Laser-generated focused ultrasound for thrombus characterization and fragmentation: ex vivo feasibility study Authors: Kyu Kwan Park, Pilgyu Sang, Min Gyu Joo, Deblina Biswas, Hyoung Won Baac Published: 2026 ABSTRACT Intravascular thrombus poses a significant risk for ischemic events, yet precise characterization and selective mechanical intervention remain challenging. Here, we present a laser-generated focused ultrasound (LGFU) platform incorporating a long-focused photoacoustic lens (focal length: 28 mm), which achieved a 1.79-fold increase in penetration depth compared with previously reported LGFU systems while maintaining a tight focal spot (60 μm lateral, 150 μm axial) and broadband acoustic output (− 6 dB bandwidth: ~ 15 MHz). Unlike previously reported LGFU systems developed primarily for therapeutic applications, the present platform combines LGFU-based ultrasound generation with co-aligned hydrophone reception for spectral characterization of thrombi under vessel-mimicking phantom conditions. Empirical mode decomposition (EMD) applied to complex backscattered signals enabled discrimination of thrombus thicknesses down to ~ 100 μm, highlighting the utility of broadband photoacoustic signals for sub-millimeter spectral characterization. Finite-element simulations supported the thickness-dependent spectral shift underlying this discrimination. Then, we confirmed that thrombus fragmentation linewidths were dependent on an input laser energy, which was evaluated by using a photoacoustic lens with a 15 mm focal length. Finally, fragmentation of the vessel-mimicking phantom was demonstrated with a longer focal length (28 mm). The observed fragmentation behavior was consistent with cavitation-mediated mechanical effects under the tested conditions. These results support the feasibility of an LGFU platform for combined thrombus characterization and localized fragmentation under proof-of-concept phantom and ex vivo conditions.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58064-2. Introduction A thrombus is an intravascular coagulated blood mass that obstructs blood flow and leads to thrombotic vascular disorders1. Thrombi can develop in both arteries and veins. In the carotid artery, intraluminal thrombus can contribute to luminal narrowing and embolic risk, potentially leading to ischemic stroke and other severe neurological outcomes2–4. In particular, the ability to detect and characterize sub-millimeter thrombi at an early stage is significant to support timely intervention before progressive luminal narrowing or embolization occurs. Given these risks, early diagnosis and treatment of carotid artery stenosis are critical, and numerous diagnostic methods have been developed and implemented for this purpose5–10. Representative diagnostic methods include angiography, computed tomography (CT)11,12, magnetic resonance imaging (MRI)13,14, and vascular ultrasound5,15–17. Although angiography, CT, and MRI provide high diagnostic accuracy, they require contrast agents or costly equipment, which imposes practical limitations. In contrast, vascular ultrasound is non-invasive and allows for real-time evaluation of blood flow and stenotic lesions, making it a widely adopted first-line diagnostic tool18–21. Clinical carotid ultrasound imaging is commonly performed using piezoelectric transducers in the several-megahertz range, often with linear array configurations extending above 5 MHz 22–24. Lower frequencies enable deeper tissue penetration but provide limited spatial resolution, whereas higher frequencies offer improved spatial resolution at the expense of greater attenuation at the typical carotid artery depth. This tradeoff arises primarily from frequency-dependent acoustic attenuation in tissue and overall system design. Consequently, conventional ultrasound approaches face practical difficulty in achieving both deep penetration and high spatial resolution simultaneously when using a single transducer configuration25,26. Clinical approaches for intravascular thrombus include pharmacological therapy, catheter-based mechanical removal, and surgical intervention27–29. While thrombolytics and anticoagulants are minimally invasive, they carry risks such as hemorrhage. Mechanical thrombectomy enables direct thrombus extraction but remains invasive and may cause vascular injury. Surgical approaches provide definitive treatment yet involve considerable procedural risk. Collectively, these limitations motivate the exploration of non-invasive approaches for localized thrombus