Maximum fluence for accurate functional photoacoustic microscopy Authors: Jingyi Zhu, Zhiwei Yao, Lidai Wang Published: 2026 ABSTRACT AbstractPhotoacoustic microscopy (PAM) can map blood oxygen saturation (sO2) in the brain and other organs at high spatial and temporal resolution, enabling many preclinical and clinical applications. In fast functional PAM, even when the laser power is within the safety limit, transient local thermal accumulation might adversely affect the accuracy of sO2 measurement, which, however, has not been thoroughly studied. Here, we investigated the relationship between laser irradiation and sO2 accuracy and established a numerical model to estimate the maximum laser fluence in typical fast functional PAM. In vivo experiments show that, when using 532-nm and 558-nm wavelengths for functional imaging, the maximum fluence (Fmax) for accurate sO2 measurement decreases with repetitive laser pulse number (N) on one spot as Fmax=45N−0.41 mJ·cm−2. We validated the model in a blood phantom and the mouse brain and tested its stability under long-term scanning. The maximum-fluence model provides practical guidance for selecting laser fluence in fast functional PAM.Keywords: Photoacoustic microscopy, Blood oxygen saturation, Functional imaging, Optical fluence limit Introduction Blood oxygen saturation (sO2) is a key physiological indicator reflecting tissue oxygenation status and metabolic level, and has broad biomedical applications, such as brain function monitoring and tumor microenvironment assessment [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]. Achieving noninvasive quantitative imaging of sO2 at high spatiotemporal resolution can reveal oxygen transport patterns at the microvascular level and provide direct physiological evidence of changes in functional states [11], [12], [13], [14], [15], [16], [17], [18], [19]. Photoacoustic microscopy (PAM) combines optical absorption contrast with acoustic detection, enabling high-resolution imaging of microvascular networks and their blood oxygenation information without labeling [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36]. To increase the signal-to-noise ratio and imaging speed, PAM often uses lasers with high pulse energy and high pulse repetition rate. Short-term local thermal accumulation under certain irradiation conditions may affect the accuracy of sO2 measurement, especially in fast PAM scanning. Even if the laser fluence is generally safe [37], [38], [39], [40], [41], repetitive irradiation with multiple pulses may cause short-term local heat accumulation, potentially triggering oxygen dissociation or other complicated changes. Currently, although most fast-functional PAM systems use laser pulses within safety limits [32], [36], [42], there remains a lack of quantitative research on how fast-pulsed laser irradiation affects the accuracy of oxygen saturation measurement. Here, we investigate how nanosecond laser irradiation affects sO2 accuracy and establish a numerical model that describes this relationship. In vivo experiments show that when using 532 nm and 558 nm nanosecond pulsed lasers, the maximum fluence (Fmax) for accurate measurement of sO₂ decreases with the laser pulse number (N) that repetitively irradiates the same location within a fixed irradiation time. This was further validated in blood phantom and mouse brain imaging. When we use a laser fluence under Fmax, functional PAM shows great stability during long-term mouse brain imaging. Method Fast functional photoacoustic microscopy To investigate the effect of laser irradiation on oxygen saturation accuracy, we used a fast-scanning dual-wavelength photoacoustic microscope. Fig. 1a shows the schematic diagram of a dual-wavelength stimulated Raman scattering laser. The two wavelengths (532 nm and 558 nm) have a sub-microsecond delay between their pulses. The pulse energies at 532 nm and 558 nm can be adjusted up to 420 nJ and 195 nJ, respectively. Detailed design of the optical path and the imaging probe can be referred to our previous report [43]. The imaging probe is mounted on a millimeter-range voice-coil stage (VC) to scan along the x-axis. A ball-screw translation stage (BST) was used to scan the whole voice-coil stage in the y-axis. One laser pulse generates a one-dimensional time-resolved PA signal, and volumetric imaging is achieved by fast scanning in the x- and y-axes. Fig. 1b shows a schematic of a laser beam irradiating the tissue. The waist diameter of the focused beam in water is ∼3 µm. Beam diameter on the tissue surface is ∼50 µm when the beam was focused in the tissue at 250 µm. The voice-coil stage drives the probe to move at a constant speed. When we use different laser repetition rates and multiple laser pulses (e.g., 10, 13, 20, 40, 50, 100, and 200) will repetitively irradiate the same effective area defined by the surface equivalent spot diameter of 50 μm. This laser pulse number for each wavelength is denoted as N. Fig. 1c and d show the pulse distribution