Open Access
Add to BibSonomy
Abstract
Photoacoustic imaging (PAI) was performed to evaluate oxygen saturation (sO2) of blood-mimicking phantoms, femoral arteries in beagles, and radial arteries in humans at various sO2 plateaus. The accuracy (root mean square error, RMSE) of PAI sO2 compared with reference sO2 was calculated. In blood-mimicking phantoms, PAI achieved an accuracy of 1.49% and a mean absolute error (MAE) of 1.09% within 25 mm depth, and good linearity (R = 0.968; p < 0.001) was obtained between PAI sO2 and reference sO2. In canine femoral arteries, PAI achieved an accuracy of 2.16% and an MAE of 1.58% within 8 mm depth (R = 0.965; p < 0.001). In human radial arteries, PAI achieved an accuracy of 3.97% and an MAE of 3.28% in depth from 4 to 14 mm (R = 0.892; p < 0.001). For PAI sO2 evaluation at different depths in healthy volunteers, the RMSE accuracy of PAI sO2 increased from 2.66% to 24.96% with depth increasing from 4 to 14 mm. Through the multiscale method, we confirmed the feasibility of the hand-held photoacoustic/ultrasound (PA/US) in evaluating sO2. These results demonstrate the potential clinical value of PAI in evaluating blood sO2. Consequently, protocols for verifying the feasibility of medical devices based on PAI may be established.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Oxygen is vital for human survival and normal functioning of organs. In the occurrence and development of many diseases, abnormal tissue metabolism of oxygen can occur [1–9]. Therefore, oxygen saturation (sO2) is an important physiological parameter that characterizes the metabolism and functioning of human tissues, and it may serve as a potential predictor for the diagnosis and progression of diseases [7–14]. Hypoxia can cause functional damage to tissues as well as organ damage along with a series of adverse effects, such as cell death and tissue necrosis, and can even be life-threatening in severe cases [15]. Furthermore, it is involved in some pathogenetic processes of tumors. Previous studies have reported that malignant tumors, such as breast cancer, exhibit higher blood supply and lower sO2 in the tumor area compared with the tumor surrounding area; this blood-rich hypoxia feature may result from the higher metabolism of tumor cells with increased oxygen consumption [16,17]. Hypoxia of the tumor microenvironment is also related to tumor progression, aggressiveness, resistance to radiotherapy and chemotherapy [9]. Furthermore, hyperoxia also plays an important role in diseases. An increase in the tissue O2 level can help generate excessive reactive oxygen species (ROS), which causes cellular damage and multiple organ dysfunctions. Several studies have reported that oxygen, especially in the hyperoxic state, plays a major role in many lung diseases [1,18]. Cardiac effects due to hyperoxia, such as vasoconstriction and subsequent decrease in heart rate, stroke volume, and cardiac output, have also been noted [2]. Therefore, accurate assessment of blood and tissue oxygen levels is crucial for the diagnosis, treatment, and prognostic evaluation of diseases.
Currently, human blood sO2 can be measured both invasively and noninvasively. For invasive testing, arterial blood has to be drawn, and sO2 can be evaluated based on blood gas analysis or through spectrophotometric measurement of the optical density, which is considered the gold standard for sO2 measurement. However, the invasive method cannot provide continuous real-time data. Conversely, a noninvasive method, such as pulse oximetry (the most common method) [19–21], can provide safe, effective, and continuous monitoring of sO2. However, clinical pulse oximetry is often placed in limited areas such as fingertip (or toe tip), earlobe, and nose wing, and lacking of spatial resolution limits the clinical application of this technique [22–25].
Near-infrared spectroscopy imaging (NIRS) is suitable for monitoring the local blood oxygen level of tissues, such as the brain, skin flaps, and muscles [26–28]. Unlike pulse oximetry, NIRS is not restricted by hypovolemia and vasoconstriction [19,23], but it can be affected by factors such as blood volume and hemoglobin concentration. Furthermore, although pure optical imaging methods achieve high accuracy in terms of blood oxygen measurement, their spatial resolution and imaging depth are limited [29]. Blood-oxygen-level–dependent functional magnetic resonance imaging (BOLD-fMRI) has a high spatial resolution, which is also sensitive to the oxygen level changes in blood flow and tissue. BOLD-fMRI has been recognized as a robust imaging modality for brain function evaluation [30–32]. However, fMRI has limitations such as being time-consuming and expensive, exhibiting imaging restrictions in case of metal implants, and being unable to achieve real-time dynamic monitoring.
Recently, photoacoustic imaging (PAI), a novel imaging technique that combines the merits of high-contrast optical imaging and deep-penetrating acoustic imaging, has advanced rapidly [33–38]. PAI can provide real-time quantitative functional information of tissues based on multispectral absorption of tissue biomolecules, such as deoxygenated hemoglobin, oxygenated hemoglobin, melanin, and lipids [39–41]. Therefore, PAI can evaluate human blood sO2 and reflect the pathophysiological characteristics of tissues in many diseases, such as ischemia, inflammation, and cancer [33,42–47]. Thus, the accuracy and stability of PAI in estimating sO2 should be verified urgently.
Several studies have validated the ability of PAI in evaluating blood sO2 [43,48–52]. In 2017, Mitcham et al. used bovine prostate tissue to prove the accuracy of PAI sO2 and demonstrated that the PAI sO2 was well matched with CO-oximeter sO2 [51]. Studies on animal organs and tissues have also validated that PAI is capable of evaluating blood sO2 in vivo [53,54]. In 2020, Yang et al. reported that PAI and NIRS exhibited similar results in evaluating the blood hemodynamics and sO2 on the right arm of a human [43]. In 2021, Merdasa et al. compared PAI with a commercial sO2 monitor in estimating the spatial distribution of sO2 during the onset of ischemia; their study demonstrated the capability of PAI in the sO2 evaluation [55].
Although several studies have demonstrated the capability of PAI in assessing sO2, few of them involved real-time and dynamic sO2 assessment in vivo, which is closer to clinical application. In addition, there is no well-validated standardized performance test method to promote translation and maturation of PAI devices. The configuration and parameters of PAI devices vary widely in many studies of monitoring blood sO2 (e.g., linear-array/circular scanning systems, optical wavelengths, and acoustic frequency range). As the hand-held photoacoustic/ultrasound (PA/US) probe has become the mainstream and hotspot of clinical application research in PAI, we employed a hand-held PA/US dual-imaging system to verify the feasibility of PAI in evaluating sO2 in this study. In our previous studies, the PAI system was used in rheumatoid arthritis, thyroid diseases, and breast diseases, where the accuracy of PAI sO2 evaluation received widespread attention [38,41,47,56–58]. In this study, we performed multiscale verifications under multiple blood sO2 plateaus to explore the accuracy of PAI in evaluating sO2 in different experimental subjects (specifically, phantoms, animals, and humans). The PAI sO2 was compared with the reference sO2 to verify the feasibility and stability of PAI in evaluating blood sO2.
