Exploring the Principles and Applications of Photoacoustic Imaging Through Signal Processing and Experimental Techniques

Abstract

Photoacoustic imaging (PAI) is an emerging biomedical imaging technique that combines the high optical contrast of light-based imaging with the depth resolution of ultrasound. In this lab, we investigated the principles and practical applications of PAI through two complementary modules: MATLAB-based signal processing and experimental imaging using the Vevo LAZR system. The first module focused on digital signal analysis using radiofrequency (RF) data collected from ultrasound and photoacoustic sources. Fast Fourier transforms were used to determine frequency characteristics of a 25 MHz transducer, and Hilbert transforms were applied to construct envelope-detected images that mimic conventional B-mode imaging.

The experimental portion involved imaging phantom tubes containing indocyanine green (ICG) and hemoglobin. Spectral scans of ICG confirmed strong photoacoustic signal generation near 800 nm, with peak signal occurring when the sample was positioned within the acoustic and optical focal plane. Continuous irradiation revealed photobleaching effects, as ICG signal intensity decreased over time. Blood oxygen saturation imaging demonstrated PAI’s capability to distinguish between oxygenated and deoxygenated hemoglobin using multi-wavelength excitation and spectral unmixing techniques.

Across both lab components, photoacoustic signal formation, tissue-mimicking contrast, and the limitations of depth, resolution, and signal degradation were explored. Together, these findings reinforce the utility of PAI for non-invasive molecular imaging and provide foundational experience in both signal processing and experimental setup.

Introduction

​​The history of photoacoustic imaging can be traced back to 1880, with Alexander Graham Bell’s discovery that modulated light produces acoustic waves. Starting in the mid-1900s, similar physical mechanisms were applied to gas analysis studies through the development of gas spectroscopy and photothermal material inspection applications. Photoacoustic technology further developed at the beginning of the 21st century where it further developed into a well-accepted biomedical imaging modality.1,2

The basic principle of photoacoustic imaging relies on the illumination of a region of interest with pulsed light that is then absorbed by photo-absorbers within the tissue, resulting in a small rise in temperature. The increase in temperature results in thermal expansion, which generates an acoustic pressure wave. This pressure wave can be received by an ultrasonic transducer and reconstructed into an image. The source of contrast in these photoacoustic images is primarily derived from the optical absorption properties of different tissues. Photoacoustic imaging system components generally include a light source, typically a nanosecond pulsed laser, and an ultrasound system. Targets for imaging may include blood, tissue, fat, and metallic tool tips. The methods for photoacoustic imaging are described in greater detail in this report below.3

Because traditional optical imaging methods cannot maintain high-resolution imaging in deep biological tissues due to optical scattering, the capability to achieve high resolution optical contrast images in biological tissues a few millimetres to centimetres in depth makes photoacoustic imaging particularly appealing in clinical applications.4,2 Current uses of photoacoustic imaging include photoacoustic-guided surgery, the goal of which is to visualize both the tool tip and a structure that needs to be either targeted or avoided in the same photoacoustic image. This concept has been used to visualize biopsy targets such as tumors and identify nerves and vasculature during interventional procedures.3 PAI can also be used in breast imaging for cancer detection, vascular imaging in extremities, as well as imaging of carotid vessels, and dermatologic, musculoskeletal, gastrointestinal, and adipose tissues.4

The experiments in this report used PAI to observe a photoacoustic dye and a blood sample in a phantom box. This report begins with an overview of the operation of a PAI system, followed by the imaging and analytic methods used in this lab. Results and associated quantitative and qualitative analyses are presented. Key findings, limitations, and advantages of this lab are summarized in the conclusion.  

Background

Photoacoustic imaging is a hybrid biomedical imaging modality that combines the molecular contrast of optical imaging with the spatial resolution and depth penetration of ultrasound. It relies on the photoacoustic effect, where short laser pulses are absorbed by chromophores (like hemoglobin or dyes such as indocyanine green), leading to localized thermoelastic expansion and the emission of ultrasound waves. These acoustic waves are detected by transducers and reconstructed into images that reveal optical absorption patterns at greater depths than purely optical techniques allow.7​ Compared to other modalities, PAI is label-free, non-ionizing, and provides functional information such as oxygenation and dye concentration in real time​.5

Throughout this lab, PAI was used to explore multiple principles relevant to translational imaging: contrast dye behavior, photobleaching, spectral specificity, and signal processing. The dye indocyanine green (ICG) was used due to its strong absorption in the near-infrared region and low fluorescence quantum yield, which makes it an efficient photoacoustic agent​.6 Spectral unmixing was also employed to differentiate oxygenated and deoxygenated hemoglobin, demonstrating how PAI can be extended to functional imaging of blood oxygen saturation. The Hilbert transform was introduced to enhance image interpretability by isolating the amplitude envelope from raw RF data, mimicking clinical ultrasound display techniques and improving the visibility of structural features.

