Introduction

In contemporary medical assessments, imaging provides non-invasive insights for early diagnosis, treatment planning, and real-time monitoring of physiological processes1. Optical imaging (OI) is particularly valuable in differentiating between tissue types and identifying molecular or cellular activity, allowing us to monitor functional and molecular dynamics2,3,4. Nevertheless, due to the strong light scattering in tissue, the penetration depth of OI is constrained to the subsurface-level5,6. On the other hand, ultrasound imaging (USI) can penetrate deeply into tissues, providing anatomical visualization7,8. USI is widely employed in clinical practice for guiding procedures and monitoring dynamic physiological changes, making it indispensable across both primary care and specialized settings9,10,11,12. However, USI lacks the molecular and functional specificity needed for targeted disease detection and assessment6,12.

Multimodal imaging combines complementary modalities to overcome individual limitations, synergizing their strengths to deliver comprehensive diagnostic insights. For example, combining positive emission tomography (PET) with computed tomography13,14 or magnetic resonance imaging (MRI)15,16 yields both metabolic and anatomical information, overcoming the limited spatial resolution of PET alone while enhancing diagnostic precision through detailed structural imaging. Similarly, the combination of OI and USI seamlessly merges functional and molecular insights with high-resolution, deep-tissue anatomical visualization. Fluorescence (FL) imaging (FLI) with USI offers molecular-targeted en-face guidance with anatomical localization, enhancing surgical accuracy17,18,19,20,21,22. Moreover, photoacoustic (PA) imaging (PAI) with USI provides detailed microvascular mapping, oxygen saturation levels, and molecular agent detection along with structural imaging, enabling comprehensive real-time diagnostics23,24,25,26,27.

Despite the advantages, integrating USI and OI faces several challenges arising from the opacity of conventional ultrasound transducers (UTs). Early solutions involved oblique light paths, which increased the system size and complexity and reduced signal-to-noise ratios (SNRs)28,29,30,31,32. An opto-US combiner can address this issue, but it still complicates the system design33,34,35,36. Another approach uses ring-shaped UTs with central holes for light passage, but this design compromises electrical properties and the image quality37. Recently, transparent ultrasound transducers (TUTs) have emerged, allowing light transmission, simplifying system design, and improving SNRs37,38,39,40,41,42. However, most TUTs still rely on single-element designs, requiring mechanical scanning and limiting clinical uses. Although recent developments in TUT-array43 and transparent capacitive micromachined ultrasound transducers44 have demonstrated potential, they have yet to be adapted into handheld forms suitable for clinical uses.

In this study, we present a clinical triple-modal USI, PAI, and FLI system, fully integrated in a handheld opto-US probe. This combined multimodal probe incorporates a 64-element TUT-array for USI/PAI, a miniaturized light source for PAI/FLI, and an optical camera for FLI, all in a small handheld device. Designed for a wide range of applications, the TUT-array operates at a 7 MHz center frequency with 45% bandwidth, and its transparency allows for coaxial alignment of light and US beams. It demonstrates feasibility in dual-modal USI/PAI on human arms, mapping vascular structures and hemoglobin oxygen saturation (sO2) from PAI overlaid on the conventional B-mode US image. In mice, triple-modal USI/PAI/FLI imaging before and after FDA-approved indocyanine green (ICG) injections quantifies signal enhancement in blood and lymphatic vessels. Finally, the triple-modal USI/PAI/FLI demonstrates clinical potential by guiding microsurgical treatment for lymphedema patients, particularly during preoperative planning. In microsurgical lymphaticovenous anastomosis (LVA) procedures, it is essential for clinicians to accurately observe the patient’s lymphatic and vascular conditions, as this information is key to planning personalized surgical treatment45. Although conventional imaging modalities, such as MRI46, USI47, and OI48,49, have been used for lymphatic assessments, no clinical device has yet been developed for multimodal imaging to allow real-time diagnosis of lymphatics, microvessels, and surrounding tissues. Our triple-modal imaging system provides a comprehensive solution by providing structural information through USI, mapping microvessels and lymphatics with PAI, and visualizing lymphatics en-face with FLI, all within a single, seamless system. This triple-modal imaging system not only overcomes the limitations of conventional approaches in reconstructive surgery field but also potentially opens avenues for broader clinical applications, providing a versatile platform for both surgical planning and non-invasive diagnostics.