2. Methods and material
2.1 Experimental imaging system
2.1.1 Composition and principle
PA effects refer to a phenomenon wherein a biological tissue absorbs a part of the energy of the laser pulse and then generates thermal expansion and ultrasonic waves. The intensity of PA signals is proportional to light energy deposition, which is the product of the absorption coefficient and the local light fluence. Therefore, PAI reflects the optical characteristics of tissues, such as hemoglobin content, which further reflects the blood sO2.
The PA/US dual-imaging system was based on a commercial ultrasonic platform (Resona 7, Shenzhen Mindray Biomedical Electronics Co., Ltd., Shenzhen, China). Considering the imaging depth of clinical application and the size of the object, we used a hand-held linear probe (L9-3PAU). The probe had a total of 192 arrayed elements; the width of each arrayed element was 0.2 mm; and the center frequency was 5.8 MHz, with a fractional bandwidth (−6 dB) of 65%. This integrated probe was provided with an optimized structure of an optical fiber and a lens. This helped to improve the stability of the internal interface of the probe, enhance the scattering ability of the lens to the laser at the transmitter, and reduce the probe’s absorption of the reflected PA signals from the tissue surface. The system retained high horizontal and vertical resolutions (less than 1 mm), and the signal noise ration (SNR) of the system was 27.5 dB in the depth range from 5 mm to 20 mm. A tunable optical parametric oscillator laser (SpitLight 600-OPO, InnoLas Laser GmbH, Krailling, Germany) was used, which generated signal light with a tuning range of 680–980 nm at 10 Hz. The luminous flux on the surface of the inspection object was less than 10 mJ/cm2, and the fluctuation was less than 5%. The real-time recorded laser energy was used to correct the PA signal to reduce the influence of the laser energy fluctuation. In our study, we applied a threshold of 35% of the maximum pixel value for all PA images to reject signals with quite low intensity and reduce the interference. In PA/US dual imaging, both modes are synchronized and imaged in real time at a frame rate of 10 Hz. For the calculation of sO2, photoacoustic is alternately imaged at two wavelengths, and the calculated distribution is displayed at a frame rate of 5 Hz.
2.1.2 Oxygen saturation (sO2) concept
Blood perfusion and metabolism significantly vary between organs and even within different tissues in an organ under various physiological conditions. To ensure healthy functioning of the human body, proper blood circulation is extremely important as it provides sufficient oxygen to organs and tissues. Oxygen transport is achieved through hemoglobin, which has two forms, namely oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb). Blood sO2 refers to the percentage of HbO2 volume in the blood to the total hemoglobin that can be bound (i.e., the concentration of oxygen in the blood). Currently, sO2 is used to estimate the oxygen-carrying capacity of hemoglobin. The formula used is expressed as follows:
In the above equation, [HbO2] and [Hb] represent the concentrations of HbO2 and Hb, respectively. sO2 depends on the oxygen partial pressure of the blood (SpO2). Under normoxic conditions, arterial blood sO2 (SaO2) and venous blood sO2 (SvO2) are 93%–98% and 70%–75%, respectively. Clinically, the oxygen content of human blood and tissues is reflected by sO2. Unlike normal tissues, malignant tumors often present abundant blood vessels and low oxygen saturability. Therefore, PAI can distinguish benign from malignant masses. The principle of PAI in measuring sO2 is similar to that of a pulse oximeter, which depends on the absorption of optical by HbO2 and Hb, and has similar limitations. COHb (carboxyhemoglobin), MetHb (methemoglobin), HbO2, and Hb have common absorption characteristics in the same spectral region. From the above formula, COHb and MetHb are not considered. The results of sO2 measurement are affected by the concentrations of COHb and MetHb. However, the pulse oximeter currently used in the clinic usually assumes that there is no nonfunctional hemoglobin in the blood or that its concentration is very low, which has no obvious effect on the detection results of blood oxygen saturation.
2.1.3 Image reconstruction and algorithm
As shown in Fig. 1, the absorption of light by hemoglobin greatly increases, while the scattering effect of tissue is quite strong in the visible light range below 700 nm [59]. In the red spectral region (600–700 nm), the absorption coefficients of HbO2 and Hb differ substantially; the degree of light absorption largely depends on the saturation of O2. As the absorption of HbO2 and Hb is similar in the infrared spectral region (800–1000 nm), the absorption of light mainly reflects the total amount of HbO2 and Hb. Therefore, sO2 can be quantified by exploiting the absorption spectra of human arteries.
Laser propagating in tissues is not only absorbed but also scattered by tissues. The algorithm used in this manuscript calculates the sO2 according to the absorption coefficient. To obtain accurate values, it is necessary not only to ensure that the initial optical fluence of lasers with different wavelengths is consistent but also to consider the differences in the variation of optical fluence transmission along depth in tissue at different wavelengths. Thus, we selected two relatively close wavelengths (λ1 = 750 nm and λ2 = 830 nm) to reduce the influence of the difference at the same depth. The deoxygenated hemoglobin has stronger absorption at 750 nm, and the oxygenated hemoglobin has stronger absorption at 830 nm. In the depth of 5–20 mm, the difference between the reduced scattering coefficients and absorption coefficient between the two wavelengths is small, so that more accurate sO2 can be obtained [60]. Besides, the pulse energy of the OPO laser maintains a high level at these two wavelengths, which could benefit the overall SNR.
In circumstances where the device emits light at two wavelengths in the near-infrared spectral region, we assume that only changes in Hb and HbO2 are considered. The absorption coefficient under two wavelengths can be expressed as Eq. (2):
In the above equations, CHb and CHbO2 represent the Hb and HbO2 content, respectively, and $\varepsilon _{Hb}^{\lambda 1}$, $\varepsilon _{Hb}^{\lambda 2}$, $\varepsilon _{HbO2}^{\lambda 1}$, and $\varepsilon _{HbO2}^{\lambda 2}$ represent the extinction coefficients of the Hb and HbO2 at wavelengths λ1 and λ2.
Assuming that the intensity of the PA signal is proportional to the absorption coefficient of the tissue, a simplified equation can be described as Eq. (3):
In this equation, Aλ1 and Aλ2 represent the intensity information of the PA signal obtained at wavelengths λ1 and λ2, respectively.
The process of obtaining numerical results of sO2 was as follows. According to the US image, the region of interest (ROI) was selected, and then the mean sO2 in the region was calculated on the PA image. Finally, these means were averaged on the time scale.
After comprehensive consideration of reconstruction accuracy and imaging speed, the delay-and-sum algorithm was used for image reconstruction for both ultrasound and photoacoustic imaging in our system. The images were aligned with ultrasound and photoacoustic imaging results to ensure the system performs dual mode at a frame rate of 10 Hz.