Methods

Lab 1: MATLAB-Based Processing of Ultrasound and Photoacoustic Data

This portion of the lab involved offline signal processing of pre-recorded ultrasound and photoacoustic data using MATLAB. The goal was to understand how to extract spatial and frequency-domain information from A-line data, distinguish between photoacoustic and ultrasound signals, and apply signal processing techniques to enhance image quality.

In Exercise 1, an A-line signal was acquired using a 25 MHz single-element transducer to detect a human hair phantom. The signal was preloaded from Data2.mat, and its frequency content was analyzed using the fast Fourier transform (FFT). The sampling frequency was 400 MHz, and the peak frequency content was expected around 22–25 MHz.

Exercise 2 involved identifying photoacoustic and ultrasound signals within a single A-line acquired sequentially, preloaded from Aline.mat. Using known sampling rates and the speed of sound in tissue (1540 m/s), the distance to the phantom was calculated based on both the PA and US signals. A correction factor was applied for the two-way travel time in ultrasound.

Exercise 3 explored the use of the Hilbert transform for envelope detection in RF data. An image matrix from Data1.mat was displayed before and after applying the Hilbert transform. A single A-line (20th channel) was plotted to demonstrate how the transform creates a smooth amplitude envelope. The entire RF matrix was transformed and visualized using imagesc(abs(hilbert(Data1))) to simulate a conventional B-mode image.

Lab 2: Experimental Photoacoustic Imaging with the Vevo LAZR System

Lab 2 was conducted using the VisualSonics Vevo LAZR system with a high-frequency linear array transducer and pulsed laser excitation. Real-time photoacoustic and ultrasound imaging was performed on tube phantoms containing contrast agents and hemoglobin preparations.

In Exercise 1, 200 µL of 5 µM Indocyanine Green (ICG) was pipetted into a tube submerged in deionized, degassed water. A spectral scan from 680 to 970 nm in 5 nm increments was performed using the NIR1 mode to identify the peak photoacoustic absorption, which was expected around 800 nm. Ultrasound was used to align the tube within the focal zone, and then the tube was shifted below focus for a second scan to observe spectral signal differences.

Exercise 2 examined the photobleaching effect of ICG. With the tube in focus, continuous imaging was performed at 800 nm for five minutes in single-wavelength mode. A region of interest (ROI) was drawn around the tube, and the photoacoustic signal was monitored over time to assess signal decay due to dye degradation.

Exercise 3 assessed oxygen saturation by imaging two hemoglobin solutions side-by-side: one oxygenated and one deoxygenated with sodium dithionite. The system was set to “OxyHemo” mode, which utilizes dual-wavelength excitation for spectral unmixing. Provided MATLAB-processed images were used to display blood oxygen saturation maps.

Results and Analysis

Part 1

The first exercise focused on deducing the frequency content of an ultrasound signal. Data was entered into Matlab as described in the methods, and the results produced the image shown in Figure 1A. As shown, there is a large peak from about 0 to 500 data points, which can be best described as the transmitted pulse. At about 8500 data points, another signal is detected. This point would coincide with the front face of the hair strand as the signal hits. A second, smaller peak is shown as the signal meets the back face of the hair, but it is difficult to visualize here. This could be due to the signals happening at very close data points, so it is difficult to pick out the individual peaks in this image. Figure 1B shows the plot of the maximum peak of the transducer. It peaks at around 22MHz, which is within the expected operating range of 22-25MHz of the instrument.

In ultrasound imaging, the distance can be calculated as D = (time * velocity)/2. Velocity in this equation would be considered the speed of sound, or 1540 m/second. The sampling frequency of this dataset was 400 MHz, or 400 million sample points taken per second. The data point at 8500 would be taken in 21.25 microseconds. Using the equation, the distance is calculated to be 0.0164 meters, or 16.4 millimeters. This means that the strand of hair was 16.4 millimeters away from the transducer when the ultrasound signal was  measured.

