We have developed spatially Fourier-encoded photoacoustic microscopy using a digital micromirror device. The spatial intensity distribution of laser pulses is Fourier-encoded, and a series of such encoded photoacoustic measurements allows one to decode the spatial distribution of optical absorption. The throughput and Fellgett advantages were demonstrated by imaging a chromium target. By using 63 spatial elements, the signal-to-noise ratio in the recovered photoacoustic signal was enhanced by ~4×. The system was used to image two biological targets, a monolayer of red blood cells and melanoma cells.

Optical-resolution photoacoustic microscopy (OR-PAM) [^{3}) is to increase the local light fluence. However, the light fluence cannot exceed the damage threshold. For biological tissues, the maximum light fluence is usually further restricted by the ANSI safety standard [

The SNR can be significantly improved by multiplexing. By delivering more energy to the target per laser pulse, multiplexing illumination gains a throughput (Jacquinot) advantage over single-element illumination. The signal of each individual element can be recovered by decoding the measured signal series. If the measurement contains only signal-independent noise sources,

Fourier and Hadamard transformations are commonly used in multiplexing methods [^{2}. However, this work cannot be readily implemented in microscopic biomedical imaging for two reasons. First, the utilization of the one-dimensional mask required scanning the target, resulting in a reduced imaging frame rate. Second, the lateral resolution (100 µm) was insufficient to distinguish micrometer-scale biological features.

In this letter, we present a spatially Fourier-encoded PAM (SFE-PAM) system using a digital micromirror device (DMD). This technique is inspired by the spectral Fourier-multiplexing method. The DMD was used as an optical encoder to produce time-domain discrete Fourier modulation patterns for each individual spatial element. Therefore, the SFE-PAM system can simultaneously deliver modulated light fluence to multiple locations of the target, thereby significantly improving the SNR of the PA signal over that of a single-element raster scan.

Herein we briefly review the Fourier multiplexing theory and derive the expected SNR enhancement for the SFE-PAM system. A 2D area of the target is divided into _{a} (^{th} element, the generated PA A-line detected by the ultrasonic transducer, ν_{r} (_{r} (_{0} (_{a} (_{0} (^{3} or another chosen unit]. _{r}.

The principle of SFE-PAM is illustrated in ^{th} measurement (_{f} (_{f} (

This sequence is then decoded using the inverse Fourier transformation, which extracts the magnitude at each modulation frequency. The Fourier-decoded PA A-line, ν_{f} (_{mn} is the Kronecker delta function. _{r}_{f}

In addition, we compared the standard deviation of the PA signal amplitude in raster-scanned and Fourier-decoded PA A-lines. The signal’s standard deviation in the Fourier-decoded PA A-line, σ_{f}, is derived from

The Fellgett advantage η is defined as the ratio of the SNR of the decoded PA A-line, _{f}_{f}_{f}, to that of the raster-scanned PA A-line, _{r}_{r}_{r}'. Combining

The DMD-based SFE-PAM system (_{4} laser (SPOT 10-200-532, λ = 532 nm, Elforlight) as the illumination source. The generated laser pulses have a 2 ns pulse duration with a repetition rate of 10 kHz. The pulse energy is monitored by a photodiode detector (SM05PD1A, Thorlabs) to compensate for energy fluctuation. After expansion and collimation, the laser pulses impinge on the DMD (Discovery 4100, Texas Instruments) at an incident angle of 24° with respect to the surface normal. For each individual Fourier-modulation pattern, we use the two-dimensional Floyd-Steinberg error diffusion algorithm [^{2}, with a lateral resolution of ~3.5 µm [

To demonstrate spatial Fourier-encoding in PAM, we first imaged a chromium target of the letters “WU”, made in-house by vacuum-deposition on a microscope coverslip substrate [_{r} (_{f}

A volumetric image of the target’s optical absorption was produced by allocating each individual PA A-line in the stack to the corresponding spatial element position. Then, this volumetric 3D image was rendered as a 2D maximum amplitude projection (MAP) image along the depth direction. The PA image of the letters “WU” demonstrated that the Fourier encoding had significantly enhanced image quality compared to the raster scanning (

To test the SFE-PAM system performance with biological contrasts, we imaged red blood cells (RBCs) and melanoma cells. The average pulse energy was set to be 2.8 nJ per element. A monolayer RBC target was prepared by spreading a drop of whole bovine blood (910-250, Quad Five) across a coverslip. Due to low PA SNR, the raster scanning resulted in a low quality image [

Although implemented in OR-PAM, the conceptual design of spatial Fourier encoding is applicable to many other imaging modalities,

In summary, we report the first experimental demonstration of a SFE-PAM system based on a DMD, which is used to modulate the spatial light distribution of the laser beam. Each spatial element of the target is illuminated by modulated light fluence with a different frequency. The spatial optical absorption distribution is recovered by decoding a series of Fourier-encoded PA measurements. Compared to raster scanning with the same number of measurements, the SFE-PAM system enables more energy-efficient delivery of the laser illumination. In addition, this system possesses the Fellgett advantage, in terms of PA SNR, in the Fourier-decoded PA A-lines. The enhanced SNR benefits PA images by increasing image contrast-to-noise ratio and target identifiability. The SFE-PAM system is an attractive tool for the accurate PA measurement of biological targets with low optical absorption coefficients or low damage threshold.

The authors thank Lijun Ma, Yong Zhou, Chi Zhang, and Lei Li for helpful discussions, and Professor James Ballard for close reading of the manuscript. This work was sponsored in part by National Institutes of Health grants DP1 EB016986 (NIH Director’s Pioneer Award), R01CA134539, R01CA157277, and R01 CA159959. L. W. has a financial interest in Microphotoacoustics, Inc. and Endra, Inc., which, however, did not support this work.

Principle of spatially Fourier-encoded photoacoustic microscopy. A 3-pixel encoding of (_{f}

Schematic overview of the spatially Fourier-encoded photoacoustic microscopy system (not to scale). AMP, signal amplifiers and filters; BS, beam sampler; DAQ, data acquisition system; DMD, digital micromirror device; L1–L2, lenses; M1–M2, mirrors; OL1, objective lens (Mitutoyo, M PLAN APO 10×/0.28); OL2, objective lens (Olympus, LUCPlanFLN 40×/0.60); PD, photodiode detector; UT, ultrasonic transducer.

(a) Photograph of a chromium target of the letters “WU”. (b) Stack of PA A-lines of the letter “U” acquired with raster scanning. The measurement result from the raster scanning was acquired twice and averaged to match the total number of signals acquired with the Fourier encoding. (c) Stack of radiofrequency PA signals of the letter “U” acquired by Fourier encoding. (d) Comparison of PA signals (between red triangles) in (b) and (c). (e) Stack of Fourier-decoded PA A-lines of the letter “U”. (f) Comparison of PA A-lines (between red triangles) in (b) and (e).

PA images of a “WU” logo made of deposited chromium acquired by (a) raster scanning and (b) Fourier encoding.

Demonstration of spatial Fourier-encoding method in biological targets. PA images of monolayer RBCs acquired by (a) raster scanning and (b) Fourier encoding. RBCs are identified by white dashed circles. A melanoma cell was also imaged using (c) raster scanning and (d) Fourier encoding.