2.1.4 Reference monitor
The reference device of the animal was a patient monitor (BeneView T5; Mindray, Shenzhen, China), with its pulse oximeter probe 512E. The reference device of humans was also a patient monitor (BeneVision N12; Mindray, Shenzhen, China) and its pulse oximeter probe 512E.
2.2 Material
The breast tissue phantom consisted of 0.1% TiO2 (Sigma-Aldrich, MO, USA) and gel wax (Yaley Enterprises, CA, USA). The gel wax was heated to the liquid state at 170°C and placed into a sonicator (KQ3200ES, KunShan Ultrasonic Instruments, JiangSu, China), which sustained a water bath at 100°C. Subsequently, TiO2 was added to the gel wax solution for mixing and degassing. When TiO2 particles or bubbles were no longer evident, the mixed solution was slowly poured into an acrylic container at 70°C. Four silicon tubes with outer and inner diameters of 1 mm and 0.5 mm, respectively, were used to simulate blood vessels, and they were placed at the following four depths: 10, 15, 20, and 25 mm (Fig. 2). To mimic blood, we used a mixture solution of NiSO4 (pseudo-oxygenated hemoglobin) and CuSO4 (pseudo-deoxygenated hemoglobin) [61]. The solution was mixed in different proportions and presented as different blood sO2 levels in PAI. The sO2 levels of blood mimics were set to four levels, namely, 100%, 90%, 80%, and 70%, and ink was used as a normalized energy reference. The reference sO2 values of the pseudo-blood solution were calculated based on the following equation [62]:
According to the formula, the concentration ratio of copper sulfate and nickel sulfate under different blood oxygen levels can be calculated. Then, the volume ratio can be obtained according to the concentration. We first prepared the original solution of 0.5 M/L copper sulfate and 2.2 M/L nickel sulfate, and then mixed them according to the calculated volume ratio. Finally, we obtained the required imitation blood oxygen concentration solution.
Fig. 2. Schematic diagram of breast tissue and vessel phantom, showing the silicon tubes at four depths.
2.3 Animals
The animals involved in this study were six 1–2-year-old healthy beagles (four males and two females), weighing 10–10.5 kg. This study was approved by the Ethics Committee of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences.
2.4 Subject enrollment
Eight healthy volunteers, comprising five men and three women, who met the ASA classification I of the Anesthesia Association, were enrolled as human subjects in this study. The volunteers were healthy adults with carboxyhemoglobin (COHb) < 3%, methemoglobin (MetHb) < 2%, and total hemoglobin concentration (ctHb) > 10 g/dL. This study was approved by the Ethics Committee of Peking Union Medical College Hospital, Chinese Academy of Medical Sciences. All volunteers signed informed consent forms before the experiment.
2.5 Methods
2.5.1 Phantom study
First, a pseudo-blood solution of 100% sO2 level was poured into the tubes. Then, gray-scale US imaging was performed to locate the longitudinal section of the silicon tubes. Moreover, PAI (laser wavelengths of 750 nm and 830 nm) was performed in both the longitudinal and transverse sections. Next, pseudo sO2 was measured three times at four 5-equal points of the tubes, and the average values were calculated. Finally, the pseudo-blood solution of other pseudo sO2 levels was replaced (90%, 80%, and 70%), and the above operations were repeated (Fig. 3).
Fig. 3. Photograph of breast tissue and vessel phantom (yellow asterisk: breast phantom; yellow arrow: hand-held PA/US transducer).
2.5.2 Animal study
The beagles were given intramuscular anesthesia and then wore masks to inhale mixed gas. The probe of the patient monitor was fixed on each dog’s tongue to obtain real-time arterial blood sO2 as the reference sO2. The right hind leg of each dog was shaved, and PAI was performed to estimate the femoral arteries’ blood sO2 as PAI sO2. The femoral artery depth of beagle dogs is approximately 5–8 mm, which is within the effective depth of the device. First, to maintain the reference sO2 beyond 90%, mixed gas with oxygen concentration above 21% was provided through masks. By increasing the proportion of nitrogen (N2), the reference sO2 decreased to approach 70%. Then, we scaled down the proportion of nitrogen to ensure that the reference sO2 rises to the baseline level. The blood sO2 estimated by two devices in real time was recorded (Fig. 4). PAI sO2 was measured twice at each reference sO2 plateau, and the average value was recorded. Each dog's experiment was carried out independently. Under anesthesia, the respiration was relatively gentle, and the error caused by respiration was relatively low.
Fig. 4. PA/US dual imaging system. a. sO2 evaluation of beagles. b. sO2 measurement of dog’s tongue artery with pulse oximetry, the probe of the monitor was fixed on dog’s tongue to obtain real-time arterial blood sO2 as reference sO2. c. sO2 measurement of dog’s femoral artery with PAI, the probe was placed on the dog’s shaved right hind leg. d. Doppler ultrasound image of a beagle’s femoral artery (yellow asterisk: patient monitor; yellow arrows: PAI system).
2.5.3 Clinical studies in humans
Clinical studies in humans were conducted in accordance with Medical electrical equipment: Particular requirements for basic safety and essential performance of pulse oximeter equipment (ISO 80601-2-61:2017) [63]. We followed the Guideline for evaluating and documenting SpO2 accuracy in human subjects. For enrollment of subjects, we followed the inclusion criteria of the Guideline. Regarding the range of sO2, the Guideline states that to evaluate the performance of a tested device in humans, the accuracy (represented as RMSE) should be less than or equal to 4.0% in the range of 70%–100%. In the design of the experiment, we strictly followed the Guideline. During the experiment, we closely monitored the volunteers’ condition and developed an emergency plan for oxygen inhalation. No adverse reactions occurred during the whole experiment.
In this study, we validated the sO2 accuracy of PAI in comparison with secondary standard pulse oximeter equipment (patient monitor with pulse oximeter probe). Before scanning, the eight volunteers were asked to remove anything that may affect the measurement results on arms and fingers, such as nail polish. Four volunteers, comprising three men and one woman, underwent real-time sO2 evaluation. They were instructed to lie supine on the examination table, wear masks, and breathe normally. The concentration of O2 in the mixed gas inhaled was reduced by scaling up the proportion of N2. A total of five sO2 plateaus (i.e., 100%–97%, 96%–92%, 91%–85%, 84%–78%, and 77%–70%) were set up [63]. To ensure that two different devices were evaluating the same artery branches, we examined the volunteers’ right arms and fingers, which are less affected by breathing. In addition, the depth of objects was approximately 4–14 mm, which was within the effective depth of the PAI device. The probe of the patient monitor was placed on the volunteer’s right index finger to provide reference sO2, and that of PAI was placed on the right radial artery. Gray-scale and color Doppler US were first performed to detect the vessel and visualize the blood flow within it to verify that it was an artery. Then, PA/US imaging was performed in the same site. The room temperature was set at 25°C and was held constant.
