A Review on the Roles of Photoacoustic Imaging for Conventional and Novel Clinical Diagnostic Applications

Abstract

Photoacoustic imaging is a promising medical imaging modality that enables the visualization of molecular functional and morphological information of biological tissues. Its clinical potential has been widely investigated for assessing and diagnosing various diseases. Currently, several research groups are developing photoacoustic imaging systems for translation from the laboratory to the clinic. In particular, the integration of photoacoustic imaging into existing diagnostic ultrasound applications, such as cancer diagnosis, has shown promising results. Additionally, recent research has explored the application of photoacoustic imaging for novel clinical uses. In this review paper, recent trials of photoacoustic imaging in both conventional and novel clinical applications are summarized and evaluated. Additionally, current limitations and future directions of photoacoustic imaging for successful translation into the clinical world are discussed. The aim of this review is to provide a comprehensive overview of the recent advancements in photoacoustic imaging and highlight its potential for clinical diagnosis and treatment. It is hoped that this review will contribute to the development of improved diagnostic and therapeutic approaches for a wide range of diseases using photoacoustic imaging.

1. Introduction

Visualization of biological tissue is a critical tool for diagnosing diseases, particularly for the early detection of cancer [1,2]. Imaging techniques including magnetic resonance imaging (MRI) [3,4,5], positron emission tomography (PET) [6,7,8], and X-ray computed tomography (CT) [9,10] have been extensively used in clinics for the characterization, assessment, and monitoring of tumor pathologies. However, these imaging techniques have some drawbacks, such as high costs, slow imaging speed, immobility, or potential side effects from ionizing radiation.

For routine screening, ultrasound imaging (USI) has been widely used due to its safety, mobility, and ease of use [11,12,13]. USI provides excellent morphological information on biological tissues, allowing for the identification of tumor masses and their spatial relationships with surrounding tissues. However, the lack of molecular functional information in USI can degrade diagnostic performance, limiting its accuracy in detecting small or early stage tumors.

Photoacoustic imaging (PAI) is an emerging medical imaging technique that integrates the principles of USI and optical imaging techniques. Instead of ultrasound (US) transmission in USI, PAI illuminates biological tissue with a short-pulsed laser (typically a few nanoseconds), which is absorbed by optically absorbing chromophores in the tissue. This absorbed light energy is converted into heat energy, which generates acoustic waves through thermoelastic expansion [14]. These acoustic waves are called photoacoustic (PA) waves, which can be captured by conventional US transducers and imaged using the US image generation procedure.

The core advantage of PAI is the visualization of optical absorption properties of biological tissue with a spatiotemporal resolution of USI [15]. PAI can provide strong optical contrast with both endogenous (e.g., oxyhemoglobin, deoxyhemoglobin, melanin, and lipid) [16,17,18,19] and exogenous [20,21,22,23,24] chromophores. From a clinical perspective, PAI can complement conventional USI with its unique advantages. Firstly, PAI and USI can be implemented in a single imaging platform, reducing the need for multiple imaging modalities [25,26,27,28,29,30]. Secondly, the resolution and imaging depth can be adjusted according to the target tissues [31,32,33,34,35,36]. Lastly, various geometries of imaging probes are applicable for optimized data acquisition, enabling the customization of imaging systems for specific clinical applications [37,38,39,40].

Recently, PAI techniques have been improved with advanced light sources [41,42] and novel image reconstruction algorithms [43,44,45]. Many research groups are translating their PAI systems from the laboratory bench to the clinical bedside [46]. The main purpose of PAI is to provide complementary molecular functional information of the region, augmenting the morphological information from USI. By adding PAI to existing USI, diseased tissues can be accurately delineated before biopsy or surgical operation, improving diagnostic accuracy. In addition, multispectral PA responses would be more helpful to diagnose or classify diseased tissues in vivo. Furthermore, recent research has investigated the potential application of photoacoustic imaging in novel clinical uses that have not yet been explored with USI.

In this review, we provide an overview of the clinical trials of PAI systems focusing on the potential of photoacoustic imaging for clinical diagnosis and treatment. Particularly, we aimed to review PAI studies for clinical application in terms of two aspects: (1) PAI studies improving the conventional USI-based diagnosis; and (2) PAI studies introducing unconventional clinically viable information. The former refers to clinical applications where PAI could be added to an existing USI procedure in the current diagnostic protocol, and thus more likely to be clinically approved in the future. The latter proposes new forms of diagnosis, which may not fit in the current protocol but provided promising findings that encourage the development of new clinical workflow. We highlight the integration of PAI into existing diagnostic USI applications and the exploration of novel clinical uses. The systemic configurations are also overviewed according to the geometric structures of US transducers, which determine the imaging region and applicable area. The goals of this review are to survey the recent advances of PAI toward clinical translation, discuss the current limitations, and present the future direction of research for successful clinical translation.

2. Imaging Probes for Photoacoustic Imaging

Various types of probes have been used in PAI to provide functional information in addition to structural information obtained from conventional USI (

Table 1. Typical performance of PAI probes with respect to the types of transducers. LAT, linear array transducer; TRT, transrectal transducer; CAT, curved array transducer; HAT, hemispherical array transducer; 2D, two-dimensional; 3D, three-dimensional; PA, photoacoustic; US, ultrasound; FOV, field of view; PAI, photoacoustic imaging; USI, ultrasound imaging.