characterization and controlled mechanical intervention. Laser-generated focused ultrasound (LGFU) offers a potential approach for localized thrombus characterization and mechanical intervention. In LGFU, a CNT-PDMS thin film fabricated on a concave lens substrate absorbs pulsed laser energy and generates broadband acoustic emission via the photoacoustic effect30. The generated acoustic waves are geometrically concentrated at the acoustic focus, enabling localized high-pressure ultrasound generation at increased focal depths. Combined with the impulse-like broadband output of photoacoustic generation, this focusing architecture provides a basis for backscatter-based spectral discrimination of thrombi under simplified phantom conditions. Previously reported LGFU systems have demonstrated diverse biomedical applications, including cell detachment, ultrasonic scalpels, and tissue characterization31–35. Because the photoacoustic lens generates broadband acoustic pulses, it can be integrated with a broadband ultrasound receiver to form a pulse-echo configuration which is suitable for scanning-based assessment. Such systems have been demonstrated in prior studies using all-optical ultrasound architectures34,36. However, their short focal length (< 15 mm) has limited their applicability to deeper targets. To address this limitation, we investigated whether a long-focus LGFU architecture could support localized thrombus characterization through backscatter-based spectral analysis and localized mechanical fragmentation under vessel-mimicking phantom and ex vivo conditions. In this sense, the main contribution of this study lies in integrating these two functions within a unified proof-of-concept LGFU platform. In this study, we present an LGFU platform based on a long-focus photoacoustic lens and evaluate its feasibility for localized thrombus characterization and localized fragmentation under vessel-mimicking phantom and ex vivo conditions. Through numerical simulations and experimental validation, thrombi of varying thickness were investigated within vessel-mimicking phantoms. Empirical mode decomposition (EMD) was employed to analyze complex backscattered signals, enabling frequency-based discrimination of thrombi as thin as 100 μm. Furthermore, focal cavitation was utilized to achieve localized mechanical fragmentation, with fragmentation width controlled through laser pulse energy adjustment. This study provides a proof-of-concept demonstration that a long-focus LGFU platform can support localized thrombus characterization and energy-tunable mechanical fragmentation under simplified phantom and ex vivo conditions, thereby establishing a basis for future translational development. Materials and methods Fabrication of CNT-PDMS photoacoustic lens A photoacoustic lens was fabricated by coating a plano-concave lens substrate (diameter: 44 mm, radius of curvature: 28 mm) with a thin film of carbon nanotube (CNT)–polydimethylsiloxane (PDMS) composite. First, PDMS (Sylgard 184, Sigma-Aldrich, USA) was diluted by mixing with 95% n-hexane in a 1:1 weight ratio and then coated onto the concave surface of the lens using a drop-casting method. Multi-walled CNTs (Sigma-Aldrich, USA) were dispersed in toluene at a weight ratio of 1:500 to make a solution, which was subsequently spray-coated onto the PDMS layer. The coated CNT–PDMS photoacoustic lens was then cured on a hot plate at 95 ℃ for 4 h. Acoustic pressure field simulation The acoustic pressure field was numerically computed by solving the linear wave equation using the finite element method (COMSOL Multiphysics 5.5, COMSOL Inc., USA). Material properties of the vessel phantom and thrombus were incorporated into approximate physiologically relevant acoustic conditions, while blood was modeled as water for simplicity. The simulated lens geometry (diameter and focal length) matched the fabricated lens. Thrombus thicknesses of 100 μm and 1000 μm were evaluated. Time-domain backscattered signals were recorded at a hydrophone position oriented 45° relative to the acoustic axis. Thrombus and phantom preparation for ex vivo experiments Thrombus samples were prepared by mixing 1 g of defibrinated sheep blood (MB Cells, Republic of Korea) with 0.003 g of fibrinogen from bovine plasma (Sigma-Aldrich, USA) and one unit of thrombin from bovine plasma (Sigma-Aldrich, USA), followed by incubation in water bath maintained at 36.5 ℃ to mimic physiological coagulation conditions. The fabrication process is illustrated in Supplementary Fig. S1. A