First, an O2 concentration of more than 21% was provided so that the reference sO2 was higher than 97% and remained stable for more than 30 s. The PAI sO2 was recorded. Then, the proportion of N2 was scaled up to reduce the reference sO2 to approximately 92%, and remained stable for more than 30 s. The PAI sO2 was subsequently recorded. The above steps were repeated to continuously increase the N2 ratio to the critical value until the data collection process was complete. To decrease the chances of errors, an examination was performed twice at each plateau, and each set of data was measured twice. Each volunteer's experiment was carried out independently.
Four other healthy volunteers, comprising two men and two women, were instructed to lie supine on the examination table. The radial artery was first displayed under gray-scale ultrasound and confirmed by color Doppler. The depths of artery were measured in gray-scale US mode, by which six sites were settled where the depth of the radial artery (the distance from the skin surface to the anterior wall of the artery) was 4 mm, 6 mm, 8 mm, 10 mm, 12 mm, and 14 mm. Then PAI sO2 measurement was performed at these six sites of the the right radial artery. PAI sO2 was measured three times at each depth, and the average values were calculated. The room temperature was set at 25°C and held constant.
During the deoxygenation process, it was necessary to ensure that the blood oxygen level of the subject reached the designated plateau and remained stable for more than 30 s. Moreover, the interval time between two measurements had to be more than 20 s to ensure data independence.
2.5.4 Statistical analysis
We used mean ± standard deviation to describe the data distribution. Correlations between the reference sO2 and PAI sO2 were determined by computing Pearson’s correlation coefficient R. We used the root mean square error (RMSE) and the mean absolute error (MAE) to evaluate PAI performance, where RMSE expressed the accuracy between the reference sO2 and PAI sO2. Statistical analysis was performed using SPSS for Windows 22.0 (SPSS Inc., Chicago, IL, USA) and MedCalc 19.0.4 (Medical, Ostend, Belgium) software.
3. Results
3.1 Phantom study
The PAI sO2 values of four pseudo-sO2 levels blood-mimicking phantoms (i.e., 100%, 90%, 80%, and 70%) at different depths are presented in Tables 1 and 2. Figure 5 shows the PAI sO2 of the blood-mimicking phantoms at four depths in the transverse section.
Table 1. sO2 measurement results of blood-mimicking phantoms by PAI at different depths
Table 2. sO2 measurement results of blood-mimicking phantoms of different pseudo-sO2 levels via PAI
Fig. 5. Results of pseudo-blood sO2 estimation of PAI. The green box is the ROI, and the red box is the result of sO2. a–d. 20 mm depth: for 100%, 90%, 80%, and 70% sO2 levels, the PAI sO2 were 99.13%, 87.65%, 77.01%, and 65.60%, respectively; e. 15 mm depth: for the 80% sO2 group, the PAI sO2 was 81.66%; and f. 25 mm depth: for the 80% sO2 group, the PAI sO2 was 76.18%.
The mean values of the PAI sO2 of the four pseudo-sO2 level blood-mimicking phantoms were 98.86% (100%), 90.39% (90%), 78.63% (80%), and 69.69% (70%). The overall RMSE and MAE of the PAI sO2 were 1.49% and 1.09%, respectively. Pearson's correlation coefficient (R) between the pseudo sO2 and the PAI sO2 was 0.968 (P < 0.001), indicating a strong correlation. The RMSE of the PAI sO2 at four depths was 0.67% (10 mm), 1.23% (15 mm), 1.76 (20 mm), and 1.97% (25 mm). The MAE of the PAI sO2 at different depths was 0.53% (10 mm), 0.96% (15 mm), 1.46 (20 mm), and 1.42% (25 mm). The RMSE of the PAI sO2 of the four pseudo-sO2 level blood-mimicking phantoms was 0.18% (100%), 1.31% (90%), 2.32 (80%), and 1.31% (70%). The MAE of the PAI sO2 of the four pseudo-sO2 level blood-mimicking phantoms was 0.14% (100%), 1.29% (90%), 1.82 (80%), and 1.11% (70%). From 100% to 80%, with a decrease in the sO2 level, the standard deviation, RMSE, and MAE of the PAI sO2 increased. However, this trend did not apply to the blood-mimicking phantoms at 70% sO2 level, which showed lower standard deviation (1.32%), RMSE (1.31%), and MAE (1.11%) compared with those at 80% sO2 level (standard deviation, 1.93%; RMSE, 2.32%; MAE, 1.82%).
Figure 6 shows the PAI sO2 at different depths. The average RMSE of the PAI sO2 in each depth showed an increasing trend with depth. For the blood-mimicking phantoms at 100% pseudo-sO2 level, the RMSE of the four depths remained at a low level with only slight fluctuations. For the blood-mimicking phantoms at 90% and 80% pseudo sO2, the accuracy of the PAI sO2 (RMSE) showed a decreasing trend (an increasing trend in RMSE) with depth. However, for the 70% sO2 level, the RMSE showed an abnormal relative decrease at 20 mm and 25 mm depths, and a more accurate result was obtained at a depth of 25 mm, compared with that of the 90% and 80% sO2 levels.
3.2 Results of the animal study
The results of the PAI sO2 in beagles are presented in Table 3. The overall RMSE and MAE were 2.16% and 1.58%, respectively. In animal study, each dog's experiment was carried out independently. The initial blood sO2, trend of blood sO2 changing, and measurement time had individual differences. The PAI sO2 and the reference sO2 of beagle #1 is shown in Fig. 7. The results of the other five animals are shown in Figs. S1-S5. Pearson's correlation coefficient (R) between the reference sO2 and the PAI sO2 of the whole group was 0.965 (P < 0.001) (Figs. 8 and 9).
Table 3. RMSE and MAE of the PAI sO2 of six beagles’ femoral arteries
Fig. 9. Results of canine (beagle #5) femoral artery blood sO2 estimation of PAI. The green box is the ROI, and the red box is the result of sO2. a–c. The monitor-tongue sO2 values were 94%, 85%, and 78%, and the PA-estimated sO2 values were 93.1%, 84.8%, and 78.2%.
3.3 Results of the clinical studies in humans
In real-time sO2 evaluation in humans, the results showed that PAI achieved an RMSE of 3.97% and an MAE of 3.28% (R = 0.892; P < 0.001). According to ISO 80601-2-61:2017, the accuracy was less than 4.0% over the range of 70% to 100% reference sO2 [63]. The results are presented in Figs. 10 and 11. The rest of the volunteer data are shown in Figs. S5-S8.