	Conventional Transducers		Customized Transducers
	LAT	TRT	CAT	HAT
Dimension (image)	2D	2D	2D	3D
Number of elements	128–256	64–92	64–256	128–1064
Advantages	Dual-modal PA/US Good spatial resolution (USI) Affordable cost	Dual-modal PA/US Wide FOV (~170 degree view) Affordable cost	Optimal for 2D PAI due to wide reception angle	Optimal for 3D PAI due to wide reception angle Volumetric imaging (up to real-time) Near-isotropic spatial resolution
Limitations	Limited reception angle for PAI Anisotropic spatial resolution (axial < lateral < elevation)	Low spatial resolution Highly limited reception angle Anisotropic spatial resolution (axial < lateral < elevation)	Not appropriate for USI High cost Relatively small FOV	Not appropriate for USI Very high cost Relatively small FOV
Clinical applications	Thyroid, breast, skin, cervical vasculature, peripheral vasculature	Prostate, uterine	Thyroid, breast, cervical vasculature, peripheral vasculature, muscle, gastrointestinal	Breast, brain, skin, cervical vasculature, muscle

). Linear array transducers (LATs) are widely used in clinical studies due to their cost-efficiency and simplicity of implementation [47,48,49]. The piezoelectric elements in the LATs are aligned in a row, resulting in a rectangular field of view (FOV) in the cross-sectional plane. LATs offer an excellent near-field resolution, which is advantageous for PAI, as the target depth in PAI is usually shallow compared to USI. Since LATs have been used in the initial trials of clinical PAI systems, there have been numerous preliminary studies for evaluating imaging performance in both preclinical and clinical applications [50,51,52]. However, LATs have a finite aperture, which limits the signal acceptance angle, leading to radial artifacts [53].

For dedicated usage for transrectal USI, transrectal transducers (TRTs) have been designed to be inserted through the rectum and to provide wide radial FOV at the same time. TRTs are basically convex array transducers where the US transducer elements are facing away from the center of the arc, which is beneficial for wide FOV with the small footprint of the transducer. For PAI, however, the signal acceptance angle is even more limited compared to LATs because the convex geometry allows only a small portion of transducer elements to receive PA signals at effective reception angles.

To overcome the limited view of LAT and TRT, curved array transducers have been widely used in both preclinical and clinical studies [54,55,56,57,58,59]. Curved array transducers (CATs) with arc or ring shapes can cover a wider angle of view compared to LATs, producing high-quality tomographic images of biological tissues in vivo. However, from the perspective of human imaging, CAT-based PAI systems need to overcome their relatively small FOV which may limit the versatility of the system. Currently, CATs have mainly been applied in breast imaging [60].

To achieve high-quality three-dimensional (3D) images, hemispherical array transducers (HATs) have also been utilized for biological imaging [61,62,63,64]. HATs typically consist of 128–512 elements at wide detecting angles, enabling the acquisition of volumetric images without the need for scanning the transducer. However, the acquisition and processing of large volumetric data require a specialized data acquisition system and processing device, which are typically expensive. Currently, emerging research using HATs has focused on diagnosing breast and skin conditions as well as detecting tumors.

3. Photoacoustic Imaging Amended to Conventional Diagnostic Procedures

3.1. Thyroid Imaging

The thyroid has been widely studied using PAI [65]. The current standard diagnostic method for thyroid nodules is USI-based risk stratification, which relies solely on the shape of the nodules. However, this approach often leads to overdiagnosis of benign cases and unnecessary biopsies, resulting in low specificity [66,67]. To address this limitation, researchers have turned to PAI to provide additional pathological and metabolomic information. The results of recent studies using PAI-aided analysis techniques have demonstrated its potential to complement the existing USI-based risk stratification and improve the diagnostic accuracy of thyroid nodules.

Roll et al. conducted a clinical study to evaluate the potential of multispectral PAI in distinguishing benign and malignant nodules in the thyroid gland [68]. They employed a dual-modal PA and US imaging (PAUSI) system equipped with a tunable pulsed laser with a PRF of 25 Hz and a fluence of 20 mJ/cm². A 256-element CAT with a center frequency of 3 MHz and a coverage angle of 125° provided cross-sectional images with a field of view of 40 × 40 mm² (Figure 1a). Multispectral PA images were acquired from six Graves’ disease lobes, eight healthy lobes, thirteen benign nodules, and three malignant nodules using eight wavelengths (700, 730, 760, 800, 850, 900, 920, and 950 nm). The results indicated that Graves’ disease lobes exhibited higher deoxygenated hemoglobin (p = 0.0055) and total hemoglobin concentration (p = 0.0084), and lower lipid content (p = 0.0293). The multispectral PA response also demonstrated lower sO₂ levels (p = 0.0393) and fat content (p = 0.1295) in malignant nodules compared to benign nodules. Although the study population was limited, the findings demonstrated the potential of multispectral PAI for identifying thyroid diseases.

Noltes et al. demonstrated the characterization of malignant nodules using spectrally unmixed PA images with high spatial resolution and spectral contrast [69]. They analyzed data from 27 nodules obtained from 17 patients, including 11 malignant and 16 benign nodules, with final histopathology. Multispectral PA responses from 14 optical wavelengths (680, 700, 730, 760, 800, 850, 900, 930, 970, 1000, 1030, 1064, 1100, and 1195 nm) were analyzed to identify patterns for malignancy stratification, particularly vascular patterns in the thyroid region. The resulting images resolved blood vessels with a diameter of approximately 250 μm at a depth of 2 cm. The findings showed a matched (R²-value of 0.9426) intramodular microvascularity and extrathyroidal extension between the unmixed PA images and histopathological images (Figure 1b).