carotid artery model for thrombus evaluation was implemented using a Carotid Artery Blue Phantom (CAE Healthcare, Canada). To clearly distinguish the effects of thrombus thickness, coagulated thrombi were fabricated with thickness differences exceeding 200 μm and classified into four groups (Sample A: ~ 130 μm, Sample B: ~ 330 μm, Sample C: ~ 530 μm, and Sample D: ~ 760 μm). Each thrombus was affixed to the inner wall of the vessel within the phantom. Photographs of the fabricated thrombi are shown in Supplementary Fig. S2. In addition, a soft-tissue phantom was fabricated using a 1% agarose gel containing a 5-mm-diameter channel to replicate an ex vivo carotid artery–like environment for thrombus fragmentation experiments. General experimental setup A Q-switched Nd:YAG pulsed laser (λ = 532 nm, 7 ns pulse width, 10 Hz repetition rate; EVG 00200, Quantel, USA) was used for optical excitation. The laser pulse energy was controlled in the range of 30–200 mJ per pulse. The laser beam diameter was adjusted using a beam expander to match the diameter of the photoacoustic lens. The acoustic output of the fabricated photoacoustic lens was characterized using a custom-built fiber-optic hydrophone (FOPH) based on a single-mode fiber37 and connected to a photodetector with a − 3 dB bandwidth of 75 MHz (PDB420C, Thorlabs, USA). Acoustic pressure was measured by step scanning, and the focal region size was defined as the full width at half maximum (FWHM) of the measured pressure distribution. Backscattered ultrasound signals from thrombus phantoms were acquired using a polyvinylidene fluoride (PVDF) needle hydrophone (3-dB bandwidth: 1–15 MHz, 1 mm sensing diameter; Precision Acoustics Inc., UK). The hydrophone bandwidth was selected to capture the dominant spectral components of the backscattered signals relevant to thrombus characterization. The acquired signals were digitized using a digital oscilloscope (5 GS/s sampling rate; Wavesurfer 454, LeCroy, USA). Real-time images of thrombus fragmentation were recorded using a handheld microscope (Dino-Lite AM4113, AnMo Electronics, Taiwan). Results and discussion Long-focused laser-generated focused ultrasound system We developed a photoacoustic lens capable of delivering focused ultrasound to vascular targets at clinically relevant depths. Figure 1a illustrates the LGFU system designed for the diagnosis and fragmentation of intravascular thrombi. Upon pulsed-laser irradiation, thermoelastic expansion at the concave CNT–PDMS layer generates broadband ultrasound, and the resulting wavefront converges toward the acoustic focus. By concentrating acoustic energy into a tightly confined focal region, the system enables high spatial precision for both diagnostic and therapeutic applications. The measured focal spot size was 150 μm (axial) and 60 μm (lateral), corresponding to spatial resolution on the order of tens of micrometers. Figure 1b shows the fabricated photoacoustic lens consisting of a plano-concave lens substrate (diameter: 44 mm, radius of curvature: 28 mm) coated with a CNT–PDMS nanocomposite layer. The acoustic focal length is defined by the radius of curvature of the substrate. Table 1 summarizes the geometric and focal characteristics of the proposed lens in comparison with reported photoacoustic lenses.Fig. 1Long-focused photoacoustic lens and the corresponding LGFU system. (a) Graphical abstract of the LGFU-based thrombus evaluation and fragmentation platform. (b) Photograph of the fabricated long-focused photoacoustic lens. (c) Experimental setup for ex vivo thrombus diagnosis. M, mirror; BE, beam expander.Table 1Comparison of photoacoustic lens characteristics.D [mm]R [mm]f-numberFocal spot size [μm]Referencelateralaxial64.70.78100430311211.50.9610065032159.20.6190200332515.570.62702703544280.6360150This work* D, diameter; R, radius of curvature ≒ focal length of ultrasound; f-number = R/D. Long-focused photoacoustic lens and the corresponding LGFU system. (a) Graphical abstract of the LGFU-based thrombus evaluation and fragmentation platform. (b) Photograph of the fabricated long-focused photoacoustic lens. (c) Experimental setup for ex vivo thrombus diagnosis. M, mirror; BE, beam expander. Table 1Comparison of photoacoustic lens characteristics.D [mm]R [mm]f-numberFocal spot size [μm]Referencelateralaxial64.70.78100430311211.50.9610065032159.20.6190200332515.570.62702703544280.6360150This work* D, diameter; R, radius of curvature ≒ focal length of ultrasound; f-number = R/D. Comparison of photoacoustic lens