Two-step proximal gradient descent algorithm for photoacoustic signal unmixing.

Area of Science:

• Biomedical optics
• Photoacoustic imaging
• Computational imaging

Background:

• Photoacoustic microscopy (PAM) employs multiple wavelengths to quantify absorber concentrations.
• The speed of sound imposes a sub-microsecond limit on wavelength switching, hindering high-speed, broad-spectrum imaging.
• Overlapping signals from rapid wavelength switching present a significant challenge in PAM.

Purpose of the Study:

• To develop a novel computational method for separating overlapped photoacoustic signals.
• To overcome the sound-speed limitation on wavelength switching time in PAM.
• To enhance the speed and spectral range of photoacoustic imaging.

Main Methods:

• Introduction of a new signal unmixing algorithm: two-step proximal gradient descent.
• Utilizing computational separation to deconvolve temporally overlapped multi-wavelength photoacoustic signals.
• Simulations and in vivo experiments to validate the algorithm's performance.

Main Results:

• The two-step proximal gradient descent algorithm successfully unmixed up to nine overlapped signals in simulations.
• Separation of three overlapped signals with a 12-ns delay and 15.9-dB signal-to-noise ratio was achieved.
• In vivo application demonstrated successful unmixing of three-wavelength photoacoustic signals in microvessels.

Conclusions:

• The developed algorithm breaks the sound-speed bottleneck for wavelength switching in photoacoustic microscopy.
• This method enables high-speed, broad-spectrum imaging by computationally separating overlapped signals.
• Improved accuracy in oxygen saturation imaging was observed using the unmixed data from microvessel imaging.