In Figure 2, the A-line displays a US pulse signal around 7500 samples and a PA signal around 16000 samples. Using the same equation D = (time * velocity)/2 as in Figure 1, we can verify the distance calculated from the transducer to the PA signal and to the US signal. Velocity in this equation would be considered the speed of sound, or 1540 m/second. The sampling frequency of this dataset was 200 MHz, or 200 million sample points taken per second. The data point at 7500 would be taken in 37.5 microseconds and the PA signal at 80 microseconds. Using the equation, the distance is calculated to be 2.8875 cm for the US signal and 6.16 cm for the PA signal. 

In Figure 3a, the image generated from the raw RF data using imagesc(Data1) displays alternating dark and light bands caused by oscillating positive and negative components of the signal. While the structure of the phantom is somewhat visible, the contrast is low and the image is not easily interpretable as a conventional B-mode image.

After applying the Hilbert transform, the envelope of each A-line is extracted and used to produce the image shown in Figure 3b. This envelope-detected image shows significantly improved contrast and clarity. The oscillations are replaced by smooth intensity values, making the structural content of the phantom (e.g., letters “BME”) much clearer and more representative of what we expect in an ultrasound.

Figure 3c shows a plot of the 20th A-line signal before and after applying the Hilbert transform. The original RF signal fluctuates around zero, whereas the envelope (in red) forms a smooth curve outlining the signal’s intensity over time. This demonstrates how the Hilbert transform helps isolate the meaningful information in the signal while discarding phase-related noise, which proves that the Hilbert transform enhances signal visualization by converting the real-valued RF data into a form that reveals amplitude variations more clearly.

Part 2

Indocyanine green (ICG) is a clinically approved dye widely used for applications such as liver function testing, cardiovascular diagnostics, and intraoperative imaging due to its near-infrared absorption properties. At 800 nm, ICG exhibits a high molar extinction coefficient of approximately 1.22 × 10⁵ M⁻¹·cm⁻¹, making it an effective optical absorber. ICG is both an absorber and a fluorophore, though it has a low quantum yield (~0.016) in aqueous environments, meaning that most of the absorbed energy is released non-radiatively. This characteristic makes it an excellent photoacoustic contrast agent, as it converts optical energy into acoustic waves efficiently rather than emitting fluorescence.

Ultrasound imaging was used in this experiment to localize and position the ICG-filled tube within the phantom (Figure 4a). This was important for ensuring the tube was within the optical and acoustic focus of the imaging system before collecting spectral data. When positioned correctly, a strong photoacoustic signal was observed at 800 nm (Figure 4b), with the spectrum (Figure 4c) clearly peaking at this wavelength—consistent with the known optical absorption profile of ICG.

After repositioning the tube below the photoacoustic focus, signal strength at 800 nm was notably reduced (Figure 4d), and the corresponding spectrum (Figure 4e) exhibited a broader, less intense peak. This attenuation is attributed to the tube falling outside the focal zone, where energy delivery and signal detection are less efficient. The ultrasound image (Figure 4f) confirmed that the tube was no longer aligned with the transducer’s focal plane.

Photobleaching is the process by which a dye like ICG loses its ability to absorb light after prolonged exposure. This affects the dye's ability to generate a photoacoustic signal. In this experiment, we imaged the ICG-filled tube continuously at 800 nm over a five-minute period to observe this effect.

In Figure 5a, the first photoacoustic image taken at the start of the scan shows a strong, concentrated signal from the ICG solution, indicating high optical absorption and efficient photoacoustic conversion. By the end of the 5-minute period, as shown in Figure 5b, the signal has visibly decreased in intensity, demonstrating the impact of photobleaching on the dye’s absorption properties.

This change is more clearly visualized in Figure 5c, where the photoacoustic signal intensity is plotted over time. The curve shows a steady decline, confirming that the ICG’s signal is diminishing with each frame. This suggests that ICG is being gradually photodegraded, reducing its effectiveness as a contrast agent during continuous laser exposure. The trend is consistent with photobleaching behavior and highlights a key limitation when using ICG in long-duration imaging.

In Figure 6a, the photoacoustic image at 750 nm reveals a stronger signal from the tube containing deoxygenated hemoglobin, as this wavelength is closer to the absorption peak of Hb. In Figure 6b, the signal intensity shifts, with the oxygenated hemoglobin tube showing greater contrast. This shift occurs because 850 nm is more strongly absorbed by HbO₂, demonstrating how dual-wavelength photoacoustic imaging can distinguish between oxygenation states.