Kim et al. presented a novel approach to improve the diagnostic accuracy of thyroid nodules by combining multispectral PAI analysis with conventional US stratification [70]. They utilized a PAUSI system with a 128-element LAT and a laser source to collect data from 23 malignant and 29 benign patients. Multispectral PA images were obtained from five wavelengths (700, 756, 796, 866, and 900 nm), and three parameters (sO₂ level, spectral gradient of the PA signal, and skewness of the sO₂ distribution) were defined from the multispectral data to determine a machine learning trained decision boundary, which provided a probability of malignancy. By combining this machine learning score with the conventional American Thyroid Association (ATA) Guidelines, they demonstrated a new risk stratification metric, named ATAP (ATA Guidelines with PA-indicated-probability of PTC) score. The ATAP score distinguished malignant from benign nodules with a sensitivity of 83% and a specificity of 93%, addressing the low specificity of the conventional ATA guideline score.

3.2. Breast Imaging

Breast USI has been used as a diagnostic tool to complement X-ray mammography by observing abnormalities in the breast such as lumps, cysts, or tumors. It is particularly useful for women who should not be exposed to X-rays or for women with dense breasts [71]. LATs are typically used for diagnosing breast cancer with handheld scanning. However, the diagnostic accuracy of handheld US scans depends on the proficiency of the operator. To address this limitation, the automated breast ultrasound system (ABUS) provides motorized scanning of a large-aperture LAT (~768 elements) to obtain 3D US images of the entire breast with consistent image quality, which enables the detection of suspicious masses. Although US Doppler can also provide blood flow information near breast masses, breast USI suffers from low specificity and requires other imaging modalities to effectively characterize the distinct features of benign or malignant tumors. PAI is an attractive technique for imaging breast cancer, as it is compatible with USI and allows multispectral measurement of hemoglobin oxygen saturation (sO₂).

In a study by Wang et al., a fast and miniaturized system was developed to achieve deep tissue imaging of human breasts [72]. The system utilized a compact laser with a fixed wavelength of 1064 nm and a tunable pulse repetition rate (PRF) of up to 50 Hz to observe the distribution of blood vessels in human breasts (Figure 2a). The imaging probe consisted of a 128-element LAT operating at a center frequency of 5 MHz and a bifurcated fiber bundle. By scanning the probe in the elevational direction, 3D data were acquired from stacks of cross-sectional B-mode data, and landscape images were reconstructed using maximum amplitude projection (MAP) (Figure 2b). The total imaging time for one breast was 75 s with a scanning range of 75 mm. Nyayapathi et al. further developed the system by equipping it with two 128-element LATs with a center frequency of 2.25 MHz [73]. A translational scan was performed to achieve mammography-like images with the two LATs sandwiching the breast from both sides. The resulting images successfully identified blood vessels and their depth from the surface. The total scanning time for the whole breast was 80 s, making it feasible for clinical application. The same research group obtained PA images from 38 patients with breast tumors and found higher vascular activity in tumor-bearing breasts compared to healthy breasts [74]. These studies demonstrate the potential of PAI for breast imaging, as it allows for non-invasive and non-ionizing imaging of deep tissue structures. The development of compact and miniaturized systems with improved imaging speed and quality provides promising opportunities for the early detection and diagnosis of breast cancer.

Diot et al. conducted a study on multispectral analysis of breast tissues using a CAT [75] and designed a hand-held PAI probe by integrating a 256-element CAT with a laser-delivering fiber bundle (Figure 2c). The CAT consisted of piezoelectric elements with a center frequency of 5 MHz, arranged in an arc surface with an entire angle of 174° and a diameter of 120 mm. The system was equipped with a 256-channel data acquisition module, facilitating fast data acquisition and real-time image generation at a remarkable rate of 50 fps. For their investigation, ten patients with malignant breast masses and three healthy volunteers were recruited and imaged using multispectral illumination at wavelengths ranging from 700 to 970 nm. By applying the linear spectral unmixing technique, the study explored the components of oxyhemoglobin, deoxyhemoglobin, total blood volume, lipid, and water. The results revealed remarkable differences in the total blood volume and the ratio of oxyhemoglobin to deoxyhemoglobin between cancerous and control tissues, indicating the diagnostic potential of this technique. In a subsequent study, Kukacka et al. demonstrated the visualization of overlaid PA and US images of breast lesions in vivo [76]. They simultaneously achieved multispectral PA images and corresponding US images from sixteen malignant and six benign patients (Figure 2d). The spectrally unmixed image showed the diagnostic potential of this technique in assessing vasculature abnormalities, hypoxia, and inflammation around the nodules in vivo (Figure 2e).

Neuschler et al. conducted a multi-site study with 2105 patients to test the diagnostic utility of multispectral PAI for breast cancer diagnosis [77]. They used a 128-element LAT with a center frequency of 10 MHz and two laser wavelengths (757 and 1064 nm) to assess the sO₂ in the breasts. The spectrally unmixed PA images were overlaid on the corresponding US B-mode images, and five features in PA images were defined to score the malignancy of breast masses. These PA features were subsequently incorporated into the conventional Breast Imaging Reporting and Data System (BI-RADS) categories, which provide a risk stratification score ranging from 1 (negative) to 5 (highly suspicious for malignancy). The results showed that multispectral PA analysis has the potential to enhance diagnostic specificity when combined with conventional ultrasound B-mode images. Menezes et al. also demonstrated the feasibility of multispectral PA analysis in a similar study utilizing the same device [78]. By combining the PA scoring method with the BI-RADS score, the researchers successfully downgraded the scores of incorrectly diagnosed benign nodules with initial scores of 4 (suspicious for malignant) to 3 (probably benign) or 2 (negative but with benign findings). These studies highlight the potential of multispectral PAI as a complementary diagnostic tool to existing ultrasound stratification methods, ultimately improving the specificity of breast cancer diagnosis.