Results

Fabrication of the TUT-array

As shown in Fig. 1a, the TUT-array is fabricated in five steps: preparing the individual layers, laminating the layers into a stack, curing and stabilizing the stack, slicing the stack into channels, and overcasting the acoustic lens. (1) The first step is preparing the individual components that form the TUT-array. As the piezoelectric material, lead magnesium niobate-lead titanate (PMN-PT) is used for its high electromechanical coupling coefficient (~ 0.6), high piezoelectric constant (~ 1190 pC/N), and low longitudinal velocity (~ 4600 m/S), properties that provide superior acoustic performance, especially in diced small elements within the TUT-array50. While PMN-PT inherently possesses optical transparency51, high-precision lapping and polishing are applied to achieve a clear surface for enhanced transparency. Both sides are fully coated with indium-tin-oxide (ITO) as a transparent electrode, followed by gold stripes along the edges to improve electrical connection efficiency with electrodes of other layers. The first matching layer (ML1), made of a SiO₂-epoxy composite39, is coated with gold applied as stripes along its top, bottom and side edges. The second matching layer (ML2) is made of Epotek 301. The PMN-PT and two MLs are cut with a dicing saw into rectangles measuring 23 × 5.5 mm in width and length. The TUT-array uses two flexible printed circuit boards (FPCBs): one has connectors for each channel control, and the other has a connector for the common voltage ground. The FPCBs feature a central opening, measuring 20 × 3.5 mm along the x and y axes, respectively, creating a physical void to allow the laser beam to pass through. A polycarbonate backing block (BB) is prepared, with its top surface having the same area as the PMN-PT rectangle. (2) After a small amount of bonding epoxy is applied between each layer, all of the layers are stacked in the bottom-to-top order of the BB, FPCB1, PMN-PT, ML1, FPCB2, and ML2. (3) The stacked assembly is pressed and heat-cured at 65 °C for 4 h, establishing firm physical contact between the layers. It is then cooled at room temperature for 24 h to ensure thermally stabilization. As a result, the gold stripe on the bottom surface of the PMN-PT is electrically connected to the gold on FPCB1, while the gold stripe on the top surface of the PMN-PT is electrically connected to the gold on FPCB2. (4) The bonded stack assembly is then diced using a 25 μm thick blade, forming 64 individual elements. (5) The diced stack is secured in a customized steel acoustic lens mold, then highly transparent silicone is poured, degassed, and cast into the mold. In a cross-sectional view cut through its center (Fig. 1b), all elements are precisely aligned at a 300-μm pitch. Figure 1c shows a photograph of the fully fabricated TUT-array at front- and back-side views. Notably, in the back-side view, the transparency of the US effective footprint in the TUT-array allows the background text to be clearly visible. Additionally, in the close-up image, the separated elements can be faintly distinguished.

Fig. 1: Fabrication process of a TUT-array.

a A step-by-step overview of the TUT-array fabrication process. b Photograph of a cross-section of the TUT-array. All the elements are precisely aligned. c Photographs of the TUT-array in front- and back-side views. ML matching layer, FPCB flexible printed circuit board, P PMN-PT, BB backing block, ITO indium-tin-oxide, and Au gold.

Acoustic and optical characteristics of the TUT-array

Prior to operating the TUT-array, we applied a high re-poling voltage (~ 180 V) to each element for 60 s to enhance dielectric performance, as the polarization of the PMN-PT may have degraded during fabrication, potentially compromising acoustic performance (Fig. S1, Supplementary information). To assess the acoustic performance, we used a pulse-echo test to examine the impulse response and verify the acoustic signal and its frequency response. The acoustic waveform was appropriately attenuated, with a duration of less than 1 μs, and the frequency range exhibited a center frequency of 7 MHz and a bandwidth of 45%, suitable for imaging purposes (Fig. 2a). The electrical impedance and phase angle spectra were measured to determine the electrical characteristics of the TUT-array across different frequencies (Fig. 2b). The resonance (fr) and anti-resonance (fa) frequencies were observed as distinct peaks at 6.53 MHz and 7.54 MHz, respectively. The effective electromechanical coupling coefficient (keff) was calculated to be 0.5. These results were consistent with Krimholtz-Leedom-Matthaei (KLM) modeling-based simulations (Fig. S2, Supplementary information), confirming that the TUT-array had been fabricated as designed. In addition, the acoustic beam field generated by the TUT-array matched well with the simulation results (Fig. S3, Supplementary information). The averaged center frequency and bandwidth of all elements were calculated to be 7 MHz and 45%, respectively (Fig. 2c). Based on the pulse-echo waveform of each element, the averaged SNR was 45 dB (Fig. 2d). In addition, crosstalk measurements demonstrated the effectiveness of the appropriately spaced pitch and the damping effect of the filled silicone between the elements (Fig. 2e). The measured values were lower than the standard threshold of –30 dB in the 1–9 MHz frequency range, with –35.5 dB for adjacent elements and –41.1 dB for the second nearest elements.

