three layers [ Fig. 2(b)]. The ring beam was adjusted on each cross-sectional slice by translating the phantom in the vertical direction. The results from this experiment demonstrated the uniqueness of the all-reflective PAT system in creating co-registered PA-US tomographic images with significant depth and will be discussed further in the results section.

Results and Discussion
3.1 Photoacoustic Results Using Full-Ring Illumination with a Linear Array Transducer Figure 3 plots the PA image for the diagonal graphite absorber in gelatin for the experimental setup shown in Fig. 1(a). A clear PA signal is detected within the field of view of the linear array US transducer, and a depth of 25 mm inside the phantom is imaged. As anticipated, the PA signal intensity is the largest superficially and is attenuated with horizontal depth [ Fig. 3(a)] due to absorption and scattering. The normalized PA amplitude versus horizontal depth for this image is plotted in Fig. 3(b), demonstrating the depth dependence of the fluence and the relative uniformity of the PA signal within the middle portion of the phantom. The uniform PA signal between 10 and 20 mm underscores the effectiveness of omnidirectional ring illumination in providing a uniform fluence map within a targeted cross section.
In an ideal PA platform, the fluence is depth-independent, and the proposed ring illumination system (Sec. 3.3) can enhance the PAT imaging by providing a more uniform illumination map. The PA signal after 25 mm disappears as the angled optical absorber moves outside of the selected illuminated cross-sectional area. Figure 4 plots the results from the second experiment, which used two graphite absorbers in separate planes, as shown in Fig. 1(c). Since the linear array transducer has a limited field of view, only a section of the horizontal graphite absorber is shown in Fig. 4(a). As anticipated, both objects can be seen in the US image, while in the PA image, the targeted layer shows a high PA signal which demonstrates the selectivity of the full-ring illumination mode in imaging a targeted crosssectional slice. The full-ring illumination is also able to deliver sufficient energy to the core of the 70-mm tissue-mimicking phantom [ Fig. 4(b)]. This finding highlights the benefit of the ring-illumination system which allows for a lower fluence and a greater cross-sectional illumination depth. The limited field of view of the linear transducer used in these experiments would also not be a problem for the full-ring US transducer, where all objects would be visualized regardless of the location.
It is important to mention that there was an observed PA signal from the middle layer of the phantom (at a depth of 30 mm) which happened due to two reasons. First, the phantom was created in two parts by initially pouring the first layer and then allowing it to set so it could support the graphite rod, after which a second layer of the same material was poured to finish  the phantom. The setting process of the first layer and the possible settling of the cellulose scattering material could have created the PA signal seen at the interface. The second reason is related to the incident position of the ring beam on the interface surface between two layers of gelatin phantom which is located at 20 mm above the bottom graphite. Figure 5 shows the UST and PAT images of the full-ring tomography system shown in Fig. 2(b). The UST images are reconstructed using the waveform method, 34 and PAT images are reconstructed using the backprojection method. 35 For the backprojection reconstruction, the RF signals received at each transducer across multiple acquisitions of the same slice are averaged together to increase the signal-to-noise ratio (SNR). In Fig. 5, the top and bottom cross-sectional slices, which are separated by 4 cm of gelatin, are shown. In general, the results demonstrate the vertical depth-independent imaging capability of the full-ring illumination system utilizing a ring US transducer. The normalized PA amplitude as a function of the horizontal depth for the three different cross sections is shown in Fig. 6. As anticipated, all three slices showed high PA signal close to the outer surface, and the signal degraded slightly as the horizontal depth increased. In general, the results once again showed the deposited energy uniformity within the phantom, irrespective of vertical imaging depth.

Large-Scale Omnidirectional Illumination for Full-Ring Photoacoustic Tomography Experiment
To increase the fluence in the central part of the scanned slice, it is important to improve the incident angle of the ring beam and the incident location of the ring beam with respect to the targeted cross-sectional area. In the previous experiment, the ring beam was positioned 10 mm below the targeted cross-sectional slice. Future works will examine the effects of target illumination as a function of beam position below a cross-sectional area and the resulting PA image quality.

Future Work: An Optimal, Adjustable, All-Reflective, Full-Ring Illumination Photoacoustic Tomography System
For the full-ring illumination and US transducer system, the single cone mirror and parabolic reflector imaging system is able to create a ringed beam with a 39-deg angle with respect to the phantom surface. To improve the efficiency of the illumination, it is important to enhance the incident angle of the ring beam. The proposed three mirror system (Fig. 7) will have a tissue incidence angle of 66-deg with respect to the object surface, as opposed to 90-deg or normal incidence, which is anticipated to improve the PAT results. This system consists of a coneshaped reflector and two conical reflectors which can be adjusted independently for scanning the length of a cylindrical object. As shown in Fig. 7, the collimated beam from the laser source is directed normally to the cone-shaped reflector, and upon reflection, a circular beam is directed toward the stationary conical ring reflector. The stationary conical ring reflector then transmits the cylindrical-shaped beam to the mobile conical mirror which focuses the beam onto the targeted slice of the object. The cone-shaped reflector and the first stationary mirror are external to the water tank that houses the second mobile mirror and the ring US transducer.
During the scanning, the mobile conical reflector moves synchronously with the ring US transducer. Since the mobile reflector is placed below the transducer level, the breast tissue coverage will be identical to the UST and imaging areas close to the chest wall is feasible. The size and profile of the breast are defined by the UST images, which will be used to find the optimal position of the mobile reflector and adjust it to match the illumination and the acoustic acquisition planes. The design is verified by using ray tracing, which showed that it could illuminate breast tissue of 140 mm in diameter.

Conclusions
Multiple, proof-of-concept experiments using mirror-based, full-ring illumination systems were presented in this work for PAT imaging. A diagonal graphite absorber in gelatin and planar graphite absorbers were used to demonstrate uniformity of the illumination system, irrespective of vertical imaging depth. These experiments were followed by a full-ring illumination and full-ring US transducer system, which again demonstrated the energy uniformity within the phantom even with a non-optimal 39-deg illumination angle. Last, a proposed three-reflector, ring illumination US/PAT system was presented with a 66-deg illumination angle, which promises exciting future results for medical diagnosis due to its practicality of design and ease of scalability. All of the presented setups achieve a low fluence due to their inherent ring design and will no doubt benefit patients and clinicians in future imaging and diagnostic needs.

Disclosures
The authors have no relevant financial interests in the manuscript and no other potential conflicts of interest to disclose.