**3. Results**

#### *3.1. Performance Test of the 3D-PACT*

#### 3.1.1. Feasibility Verification of Quartz Bowl in the System

First of all, the photoacoustic experiments were carried out in the ultrasonic transmission mode and ultrasonic reflection mode to verify the feasibility of the quartz bowl. The schematic diagrams of the two experimental devices are shown in Figure 4a,b, respectively. Black tape was selected as the target due to the strong laser absorption. The energy at the fiber exit was around 100 μJ. The energy E in Table 1 refers to the laser energy of the imaging target surface.

**Figure 4.** (**a**) Schematic diagram of photoacoustic imaging device in ultrasonic transmission mode. (**b**) Schematic diagram of photoacoustic imaging device in ultrasonic reflection mode.

The ultrasonic reflection efficiency *H* of the quartz bowl is defined as the ratio of the photoacoustic signal amplitude in the above two signal acceptance modes in the research. The specific expression is:

$$H = \mathcal{G}\_{ref} / \mathcal{G}\_{trm} \tag{6}$$

where *Gref* is the signal amplitude per microjoule laser energy in the ultrasonic reflection mode and *Gtrm* is the signal amplitude per microjoule laser energy in the ultrasonic transmission mode. Furthermore, *Gtrm* was about 0.01232 V/μJ, found by analyzing the photoacoustic signal obtained from photoacoustic experiment in ultrasonic transmission mode. The quartz bowl was divided into eight areas (Q1–Q8), and the quartz bowl's ultrasonic reflection efficiency was tested after subtracting the light attenuation of each area.


**Table 1.** The ultrasonic reflection efficiency of the quartz bowl.

The ultrasonic reflection efficiency of quartz bowl was between 198.70% and 204.71%, which could be seen by noting that the amplitude of the photoacoustic signal was nearly doubled in ultrasonic reflection mode. Experiments showed that the quartz bowl used in the system had a higher efficiency of ultrasonic signal reflection, which was helpful for the collection of photoacoustic signals in PACT.

#### 3.1.2. Resolution Test of the System

In the resolution test experiment, a black sphere with a diameter of 0.3 mm in 2% agar was used as an imaging target and photoacoustic signal source, and was placed in the center of the circular illumination zone, as shown in Figure 5a. Because of its all-black color, the sphere was a good absorber at the laser wavelength of 532 nm used in the system. In the process of data acquisition, the sampling rate was 50 MHz, the number of sampling points in a B-scan was 180 and the total laser energy reaching the surface of the phantom was about 3.2 mJ (the optical fluence of each bundle was 0.7 mJ/mm2). During the experiment, the imaging target, quartz bowl, and the ultrasonic transducer were all immersed in distilled water to achieve acoustic coupling. The reconstructed photoacoustic image is shown in Figure 5b. It should be noted that the imaging resolution in this research was defined as the full width at half maximum of the normalized absorption intensity at the centerline of photoacoustic image Figure 5b. As shown in Figure 5c, the imaging resolution of the system was 0.31 ± 0.02 mm based on quantitative calculations. In addition, according to the raw data analysis of the photoacoustic signal shown in Figure 5d, it can be seen that the intensity of the photoacoustic signal of the imaging target was relatively uniform at the cross section, and there was no phenomenon showing that the photoacoustic signal was submerged. These also reflected the ability of the system to maintain the photoacoustic coplanar property.

**Figure 5.** (**a**) Phantom of resolution test, (**b**) reconstructed PA image of target, (**c**) the profile along the dotted line in (**b**), and (**d**) raw data of a B-scan.

#### 3.1.3. The Superiority of Photoacoustic Coplanar Configuration and Virtual Point Detection

In the photoacoustic tomography system proposed herein, the narrow sound beam of dual-foci virtual point ultrasonic transducer and the annular illumination beam were coupled to form a photoacoustic coplanar mode based on the quartz bowl with the characteristics of laser transmission and ultrasonic reflection. To prove the advantages of the virtual-point ultrasonic detection and photoacoustic coplanar mode, a comparative experiment was performed using a conventional unfocused ultrasonic transducer (V310, Olympus, Tokyo, Japan). The results of the comparative experiment are shown in Figure 6. In the experiment, the black tape was evenly wrapped around a steel column with a diameter of 25 mm as an imaging phantom, as shown in Figure 6a. The number of sampling points per imaging section was 90, and it took about 9 s to complete a photoacoustic data acquisition of a cross section. Figure 6c,d showed the reconstruction images from the dual-foci virtual point ultrasonic transducer and unfocused ultrasonic transducer, respectively. From a qualitative point of view, it is obvious that the quality of the reconstructed image with the dual-foci transducer was better than its counterpart in terms of qualities such as shape-similarity and image-purity.