The effects of depth on PAI evaluation showed that the standard deviations of the PAI sO2 of radial arteries at six depths from 4 mm to 14 mm were 1.98%, 4.00%, 4.23%, 8.05%, 5.72%, and 11.87%. The RMSE of the PAI sO2 at six depths was 2.66%, 6.89%, 13.01%, 12.13%, 13.63%, and 24.96%, showed a general increasing trend with depth, especially within the depth from 4 to 8 mm. The standard deviation also increased significantly at depths above 8 mm, indicating that the accuracy of PAI sO2 of radial arteries decreased with the increasing depth (Table 4 and Figs. 12 and 13).
Table 4. Errors and RMSE of the PAI sO2 measurement of four volunteers’ radial arteries at different depths
Fig. 12. Results of human radial artery blood sO2 estimation of PAI. The green box is the ROI, and the red box is the result of sO2. a–c. The reference sO2 values were 95%, 86%, and 69%, and the PAI sO2 values were 92.19%, 86.79%, and 75.83%.
4. Discussion
Accurate assessment of blood sO2 is crucial for the early diagnosis, treatment, and prognostic evaluation of diseases [9,43,51,53,54]. PA/US dual imaging is a novel technique with reasonably good spatial resolution, which allows both real-time qualitative imaging and measurement of sO2 in deeper tissues, compared with traditional oximeters. The ability of PAI in providing both structural and metabolic information of blood sO2 exhibits great potential in clinical application and has attracted significant attention [33,42–47]. However, only a few studies have been published on validating the PAI performance in evaluating sO2. Among these studies, some did not consider large and specific sO2 ranges and others only validated in animals or phantoms. In this study, a multiscale verification was performed in phantoms, animals, and humans with a large span of blood sO2 plateaus, to validate the feasibility and stability of the hand-held PA/US imaging system in real-time in vivo sO2 measurement.
The phantom results showed that PAI achieved an accuracy of 1.49% and a mean absolute error (MAE) of 1.09% within 25 mm depth, and good linearity (R = 0.968; p < 0.001) was obtained between PAI sO2 and reference sO2. The in vivo studies were conducted in dogs and human volunteers with various blood sO2 plateaus. The results showed that the RMSE and MAE of PAI sO2 in canines were 2.16% and 1.58%, respectively. Pearson's correlation coefficient (R) between the reference sO2 and the PAI sO2 was 0.965 (P < 0.001), and the sO2 curves of the reference sO2 and PAI sO2 were consistent. The in vivo human study showed that with changes in different sO2 levels, PAI sO2 and reference sO2 were mostly consistent. PAI achieved an RMSE of 3.97% and an MAE of 3.28% (R = 0.892; P < 0.001). Finally, we designed a study to explore the effects of depth on PAI sO2 evaluation, where both the average errors and RMSE of the PAI sO2 of radial arteries at six depths from 4 to 14 mm were calculated. The average errors showed a general increasing trend with depth. Above all, the results demonstrated that PAI can estimate blood sO2 with agreeable accuracy.
The Fig. 8 and 11 shows the linear correlation between the PAI sO2 and reference sO2 in both animals and humans, of which the two fitting equations demonstrate fitting slopes less than 1 and intercepts of 10.02 and 20.31. This result may be caused by the following three reasons: (1) The system uses the method of monitoring the output energy of laser to correct the fluctuation of laser energy. In fact, the light output by the laser needs to be transmitted through the optical fiber to reach the area to be illuminated by the sample. However, fiber has different coupling efficiency and transmission efficiency for lasers of different wavelengths, resulting in a deviation of optical fluence when lasers of different wavelengths finally reach the illumination area. So, we can reduce the intercept generated by the optical fiber by measuring the coupling efficiency and transmission efficiency of the fiber at the wavelengths used. (2) Similarly, different wavelengths have different fluence distribution in tissues, and the difference of fluence will also lead to deviation of calculation results. (3) In addition to the influence of scatterers in tissues on laser, the absorption of blood will also affect the distribution of optical fluence. When there are abundant blood vessels in the near field, the absorption of laser by blood in the near field will affect the far field optical fluence, while the absorption of laser by blood with different sO2 is different. The blood with low sO2 has a strong absorption of 750 nm laser, which makes the far field 830 nm laser have a large fluence, and the calculation result of sO2 will be on the high side. On the contrary, the blood with high sO2 has a strong absorption of 830 nm laser, which leads to a low calculation result in the far field, so the fitting slope will be less than 1.
However, the capability of in vivo PA/US dual imaging still has some limitations. Due to the complexity of light absorption and scattering in tissue, it is difficult to accurately calculate the actual value of sO2 by PAI, which indicates that PAI reflects the relative in vivo sO2 level of blood. Regarding the effects of depth on PAI sO2 evaluation, we found that the errors of PAI sO2 increased with depth, and the errors and standard deviations were maintained at a low level within 8 mm, indicating that PAI sO2 may be able to achieve a relatively accurate and precise result in shallow tissues. However, in deep tissues, the accuracy of PAI sO2 dropped markedly, which is the main challenge for PAI to overcome. In this study, the error range of PAI sO2 at different depths after multiple measurements was given, which may provide the basis and direction for a further study of depth compensation in PAI.
Several studies have assessed the influence of depth on PAI sO2 evaluation. Rianne et al. placed four tubes filled with human blood at different depths in a two-slab soft tissue phantom and performed PAI to evaluate the sO2 of the blood. Only three shallower tubes (within 20 mm) were visible in the PAI, while the location of the fourth tube was not visible [50]. William et al. compared the accuracy of the PAI sO2 estimation of different blood tube depths from 5 to 35 mm; the maximum depth was deeper than that in our study. The RMSE of the PAI sO2 increased with the depth from 4% to 14% and finally achieved 5%–10% through fluence correction [52]. In our study, the average RMSE of the PAI sO2 at the four depths was 0.67% (10 mm), 1.23% (15 mm), 1.76% (20 mm), and 1.97% (25 mm). This suggests that the accuracy of the PAI sO2 decreased with depth, thus exhibiting a depth dependence from 10 to 25 mm. Compared with the previous studies of tissue phantoms, our PAI system achieved a relatively better accuracy, which can be attributed to the post-processing algorithm we used in this system. The received PA signals could be corrected through comparison of the actual output energy at the transmitter with the theoretical output laser energy, which made the PAI sO2 closer to the real value. A multiple-frame average method in PAI sO2 estimation was also used to obtain relatively stable sO2 results. Moreover, the integrated probe improved the reliability and stability of the detection of the PA signals.
