A comprehensive review of high-performance photoacoustic microscopy systems

Abstract

Photoacoustic microscopy (PAM), an imaging modality with emerging importance in diverse biomedical applications, provides excellent structural and functional information at the micro-scale. Technological innovations have significantly enhanced PAM's performance, including sensitivity and contrast, making it a powerful tool. This review explores high-performance PAM, focusing on its signal-to-noise ratio, imaging speed, resolution, depth, functionality, and practicality, and commenting on the role of artificial intelligence in enhancing each feature. After providing comprehensive insights, the review concludes with future directions for developing high-performance PAM for advanced biomedical imaging and clinical applications.

Keywords: artificial intelligence; depth; high performance; imaging speed; photoacoustic microscopy; resolution; signal-to-noise ratio; system development.

Fig. 1

(a) Configuration diagrams for opto-ultrasound alignments. (b) System schematic of OUC-based PAM. Reprinted with permission from Ref. . (c) System schematic of RUT-based PAM. Reprinted with permission from Ref. © Springer Nature. (d) System schematic of TUT-based PAM. Reprinted with permission from Ref. .

Fig. 2

AI-based image processing in PAM. (a) Distortion correction. Reprinted with permission from Ref. . (b) Low laser dosage excitation. Reprinted with permission from Ref. .

Fig. 3

Voice coil-based PAM systems. (a) System schematic of voice-coil-based OR-PAM. (b) PA MAP images of mouse vasculature. Reprinted with permission from Ref. © Optica Publishing Group. (c) System schematic of voice-coil-based UV-PARS. (d) UV-PARS image of a tissue section of human prostate and the adjacent true H&E image. Reprinted with permission from Ref. .

Fig. 4

MEMS-based PAM systems. (a) System schematic of reflection-mode MEMS-UV-PAM. (b) PA MAP image of a frozen mouse brain tissue section and a standard H&E stained image of the tissue section. Reprinted with permission from Ref. © John Wiley and Sons. (c) System schematic of MEMS-based dual-scale multi-wavelength PAM. (d) PA MAP and depth-encoded PA images of normal mouse brain cortex. Reprinted with permission from Ref. .

Fig. 5

Galvanometer-based PAM systems. (a) System schematic of PAM using a cylindrically focused transparent ultrasound transducer (CFT-UT). (b) PA images of a mouse ear and mouse cortex vasculature. Reprinted with permission from Ref. . (c) System schematic of parallel UV-PAM. Reprinted with permission from Ref. .

Fig. 6

Polygon-based PAM systems. (a) System schematic of ultrafast functional PAM. (b) PA images of mouse brain vasculature and the oxygen saturation (sO2) of hemoglobin. Reprinted with permission from Ref. . (c) System schematics of two generations of ultrafast functional PAM, in the green and red dashed boxes, respectively. (d) PA images of the vessel-by-vessel oxygen saturation mapping of a mouse placenta. Reprinted with permission from Ref. .

Fig. 7

Lateral multifocal implementation. (a) System schematic of multifocal structured illumination microscopy using a beamsplitting grating. (b) Maximum amplitude projection images of a mouse ear. Reprinted with permission from Ref. . (c) System schematic of acoustic ergodic relay-based PAM. (d) Multifocal optical resolution PA image of a mouse ear. Reprinted with permission from Ref. .

Fig. 8

Deep learning-assisted reconstruction of undersampled PAM data. Demonstration of in vivo PAM images of blood vessels in a (a) mouse ear and (b) mouse brain. Reprinted with permission from Refs. , .

Fig. 9

Nonlinear effect-based PAM. (a) Principle of PI-PAM. (b) Image and line profile comparisons between conventional PAM and PI-PAM. Reprinted with permission from Ref. © American Physical Society. (c) Principle of high-order coefficient PAM. (d) Image and line profile comparisons between conventional PAM and high-order PAM. Reprinted with permission from Ref. .

Fig. 10

Localization-based PAM. (a) Principle of RBC-localization PAM. (b) Image comparison between conventional PAM and RBC-localization PAM. Reprinted with permission from Ref. . (c) Principle of UV-localization PAM. (d) Image comparison between MIR-PAM and UV-localization PAM. Reprinted with permission from Ref. © Springer Nature.

Fig. 11

AI-based resolution enhancement in PAM (a) AR-PAM to OR-PAM transformation in a mouse ear blood vessel image. Reprinted with permission from Ref. . (b) PAM to confocal fluorescence microscopy transformation in a fibroblast image. Reprinted with permission from Ref. .

Fig. 12

Axial multifocal implementations in PAM. (a) System schematic of spatially invariant resolution (SIR-) PAM. (b) SIR (upper) and conventional (lower) PA images of zebrafish embryos. Reprinted with permission from Ref. . (c) System schematic of DOE-based PAM. (d) Gaussian beam (GB, left) and needle beam (NB, right) PA images of a mouse brain without a skull. Reprinted with permission from Ref. . © Springer Nature. (e) System schematic of metasurface-based PAM. (f) PA histological images of a thick brain sample without (upper) and with (lower) a UV metasurface (UVM). Reprinted with permission from Ref. .

Fig. 13

Deep learning-assisted DOF extension in PAM. (a) In vivo PAM images of blood vessels in the eye and brain. Reprinted with permission from Ref. . (b) Demonstration of in vivo AR-PAM images of blood vessels in the abdomen. Reprinted with permission from Ref. .

Fig. 14

Multispectral PAM. Functional analysis for studying (a) oxygen saturation, (b) blood flow speed, and (c) metabolism (lipogenesis). Reprinted with permission from Refs. , , . Contrast agent-aided imaging with (d) organic and (e) inorganic dyes. CV, cerebral vessels; mLV, meningeal lymphatic vessels; Hb, hemoglobin. Reprinted with permission from Refs. , .

Fig. 15

Multimodal PAM. (a) System schematic of Co5M. (b) Multimodal images of signals from optoacoustic (OA), second harmonic generation (SHG), two-photon excitation fluorescence (2PEF), and third harmonic generation (THG), and brightfield (BF) images in a mouse ear. Reprinted with permission from Ref. . (c) Schematic diagram of the quadruple imaging system. (d) Multimodal images of signals from PA, US, OCT, and fluorescence (FL) in a mouse eye. Reprinted with permission from Ref. .

Fig. 16

AI-based functional processing. (a) DL-based vessel segmentation in a human palm. Reprinted with permission from Ref. . (b) DL-based high-resolution oxygen saturation mapping in a mouse ear. Reprinted with permission from Ref. .

Fig. 17

Laser diode-based PAM. (a) System schematic of PLD-based PAM. (b) PA images of mouse ear vasculature. Reprinted with permission from Ref. © Chinese Laser Press. (c) System schematic of CWLD-based PAM. (d) A set of in vivo CWLD-PA images of a mouse ear. Reprinted with permission from Ref. .

Fig. 18

Handheld PAM systems. System schematics of (a) a MEMS-based handheld PAM, (b) an optical scanning handheld PAM, and (c) a galvanometer-based freehand scanning PAM. Reprinted with permission from Refs. , , and © Optica Publishing Group.

Fig. 19

Implanted PAM systems. System schematic of (a) wearable OR-PAM on freely moving rats, and (b) dual-wavelength head-mounted PAM on a freely moving mouse. Reprinted with permission from Refs. © John Wiley and Sons, and .

Fig. 20

AI-based processing. (a) Virtual H&E staining. Reprinted with permission from Ref. . (b) Feature-fusion cancer classification. Reprinted with permission from Ref. .