characteristics. * D, diameter; R, radius of curvature ≒ focal length of ultrasound; f-number = R/D. Compared with previously reported photoacoustic lenses, the proposed lens achieves a focal length 1.79 times longer than that reported by Heo et al.35. However, their short focal length (< 16 mm) has limited applicability to deeper targets, and all reported demonstrations have been conducted under free-space conditions without intervening tissue layers. This extended focal length enables significantly greater penetration of focused photoacoustic ultrasound, thereby broadening the therapeutic and diagnostic potential of LGFU systems for deep-seated vascular targets. Based on this photoacoustic lens, we implemented an LGFU system for localized characterization and mechanical fragmentation of thrombi under vessel-mimicking phantom conditions. Vessel-mimicking phantoms replicating the depth and lumen diameter of the human carotid artery were employed for validation (Fig. 1c). The confined acoustic focus enabled reliable discrimination and mechanical disruption of thrombi with thicknesses down to approximately 100 μm. Figure 2 shows the measured acoustic output characteristics of the fabricated long-focused photoacoustic lens. Figure 2a shows the measured positive and negative peak pressure amplitudes as a function of laser pulse energy. The negative peak is measurable up to ~ 17 mJ/pulse, beyond which cavitation onset prevents further measurement. The cavitation onset at ~ 17 mJ/pulse provides a physical reference point, corresponding to a peak negative pressure of approximately − 13 MPa37,38. Figure 2b shows the normalized frequency spectrum of the focused ultrasound at the focal spot, characterized using a custom-built FOPH based on a cleaved single-mode fiber. Although the SMF-based FOPH exhibits a non-uniform frequency response that attenuates at lower frequencies39, the measured − 6 dB bandwidth of ~ 15 MHz demonstrates the broadband acoustic output of the fabricated CNT-PDMS lens. These results are consistent with previously reported CNT-PDMS-based photoacoustic transmitters40. Further characteristics of the fabricated lens (waveform at focus, focal zone size) are shown in Supplementary Fig. S3.Fig. 2Acoustic characteristics of the fabricated long-focused photoacoustic lens. (a) Laser pulse energy–pressure amplitude relationship measured at the focal spot. The negative peak pressure at cavitation onset (~ 17 mJ/pulse) corresponds to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{th} \approx$$\end{document}−13 MPa, based on the reported cavitation threshold at a rigid boundary in water. (b) Measured frequency spectrum of the focused ultrasound at the focal spot, exhibiting a − 6 dB bandwidth of approximately 15 MHz, including the detector frequency response. Acoustic characteristics of the fabricated long-focused photoacoustic lens. (a) Laser pulse energy–pressure amplitude relationship measured at the focal spot. The negative peak pressure at cavitation onset (~ 17 mJ/pulse) corresponds to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{th} \approx$$\end{document}−13 MPa, based on the reported cavitation threshold at a rigid boundary in water. (b) Measured frequency spectrum of the focused ultrasound at the focal spot, exhibiting a − 6 dB bandwidth of approximately 15 MHz, including the detector frequency response. Numerical simulation of backscattered ultrasound Prior to the experimental validation, finite-element simulations were conducted to investigate the relationship between thrombus thickness and the resulting backscattered ultrasound signals (Fig. 3a). A carotid artery–mimicking vessel containing an intravascular thrombus was modeled, and acoustic pressure signals were recorded at a detection angle of 45° relative to the specular reflection direction. The simulation employed pressure acoustics and thermoviscous acoustics interfaces, and the photoacoustic excitation was modeled as a volumetric heat source,1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{*{20}c} {Q = G\left( t \right)e^{{ - \alpha_{l} L}} ,} \\ \end{array}$$\end{document}where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G\left(t\right)$$\end{document} denotes a temporal Gaussian pulsed laser source, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha}_{l}$$\end{document} is the absorption coefficient of the CNT–PDMS nanocomposite, and L is the thickness of film, respectively. The simulated time-domain backscattered signals revealed that acoustic reflections occurred at both the