The limitation of this study is that we used the original raw results of the PA system to evaluate sO2 without a corresponding luminous flux compensation method to correct the influence of the depth, which has been proven to be significant in the PAI sO2 measurement in previous studies [52,60,64]. We have paid attention to the work of fluence correction. For example, for the breast region, through analyzing the lighting mode and breast anatomy, we used COMSOL software to establish a model, simulate the change of fluence in the tissue, and correct the sO2 calculation results [60]. In this study, the experimental results revealed that in a certain depth range, without fluence correction, the measurement accuracy was also reliable, which could help doctors qualitatively diagnose the regional blood oxygen saturation of patients. The purpose of this paper was not to provide an accurate result of blood oxygen saturation but to provide the real clinical performance of photoacoustic function in measuring blood oxygen saturation, and also to provide the basis for designing different models for fluence correction in different tissues in the future. We also found that at the lower sO2 level (70%), the RSME of the PAI sO2 decreased abnormally, which might be due to a low signal-to-noise ratio. In the future, an improved algorithm will be developed to reduce this error, and the accuracy of the PAI sO2 assessment can be further improved. Finally, we applied two wavelengths mode PAI to calculate the blood sO2 in this study, which allowed us to achieve a real-time imaging system combing conventional US and PAI. It has been reported that multiple wavelengths PAI achieves a better performance in sO2 evaluation [65], while it requires a high repetition frequency of laser and the imaging speed is limited. In follow-up study, we will concentrate on a multiple wavelengths imaging system, which may achieve a good balance on the accuracy of sO2 calculation and the imaging speed.
5. Conclusions
In this study, we verified the feasibility and stability of a hand-held PA/US dual-imaging system in evaluating sO2 through a multiscale method, which encompassed phantoms with different sO2 concentrations, animals, and humans with a large span of in vivo blood sO2 plateaus. By comparing the results of the PA/US dual-imaging system and patient monitors, we showed that sO2 evaluation by the PAI system was accurate and stable under multiple sO2 plateaus in phantoms and in vivo, which not only demonstrated the technical stability and interpretability of this PAI device but also reflected the integrity and innovation of this study. Furthermore, our study might also provide a multiscale and standardized protocol for evaluating the performance of medical devices based on PAI, along with further optimizations of the system. The continuous improvement and optimization of technology is expected to enable better real-time, noninvasive, and accurate evaluations of sO2 in human tissues, which could provide a more reliable evidence-based evaluation method for the occurrence, diagnosis, and treatment of oxygen metabolism–related diseases such as tumors and inflammation.
Funding
CAMS Innovation Fund for Medical Sciences (CIFMS) (2020-I2M-C&T-B-035); National Natural Science Foundation of China (61971447); Beijing Municipal Natural Science Foundation (JQ18023); Beijing Nova Program Interdisciplinary Cooperation Project (xxjc2018023); International Science and Technology Cooperation Programme (2015DFA30440); National Key Research and Development Program from the Ministry of Science and Technology of the People’s Republic of China (2017YFE0104200).
Acknowledgments
We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
Data availability
Data underlying the results presented in this paper are not publicly available now but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. C. R. Carvalho, G. de Paula Pinto Schettino, B. Maranhão, and E. P. Bethlem, “Hyperoxia and lung disease,” Curr. Opin. Pulm. Med. 4(5), 300–304 (1998). [CrossRef]
2. J. L. Rodgers, D. Iyer, L. E. Rodgers, S. Vanthenapalli, and S. K. Panguluri, “Impact of hyperoxia on cardiac pathophysiology,” J. Cell. Physiol. 234(8), 12595–12603 (2019). [CrossRef]
3. D. M. Gilkes, G. L. Semenza, and D. Wirtz, “Hypoxia and the extracellular matrix: drivers of tumour metastasis,” Nat. Rev. Cancer 14(6), 430–439 (2014). [CrossRef]
4. K. DePeaux and G. M. Delgoffe, “Metabolic barriers to cancer immunotherapy,” Nat. Rev. Immunol. 21(12), 785–797 (2021). [CrossRef]
5. M. W. Dewhirst, Y. Cao, and B. Moeller, “Cycling hypoxia and free radicals regulate angiogenesis and radiotherapy response,” Nat. Rev. Cancer 8(6), 425–437 (2008). [CrossRef]
6. T. G. Graeber, C. Osmanian, T. Jacks, D. E. Housman, C. J. Koch, S. W. Lowe, and A. J. Giaccia, “Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours,” Nature 379(6560), 88–91 (1996). [CrossRef]
7. H. K. Eltzschig and P. Carmeliet, “Hypoxia and inflammation,” N Engl J Med 364(7), 656–665 (2011). [CrossRef]
8. R. A. Cairns, I. S. Harris, and T. W. Mak, “Regulation of cancer cell metabolism,” Nat. Rev. Cancer 11(2), 85–95 (2011). [CrossRef]
9. W. R. Wilson and M. P. Hay, “Targeting hypoxia in cancer therapy,” Nat. Rev. Cancer 11(6), 393–410 (2011). [CrossRef]
10. G. L. Semenza, M. K. Nejfelt, S. M. Chi, and S. E. Antonarakis, “Hypoxia-inducible nuclear factors bind to an enhancer element located 3’ to the human erythropoietin gene,” Proc. Natl. Acad. Sci. U. S. A. 88(13), 5680–5684 (1991). [CrossRef]
11. G. L. Wang, B. H. Jiang, E. A. Rue, and G. L. Semenza, “Hypoxia-inducible factor 1 is a basic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension,” Proc. Natl. Acad. Sci. U. S. A. 92(12), 5510–5514 (1995). [CrossRef]
12. P. H. Maxwell, M. S. Wiesener, G. W. Chang, S. C. Clifford, E. C. Vaux, M. E. Cockman, C. C. Wykoff, C. W. Pugh, E. R. Maher, and P. J. Ratcliffe, “The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis,” Nature 399(6733), 271–275 (1999). [CrossRef]
13. M. Ivan, K. Kondo, H. Yang, W. Kim, J. Valiando, M. Ohh, A. Salic, J. M. Asara, W. S. Lane, and W. G. Kaelin, “HIFalpha targeted for VHL-mediated destruction by proline hydroxylation: implications for O2 sensing,” Science 292(5516), 464–468 (2001). [CrossRef]
14. P. Jaakkola, D. R. Mole, Y. M. Tian, M. I. Wilson, J. Gielbert, S. J. Gaskell, A. von Kriegsheim, H. F. Hebestreit, M. Mukherji, C. J. Schofield, P. H. Maxwell, C. W. Pugh, and P. J. Ratcliffe, “Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation,” Science 292(5516), 468–472 (2001). [CrossRef]
15. J. W. Lee, J. Ko, C. Ju, and H. K. Eltzschig, “Hypoxia signaling in human diseases and therapeutic targets,” Exp Mol Med 51, 1 (2019).