Figure 2. PA images acquired from human breasts in vivo. (a) Schematic illustration of the configuration of PA imaging system for breast imaging. (b) A depth-encoded PA MAP image of the breast from a healthy volunteer. (c) Schematic illustration of the dual-modal PA and US image acquisition from breast cancer patients. (d) Overlaid PA and US image around the breast mass in a malignant patient. (e) Spectrally unmixed image by applying a linear unmixing technique. PA, photoacoustic; US, ultrasound; LAT, linear array transducer; CAT, curved array transducer; DAQ, data acquisition module; MAP, maximum intensity projection; HbO, oxyhemoglobin; HbR, deoxyhemoglobin. The images are reproduced with permission from [72,76].

Figure 2. PA images acquired from human breasts in vivo. (a) Schematic illustration of the configuration of PA imaging system for breast imaging. (b) A depth-encoded PA MAP image of the breast from a healthy volunteer. (c) Schematic illustration of the dual-modal PA and US image acquisition from breast cancer patients. (d) Overlaid PA and US image around the breast mass in a malignant patient. (e) Spectrally unmixed image by applying a linear unmixing technique. PA, photoacoustic; US, ultrasound; LAT, linear array transducer; CAT, curved array transducer; DAQ, data acquisition module; MAP, maximum intensity projection; HbO, oxyhemoglobin; HbR, deoxyhemoglobin. The images are reproduced with permission from [72,76].

3.3. Transrectal Imaging

Transrectal USI is a widely used technique for assessing the structural features of prostate glands, including the identification of abnormal nodules and guiding needle biopsy. In recent years, PAI has emerged as a promising tool for providing functional information on suspicious regions. However, transrectal USI presents a unique challenge due to the use of a narrow transrectal US transducer (TRT) that must fit within the colorectal region. This constraint imposes limitations on the size of the optical delivery component. Moreover, the convex shape of the TRT, in which elements are aligned on the surface, presents further challenges to the reception of PA signals due to the limited aperture. Therefore, an exploration of prostate research that addresses these challenges and examines the significance of PA indices under harsh PA signal reception conditions is of interest.

Horiguchi et al. conducted a pilot study to explore the potential of transrectal PAUSI in visualizing microvasculature in periprostatic tissues from seven prostate cancer patients [79]. The study aimed to detect the location and extent of neurovascular bundles (NVBs) in prostate cancer patients. The PAUSI system utilized a 128-element TRT with a center frequency of 6.5 MHz, offering a detectable FOV of 170° (Figure 3a). The illumination system delivered a pulsed laser with a wavelength of 756 nm, a PRF of 10 Hz, and a fluence of 11 mJ/cm², utilizing two convex apertures (2.5 mm × 180° each) positioned beside the TRT. The transrectal part of the imaging probe had a thickness of ~26 mm, allowing for effective transrectal imaging. During radical prostatectomies, PA and US images of the periprostatic tissues were acquired (Figure 3b). The results demonstrated the ability to identify the location of periprostatic NVBs, suggesting the feasibility of incorporating this technique in nerve-sparing radical prostatectomy procedures. In a follow-up study, they modified the probe to be smaller and lighter to facilitate transrectal delivery [80]. Microvascular density (MVD), total vessel area (TVA), and total vessel length (TVL) were investigated from PA images acquired from three prostate cancer patients. The result showed higher TVA and TVL in the peripheral region compared to the inside of the tumors. By comparing with the histopathological results, PAI can potentially reflect the microvascularity, suggesting a novel modality for imaging prostate cancer angiogenesis.

In a recent study by Kothapalli et al., a clinical transrectal PAUSI technique was employed to image prostate cancer patients using a miniaturized 64-element capacitive micromachined TRT with a center frequency of 5 MHz [81]. For intrarectal irradiation, a tunable pulsed laser with a PRF of 10 Hz, emitting wavelengths ranging from 680 to 950 nm, and a fluence of 10 mJ/cm², was coupled to three fiber optic paths (Figure 3c). The imaging probe had a thickness of approximately 23 mm. Among the 20 patients who participated in the clinical trial, 10 patients received intravenous administration at a concentration of 2.5 mg/mL. The use of ICG enabled the acquisition of high-contrast and high-resolution PA images through transrectal PAI (Figure 3d), facilitating the visualization of prostate vascularity and surrounding tissues. The results demonstrated the potential of this approach in differentiating prostate cancers, reducing the need for random biopsies through targeted biopsies, and aiding in the selection of suitable treatment options.

