Painting a Better (Diagnostic) Picture: Tufts Duo Advances the Quality of Photoacoustic Images
Early detection of disease can save lives and prevent chronic health problems—but the technology that allows doctors to see what’s going on inside a person’s body can be limited, expensive, and time-consuming. These limitations act as a barrier as to what can be detected and who can access these important health-screening tools.
Photoacoustic (PA) imaging is a more portable, affordable, and promising diagnostic tool that uses light waves to craft internal images of the human body. Ph.D. student Avijit Paul and Associate Professor Srivalleesha Mallidi developed a method to improve light-emitting diode (LED)-based PA images, a meaningful step forward in making photoacoustic imaging a practical tool that can identify and prevent health issues. Their study, published in the Journal of Biophotonics, was recently recognized as one of the top 10 most cited papers of the journal in 2024.
Microscopic vibrations that create images
Photoacoustic imaging starts by shooting rays of light onto an object—for example, tissue in a person’s leg. The light energy is absorbed by the tissue, causing it to heat and vibrate on a microscopic level due to thermoelastic expansion. These microscopic vibrations create ultrasound waves that are recorded by a device called an ultrasound transducer. The time it takes for the waves to travel to the device indicates how far the waves traveled, allowing scientists to work backwards to piece together the structure of the tissue and construct a 3D image.
“In optical imaging–a common imaging technique that uses fluorescent light (low pulse-width, <150 nanoseconds)–the light often scatters, limiting how deep you can see into, for example, a tissue,” explained Paul, who was the first author on the Journal of Biophotonics paper. “PA imaging bypasses this with ultrasound waves received by the transducer, which enable deep, accurate imaging without straight-line photon paths that scatter. It’s also safer and less expensive compared to the X-ray radiation used in CT scans and magnetic resonance imaging (MRI).”
PA imaging typically uses light from a laser. While this technology remains cheaper and more portable compared to other imaging techniques like MRIs and CT scans, it still requires some bulky and expensive hardware. LEDs are less expensive and much smaller, but because they use less energy, imaging with this technology has worse interference from background signals (i.e., interference from other instruments or environment), reducing the image quality.
Paul and Mallidi created a method to develop high-quality LED-based images by analyzing original, lower-quality images with a U-Net—a type of AI model used for identifying or outlining specific objects in an image, such as blood vessels or cells.
Bringing the imaging method to new problems and environments
The Tufts duo tested this process on LED-based images, observing a four-fold increase in the signal-to-noise ratio. This means that the AI model was able to separate the interference from the environment or other instruments and create a more clear, accurate image.
They also developed these high-quality images in real-time, uploading results and generating the images almost instantly, which sped up the typically slow imaging process.
“This study serves as a foundation for addressing the limitations of PA imaging,” Paul explained. “Our method could help to lower costs, increase accessibility, and create high-quality images faster, speeding up the detection of diseases and dynamic biological processes like blood pumping through a heart. More efficient PA imaging could increase access to important diagnostic tools, and we are hoping this method can be implemented in a clinical setting in the future.”
Paul received a Tufts Launchpad Accelerator Grant to help bring this method out of the lab and into a clinical environment. He’s using the support to launch a startup called ClAIrVuE, teaming up with Mallidi and doctors in the Tufts University School of Medicine to commercialize the diagnostic image-enhancement process.
This research is a part of the Integrated Biofunctional Imaging and Therapeutics Laboratory led by Professor Mallidi, which uses AI and imaging technologies involving light and sound to better understand and treat pathologies like cancer.
Their new approach to PA imaging could also be applied to identifying tumors, which are energy intensive and stimulate new blood vessel growth for nutrients. These blood vessels appear differently, and PA imaging can capture them to indicate tumor presence without the need for reporter elements–dyes or fluorescent tags that are commonly used to detect internal aspects of the human body.
Paul and Mallidi’s innovative approach to PA imaging could serve as a valuable tool for both clinical workers and patients alike, increasing the affordability and efficiency of medical equipment and reducing barriers to early identification and treatment of health issues.