16. C. M. Sehgal, P. H. Arger, S. E. Rowling, E. F. Conant, C. Reynolds, and J. A. Patton, “Quantitative vascularity of breast masses by Doppler imaging: regional variations and diagnostic implications,” J Ultrasound Med 19(7), 427–440 (2000). [CrossRef]
17. S. Mallidi, G. P. Luke, and S. Emelianov, “Photoacoustic imaging in cancer detection, diagnosis, and treatment guidance,” Trends Biotechnol. 29(5), 213–221 (2011). [CrossRef]
18. F. Dias-Freitas, C. Metelo-Coimbra, and R. Roncon-Albuquerque, “Molecular mechanisms underlying hyperoxia acute lung injury,” Respir Med 119, 23–28 (2016). [CrossRef]
19. E. D. Chan, M. M. Chan, and M. M. Chan, “Pulse oximetry: understanding its basic principles facilitates appreciation of its limitations,” Respir Med 107(6), 789–799 (2013). [CrossRef]
20. J. Y. A. Foo, K. P. Chua, and X. J. A. Tan, “Clinical applications and issues of oxygen saturation level measurements obtained from peripheral sites,” J Med Eng Technol 37(6), 388–395 (2013). [CrossRef]
21. K. D. Torp, P. Modi, and L. V. Simon, “Pulse Oximetry,” in StatPearls (StatPearls Publishing Copyright © 2021, StatPearls Publishing LLC., 2021).
22. M. Nitzan, A. Romem, and R. Koppel, “Pulse oximetry: fundamentals and technology update,” Med. Devices: Evidence Res. 7, 231–239 (2014). [CrossRef]
23. A. Jubran, “Pulse oximetry,” Crit Care 19(1), 272 (2015). [CrossRef]
24. W. Ottestad, J. I. Kåsin, and L. Ø. Høiseth, “Arterial oxygen saturation, pulse oximetry, and cerebral and tissue oximetry in hypobaric hypoxia,” Aerosp Med Hum Perform 89(12), 1045–1049 (2018). [CrossRef]
25. V. De Santis and M. Singer, “Tissue oxygen tension monitoring of organ perfusion: rationale, methodologies, and literature review,” Br J Anaesth 115(3), 357–365 (2015). [CrossRef]
26. J. Steppan and C. W. Hogue, “Cerebral and tissue oximetry,” Bailliere’s Best Pract. Res., Clin. Anaesthesiol. 28(4), 429–439 (2014). [CrossRef]
27. P. Bickler, J. Feiner, M. Rollins, and L. Meng, “Tissue oximetry and clinical outcomes,” Anesth. Analg. 124(1), 72–82 (2017). [CrossRef]
28. M. Nitzan, I. Nitzan, and Y. Arieli, “The Various Oximetric Techniques Used for the Evaluation of Blood Oxygenation,” Sensors (Basel, Switzerland) 20 (2020).
29. L. V. Wang, “Prospects of photoacoustic tomography,” Med. Phys. 35(12), 5758–5767 (2008). [CrossRef]
30. C. Baudelet and B. Gallez, “How does blood oxygen level-dependent (BOLD) contrast correlate with oxygen partial pressure (pO2) inside tumors?” Magn. Reson. Med. 48(6), 980–986 (2002). [CrossRef]
31. A. M. Dale and E. Halgren, “Spatiotemporal mapping of brain activity by integration of multiple imaging modalities,” Curr. Opin. Neurobiol. 11(2), 202–208 (2001). [CrossRef]
32. S. Ogawa, T. M. Lee, A. R. Kay, and D. W. Tank, “Brain magnetic resonance imaging with contrast dependent on blood oxygenation,” Proc. Natl. Acad. Sci. U. S. A. 87(24), 9868–9872 (1990). [CrossRef]
33. F. Knieling, C. Neufert, A. Hartmann, J. Claussen, A. Urich, C. Egger, M. Vetter, S. Fischer, L. Pfeifer, A. Hagel, C. Kielisch, R. S. Görtz, D. Wildner, M. Engel, J. Röther, W. Uter, J. Siebler, R. Atreya, W. Rascher, D. Strobel, M. F. Neurath, and M. J. Waldner, “Multispectral Optoacoustic Tomography for Assessment of Crohn's Disease Activity,” N Engl J Med 376(13), 1292–1294 (2017). [CrossRef]
34. I. Steinberg, D. M. Huland, O. Vermesh, H. E. Frostig, W. S. Tummers, and S. S. Gambhir, “Photoacoustic clinical imaging,” Photoacoustics 14, 77–98 (2019). [CrossRef]
35. A. B. E. Attia, G. Balasundaram, M. Moothanchery, U. S. Dinish, R. Bi, V. Ntziachristos, and M. Olivo, “A review of clinical photoacoustic imaging: Current and future trends,” Photoacoustics 16, 100144 (2019). [CrossRef]
36. M. Erfanzadeh and Q. Zhu, “Photoacoustic imaging with low-cost sources; A review,” Photoacoustics 14, 1–11 (2019). [CrossRef]
37. Z. Hosseinaee, M. Le, K. Bell, and P. H. Reza, “Towards non-contact photoacoustic imaging [review],” Photoacoustics 20, 100207 (2020). [CrossRef]
38. C. Zhao, R. Zhang, Y. Luo, S. Liu, T. Tang, F. Yang, L. Zhu, X. He, M. Yang, and Y. Jiang, “Multimodal VEGF-Targeted Contrast-Enhanced Ultrasound and Photoacoustic Imaging of Rats with Inflammatory Arthritis: Using Dye-VEGF-Antibody-Loaded Microbubbles,” Ultrasound Med Biol 46(9), 2400–2411 (2020). [CrossRef]
39. S. Zackrisson, S. van de Ven, and S. S. Gambhir, “Light in and sound out: emerging translational strategies for photoacoustic imaging,” Cancer Res. 74(4), 979–1004 (2014). [CrossRef]
40. P. Beard, “Biomedical photoacoustic imaging,” Interface Focus. 1(4), 602–631 (2011). [CrossRef]
41. C. Zhao, Q. Wang, X. Tao, M. Wang, C. Yu, S. Liu, M. Li, X. Tian, Z. Qi, J. Li, F. Yang, L. Zhu, X. He, X. Zeng, Y. Jiang, and M. Yang, “Multimodal photoacoustic/ultrasonic imaging system: a promising imaging method for the evaluation of disease activity in rheumatoid arthritis,” Eur. Radiol. 31(5), 3542–3552 (2021). [CrossRef]
42. L. V. Wang and S. Hu, “Photoacoustic tomography: in vivo imaging from organelles to organs,” Science 335(6075), 1458–1462 (2012). [CrossRef]