Fig. 2: Acoustic and optical characteristics of the TUT-array.

a Experimental pulse-echo impulse response waveform (black) and corresponding frequency spectrum (red) of the TUT-array. b Experimental electrical impedance (black) and phase angle (red) of the TUT-array. c Quantified center frequency and –6 dB bandwidth for each element. d Quantified ultrasound (US) A-line signal-to-noise ratio (SNR) for each element. e Crosstalk between adjacent and second nearest elements. f Light transmittance of layers composing the TUT-array. g Comparison of the acoustic beam fields of the commercial opaque US transducer and the TUT-array. ML matching layer, ITO-P ITO coated PMN-PT, BB backing block, and FWHM full-width-half-at-maximum.

Regarding the optical characteristics of the TUT-array, we spectroscopically examined each layer individually across the wavelength range of 400–900 nm. As seen in Fig. 2f, the near-infrared (NIR) optical wavelength region, widely used in PAI and FLI, is highlighted in a yellow box, and each TUT-array component exhibits light transmittance above 76.8% in this region. The fabricated TUT-array demonstrates a peak transmittance of 72.7% at 720 nm, with relatively constant optical transmittance across the NIR region. We compared the acoustic beam field of a commercial opaque UT (64 elements; center frequency of 8 MHz; bandwidth of 60%; L12-5A; and SonicLab Co., Ltd.) with that of the prototype TUT-array (Fig. 2g). Both the TUT-array and the commercial opaque UT operate central 32 elements, each with a driving voltage of 3.5 V. The US beam was focused at a depth of 25 mm, and detection was performed using a hydrophone. The SNR and acoustic pressure of the TUT-array were respectively 1 dB and 22 kPa lower than those of the commercial UT. Additionally, when measuring the full-width-at-half-maximum (FWHM) values of the acoustic fields, the TUT-array showed only a slight difference of 0.03 mm in the x-axis (lateral direction) and 0.14 mm in the y-axis (axial direction). These results indicate that the TUT-array’s performance was comparable to that of the commercial UT, with only minimal differences.

Performance benchmarking of triple-modal US/PA/FL imaging with an TUT-array opto-US probe

To investigate the potential of a triple-modal USI/PAI/FLI system using the TUT-array, we benchmarked the system’s performance in phantoms (Fig. 3). A schematic of the multimodal opto-US probe is illustrated in Fig. 3a. Similar to conventional UTs, the handheld, compact, and user-friendly probe is designed to be easy for clinicians to adopt. The opto-US probe is composed of the TUT-array and an optical module. Within the optical module, light is delivered through an optical fiber and transformed into a line beam by an engineered diffuser (ED). The line beam is then reflected by the mirror and dichroic mirror (DM), passing through the TUT-array and a US gel-pad to reach the target. The TUT-array is positioned coaxially at the end of the light path, enabling both the transmission and reception of US waves in a coaxial configuration. An NIR camera placed behind the DM captures a clear field of view (FOV) beyond the TUT-array for wavelengths above 800 nm. The photographs show the internal view of the opto-US probe and clearly illustrate how the laser illumination passes through when it is turned on and off. More detailed information is provided in the “Methods” section. Altogether, our new opto-US probe is much thinner, smaller, and lighter than the previously reported handheld PAUS probe with an opaque UT24,52,53,54,55. These improvements result from both the coaxial alignment of the acoustic and optical planes, and the significantly smaller laser delivery unit, achieved by replacing a bulky fiber bundle with a single optical fiber, which further enhances the optical coupling efficiency.

