Deep computational photoacoustic mesoscopy through heterogeneous tissues enabled by scanning compensation and angular-spectrum enhancement.

Optical microscopy enables high-resolution visualization of biological structures but is fundamentally limited in vivo by strong scattering, restricting penetration to superficial depths. Photoacoustic microscopy (PAM) overcomes this optical diffusion barrier yet remains constrained by a depth-resolution trade-off: high-frequency transducer (HF-UT) provide fine detail but fails through heterogeneous tissues, whereas low-frequency transducer (LF-UT) penetrate deep but blurs structure. Breaking this limitation requires a new imaging strategy that decouples penetration from resolution. Here, we overcome this long-standing limitation by introducing a computational PA mesoscopy (CPAMe) framework built upon an LF-UT. A real-time hardware stabilization strategy-combining pulse-by-pulse laser-energy compensation with point-by-point encoder correction-ensures uniform, high-speed volumetric sampling. Building on multi-layer speed-of-sound modeling, we further developed a directionally weighted angular-spectrum synthetic-aperture focusing technique (DWAS-SAFT) that restores off-focus resolution and suppresses heterogeneity-induced artifacts using only a single volumetric scan. CPAMe markedly improves lateral resolution by 32.5% (from 673 to 454 µm), 40% (from 705 to 423 µm), and 46% (from 1368 to 733 µm) in tissue phantoms, through the mouse skull, and through human cranial repair PMMA, respectively. These capabilities enable high-resolution transcranial brain imaging, whole-body and molecular small-animal imaging, tumor visualization, and proof-of-concept human vascular imaging. Together, CPAMe provides a practical and scalable route to deep-tissue, high-resolution imaging, opening opportunities for non-invasive transcranial monitoring and advancing mesoscopy toward clinical translation.