3.4. Cervical Vasculature Imaging

PAI has been also applied in the imaging of human carotid arteries, offering insights into potential disorders within these blood vessels by integrating with the conventional USI. Ivankovic et al. demonstrated a 3D multispectral PAI system for monitoring the bifurcation area of the human carotid artery, particularly in relation to ischemic stroke (Figure 4a) [82]. Their system utilized a hand-held HAT array capable of capturing 10 multispectral data points per second. A higher contrast-to-noise ratio was observed with wavelengths in the range of 800–900 nm (18.8 ± 4.28 dB) compared to shorter wavelengths of 730–760 nm (9.18 ± 3.8 dB), indicating that the carotid artery is a highly oxygenated vessel. Karlas et al. acquired multispectral PA images of the human neck area from five atherosclerosis patients, with the goal of comparing the levels of lipid and hemoglobin in healthy volunteers (Figure 4b) [83]. They used 28 wavelengths ranging from 700 to 970 nm with a 10 nm step, enabling the generation of a multispectral stack of PA images (Figure 4c). They also acquired US images co-registered to each PA image, and then manually segmented the regions of arterial cross-section, plaque, and lumen. Using spectral unmixing techniques, lipid and hemoglobin concentrations within the carotid artery were determined. The results demonstrated distinguishable differences in the fat–blood ratio between atherosclerosis patients and healthy volunteers, illustrating the feasibility of molecular imaging for human carotid atherosclerosis and the potential clinical assessment of carotid plaques. These studies showcase the valuable role of PAI in investigating carotid artery pathologies and highlight its potential for non-invasive assessment of vascular disorders.

4. Photoacoustic Imaging as Novel Diagnostic Methods

4.1. Peripheral Vasculature Imaging

Vascular imaging has prominent applications of PAI in biomedical studies. Several research groups have achieved PA images for the evaluation of vascular function, especially peripheral vasculature assessment. Yang et al. demonstrated a vascular dysfunction monitoring PAI system for diabetic patients, employing a 128-element CAT array and a tunable pulsed laser with a repetition rate of 20 Hz (Figure 5a) [84]. Multispectral PA images of the dorsalis pedis artery and arterioles in the distal hallux region were acquired from seven diabetic patients and seven healthy volunteers, while a pressure cuff was applied to induce venous and arterial occlusions alternately (Figure 5b). The recorded PA signals indirectly reflected changes in blood volume (Figure 5c). Arterial occlusion led to a decrease in PA signals due to reduced blood volume resulting from arterial compression. Upon release of the pressure, the PA signal gradually recovered, with distinguishable differences in recovery time observed between the diabetic and normal groups. A noticeable time delay was observed in the arteries of the diabetic group (healthy: 14.66 ± 1.10 s, diabetic: 24.05 ± 5.36 s). In addition, a lower variation in PA signal was observed in the veins (healthy: 5.63 ± 0.024%, diabetic: 3.38 ± 0.331%), indicating decreased blood perfusion in the feet. These findings highlighted distinct vascular responses in the diabetic group compared to the healthy controls under occlusion. In a recent study, Choi et al. conducted in vivo morphological and physiological analyses of healthy human feet using 3D PAUS images [85]. To enhance the reliability of PAI, they implemented adaptive contour scanning of a 128-element LAT, tracking the curvature of the dorsal surface of the foot (Figure 5d). Several image processing techniques were applied to achieve comprehensive multi-structural visualization of the entire foot, including the skin, bone, and vasculature (Figure 5e). Quantitative analysis of parameters such as total hemoglobin concentration, vessel density, sO₂, and vessel depth, obtained from five independent acquisitions of ten healthy volunteers, further evaluated the reliability of the system. The resulting images showed the potential of this technique to quantify the blood vessel activities and valuable insights into both morphological vessel networks and physiological sO₂ distribution.

4.2. Volumetric Vasculature Imaging in Breast

Volumetric vasculature imaging in the breast has emerged as an innovative diagnostic procedure enabled by 3D PA images using HATs. This technique offers the visualization and assessment of blood vessels within breast tissue, presenting valuable insights for the detection and characterization of breast lesions. This chapter will explore the applications of volumetric vasculature imaging in the breast using PAI, highlighting its potential impact on breast cancer diagnosis and management.

Kruger et al. conducted a notable study demonstrating human breast imaging using a 128-element HAT with a center frequency of 5 MHz [86]. The piezo elements were arranged in a spiral distribution on the probe’s hemispherical surface, which had a radius of 100 mm (Figure 6a). PA images were acquired from a healthy volunteer by illuminating a 20 mJ pulsed laser at a wavelength of 800 nm. The total data acquisition time amounted to 24 s, encompassing a FOV of 64 × 50 mm². The resulting images visualized 3D vascular networks up to a depth of 40 mm (Figure 6b). To achieve enhanced tissue penetration and improved contrast sensitivity, the system was further modified with a 512-element, 2 MHz center frequency HAT (Figure 6c) [87]. The transducer was mounted on a two-axis translational stage to enable spiral scanning, thereby achieving a larger FOV in the form of a circular region with a radius of 96 mm. Volumetric breast data were collected from four healthy female volunteers (Figure 6d), revealing clear images of the breast vasculature with the potential to provide complementary information when integrated with conventional X-ray mammography.

Toi et al. also developed and demonstrated a volumetric PAI system for breast imaging [88]. Their system employed a 512-element HAT operating at a center frequency of 2 MHz, which was moved along the spherical scanning trajectory to acquire volumetric data. Notably, the inner side of the spiral scan had a higher data collection density, resulting in improved image quality at the center of the breast. Multispectral PA data from 22 malignant breast lesions were acquired using a Q-switched Alexandrite laser with wavelengths of 755 nm and 795 nm (Figure 6e). The resulting images revealed distinctive morphological changes within the malignant vascular areas. Specifically, features such as centripetal blood vessel structure, vessel disruption, and rapid narrowing were detected in invasive breast carcinoma. In a related study, Yamaga et al. utilized the same system to investigate tumor characteristics by quantifying vessel branching points (VBP) in breast cancer patients [89]. The results demonstrated that the mean VBP level, up to a depth of 7 mm below the skin surface, was significantly higher in the cancerous breast compared to the contralateral breast (Figure 6f). Furthermore, this system exhibited the capability to provide functional sO₂ maps with three-dimensional visualization in other body regions, highlighting its potential for clinical research applications [64].