Fig. 3: Triple-modal USI/PAI/FLI imaging system with an opto-US imaging probe via a TUT-array and its performance benchmark.

a Schematic and photographs of the opto-US probe, showing all optical and US components. The laser beam passes directly through the TUT-array, illustrating its transparency, which enables the beam to be coaxially aligned with the US transmission path. USI, PAI, and FLI phantom imaging performance via the opto-US probe. b US B-scan images with wire targets and c anechoic targets in a standard US phantom. d PA B-scan image with black strings in a tissue-mimicking phantom. e FL en-face image of a tube filled with indocyanine green (ICG) beneath stacked layers of chicken tissue. f USI spatial resolution. g USI contrast-to-noise ratio (CNR). h PAI spatial resolution. i FL intensity vs depth. j Simultaneous trimodal USI/PAI/FLI of tubes filled with blood and ICG beneath stacked layers of chicken tissue. DM dichroic mirror, M mirror, ED engineered diffuser, and OF optical fiber.

The opto-US probe is connected to a laser and data acquisition (DAQ) system, establishing an integrated triple-modal imaging system (Fig. S4, Supplementary information). The US Tx/Rx system is programmable for each TUT-array element, enabling US transmission and reception along with the detection of laser-induced PA signals. Additionally, FL images are acquired in real-time through the USB-connected NIR camera, delivering en-face visualization directly to the computer. This setup streamlines the triple-modal USI/PAI/FLI imaging acquisition by following a synchronized timing sequence to ensure comprehensive imaging across modalities (Fig. S5, Supplementary information). To assess the triple-modal USI (Fig. 3b, c), PAI (Fig. 3d), and FLI (Fig. 3e) system’s performance with the opto-US probe, we conducted imaging experiments in phantoms for each modality. Initially, we focused the US transmission at a depth of 20 mm. For USI, we imaged line targets (Fig. 3b) and anechoic targets (Fig. 3c) using a standard US phantom model. Based on Fig. 3b, the USI lateral and axial resolutions were measured to be 0.31 mm and 0.39 mm, respectively, with the lateral resolution gradually decreasing as the depth increases (Fig. 3f). Additionally, as shown in Fig. 3c, the US contrast-to-noise ratio (CNR) was calculated to be 7.2 at a depth of 20 mm (Fig. 3g). The lateral and axial resolutions of our TUT-array are almost identical to those of the commercial opaque UT, and its CNR, measured using a grayscale target, shows an average difference of only 0.17 dB (Fig. S6, Supplementary information). Thus, the TUT-array shows performance quite comparable to that of the commercial opaque UT. To evaluate the PAI performance, we prepared a tissue-mimicking phantom with black nylon strings56 and acquired its cross-sectional image (Fig. 3d). The PA spatial resolution was measured from the string’s profile, yielding lateral and axial resolutions of 0.38 mm and 0.71 mm, respectively (Fig. 3h). The PAI performance of the opto-US probe was compared to that of the commercial UT with an oblique laser alignment, showing only a slight lateral resolution difference of 0.03 mm and an axial resolution difference of 0.07 mm, presumably arising from the slightly lower center frequency of the TUT-array (7 MHz) than the commercial UT (8 MHz). However, the opto-US probe with coaxial laser alignment achieved an average of 7 dB higher SNR with depth, indicating improved efficiency in PAI (Fig. S7, Supplementary information). While Fig. S7 shows comparable SNR degradation between the TUT-array and the commercial probe, this result reflects both the acoustic sensitivity and the light delivery efficiency. To clarify the specific impact of coaxial versus oblique illumination, we performed an additional phantom experiment using the same TUT-array under both configurations, demonstrating that coaxial illumination yields a higher SNR of 8 dB and more uniform signal decay with depth (Fig. S8, Supplementary information). Moreover, the opto-US probe can simultaneously provide FLI from the top view, utilizing the TUT as an optical window. To benchmark the FLI performance, we imaged an ICG-filled tube beneath stacked layers of chicken tissue (Fig. 3e). The ICG-filled tube remained clearly visible with high intensity up to a depth of 4.5 mm and detectable up to a maximum depth of 7.3 mm, both with and without the TUT-array (Fig. 3i). The FL SNRs in both configurations, where the average SNR difference between measurements with and without the TUT-array is 0.82 dB across various depths (Fig. S9, Supplementary information).