These two images serve as input for spectral unmixing, which allows for pixel-by-pixel differentiation of Hb and HbO₂. The result is shown in Figure 6c, where the unmixed oxygen saturation image (StO₂) clearly separates the two tubes. The tube with oxygenated hemoglobin appears in warmer, higher-value colors (closer to 1.0), while the deoxygenated tube is darker, indicating lower oxygenation levels.

Figure 6d provides the structural reference via ultrasound, confirming that both tubes were side-by-side and properly positioned. The spatial alignment in the ultrasound image matches the location of signals seen in the photoacoustic and unmixed images, validating the result. So,  photoacoustic imaging can estimate relative oxygen saturation in tissue-like environments, with clear spatial and spectral separation between Hb and HbO₂.

Conclusion

This lab looked into the fundamental principles and applied capabilities of photoacoustic imaging (PAI) through a combination of MATLAB-based data analysis (Lab 1) and hands-on experimentation with a high-frequency ultrasound-photoacoustic imaging system (Lab 2). The first phase of the lab focused on processing raw acoustic signals, where frequency-domain analysis and envelope detection were used to reconstruct interpretable images from radiofrequency (RF) data. The application of the Hilbert transform in particular was shown to effectively enhance image contrast by converting raw oscillatory signals into amplitude-based representations more akin to clinical ultrasound.

The experimental component, conducted with the Vevo LAZR system, enabled direct observation of photoacoustic contrast agents and physiological phenomena. Spectral imaging of indocyanine green (ICG) revealed a peak absorption at 800 nm, consistent with its known optical properties, while photobleaching studies highlighted the limitations of prolonged laser exposure in real-time imaging applications. The blood oxygenation experiment further demonstrated the ability of multispectral PAI to resolve differences between oxygenated and deoxygenated hemoglobin using spectral unmixing algorithms, underscoring the clinical relevance of this modality for functional tissue imaging.

With these together, the results confirm that photoacoustic imaging offers a powerful platform for combining optical contrast with acoustic resolution. The lab exercises provided both theoretical understanding and practical insight into how PAI can be used for high-resolution molecular imaging in biological systems. 

Author Contributions: KS, JRS, and MZ equally contributed to the conceptualization, methodology, investigation, and drafting of the manuscript. JRS was responsible for coding and data analysis in MATLAB. JRS and KS led the revision and editing process, while MZ coordinated the final submission. MS provided guidance, supervision, and project oversight. All authors reviewed and approved the final version of the report.

Funding: This work was supported by the Tufts University School of Engineering and made possible through the use of facilities at the Integrated Biofunctional Imaging & Therapeutics Laboratory (iBIT Lab). We thank both institutions for providing access to the equipment and resources as part of the BME-0156: Biophotonics Laboratory course.

Acknowledgments: We extend our sincere gratitude to Dr. Maria Savvidou, Postdoctoral Researcher in the Department of Biomedical Engineering at Tufts University, for her mentorship, guidance, and support throughout the project, especially in course advising. We also thank Andrew Langley, EG’2G, for his generous technical support with both software and hardware during the lab sessions, as well as for compiling the experimental data.

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

References

1. Manohar, S., Razansky, D. Photoacoustics: a historical review. Adv. Opt. Photon. 8, 586-617 (2016). https://doi.org/10.1364/AOP.8.000586

2. Park, J., Choi, S., Knieling, F. et al. Clinical translation of photoacoustic imaging. Nat Rev Bioeng 3, 193–212 (2025). https://doi.org/10.1038/s44222-024-00240-y

3. Muyinatu, A., Lediju, B. Photoacoustic imaging for surgical guidance: Principles, applications, and outlook. J. Appl. Phys. 128 (6): 060904 (2020). https://doi.org/10.1063/5.0018190

4. Attia, A.B.E., Balasundaram, G., Moothanchery, M., Dinish, U.S., Bi, R., Ntziachristos, V., Olivo, M. A review of clinical photoacoustic imaging: Current and future trends. Photoacoustics 16 (2019). https://doi.org/10.1016/j.pacs.2019.100144.

5. Das, D., Sharma, A., Rajendran, P., & Pramanik, M. (2021). Another decade of photoacoustic imaging. Physics in Medicine & Biology, 66(5), 05TR01. https://doi.org/10.1088/1361-6560/abd669

6. Weber, J., Beard, P. C., & Bohndiek, S. E. (2016). Contrast agents for molecular photoacoustic imaging. Nature Methods, 13(8), 639–650. https://doi.org/10.1038/nmeth.3929

7. Langley, A. (2025). BME156: Photoacoustic Imaging [Lecture slides]. Tufts University.