Photoacoustic microscopy (PAM) capitalizes on the optical absorption of blood hemoglobin to enable label-free high-contrast imaging of the cerebral microvasculature in vivo. Although time-resolved ultrasonic detection equips PAM with depth-sectioning capability, most of the data at depths are often obscured by acoustic reverberant artifacts from superficial cortical layers and thus unusable. In this paper, we present a first-of-a-kind dictionary learning algorithm to remove the reverberant signal while preserving underlying microvascular anatomy. This algorithm was validated in vitro, using dyed beads embedded in an optically transparent polydimethylsiloxane phantom. Subsequently, we demonstrated in the live mouse brain that the algorithm can suppress reverberant artifacts by 21.0 ± 5.4 dB, enabling depth-resolved PAM up to 500 µm from the brain surface.

中文翻译：

---

基于字典学习的混响消除功能使小鼠大脑皮层微脉管系统的深度分辨光声显微镜得以实现。

光声显微镜（PAM）利用血液中血红蛋白的光吸收来实现体内脑微血管的无标记高对比度成像。尽管时间分辨超声波检测为PAM配备了深度切片功能，但是深度处的大多数数据通常被来自浅层皮质层的声学混响伪影所掩盖，因此无法使用。在本文中，我们提出了一种首创的字典学习算法，可在保留基础微血管解剖结构的同时去除混响信号。使用嵌入在光学透明的聚二甲基硅氧烷体模中的染色珠在体外验证了该算法。随后，我们在活老鼠的大脑中证明了该算法可以将混响伪影抑制21.0±5.4 dB，