To demonstrate simultaneous USI/PAI/FLI, two tubes, one filled with mouse blood and the other filled with ICG dye, were placed beneath chicken tissue (Figs. 3j and S10, Supplementary information). For multispectral PAI, we used 690, 720, 756, and 796 nm wavelengths. In the PA B-scan image, both blood and ICG are well visible, with the ICG signal appearing stronger than the blood signal. A spectral unmixing algorithm, isolated and visualized the ICG signal. Although USI alone could not distinguish between blood and ICG, a PA/US overlaid image did provide depth information for blood and ICG, which were measured to be 3 mm deep below the skin. In addition, the FLI en-face image offered an intuitive top view of the location of the ICG.

3D triple-modal mapping of lymphatics in live animals

To investigate the system’s trimodal imaging capability in vivo, we performed simultaneous PA/US/FL imaging to map the lymphatics of an anesthetized Sprague Dawley rat (Fig. 4). To volumetrically scan the rat’s entire right upper limb region, we attached the opto-US probe to a motorized stage and performed mosaic mechanical raster scanning to create a 45 × 30 mm2 synthetic aperture from five overlapping linear scans (Fig. 4a). In each scan slice, we acquired PA B-mode images at 690, 720, 756, and 790 nm (Fig. S11, Supplementary information), one US B-mode image, and one FL image at 756 nm. These images were stacked to create synthetic 3D PA/US volumes and FL en-face image. After preinjection imaging, a 0.2 mL of 1 mM ICG solution was injected intradermally into the right frontal paw, followed by postinjection imaging over the same FOV (Fig. 4b).

Fig. 4: 3D triple-modal mapping of lymphatics in a live rat.

Simultaneous ultrasound (US)/photoacoustic (PA)/fluorescence (FL) imaging was performed via mosaic 3D motorized scanning over the right axillary region of an anesthetized rat a before and b after indocyanine green (ICG) injection. c Maximum amplitude projection (MAP) of PA volume visualizing the vascular network across the sagittal plane, c1 before and c2 after the ICG injection. The path of the overlaid unmixed ICG contrast (in green) in (c2) visualizes the axillary lymphatics, featuring a continuous connection from peripheral lymphatic vessels (LV) to the axillary sentinel lymph node (aSLN). d Coregistered PA/US 3D volume over the axillary region (Supplementary Movie 1). LV and superficial veins are identified underneath the dermal layer (DL). e Overlaid PA/US B-mode images e1 before and e2 after the ICG injection. Detailed musculoskeletal structures and organ anatomy are distinguishable from USI, while PAI demonstrates branching blood/lymphatic vessel distribution. f Mosaic-synthesized FLI via the TUT optical window f1 before and f2 after ICG injection. g Quantified PA and FL signal enhancement with ICG injection (mean ± standard error, n = 3, independent experiments). PM pectoral muscle, RH right humerus, BiB biceps brachii, RC rib cage, L liver. Scale bar = 1 cm. All values are expressed in a.u., which denotes arbitrary units.

Both PA volumes consistently showed abundant vessels extending into the right lateral region (Fig. 4c). The pre-injection PAI revealed subcutaneous blood vessels spanning the right forelimb and thoracic region (Fig. 4c1). The post-injection PA image shows a similar vessel distribution, but with significant local PA signal enhancement from the right frontal limb down to the shoulder. After spectral PA unmixing, the stream of unmixed ICG reveals the lymphatic network in the right upper limb, flowing from the periphery and toward the thorax (Fig. 4c2). The axillary sentinel lymph node (aSLN) appears as a major node where the ICG stream aligns following the axillary curve. In the co-registered PA/US volume, US imaging visualizes the anatomical structures of major organs and musculoskeletal features in the right scapular and thoracic regions, while PA imaging identifies subcutaneous blood and lymphatic vessels traversing the connective tissues (Supplementary Movie 1 and Fig. 4d, e). Notably, the lymphatic vessels are invisible in the US images and are distinguishable only in the PA images, where they are located approximately 3 mm beneath the dermal layer, alongside the adjacent superficial veins52. ICG serves as both a PA contrast agent and a fluorophore, allowing the lymphatic network to also be visualized in the corresponding FL images, where it exhibits a high spatial correlation with the vascular morphology in the PA lymphatic maps (Fig. 4f). The PA and FL signal enhancement after ICG injection can be further quantitatively assessed from the imaged major blood vessel and aSLN (Fig. 4g), demonstrating a mere difference from blood vessel while significant signal increase at the aSLN region, measured as 11-fold and 3.6-fold increase from PAI and FLI, respectively. Overall, the equivalent discoveries of the lymphatic network in both PAI and FLI, along with the simultaneous trimodal imaging that combines anatomical structure and molecular imaging, successfully demonstrates the seamless integration of optical and acoustic imaging using the TUT.