water–thrombus interface and the thrombus–carotid wall interface, resulting in two distinct unipolar peaks (positive and negative). The inter-peak (peak-to-peak) corresponds to the propagation time through the thrombus and can be used to estimate thrombus thickness. Here, Δt denotes the inter-peak time difference measured from the received waveform. Because the measured peak positions are influenced by waveform overlap, attenuation, and signal-shape distortion, Δt does not scale linearly with thrombus thickness. For a thrombus thickness of 100 μm (Fig. 3b), the Δt was measured to be 135 ns, whereas a 1000 μm thrombus (Fig. 3c) exhibited a significantly longer Δt of 1046 ns. These values are consistent with the reported speeds of sound in water and thrombus tissue41, validating the physical accuracy of the simulation.Fig. 3Finite-element simulation of backscattered ultrasound by thrombus. (a) Schematic of the carotid artery mimic model and focused ultrasound propagation. Simulation results for a thrombus with a thickness of (b) 100 μm, (c) 1000 μm. (d) Normalized frequency spectrum of backscattered ultrasound for different thrombus thickness. Finite-element simulation of backscattered ultrasound by thrombus. (a) Schematic of the carotid artery mimic model and focused ultrasound propagation. Simulation results for a thrombus with a thickness of (b) 100 μm, (c) 1000 μm. (d) Normalized frequency spectrum of backscattered ultrasound for different thrombus thickness. In practical biological environments, however, signal interpretation is complicated by scattering, attenuation, and measurement noise, which can obscure clear peak identification in the time domain. To improve robustness against such effects, the frequency-domain characteristics of the backscattered ultrasound were analyzed using fast Fourier transform (FFT). Figure 3d shows the normalized frequency spectra corresponding to the two thrombus thicknesses. The FFT was applied to the full backscattered waveform rather than to a single isolated pulse. This approach was intentionally adopted to capture thickness-dependent spectral interference patterns arising from multiple reflections at the proximal and distal thrombus interfaces, rather than to isolate single-echo material scattering properties. The 100 μm thrombus exhibits a spectral peak at a center frequency of approximately 19 MHz with a relatively narrow bandwidth, whereas the 1000 μm thrombus shows a dominant peak near 1 MHz with a broader spectral distribution. Detector response was not included in the simulations. These thickness-dependent spectral variations arise from the temporal spacing and interference between reflections from the proximal and distal blood-thrombus interfaces in the full backscattered waveform. Reflections from the proximal and distal interfaces arrive with a temporal separation Δ t corresponding to the two-way transit time through the thrombus layer. By Fourier duality, a shorter Δ t produces higher-frequency spectral components, whereas a longer Δ t shifts spectral energy toward lower frequencies. Consequently, the dominant spectral features vary with thrombus thickness and form the basis for thickness discrimination in the present study. The generation of tens-of-nanoseconds unipolar pulses, combined with EMD-based signal processing, enabled robust thrombus thickness discrimination under noisy and nonstationary conditions. Analyzing backscattered ultrasound signal from thrombus Next, we experimentally validated thrombus-dependent changes in backscattered ultrasound signals in both the time and frequency domains. Figure 4a shows representative time-domain waveforms backscattered from the vessel wall in the absence and presence of a thrombus. When a thrombus is present, additional scattering, at the blood–thrombus and thrombus–wall interfaces, introduces multipath reflections and waveform distortion, resulting in substantially increased signal complexity. Under these conditions, simple time-domain peak-based analysis becomes unreliable. Therefore, consistent with the simulation results, frequency-domain analysis was employed to characterize thrombus-dependent signal variations. Figures 4b and c show the normalized frequency spectra of the backscattered ultrasound without and with a thrombus, respectively. In the absence of a thrombus, the bipolar photoacoustic pulse is reflected primarily from the vessel wall, preserving high-frequency components. In contrast, propagation through a thrombus attenuates and reshapes the waveform into a