within a spot diameter travel when N = 10 and N = 200, respectively. As N increases, the degree of pulse overlap within the same spot increases accordingly. The translation stage (y-axis) step size was set to 1.25 μm, the scanning range was 2.5 × 1.56 mm², and the scanning time was 40 s. (a) Schematic of the multi-pulse two-wavelength PAM system. (b) Schematic of laser beam irradiation. Waist diameter is ∼3 µm. The beam diameter at the tissue surface is ∼50 µm. The laser pulse number N is the number of pulses irradiating the same spot on the tissue surface as the laser beam scans across it. (c) and (d) are schematic of the laser pulse distribution when N = 10 or 200. Results Maximum fluence for accurate sO2 measurement To quantitatively investigate the impact of laser irradiation on sO2 measurement accuracy, we imaged sO2 in mouse ears at different laser fluences. To quantitatively analyze sO₂, we used dual-wavelength spectral unmixing with PA amplitudes acquired at wavelengths of 532 nm and 558 nm. sO₂ was calculated as follows: Where Pλ1 and Pλ2 are the PA amplitudes at wavelengths λ1 and λ2, respectively. εde and εoxy are the molar extinction coefficients of oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb), respectively. Before spectral demixing, the PA signals at the two wavelengths were normalized to their respective pulse energies. The Animal Ethics Committee of City University of Hong Kong approved the animal experiments. All in vivo experiments used ICR mice (3–5 months old; weighing 20–30 g). A nose cone was securely attached to the mouse's head. In experiments, the mice were anesthetized with 1.5% (v/v) isoflurane, and the body temperature was maintained at 37°C using a heating pad. The experiments were conducted on 7 healthy mice. The scanning area of each mouse was 2.5 × 1.56 mm2, and the scanning step size in the y-axis was 1.25 μm. The time to acquire an image was ∼40 s. When the laser beam scans across a 50-μm distance, i.e., one laser spot diameter on the tissue surface, the laser irradiation time is ∼640 µs. When we use different laser repetition rates, e.g., 15625, 21875, 31250, 62500, 78125, 156250, and 312500 Hz, the number of laser pulses N irradiating the same spot can be adjusted. Under each repetitive pulse number N for each wavelength, to image sO2, the single-pulse laser fluence on the tissue surface was set at different levels, e.g., 1.7–21.3 mJ/cm² (Fig. S1-S7). Note that single-pulse energy density refers to the fluence of one laser pulse at one wavelength on the sample surface. The single pulse fluence and Fmax are defined by the higher specific wavelength fluence at 532 nm and 558 nm wavelengths, rather than the sum of the two fluences in our experiment. We adopt this definition as a conservative standard to avoid underestimating the irradiation levels associated with accurate sO2 measurements. We selected several arterial and venous regions on the images marked by the white dashed and solid boxes in Fig. S1-S7, respectively, and calculated the average arterial oxygen saturation (saO₂) and average venous oxygen saturation (svO₂) as shown in Fig. 2a-g. The results show that, with a specific laser pulse repetition number N for each wavelength, when the single-pulse laser fluence at one wavelength is below a threshold (Fmax), sO₂ can be accurately measured, and the results agree with normal physiological values [8], [43]. However, when the laser fluence exceeded this threshold, svO₂ showed consistent overestimation, suggesting that the short-term average laser fluence can significantly affect the accuracy of sO2 measurement. (a)-(g) Average saO2 and svO2 values marked in Fig. S1-S7 of N ranging from 10 to 200 for different single-pulse fluences. The red-dashed line represents the fitted curve for saO2. The purple line represents the fitted svO2 curve, which is divided into two segments at the Fmax. The Fmax is marked by a red dashed circle. Error bars are shown for both saO2 and svO2. (h) Fmax at different N ranging from 10 to 200. To quantitatively determine the threshold laser fluence Fmax for each laser pulse repetition number N, we analyzed the trend of svO2 variation with laser fluence. At low laser fluence, the measured value stabilized within the physiologically normal range; when the laser fluence exceeded a certain critical value, svO2 began to deviate from the normal value and exhibited an approximately linear increase. Based on this characteristic, we used a piecewise linear fitting method to determine Fmax: the svO2 data at different fluences were divided into two segments; the first segment was fitted with a horizontal straight line for the stable region, and the second segment was fitted with a sloped line to fit the rising region. A MATLAB program iterated through all possible data segmentation points, using the minimum sum of squared residuals between the two fitting segments as the optimization criterion to automatically determine the optimal inflection point. The laser fluence at this