Clinical 3D functional PA/US vascular imaging of a human arm

To investigate the system’s clinical feasibility, we employed the opto-US probe for 3D functional PA/US vascular imaging of a human arm (Supplementary Movie 2 and Fig. 5a). As in the preclinical imaging, we performed 3D mosaic scanning of a 30 × 50 mm2 area of a healthy volunteer’s forearm, capturing two lanes of overlapping linear scans. We performed multispectral PAI with 690, 756, 796, 820, and 866 nm laser excitation to precisely unmix blood contrast into deoxygenated (Hb) and oxygenated hemoglobin (HbO2) molecular contrast. Because we were not performing FLI, the NIR camera and dichroic mirror inside the optical module were replaced with a standard mirror to utilize a broader band of laser excitation. The acquired US/PA volume data were synthetically merged into a wide FOV, resulting in a 3D image volume as shown in Fig. 5b.

Fig. 5: Bimodal 3D functional PA/US imaging of a human arm in real time.

a Experimental configuration. b 3D rendering of overlaid PA/US volumes, simultaneously describing the tissue anatomy from US and registered vasculatures in PA acquired at 866 nm (PA866 nm) in the forearm (Supplementary Movie 2). Representative B-mode US (b1) and PA (b2) depict distinguishable musculoskeletal structures and adjacent blood vessels, respectively. c1–c3 Maximum amplitude projection (MAP) in the x-y plane for c1 PA866nm, c2 depth encoded PA866nm, and c3 oxygen saturation (sO2) map distinguishes the arterial network from the superficial venous network. d B-mode US/PA image cut along the dotted line in (c1). Pulsatile arterial motion identified from the B-mode US/PA images (Supplementary Movie 3). e The temporal change and corresponding frequency spectrum demonstrate 1.4 Hz (84 bpm) pulsation in both US and PA images in the arterial region-of-interest (ROI) marked with a yellow box in (d). UA ulnar artery, UV ulnar vein, UN ulnar nerve, FDS flexor digitorum superficialis, SV superficial vein, ED epidermis, RA radial artery, BR brachioradialis, and FCR flexor carpi radialis. Scale bars = 1 cm. All values are expressed in a.u., which denotes arbitrary units.

The US (Fig. 5b1) and PA (Fig. 5b2) images provide detailed cross-sectional views of upper limb tissue structures (such as the ulnar artery, vein, nerve, and flexor digitorum superficialis) and adjacent blood vessels, enabling precise anatomical localization and assessment of vascular and tissue morphology. In the PA MAP image, the intricate vascular network within the human arm is clearly visualized in a top-projection view (Fig. 5c1). We further generated a depth-encoded MAP image conveying the depth information of PA contrasts with pseudo-colors, which clearly illustrates the vascular connectivity at depths ranging from 0 to 10 mm (Fig. 5c2). Additionally, sO₂ levels calculated from the ratio between the unmixed HbO2 to the total Hb (the sum of the unmixed Hb and HbO2) are spatially mapped over the PA MAP image, displaying highly oxygenated vessels in red and less oxygenated vessels in blue (Fig. 5c3). The sO2 is measured with high consistency along the vascular continuity, highlighting a number of vessels in the deep region with an approximately 90% sO2 level. We speculate that these vessels are arteries, located in relatively deeper regions than superficial veins and containing richly oxygenated blood.