sequence of unipolar transients, shifting spectral energy toward lower frequencies and producing a pronounced low-frequency component.Fig. 4Feasibility test of backscattered ultrasound signals depending on the presence of a thrombus. (a) Time-domain waveforms measured from the blood vessel wall (black) and thrombus (red). Measured frequency spectrum of the signal from the blood vessel wall (b) and thrombus (c). Feasibility test of backscattered ultrasound signals depending on the presence of a thrombus. (a) Time-domain waveforms measured from the blood vessel wall (black) and thrombus (red). Measured frequency spectrum of the signal from the blood vessel wall (b) and thrombus (c). To experimentally assess thrombus thickness, a carotid artery–mimicking phantom containing fabricated thrombi was used. Backscattered signals were acquired using a needle hydrophone positioned at a 45° angle relative to the acoustic axis of the photoacoustic lens. To evaluate repeatability and robustness, 400 independently triggered measurements were acquired for each sample while slightly varying the focal position between acquisitions. Figure 5 presents representative backscattered waveforms from four thrombus samples with different thicknesses. Signals from Sample A (~ 130 μm) and Sample D (~ 760 μm) are clearly distinguishable, whereas those from Sample B (~ 330 μm) and Sample C (~ 530 μm) exhibit substantial overlap and are difficult to discriminate directly. These intermediate cases are characterized by nonstationary, nonlinear waveforms with low signal-to-noise ratios (6–20 dB), resulting from harmonic components induced by multiple scattering and from electrical noise.Fig. 5Measured waveform examples from each sample. Sample A (d ~ 130 μm), Sample B (d ~ 330 μm), Sample C (d ~ 530 μm), Sample D (d ~ 760 μm); d = thickness of thrombus. Measured waveform examples from each sample. Sample A (d ~ 130 μm), Sample B (d ~ 330 μm), Sample C (d ~ 530 μm), Sample D (d ~ 760 μm); d = thickness of thrombus. To suppress noise while preserving thickness-dependent features, empirical mode decomposition (EMD) was applied to decompose each backscattered signal into intrinsic mode functions (IMFs). Because the interference spans both low- and high-frequency bands, conventional band-pass filtering is ineffective and risks distorting relevant signal components. Each signal was decomposed into ten IMFs, and the frequency content of each IMF was evaluated using fast Fourier transform (FFT). Representative decomposed signals are shown in Supplementary Fig. S4, with the average center frequency and 3-dB bandwidth of each IMF summarized in Supplementary Tables S1 and S2, respectively. The first- and second-order IMFs were excluded as noise based on their atypical spectral characteristics and large standard deviations. Among the remaining modes, the fourth- to sixth-order IMFs were identified as valid modes because their frequency content lay within the hydrophone response range (1–15 MHz). Figure 6a shows the center frequency and 3-dB bandwidth of these valid IMFs averaged over 400 measurements. As thrombus thickness increased, the center frequency systematically shifted toward lower frequencies, accompanied by a reduction in bandwidth.Fig. 6Frequency-domain analysis of the valid modes (IMF 4, IMF 5, and IMF 6), averaged over 400 acquisitions. (a) Center frequency and 3-dB bandwidth of the valid IMFs, shown as mean ± standard deviation (std) over 400 acquisitions. (b) Noise-filtered frequency spectra reconstructed from IMF 4, IMF 5, and IMF 6 from representative single acquisitions with the acoustic focus accurately positioned on the thrombus. Labels A, B, C, and D correspond to thrombus samples with thicknesses of approximately 130, 330, 530, and 760 μm, respectively. Frequency-domain analysis of the valid modes (IMF 4, IMF 5, and IMF 6), averaged over 400 acquisitions. (a) Center frequency and 3-dB bandwidth of the valid IMFs, shown as mean ± standard deviation (std) over 400 acquisitions. (b) Noise-filtered frequency spectra reconstructed from IMF 4, IMF 5, and IMF 6 from representative single acquisitions with the acoustic focus accurately positioned on the thrombus. Labels A, B, C, and D correspond to thrombus samples with thicknesses of approximately 130, 330, 530, and 760 μm, respectively. Due to the small focal spot size, averaging over repeated measurements introduced errors when the acoustic focus was not precisely aligned with the thrombus. To illustrate the intrinsic thickness dependence more