**Figure 6.** (**a**) Phantom of the comparative experiment. (**b**) Reference standard image by K-Wave simulation. Reconstructed PA images obtained with (**c**) the dual-foci transducer and (**d**) the traditional unfocused transducer. (**e**) A-Line plots from dual-foci transducer and traditional unfocused transducer. (**f**) Absorption intensity distribution at the centerline of (**<sup>a</sup>**–**<sup>c</sup>**).

In order to quantitatively analyze the reconstruction quality of Figure 6c,d, the simulation result of the imaging target (shown in Figure 6b) was used as a reference standard. The photoacoustic raw data of Figure 6b was obtained using a K-wave simulation, and then reconstructed using the improved back projection algorithm mentioned above. The performance indicators, such as MSE (mean squared

error), PSNR (peak signal to noise ratio), and SSIM (structural similarity) [40–42] were calculated separately according to the following formulae. The MSE index is given as:

$$MSE = \frac{1}{mn} \sum\_{i=1}^{m} \sum\_{j=1}^{n} \left\| I(i,j) - K(i,j) \right\|^2 \tag{7}$$

The *I*(*i*, *j*) and *K*(*i*, *j*) in the formula represent the reconstructed image and the reference standard image, which refer to the reconstructed photoacoustic image and the K-Wave simulation image in this paper, and the parameters *m* and *n* represent the length and width of the image.

*PSNR* is defined as:

$$PSNR = 20\*\log\_{10}(\frac{MAX\_1}{\sqrt{MSE}}) \tag{8}$$

where *MAX*1 is the gray level of the image. Lastly, the *SSIM* is given as:

$$SSIM(\mathbf{x}, y) = \frac{(2\mu\_x \mu\_y + c\_1)(2\sigma\_{xy} + c\_2)}{(\mu\_x^2 + \mu\_y^2 + c\_1)(\sigma\_x^2 + \sigma\_y^2 + c\_2)}\tag{9}$$

where *x* and *y* represent the two images for comparison, μ*x* and μ*y* are the mean of the image, σ*x* and <sup>σ</sup>*y* represent the variance of image, <sup>σ</sup>*xy* is the covariance of *x* and *y*, *c*1 = (*k*1*L*)<sup>2</sup> and *c*2 = (*k*2*L*)<sup>2</sup> are two constants used to avoid division by zero in the formula, *L* = 2*B*–1 is the range of pixel values, and *k*1 = 0.01, *k*2 = 0.03 are default values.

The results of the above three performance indexes are shown in the following Table 2.

**Table 2.** Quality evaluation of reconstructed images.


It can be seen form Table 2 that the performance indicators *SSIM* and *PSNR* of the reconstructed images were greatly improved in the case of virtual-point ultrasound detection, while the value of *MSE* was obviously reduced. The three performance indicators of the reconstructed photoacoustic image under the dual-foci virtual point transducer were better than those under the unfocused transducer that verified the reliability of the proposed system.

The *PSNR* of the photoacoustic signal from the dual-foci virtual point ultrasound transducer was 35.9379 dB, which was much larger than that of the conventional unfocused transducer (15.2504 dB), indicating that the dual-foci virtual point transducer had a higher detection sensitivity. Moreover, the high detection sensitivity of the photoacoustic tomography system is further illustrated in Figure 6e, in which the blue and red A-line plots corresponding to the raw photoacoustic signal with the dual-foci virtual-point ultrasonic transducer and the traditional unfocused ultrasonic transducer are depicted, respectively. The two A-line plots were randomly obtained under the premise of the same scanning position and sampling points. It can be seen that the photoacoustic signal from the dual-foci ultrasonic transducer had an amplitude increase of nearly 2.5-fold, which greatly improved the signal to noise ratio of the photoacoustic image. Figure 6f shows the absorption intensity at the position of the centerline of the reconstructed image, and it can be found that the absorption intensity of the center line position of the photoacoustic image obtained by the unfocused ultrasonic transducer was weak (corresponding to the black dotted line in the Figure 6f), and there are signs that the signal was flooded by noise.