Tumor detection in the breast has often been hindered by respiratory distortion resulting from lengthy scanning times. To address this challenge, Lin et al. developed a fast PAI system capable of capturing whole breast images within a single breath-hold [90]. Their approach involved the translational scanning of a 512-element ring-shaped array transducer with a center frequency of 2.25 MHz, enabling rapid scanning in less than 15 s (Figure 6g). For PA generation, a laser beam with a wavelength of 1064 nm and a fluence of approximately 20 mJ/cm² was directed towards the breast. To achieve a uniform light distribution, the beam passed through an axicon lens and an engineered diffuser, which transformed it into a donut shape. Notably, the system utilized four 128-channel data acquisition modules to facilitate high-speed data acquisition by maintaining a one-to-one mapping between the transducer elements and the data acquisition channels. Analysis of the PA images obtained from the breasts of seven breast cancer patients revealed higher blood vessel densities associated with tumor regions. Especially, in patients with dense breast tissue, where conventional X-ray mammography often faces limitations (Figure 6h), PAI exhibited great potential for sensitively diagnosing malignant regions (Figure 6i). These findings highlight the promising role of PAI in improving breast cancer detection, particularly in cases involving dense breast tissue.

4.3. Non-Invasive Staging of Melanoma

In recent years, PAI has been explored for the non-invasive staging of melanoma, offering a range of capabilities such as visualization of melanoma lesions, precise depth measurement, and accurate boundary delineation. Kim et al. conducted a study demonstrating the feasibility of PAI with a 128-element LAT in distinguishing melanoma from normal tissue using an excised tissue specimen [91]. They further advanced the application of PAI by developing a handheld scanner for in vivo melanoma depth measurement (Figure 7a) [92,93]. Through a series of in vivo experiments involving six patients, the researchers collected multispectral PA and ultrasound data to evaluate the depth-measuring capabilities of PAI. The resulting spectrally unmixed PA images exhibited remarkable accuracy in depicting melanoma contours by visualizing the concentration of melanin (Figure 7b). Furthermore, the PA images effectively measured the depth of melanoma, demonstrating a mean absolute error of 0.36 mm when compared to histopathological results. Chuah et al. reported another noteworthy study using a handheld HAT [94]. From the multispectral PA data, subcutaneous chromophores including melanin, oxyhemoglobin, and deoxyhemoglobin were mapped (Figure 7c). These studies highlight the significant advancements in the non-invasive staging of melanoma enabled by PAI. The accurate delineation of melanoma boundaries holds tremendous potential for enhancing the diagnosis and treatment of this complex and challenging form of skin cancer.

4.4. Muscular and Gastrointestinal Assessment

The hemodynamic responses that can be measured from PAI have been also utilized for analyzing the metabolic functioning of organs. Yang et al. conducted a study utilizing PAI to characterize circulatory and metabolic functions by monitoring hemodynamic changes in the forearm skeletal muscle during cuff occlusion [95]. They acquired multispectral PA images using a 128-element CAT with a diameter of 100 mm and a spanning angle of 174°. By monitoring PA signals from veins, arteries, and muscles during alternating venous and arterial occlusions (Figure 8a), a similar trend to the conventional near-infrared spectroscopy was observed (Figure 8b). This finding demonstrated the feasibility of the PAI system for imaging hemodynamic responses in both muscles and blood vessels. In another study, Regensburger et al. demonstrated the visualization of disease progression and treatment response in Duchenne muscular dystrophy (DMD) using multispectral PAI [96]. The spectrally unmixed PA signals determined the components of collagen, lipid, and hemoglobin in the muscles of 10 DMD patients and 10 healthy volunteers (Figure 8c). The results revealed a clear difference in collagen signal intensity between healthy subjects (14.60 ± 4.42) and DMD patients (24.72 ± 5.92). Volumetric images further showcased significant differences in oxyhemoglobin, deoxyhemoglobin, and total hemoglobin levels, underscoring the potential of multispectral PAI for monitoring DMD patients. Furthermore, Knieling et al. presented a preliminary study on the application of PAI for assessing Crohn’s disease, based on the correlation between hemoglobin content and intestinal inflammation [97]. Multispectral PA images were acquired from a total of 108 patients. The PA responses were compared between active and non-active patients, as determined by conventional scoring methods including the Harvey–Bradshaw index, simplified endoscopic score for Crohn’s disease, and modified Riley score. The initial results showed significant differences in oxyhemoglobin, deoxyhemoglobin, and total hemoglobin levels. These studies highlight the valuable role of PAI in investigating hemodynamic responses and metabolic functions, providing insights into the pathophysiology of various disorders, and monitoring treatment outcomes in a non-invasive manner.

5. Summary and Discussion

In this paper, we have presented an extensive overview of the diverse applications of PAI in clinical human studies. By offering molecular and functional information on biological tissues, PAI has demonstrated remarkable potential for various clinical applications, including cancer detection, melanoma staging, vascular imaging, and metabolic function assessment. Initially, PAI has been successfully integrated with existing diagnostic USI, then expanding its scope to novel clinical applications. The findings and advancements discussed in this paper provide valuable insights into the wide-ranging clinical applications of PAI and highlight its significant contributions to non-invasive imaging in the field of medical diagnosis.