inflection point is determined as Fmax. We determined the Fmax values at different laser pulse repetition numbers N when repetitively irradiating the same effective area defined by the surface equivalent spot diameter for approximately 640 μs. The Fmax decreases with the increased N with a power law relationship: Fmax=45N−0.41(R2>0.95), as shown in Fig. 2 h. This relationship directly reflects the effect of short-term thermal accumulation. The greater the pulse repetition number, the more significant the thermal accumulation, and the lower the Fmax for accurate sO2 measurement. The maximum fluence model provides important guidance for accurate in vivo imaging of sO₂. Validation of the maximum fluence model in blood phantoms To validate the maximum-fluence model, we conducted a blood-model experiment. Two transparent plastic tubes with an inner diameter of 250 µm were used to simulate blood vessels. The model medium was commercially available anticoagulated bovine blood, with sodium heparin as the anticoagulant. To prepare blood samples with varying oxygenation levels, we chemically treated the blood with sodium dithionite as previously reported [44], [45]. We injected the prepared blood samples into the plastic tubes. Nitrogen gas was used to purge the plastic tubes of air. Subsequently, we used a blood gas analyzer (BGA−102, Wondfo Biotech Ltd.) to verify the oxygen saturation of the prepared samples at the set values of 1.0 and 0.9. In this experiment, the pulse number was set to N = 200 over a travel distance of 50 μm, and the sO₂ value of the phantom was measured at different laser fluence values. The results show that the measured Fmax is 4.3 mJ/cm2, which is close to the model prediction of 5.0 mJ/cm2. The difference between the measured threshold in the brain and the fitted value may be attributed to differences in tissue heat dissipation capacity, flow speed, and optical attenuation. When the single pulse fluence was below Fmax, the measured sO₂ in the blood phantom remained close to the reference value, with an absolute error within 3%. In contrast, the error increased noticeably once the fluence exceeded Fmax (Fig. 3). sO2 images (a) and average sO2 values (b) of tubes 1 and 2 at different fluences with N = 200. The red-dashed line represents the fitted curve for sO2 in tube 1. The purple line represents the fitted curve of sO2 in tube 2, which is divided into two segments at the Fmax. The Fmax is marked by a red dashed circle. Error bars are shown for both Tubes. Scale bar, 0.5 mm. Validation of the maximum fluence model in the brain To further verify the applicability of the maximum-fluence model in vivo, we conducted functional imaging of the mouse brain at different fluences. The number of pulses is set to N = 200, the scanning area is 1.5 × 1.7 mm2, and the volumetric imaging time is 26 s. Images were acquired within the surface fluence range from 1.0 to 6.8 mJ/cm², as shown in Fig. 4a. We marked the arteries and veins in Fig. 4a with white dashed boxes and solid boxes respectively and calculated the average saO₂ and svO₂ values under different fluence levels, as shown in Fig. 4b. The results show that the measured Fmax is 4.2 mJ/cm2, which is close to the model prediction of 5.0 mJ/cm2. The difference between the measured threshold in the brain and the fitted value may be attributed to differences in tissue heat dissipation capacity, flow speed, and optical attenuation. (a) sO2 images and (b) average saO2 and svO2 values marked by the white dashed and solid boxes at different single pulse fluence with N = 200. The red-dashed line represents the fitted curve for saO2. The purple line represents the fitted svO2 curve, which is divided into two segments at the Fmax. The Fmax is marked by a red dashed circle. Error bars are shown for both saO2 and svO2. Scale bar, 0.2 mm. Long-term monitoring of brain sO2 Based on the maximum laser fluence predicted by our model, we set N = 200, and F = 3 mJ/cm² to measure the brain oxygen release. As shown in Fig. 5a, we performed sO₂ imaging on a 2.0 × 2.0-mm² field of view of the left brain. The acquisition time for a set of dual-wavelength images was ∼41 s. To further quantify oxygen saturation, the average sO₂ values of large arteries (marked with white dashed boxes), small arteries (marked with yellow dashed boxes), large veins (marked with white solid boxes), and small veins (marked with blue solid boxes) were calculated and shown in Fig. 5b. These results were close to normal physiological values [8], [43]. (a) sO2 image of the brain (N = 200, F = 3 mJ/cm²). Scale bar, 0.2 mm. (b) Average sO₂ values in different blood vessels. Large arteries marked with white dashed boxes, small arteries marked with yellow dashed boxes, large veins marked with white solid boxes, and small veins marked with blue solid boxes. (c) sO2 images during 35 consecutive scans of the mouse brain within ∼22 min. (d) Fluctuations of the representative saO2 and svO2 are marked with the white dashed box in (c) of the