In addition, the virtual-point ultrasonic transducer has a large receiving angle, which makes the photoacoustic tomography under sparse sampling possible. In the experiment mentioned above, although the number of sampling points per cross section was 90, the system with virtual-point detection could still perform high quality photoacoustic tomography. However, there were some artifacts in the photoacoustic image under the traditional unfocused transducer. The experiment proved that the system proposed in this paper can improve the imaging speed and reduce the burden of data acquisition device using sparse sampling. The comparative experiments analyzed the photoacoustic signal acquisition mode based on photoacoustic coplanar structure and virtual-point ultrasound detection from a quantitative point of view.

#### 3.1.4. Verification Experiment on Large-Scale Imaging

In order to verify the large-scale imaging characteristics of the system, a photoacoustic tomography experiment was performed using a phantom (a hollow cylinder wrapped with black tape, the diameter is 52 mm), as shown in Figure 7a. Figure 7b shows a photoacoustic image of a cross section. It can be seen that the imaging range of the system could be up to 52 mm in diameter.

**Figure 7.** (**a**) Phantom of imaging range test, (**b**) Reconstructed PA image.

To show the 3D photoacoustic imaging in a wide area, four black polyethylene tubes with large spatial distances were designed as imaging targets, as shown in Figure 8a. Figure 8b is a reconstructed photoacoustic image of the scanning section indicated by the dotted line in the real figure. Figure 8c is one of an A-line signal of the B-scan, in which the four signal peaks correspond to the four polyethylene tubes respectively, and the distance between the signal peaks matches the distance between actual tubes. Figure 8d is a 3D photoacoustic image of the imaging target, and the quantized spatial distance is consistent with those shown in Figure 8a,b. It can be seen that the 3D photoacoustic image could more accurately reflect the information of the phantom, such as shape, size, deformation, and so on. According to the photoacoustic images of this test experiment, it can be concluded that even if there was a large space distance, the target shown in Figure 8a could still be clearly imaged.

**Figure 8.** (**a**) Phantom of confirmatory experiment on scope, (**b**) reconstructed PA image of target, (**c**) one of an A-line of (**b**), and (**d**) 3D photoacoustic image.

#### *3.2. Photoacoustic Experiments of Di*ff*erent Kinds of Phantoms*

After completing the test experiment of resolution and the verification experiment of a large imaging range, a 3D photoacoustic imaging experiment of the vascular phantom was performed to further verify the feasibility of the imaging system. Two black wires with a diameter of 1 mm were used to make a phantom that simulates blood vessels in tissue, as shown in Figure 9a. In this experiment, photoacoustic signals from 180 positions were acquired on each imaging section, that is, a photoacoustic signal was collected for every 2◦ of rotation of the ultrasonic probe; it took about 18 s to acquire the information of a B-scan. The step size of the vertical movement of the phantom was 0.2 mm, and 120 B-scans were scanned during the experiment. Figure 9b shows the photoacoustic images of three B-scans and the corresponding raw data. It can be seen that the system can accurately reflect the shape and size of the phantom. Figure 9c shows the three-dimensional photoacoustic image of the phantom, which not only clearly identifies the different parts of the phantom, but also reflects the change of it in the vertical direction, that is, the reconstructed result is highly consistent with the blood vessel phantom. This result indicates that the 3D-PACT is capable of 3D imaging small objects of different spatial distribution and orientation in high quality.

**Figure 9.** (**a**) Photograph of a blood vessel phantom, (**b**) PA images and A-lines of several B-scans, and (**c**) 3D photoacoustic image.

Subsequently, a 3D photoacoustic experiment was performed on a tumor phantom obtained by soaking the plasticine, whose dimensions are shown in Figure 10a, in black ink. In this experiment, a total of 90 images were obtained by scanning the phantom with a vertical interval of 0.1 mm, and the energy of the pulsed laser reaching the phantom surface was approximately 3.6 mJ. It can be seen that the laser could penetrate the imaging phantom and the laser absorption characteristics of each B-scan, and each point on it could be well-reflected from both two-dimensional and three-dimensional photoacoustic images demonstrated in Figure 10b,c separately. By comparing Figure 10a,c, it can be seen that the size of the absorption region on the photoacoustic image was substantially the same as the actual size of the corresponding section, which verified the feasibility and reliability of this system by showing that the shape and size of tumor analogs can be well-reflected through 3D photoacoustic imaging.