Table 2. Summary of PAI in clinical human studies. PA, photoacoustic; US, ultrasound; PRF, pulse repetition rate; $\lambda$, excitation wavelength for PA imaging; $n_E$, number of elements; $f_C$, center frequency; LAT, linear array transducer; CAT, curved array transducer; HAT, hemispherical array transducer; TRT, transrectal transducer; GI, gastrointestinal; BV, blood vessel; sO₂, hemoglobin oxygen saturation; HbO, oxygenated hemoglobin; HbR, deoxygenated hemoglobin. BI-RADS, Breast Imaging Reporting and Data System; ATA, American Thyroid Association; NVB, neurovascular bundle; MVD, microvascular density; TVA, total vessel area; TVL, total vessel length; ICG, indocyanine green; DMD, Duchenne muscular dystrophy.

Application	Subject Populations	PA Imaging System							Analyzed Parameters	Ref.
		Laser			US Transducer			Frame Rate * [fps]
		Type	PRF [Hz]	$\mathit{\lambda }$ [nm]	Type	${\mathit{n}}_{\mathit{E}}$	${\mathit{f}}_{\mathit{C}}\left[\mathbf{MHz}\right]$
Thyroid cancer	8 healthy 13 benign 4 malignant	Nd:YAG	25	700, 730, 760, 800, 850, 900, 920, 950	CAT	256	3	-	HbT, sO2	[68]
	16 benign 11 malignant	Nd:YAG	25	680, 700, 730, 760, 800, 850, 900, 930, 970, 1000, 1030, 1064, 1100, 1195	CAT	256	3.4	1.8	HbO, HbR, lipid, water	[69]
	29 benign 23 malignant	Nd:YAG	10	700, 756, 796, 866, 900	LAT	128	7.5	5	sO2-related PA scores with ATA score	[70]
Breast cancer	1 healthy	Nd:YAG	50	1064	LAT	128	5	-	Depth of BVs	[72]
	1 healthy	Nd:YAG	10	1064	LAT	128	2.25	-	Depth of BVs	[73]
	38 patients	Nd:YAG	10	1064	LAT	128	2.25	-	PA intensity in nodule	[74]
	10 patients 3 healthy	Nd:YAG	50	700–970 (10 nm step)	CAT	256	5	50	HbO, HbR, HbT, lipid, water	[75]
	6 benign 16 malignant	Nd:YAG	50	700–970 (10 nm step)	CAT	256	5	50	HbO, HbR, HbT, lipid, water	[76]
	2105 patient	-	-	757, 1064	LAT	128	10	-	PA scores with BI-RADS	[77]
	209 patients	-	-	757, 1064	LAT	128	10	-	PA scores with BI-RADS	[78]
Prostate cancer	7 patients	Alexandrite	10	756	TRT	128	6.5	5	Location of NVBs	[79]
	3 patients	Alexandrite	10	756	TRT	128	6.5	5	MVD, TVA, TVL	[80]
	20 patients	Nd:YAG	10	750, 775, 800, 825, 850, 875, 900	TRT	64	5	10	NVB signal enhancement after ICG injection	[81]
3D vasculature in breasts	1 healthy	Nd-YAG	10	800	HAT	128	5	-	PA signal of BVs	[86]
	4 healthy	Alexandrite	10	756	HAT	512	2	(12 s)	PA signal of BVs	[87]
	22 malignant	Alexandrite	-	755, 795	HAT	512	2	(51.2 s)	sO2-related PA score (S-factor)	[88]
	22 malignant	Alexandrite	-	755, 795	HAT	512	2	(51.2 s)	Vessel branching points	[89]
	7 malignant	Nd:YAG	-	1064	HAT	512	2.25	10	BV density, depth of BVs	[90]
Cervical vasculature	16 healthy	Nd:YAG	10	730, 760, 800, 850, 900	HAT	256	4	100	PA signal of BVs	[82]
	5 healthy 5 atherosclerosis 2 excised plaques	Nd:YAG	50	700–970 (10 nm step)	CAT	256	4.5	50	HbO, lipid	[83]
Peripheral vasculature	7 healthy 7 diabetics	Nd:YAG	20	760, 840	CAT	128	7.5	20	PA signal change under cuff pressure	[84]
	10 healthy	Nd:YAG	10	700, 756, 796, 866	LAT	128	7.5	50	BV density, HbT, sO2	[85]
Melanoma	1 excised lesion	Nd:YAG	10	700, 756, 796, 866, 900	LAT	128	7.5	(37.5 s)	Melanin	[91]
	6 patients	Nd:YAG	10	700, 756, 796, 866, 900	LAT	128	7.5	5	Melanin	[93]
	3 patients	Nd:YAG	100	680–980	HAT	512	10	100	Melanin, HbO, HbR	[94]
Muscular function	6 healthy	Nd:YAG	20	800	CAT	128	5	2	PA signals at vein, artery, and muscle	[95]
	10 healthy 10 DMD	Nd:YAG	25	680, 700, 730, 760, 800, 850, 920, 1000, 1030, 1064, 1100	CAT HAT	256 256	4 8	-	Collagen, lipid, HbO, HbR	[96]
GI function	88 Crohn’s disease	Nd:YAG	25	700, 730, 760, 800, 850, 900	CAT	64	4	-	HbO, HbR, HbT, sO2	[97]

* Numbers in parentheses indicate the time for a single image generation.

summarizes the key specifications of the clinical application of PAI, understanding its versatility and potential in diverse clinical settings. Further research and development efforts in this field hold promise for optimizing PAI systems, refining image processing techniques, and validating the clinical utility of PAI in improving patient outcomes across various medical conditions.