brain under anesthesia. Scale bar, 0.5 mm. We further verify the stability of functional imaging in long-term monitoring. We performed 35 consecutive sO₂ scans (∼ 22 min) on the brain cortex surface using the same laser parameters (N = 200, F = 3 mJ/cm²). The field of view for each scan was 1.7 × 2.2 mm², and the acquisition rate was approximately 38 s per sO2 image (Fig. 5c and Video 1). We selected the areas marked by the white dashed box in Fig. 5c and analyzed the changes in the saO₂ and svO₂ in these areas (Fig. 5d). The results showed that the saO₂ and svO₂ values remained stable during 35 consecutive scans, and no significant drift or fluctuation in measurements was observed. The in vivo results demonstrate that, within the maximum fluence predicted by our model, continuous and stable sO₂ measurements can be achieved in the brain, verifying the model's effectiveness for long-term functional monitoring. In the brain validation and long-term monitoring experiments, we selected N = 200 because this condition provided the highest spatial sampling density in our scanning configuration, which is beneficial for repeated imaging of small cortical vessels. To compensate for the increased thermal accumulation under this condition, the single-pulse fluence was set to 3 mJ/cm2, approximately 60% of the corresponding Fmax, thereby maintaining stable and accurate sO2 measurements during long-term imaging. Discussions and conclusions This study investigates the effect of repetitive pulsed irradiation on the accuracy of sO2 imaging. We established a quantitative empirical model between the repetitive pulse number N and the maximum fluence Fmax for accurate sO2 measurement, which provides practical guidance for functional PAM. As the repetitive-illumination pulse number increases from 10 to 200 within ∼640μs, the Fmax required for accurate sO2 measurement decreases from approximately 17mJ/cm2 to approximately 5mJ/cm2. It should be noted that the proposed Fmax−N relationship is an empirical model established under the specific experimental conditions used in this study. These conditions, including the local irradiation duration over a single 50-μm spot (∼640μs), the tested pulse repetition rates, and the scanning parameters, were explicitly defined in the experiments. Therefore, N should be interpreted as a practical descriptor of repetitive irradiation under these specific temporal and scanning conditions, rather than the sole factor determining Fmax. Changes in pulse interval, local irradiation duration, scanning speed, or tissue thermal diffusivity may alter the corresponding Fmax. The methodology presented here can be extended to investigate maximum fluences under other conditions in future studies. It should also be noted that although the observed sO2 deviation is significantly dependent on the number of repetition pulses and local irradiation conditions, its underlying mechanism should not be attributed solely to thermal accumulation. Besides short-term local thermal accumulation, possible influencing factors include nonlinear photoacoustic response, local optical absorption spectra, and Grüneisen parameters. Future research combining direct thermal characterization and more comprehensive physical modeling will help differentiate these influencing factors. We validated the model in both blood phantom and in vivo experiments. Both results agree well with the model predictions, supporting its predictive ability. Although slight differences in optical and thermal properties may exist between brain and ear tissues, the measured Fmax=4.2mJ/cm2 under N=200 remains close to the model prediction. This result suggests that the power-law relationship between Fmax and N described here may be applicable to tissues such as the brain cortex under similar conditions. The observed fluence-dependent bias in sO2 measurements likely involves multiple temperature-dependent effects. First, local heating may affect hemoglobin oxygenation, including a possible shift in oxygen dissociation. However, this effect alone does not directly explain our observations, because enhanced oxygen release would be expected to decrease rather than increase the true blood oxygen saturation. In contrast, we observed that saO2 remained largely unchanged, whereas svO2 increased with increasing fluence. Second, local heating may affect the dual-wavelength PA measurement and inversion process itself through temperature-dependent changes in PA signal generation, including possible contributions from the local optical absorption spectrum, the Grüneisen parameter, or both. These effects could introduce a bias in the measured sO2. Since the present study did not directly separate these mechanisms experimentally, we have avoided attributing the observed deviation to a single dominant cause. Instead, the observed bias in measured sO2 may arise from the combined effects of temperature-dependent physiological changes and temperature-dependent photoacoustic-response changes. Compared with saO2, svO2 