**Figure 10.** (**a**) Photograph of a tumor phantom, (**b**) PA image of a B-scan, and (**c**) three-dimensional photoacoustic image.

Finally, in order to verify the photoacoustic imaging of some irregularly shaped tissues, 3D photoacoustic experiments were carried out on complex-shaped phantoms. For example, a phantom made of knotted black wire with a diameter of about 0.6 mm is shown in Figure 11a. The sampling frequency, sampling position, laser energy, and z-axis spacing of this experiment were the same as those above. Figure 11b is a 3D photoacoustic image reconstructed by using the detected photoacoustic signals detected from 90 B-scans spaced 0.1 mm apart, indicating that the reconstructed photoacoustic image could present the actual shape and dimensions of the target clearly and with high quality.

**Figure 11.** (**a**) Photograph of a knot phantom and (**b**) 3D photoacoustic image.

#### **4. Discussion and Conclusions**

A 3D-PACT imaging system based on full-view illumination and detection was proposed and developed in this study. In the design of optical path, the system adopted a mode of circular illumination with the even arrangemen<sup>t</sup> of eight laser beams around the water tank to ensure that the imaging target received more uniform laser radiation, which improved the low-e fficiency of photoacoustic imaging whose illumination was from the top to the bottom. Second, drawing on the design of photoacoustic coaxial confocal in photoacoustic microscopy imaging technology, a photoacoustic coplanar structure was proposed by means of a quartz bowl with light-transmitting and ultrasound-reflecting properties. The system extended the photoacoustic coupling from a single dimension to a two-dimensional space, increasing the photoacoustic detection range of the imaging section and improving the signal to noise ratio of the photoacoustic signal. In addition, the dual-foci virtual point ultrasonic transducer with a large receiving angle improved the imaging speed of the system via sparse sampling. In the reconstruction method of the photoacoustic image, taking the receiving angle of ultrasound transducer into account, the improved back-projection reconstruction algorithm based on sensitivity factor was achieved, and the photoacoustic image reconstruction using the algorithm took 1.4287 s.

The practicality of the illumination method and the designed structure enable the system to perform clear and high-quality 3D photoacoustic imaging on targets with complex surfaces or requiring a large imaging range. The feasibility of the system was verified through systematic testing of resolution, imaging range, and photoacoustic coplanar property and 3D photoacoustic experiments on various phantoms. The best compromise was reached between imaging performance and system cost.

The imaging speed of the system was about 9 s in one cross section (90 sampling points per B-scan, it took about 15 min to scan 100 B-scans), and had not reached the requirements of fast imaging; however, the phantom experiments conducted in this study have demonstrated the characteristics of the system. The next step will be focused on the research of sparse circular array ultrasonic transducer based on virtual point and resolution improvement. In practical applications, the system is expected to be used for corresponding work in the pathological study of an in vitro tumor and the research of targeted photoacoustic probes for breast cancer cells. Moreover, the 3D reconstruction will be further investigated.

**Author Contributions:** Design and construction of the 3D-PACT, M.S., D.H., and L.M.; Completion of photoacoustic imaging experiments, D.H., W.Z., and Y.Q.; Control system design, M.S., D.H., and L.M.; System debugging and improvement, Y.L., Y.Q., and W.Z.; Algorithm implementation, M.S., D.H., and Y.L.; Paper writing, D.H., W.Z., and M.S.

**Funding:** This research was funded by the National Key R & D Program of China (Grant No. 2017YFE0121000 and No. 2018YFC0114800), National Natural Science Foundation of China (Grant No. 11574064 and No. 11874133), Shandong Provincial Natural Science Foundation, China (Grant No. ZR2017MF041 and ZR2018MF026), the Science and Technology Development Plan Project of Shandong Province (Grant No. 2018GGX103047 and 2016GGX103032), and the Development Plan of Chinese Academy of Sciences and Wego Group (Grant No. 2017011).

**Acknowledgments:** This work was partly supported by the Institute of Biomedical and Health Engineering at Shen Zhen Institute of Advanced Technology, Chinese Academy of Sciences (SIAT).

**Conflicts of Interest:** The authors declare no conflict of interest.
