Deep learning-enhanced transparent transducer-based photoacoustic microscopy with improved SNR and resolution

Transparent ultrasonic transducer enables compact coaxial photoacoustic microscopy, but the trade-off between optical transparency and sensitivity and electrode nonuniformity results in reduced SNR and resolution. Here, we propose a deep learning-based enhancement framework comprising two task-specific networks for SNR improvement and aberration correction. Simulation results with known ground truth demonstrate consistent improvements in structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR) across a wide range of signal levels and aberration strengths, with the most pronounced enhancement observed under low-signal conditions. Experimental results obtained from a transparent transducer photoacoustic microscopy system further validate enhanced vascular visibility and structural fidelity. These results indicate that the proposed method effectively compensates for imaging degradation while enabling reduced laser energy requirements.