showed a larger increase with increasing laser fluence. A possible explanation is that arterial blood is already close to full oxygenation under normal physiological conditions, so its measured saO2 has a relatively limited dynamic range. In contrast, venous blood starts from a lower oxygenation level and therefore has more room to exhibit an apparent change under repetitive laser irradiation. In addition, the lower flow velocity and longer residence time in veins may make transient local thermal effects and the associated measurement bias more evident in the venous compartment. Therefore, the experimentally determined Fmax in this study was mainly constrained by the onset of svO2 increase, which provides a relatively conservative threshold for accurate sO2 measurement. It should also be noted that the empirical relationship between Fmax and N reported here was established for the 532/558 nm dual-wavelength PAM configuration used in this study. Since hemoglobin absorption and laser-induced photoacoustic response are wavelength-dependent, the quantitative expression may vary for different wavelength combinations. Nevertheless, the same experimental and modeling framework can be applied to determine the corresponding Fmax under other functional imaging wavelengths. In practice, the choice of repetitive pulse number N and single pulse fluence involves a trade-off among signal strength, spatial sampling density, imaging speed, thermal accumulation, and sO2 accuracy. A smaller N allows a higher single-pulse fluence and thus stronger photoacoustic signals, but it is usually associated with a larger scanning step size and reduced spatial sampling density. In contrast, a larger N provides denser spatial sampling, which is advantageous for resolving microvascular structures and maintaining robust sO2 quantification, but it requires a lower single-pulse fluence to avoid thermally induced measurement error. Therefore, parameter selection should depend on the imaging objective. In our brain validation and long-term monitoring experiments, we chose the largest tested pulse number, N=200, because this condition corresponded to the smallest scanning step size in our system and therefore provided the highest spatial sampling density for high-speed functional imaging of cerebral vessels. This setting is favorable for repeated monitoring of small cortical vessels and for maintaining stable vessel delineation during long-term imaging. At the same time, according to the experimentally determined Fmax, we used relatively low laser fluence (F=3mJ/cm2, ∼60% of Fmax) to scan the brain for about 22 min. The results show that saO2 and svO2 remained stable without significant drift, indicating that low laser fluence enables brain imaging with minimal perturbation to tissue oxygen saturation and supports long-term in vivo functional monitoring. CRediT authorship contribution statement Lidai Wang: Writing – review & editing, Writing – original draft, Methodology. Zhiwei Yao: Software, Data curation. Jingyi Zhu: Writing – review & editing, Writing – original draft, Methodology, Data curation. Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dr. Lidai Wang has a financial interest in PATech Limited, which, however, did not support this work. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgment This work is partially supported by the Research Grants Council of the Hong Kong Special Administrative Region (Grant Nos. 11104922, 11103320), the National Natural Science Foundation of China (NSFC) (81627805, 61805102), and the Institute of Digital Medicine of City University of Hong Kong. Biographies Jingyi Zhu is a Postdoctoral Fellow at the City University of Hong Kong. She earned her Ph.D. from City University of Hong Kong, her M.S. from Zhejiang University, and her B.S. from Zhengzhou University. Her research focuses on the development and application of photoacoustic and biomedical imaging technologies. Zhiwei Yao is a Ph.D. candidate at the Department of Biomedical Engineering, City University of Hong Kong. He received bachelor’s and master’s degrees from Tianjin University. His research focuses on photoacoustic microscopy and imaging. Lidai Wang received the bachelor’s and master’s degrees from Tsinghua University, Beijing, and received the Ph.D. degree from the University of Toronto, Canada. After working as a postdoctoral research fellow in Prof. Lihong Wang’s group, he joined the City University of Hong Kong in 2015. His research focuses on biophotonics, biomedical imaging, wavefront engineering, instrumentation, and their biomedical applications. Footnotes Supplementary data associated with this article can be found in the online version at doi:10.1016/j.pacs.2026.100851. Supplementary material Supplementary material Supplementary material Data availability Data will be made available on request. References Associated Data Supplementary Materials Supplementary material Supplementary material Data Availability Statement Data will be made available on request.