Beyond its multispectral imaging capabilities, the 20-Hz frame rate of the bimodal PA/US imaging captures cardiac circulation dynamics (Supplementary Movie 3 and Fig. 5d, e). We continuously imaged a forelimb cross-section over 10 s (200 frames) at a single 866 nm wavelength (Fig. 5d). Pulsatile motion was clearly observed at an identical location in both the US and PA images. To quantitatively assess this motion, we defined a region of interest (ROI) encompassing the arterial lumen and walls, and measured the frame-by-frame intensity fluctuations from both imaging modalities (Fig. 5e). The pulsation patterns from the US and PA images were temporally correlated, exhibiting synchronized systolic peaks. Spectral analysis revealed multiple harmonic peaks with a fundamental frequency at 1.4 Hz (84 bpm), confirming the regular heartbeat of the healthy volunteer.

Lymphaticovenous anastomosis (LVA) microsurgery of lymphedema patients guided by triple-modal USI/PAI/FLI

A LVA procedure is a minimally invasive microsurgery connecting LVs (diameter <0.5 mm if normal, diameter <1 mm if lymphedematous, depth <5 mm) to nearby veins (diameter ~0.5 mm, depth <1 cm) to facilitate lymph drainage57,58. After ICG injection, surgeons use both FLI and USI to preoperatively confirm the location of veins and LVs, FLI captures the ICG stream in the top view, while USI offers depth-sensitive location information of the LVs and veins for microsurgery. However, pinpointing LV locations becomes challenging in severe lymphedema patients where FLI lymphatic contrast is only faintly visible or disrupted by abnormal patterns due to ICG retention, and USI lacks of specificity in distinguishing between veins and ICG-dyed lymphatic vessels. In these typical cases, operators often need to repetitively use two separate FLI and USI systems with different FOVs, which complicates and prolongs the final confirmation and relies on the operator’s empirical judgment to identify lymph vessels.

In contrast, the opto-US probe is particularly well-suited for guiding LVA procedures, since it simultaneously provides comprehensive triple-modal US/PA/FL guided images within a single device. Before applying the triple-modal imaging system to patients, we demonstrated its capabilities in healthy volunteers (Fig. 6a–d and Supplementary Movie 4). ICG dye was intradermally injected via the interdigital web of the volunteer’s hand (Fig. 6a). The normal perfusion of ICG through the LV network was subsequently confirmed using commercial FL lymphography (Fluobeam 800, Fluoptics, France), while the FL trace of the LVs from the top view facilitated transcutaneous inspection using a 33 MHz high-frequency clinical USI system (Aplio i800, Canon Medical Systems, Japan) (Fig. 6b). High-frequency USI enabled identification of LVs based on their shape, echogenic texture, and response to pressure, however, differentiating LVs from adjacent veins remained challenging and time-intensive. Next, using the opto-US probe, we demonstrated its effectiveness in identifying LVs and adjacent veins both before and after ICG injection. Before the ICG injection (Fig. 6c), multispectral PAI was performed, acquiring strong PA signals in the transcutaneous layer. Spectral PA unmixing allowed visualization of the Hb distribution, while the FL en-face image showed no detectable signal, as expected before ICG injection. The overlaid PA/US image confirmed the location of veins within the transcutaneous layer. After ICG injection, triple-modal images were acquired at the same location (Fig. 6d). Spectral PA unmixing distinctly separated Hb and ICG signals, precisely localizing LVs and blood vessels. The FL en-face image also identified LVs at the same lateral position as the unmixed PA ICG image. Furthermore, the overlaid PA/US image confirmed that the LVs were located approximately 3 mm beneath the skin surface, along with surrounding blood vessels.

Fig. 6: Triple-modal USI/PAI/FLI guided microsurgery to perform lymphaticovenous anastomosis (LVA), of lymphedema patients.

a Lymphography of a healthy volunteer with indocyanine green (ICG) injection (Supplementary Movie 4). b Conventional clinical imaging methods: fluorescence (FL) lymphography and ultrasound imaging (USI, 33 MHz). Red and green arrows indicate veins and potential lymphatic vessels (LVs), respectively. Triple-modal USI/PAI/FLI using the opto-US probe c before and d after ICG injection. e Preoperative lymphatic illustration and ICG