43. J. Yang, G. Zhang, W. Chang, Z. Chi, Q. Shang, M. Wu, T. Pan, L. Huang, and H. Jiang, “Photoacoustic imaging of hemodynamic changes in forearm skeletal muscle during cuff occlusion,” Biomed. Opt. Express 11(8), 4560–4570 (2020). [CrossRef]
44. S. Manohar and S. S. Gambhir, “Clinical photoacoustic imaging,” Photoacoustics 19, 100196 (2020). [CrossRef]
45. C. Yang, H. Lan, F. Gao, and F. Gao, “Review of deep learning for photoacoustic imaging,” Photoacoustics 21, 100215 (2021). [CrossRef]
46. S. Manohar and M. Dantuma, “Current and future trends in photoacoustic breast imaging,” Photoacoustics 16, 100134 (2019). [CrossRef]
47. M. Yang, L. Y. Zhao, X. J. He, N. Su, C. Y. Zhao, H. W. Tang, T. Hong, W. B. Li, F. Yang, L. Lin, B. Zhang, R. Zhang, Y. X. Jiang, and C. H. Li, “Photoacoustic/ultrasound dual imaging of human thyroid cancers: an initial clinical study,” Biomed. Opt. Express 8(7), 3449–3457 (2017). [CrossRef]
48. H. M. Heres, M. U. Arabul, M. C. M. Rutten, F. N. Van de Vosse, and R. G. P. Lopata, “Visualization of vasculature using a hand-held photoacoustic probe: phantom and in vivo validation,” J. Biomed. Opt. 22(4), 041013 (2017). [CrossRef]
49. J. Xia, A. Danielli, Y. Liu, L. Wang, K. Maslov, and L. V. Wang, “Calibration-free quantification of absolute oxygen saturation based on the dynamics of photoacoustic signals,” Opt. Lett. 38(15), 2800–2803 (2013). [CrossRef]
50. R. Bulsink, M. Kuniyil Ajith Singh, M. Xavierselvan, S. Mallidi, W. Steenbergen, and K. J. Francis, “Oxygen Saturation Imaging Using LED-Based Photoacoustic System,” Sensors 21(1), 283 (2021). [CrossRef]
51. T. Mitcham, H. Taghavi, J. Long, C. Wood, D. Fuentes, W. Stefan, J. Ward, and R. Bouchard, “Photoacoustic-based sO2 estimation through excised bovine prostate tissue with interstitial light delivery,” Photoacoustics 7, 47–56 (2017). [CrossRef]
52. W. C. Vogt, X. Zhou, R. Andriani, K. A. Wear, T. J. Pfefer, and B. S. Garra, “Photoacoustic oximetry imaging performance evaluation using dynamic blood flow phantoms with tunable oxygen saturation,” Biomed. Opt. Express 10(2), 449–464 (2019). [CrossRef]
53. S. N. Hennen, W. Xing, Y.-B. Shui, Y. Zhou, J. Kalishman, L. B. Andrews-Kaminsky, M. A. Kass, D. C. Beebe, K. I. Maslov, and L. V. Wang, “Photoacoustic tomography imaging and estimation of oxygen saturation of hemoglobin in ocular tissue of rabbits,” Exp. Eye Res. 138, 153–158 (2015). [CrossRef]
54. J. Kang, E. M. Boctor, S. Adams, E. Kulikowicz, H. K. Zhang, R. C. Koehler, and E. M. Graham, “Validation of noninvasive photoacoustic measurements of sagittal sinus oxyhemoglobin saturation in hypoxic neonatal piglets,” J. Appl. Physiol. 125(4), 983–989 (2018). [CrossRef]
55. A. Merdasa, J. Bunke, M. Naumovska, J. Albinsson, T. Erlöv, M. Cinthio, N. Reistad, R. Sheikh, and M. Malmsjö, “Photoacoustic imaging of the spatial distribution of oxygen saturation in an ischemia-reperfusion model in humans,” Biomed. Opt. Express 12(4), 2484–2495 (2021). [CrossRef]
56. R. Zhang, L.-Y. Zhao, C.-Y. Zhao, M. Wang, S.-R. Liu, J.-C. Li, R.-N. Zhao, R.-J. Wang, F. Yang, L. Zhu, X.-J. He, C.-H. Li, Y.-X. Jiang, and M. Yang, “Exploring the diagnostic value of photoacoustic imaging for breast cancer: the identification of regional photoacoustic signal differences of breast tumors,” Biomed. Opt. Express 12(3), 1407–1421 (2021). [CrossRef]
57. M. Wang, L. Zhao, Y. Wei, J. Li, Z. Qi, N. Su, C. Zhao, R. Zhang, T. Tang, S. Liu, F. Yang, L. Zhu, X. He, C. Li, Y. Jiang, and M. Yang, “Functional photoacoustic/ultrasound imaging for the assessment of breast intraductal lesions: preliminary clinical findings,” Biomed. Opt. Express 12(3), 1236–1246 (2021). [CrossRef]
58. C.-Y. Zhao, Y.-X. Jiang, J.-C. Li, Z.-H. Xu, Q. Zhang, N. Su, and M. Yang, “Role of Contrast-enhanced Ultrasound in the Evaluation of Inflammatory Arthritis,” Chin Med J 130(14), 1722–1730 (2017). [CrossRef]
59. S. Prahl, “Tabulated Molar Extinction Coefficient for Hemoglobin in Water,” http://omlc.ogi.edu/spectra/hemoglobin/summary.html Available online (1999).
60. T. Han, M. Yang, F. Yang, L. Zhao, Y. Jiang, and C. Li, “A three-dimensional modeling method for quantitative photoacoustic breast imaging with handheld probe,” Photoacoustics 21, 100222 (2021). [CrossRef]
61. B. F. Martina, A. Lu, and T. C. Benjamin, “Sulfates as chromophores for multiwavelength photoacoustic imaging phantoms,” J. Biomed. Opt. 22(12), 1 (2017). [CrossRef]
62. M. Fonseca, E. Malone, F. Lucka, R. Ellwood, L. An, S. Arridge, P. Beard, and B. Cox, “Three-dimensional photoacoustic imaging and inversion for accurate quantification of chromophore distributions,” in SPIE BiOS, A. A. Oraevsky and L. V. Wang, eds. (2017), p. 1006415.
63. “ISO 80601-2-61:2017,” pp. Medical electrical equipment Part 2-61: Particular requirements for basic safety and essential performance of pulse oximeter equipment.
64. A. Q. Bauer, R. E. Nothdurft, T. N. Erpelding, L. V. Wang, and J. P. Culver, “Quantitative photoacoustic imaging: correcting for heterogeneous light fluence distributions using diffuse optical tomography,” J. Biomed. Opt. 16(9), 096016 (2011). [CrossRef]
65. T. Kirchner and M. Frenz, “Multiple illumination learned spectral decoloring for quantitative optoacoustic oximetry imaging,” J. Biomed. Opt. 26(08), 1 (2021). [CrossRef]