clearly, a noise-filtered spectrum was reconstructed by combining the valid IMFs from a single acquisition with accurate focal alignment (Fig. 6b). Compared with the averaged results, the targeted single acquisition revealed clearer spectral separation between samples. These results demonstrate that EMD-based analysis, combined with sensor-aware IMF selection, enables robust discrimination of thrombus thickness from complex, nonstationary backscattered ultrasound signals. However, the system performance was sensitive to focal alignment, due to tiny focus, motivating future integration of real-time imaging guidance and closed-loop feedback control. Non-invasive localized thrombus fragmentation Furthermore, the non-invasive localized thrombus fragmentation capability of the LGFU system was experimentally evaluated under ex vivo conditions using carotid artery and thrombus phantoms. Figure 7a schematically illustrates the mechanism of liquid jetting generated by LGFU-induced cavitation collapse near a rigid or semi-rigid boundary. Cavitation occurs when the peak rarefactional pressure at the acoustic focus exceeds the cavitation threshold, leading to rapid bubble collapse under ambient fluid pressure40. When this collapse occurs near a boundary, the asymmetric dynamics generate a boundary-directed liquid microjet with velocities exceeding 200 m/s, producing highly localized mechanical stress capable of disrupting soft tissue. In addition, the generation of cavitation by photoacoustic pulses with durations on the order of several tens of nanoseconds is accompanied by a negligible temperature rise (10−2 °C), thereby reducing unintended damage to surrounding tissues33.Fig. 7Non-invasive thrombus fragmentation using the LGFU system. (a) Schematic illustration of tissue fragmentation by a liquid microjet, generated using PowerPoint (Microsoft Corp., USA). (b) A photograph of thrombus fragmentation linewidths as a function of laser energy obtained using a short-focused (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{l}$$\end{document}=15 mm) lens (line 1: 189 mJ/pulse, line 2: 108 mJ/pulse, line 3: 45 mJ/pulse). (c) A microscopic image of thrombus debris after fragmentation. (d) Schematic illustration of the ex vivo thrombus fragmentation experiment using long-focused (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{l}$$\end{document}=28 mm) lens, created using Inventor (Autodesk Inc., USA). (e) Images captured from real-time video showing the thrombus before, during, and after fragmentation. Non-invasive thrombus fragmentation using the LGFU system. (a) Schematic illustration of tissue fragmentation by a liquid microjet, generated using PowerPoint (Microsoft Corp., USA). (b) A photograph of thrombus fragmentation linewidths as a function of laser energy obtained using a short-focused (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{l}$$\end{document}=15 mm) lens (line 1: 189 mJ/pulse, line 2: 108 mJ/pulse, line 3: 45 mJ/pulse). (c) A microscopic image of thrombus debris after fragmentation. (d) Schematic illustration of the ex vivo thrombus fragmentation experiment using long-focused (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{l}$$\end{document}=28 mm) lens, created using Inventor (Autodesk Inc., USA). (e) Images captured from real-time video showing the thrombus before, during, and after fragmentation. To first assess the intrinsic soft-tissue fragmentation capability, thrombus fragmentation experiments were conducted under unobstructed conditions with free-space ultrasound propagation. A 15 mm focal length photoacoustic lens was used to evaluate laser-energy-dependent fragmentation behavior (Fig. 7b). The results demonstrate that the fragmentation linewidth can be precisely controlled by the input laser energy (189 mJ/pulse: 191 μm; 108 mJ/pulse: 110 μm; 45 mJ/pulse: 66 μm). As the laser energy increased, the cavitation cloud expanded and the fragmentation width increased correspondingly. The resulting thrombus debris exhibited a broad size distribution ranging from several micrometers to several tens of micrometers (Fig. 7c). Under all tested conditions, it was difficult to find any visible gross thermal damage or collateral mechanical injury to the surrounding tissue. These results suggest that the highly confined focal region of LGFU may provide an advantage for selective thrombus targeting while reducing unintended effects on surrounding structures under the tested conditions. Subsequently, thrombus fragmentation was evaluated in a vessel-mimicking