Comprehensive Summary
This study examines photoacoustic microscopy (PAM), an invaluable tool in biomedical research, and seeks to accelerate the data acquisition process and reduce the high sampling demands. The study offers a solution, called Information-Efficient Photoacoustic Microscopy (IE-PAM), that combines scanning encoding and neural network decoding to produce reconstructions from limited measurements. The system consisted of a streamlined SS-PAM design that was combined with the learning reconstruction model IE-PAM. The system was able to create high-quality images with limited data. Comparison with other models showed that IE-PAM was the best-performing, as it could reconstruct images at a sampling rate of up to 1.5%.
Outcomes and Implications
This new deep learning model, IE-PAM, could be used to make photoacoustic microscopy more accessible in biomedicine in the future, as it is more cost-effective and easier to use. PAM is often used in clinical settings to observe tissue during surgery and imaging, so it could aid physicians in making more comprehensive diagnoses.