The promising clinical applications of PAI presented in this paper open up avenues for further research and development toward its successful translation into clinical practice. In addition to the studies discussed in this paper, there are several potential clinical applications that PAI can be utilized. One such application is functional brain imaging, which has been widely investigated in preclinical animal studies. However, translating this technique to human subjects presents challenges due to its invasiveness, as it currently requires opening the scalp for imaging [98,99,100]. Recently, Na et al. reported human brain imaging by using four 256-element quarter-ring-shaped CAT [101], but they achieved images after hemicraniectomy, suffering from the invasiveness. Despite the current limitations, ongoing advancements in technology and potential improvements in non-invasive imaging approaches may offer promising possibilities for achieving functional imaging of the human brain using PAI.

Another potential clinical application is PA endoscopy (PAE), which has been widely investigated in biomedical studies [102]. In preclinical research, PAE has demonstrated its feasibility for diagnosing tumors, polyps, and other colorectal conditions using in vivo small animal models [103,104,105] and ex vivo human colorectal tissues [106]. However, PAE has not yet been applied to in vivo human imaging due to certain challenges related to inserting miniaturized imaging probes into the body for clinical use. The approval and adoption of such miniaturized devices for clinical human imaging require compliance with rigid regulatory standards, including good manufacturing practices. Despite the current limitations, the potential for future clinical applications of PAE in human analysis remains promising. To realize its clinical translation, it will be essential to focus on achieving technical improvements, as well as obtaining regulatory approval for safe and effective use in humans.

While PAI has demonstrated remarkable capabilities in various areas, there are several important considerations and future directions that need to be addressed to facilitate its clinical translation. Firstly, further advancements in PAI technology are required to enhance its clinical applicability [107]. This includes the development of compact and portable PAI systems that are user-friendly and cost-effective, enabling widespread adoption across different healthcare settings [41]. Improvements in imaging speed, spatial resolution, and depth penetration are also crucial for obtaining high-quality images and maximizing clinical utility.

Standardization of imaging protocols and data analysis techniques is another critical aspect of the clinical translation of PAI [108]. Establishing consensus guidelines and benchmarks for image acquisition, data processing, and interpretation will enable accurate and reproducible results, facilitating comparison and validation of findings across different studies and institutions. To achieve this goal, rigorous validation through well-designed clinical trials is necessary. Large-scale studies involving diverse patient populations and comparisons with established diagnostic modalities are needed to assess the clinical performance and utility of PAI in real-world scenarios.

In addition, the development and application of artificial intelligence (AI)-based analysis techniques also represent a promising avenue for enhancing the diagnostic accuracy of clinical PAI. Extensive research has recently been explored to assess AI-based analysis with PA images [109,110,111]. Through these techniques, PAI can significantly improve the diagnostic process by providing efficient guidance and supporting accurate decision-making. To establish robust and reliable AI algorithms, large-scale datasets with multi-location acquisitions may be essential, which is one of the main challenges. By using the potential of AI, PAI can overcome challenges related to limited-view issues and enhance image qualities.

Moreover, efforts should be directed toward the development of targeted contrast agents or molecular probes that can specifically enhance the PAI signals of pathological tissues or biomarkers [112,113,114]. The integration of molecular imaging capabilities into PAI can enable the non-invasive detection of molecular targets and facilitate early diagnosis, treatment planning, and monitoring of therapeutic responses. Advancements in nanoparticle-based contrast agents, biomolecular targeting strategies, and theranostic agents hold great promise for enhancing the sensitivity and specificity of PAI in molecular imaging applications. However, the application of nanoparticles as contrast agents in clinical human studies has not been realized thus far. The regulatory approval process for the usage of exogenous agents in humans poses significant challenges, requiring comprehensive safety and efficacy assessments. Nevertheless, extensive preclinical research has demonstrated the potential of nanoparticles in PAI, showcasing their capabilities in contrast-enhanced imaging, targeted drug delivery, and treatment monitoring. While the translation of nanoparticle-based PA agents to clinical human studies is challenging, they are promising for advancing the capabilities of PAI in clinical diagnosis and theranostic applications.

In conclusion, while PAI has shown remarkable promise in various clinical applications, further research and development efforts are required for its successful translation into routine clinical use. Advances in technology, standardization of protocols, validation through clinical trials, and development of targeted contrast agents are all crucial elements in realizing the full potential of PAI as a powerful clinical imaging modality. To date, there have been only a few clinically approved PAI systems. Imagio™ Breast Imaging System (Seno Medical Instruments Inc.) succeeded in obtaining the first pre-market approval (PMA) by the United States Food and Drug Administration (FDA) in 2021. The approval was based on compelling evidence presented in their report, which demonstrated that PAI could significantly enhance the specificity of conventional ultrasound-based breast cancer diagnosis, as discussed in the previous section [77]. In addition, iThera Medical GmbH succeeded in obtaining CE Marks with their PAI systems: MSOT Acuity (PAI only) in 2019 [83] and MSOT Acuity Echo (dual-modal PAI and USI) in 2021 [97]. These achievements marked a significant advancement in the clinical translation of PAI. With continued interdisciplinary collaborations and concerted efforts, PAI holds great potential to transform medical diagnostics and improve patient outcomes across a wide range of clinical scenarios.