environment using a long-focused photoacoustic lens (Fig. 7d). A 5-mm-diameter cylindrical channel, corresponding to the human carotid artery, was formed in a 1% agarose gel phantom. A thrombus sample with a thickness of 2 mm was positioned near the center of the channel to partially occlude the lumen. Although clinical thrombi do not completely obstruct blood flow, this configuration was intentionally adopted to clearly visualize the fragmentation width. The long-focused photoacoustic lens (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{l}$$\end{document} = 28 mm) was employed with a laser energy of 200 mJ, and the phantom was scanned at 0.1 mm/s using a motorized three-dimensional stage. Thrombus fragmentation was recorded in real time from the lateral side. Figure 7e shows sequential images of thrombus fragmentation during the ex vivo experiment (see video 1). Prior to laser irradiation, the thrombus remained intact within the channel. During focused ultrasound scanning, localized disruption occurred at the acoustic focus, and after scanning, a clearly defined fragmented region with a lateral width of several hundred micrometers was formed. These results confirm that spatially confined thrombus fragmentation can be achieved using a long-focused photoacoustic lens under ex vivo conditions. Supplementary Figure S5 shows laser-shadowgraphy images of thrombus fragmentation consistent with cavitation-mediated effects before, during, and after LGFU exposure. Taken together, the LGFU system enabled non-invasive thrombus fragmentation without observable gross thermal damage under the tested conditions both under free-space conditions and within a vessel-mimicking channel. Because the present experiments were conducted ex vivo without physiological flow or tissue heterogeneity, direct clinical translation requires further validation. For practical applications, the LGFU source can be integrated with optical or piezoelectric detectors to form a pulse-echo imaging system. In this configuration, conventional ultrasound imaging could first be used to identify an approximate location and extent of thrombus. The extracorporeal LGFU platform could then provide site-specific high-precision characterization followed by targeted fragmentation at the identified site. Future work will focus on enhancing acoustic pressure, higher-energy pulsed lasers, quantifying cell viability adjacent to fragmentation sites, characterizing propagation property across heterogeneous tissue environments, and validating in vivo efficacy under physiological flow conditions. Conclusion In this study, we developed a multifunctional LGFU platform for thrombus characterization and fragmentation. The platform enabled thickness-sensitive characterization of thrombi down to 100 μm and demonstrated localized mechanical fragmentation through acoustically confined focusing. Single-pulse broadband ultrasound combined with EMD-based signal processing allowed robust decomposition of complex backscattered signals and thickness discrimination under nonstationary and noisy conditions, with repeatability confirmed over 400 acquisitions. System performance was sensitive to focal alignment, underscoring the importance of real-time imaging guidance and closed-loop feedback control. In addition, the ex vivo experimental conditions, which did not incorporate physiological blood flow or tissue heterogeneity, limit direct clinical translation of the present results. Furthermore, due to the highly confined focal region, the current platform is more suitable for localized thrombus interrogation and localized fragmentation than for whole-thrombus imaging or rapid bulk removal without scanning. At the same time, this confined focal zone may offer an advantage for selective targeting of thrombus regions while minimizing unintended effects on surrounding structures. Future studies will focus on quantifying cell viability in tissues adjacent to fragmentation zones, characterizing focal ultrasonic pressure across heterogeneous tissue environments, and validating in vivo efficacy under physiological flow conditions. Integration of real-time imaging guidance and automated focal positioning is expected to further enhance system robustness and clinical applicability. This LGFU platform supports a proof-of-concept framework for localized thrombus characterization and mechanical fragmentation, with potential for future extension through scanning or image-guided targeting strategies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2