Label-free photoacoustic microscopy blood flow direction and speed combined measurement method based on space-time cross correlation

Abstract

The invention provides an unmarked photoacoustic microscopy blood flow direction and speed combined measurement method based on space-time cross correlation, which comprises the following steps: acquiring a two-dimensional structure image of a target tissue through a photoacoustic microscopy imaging platform, and carrying out image preprocessing and blood vessel enhancement on the two-dimensional structure image to obtain an enhanced tubular region; based on the enhanced tubular region, performing blood vessel segmentation, morphological optimization, center line extraction and direction calculation to obtain blood vessel direction information; path planning, scanning point sequence optimization and time sequence synchronization are carried out based on the blood vessel direction information, and a directional scanning path is obtained; a photoacoustic signal sequence is collected in the blood vessel direction, and space-time cross-correlation analysis is carried out; the blood flow direction and the blood flow speed are quantitatively calculated and visually displayed. According to the method, structural information guidance can be combined, rapid scanning in the blood vessel direction can be achieved, directional blood flow analysis is achieved through space-time cross-correlation, and the measuring speed and precision are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of photoacoustic blood flow measurement technology, and in particular to a markerless photoacoustic microscopic blood flow direction and velocity measurement method based on spatiotemporal correlation. Background Technology

[0002] Photoacoustic microscopy (PAM) is a multimodal imaging technique that combines optical absorption contrast with ultrasonic spatial resolution. Its basic principle is as follows: Ultrashort pulse laser light is applied to the surface of biological tissue. Light absorbers (such as hemoglobin and melanin) absorb the light energy and undergo instantaneous thermoelastic expansion, generating an ultrasonic signal. After being received by a transducer, amplified, and processed by a reconstruction algorithm, tissue images with high contrast and cellular-level spatial resolution can be obtained.

[0003] Photoacoustic microscopy, due to its specific absorption characteristics of hemoglobin, has been widely used in fields such as skin microcirculation monitoring, tumor angiogenesis research, brain functional activity imaging, and pharmacokinetics assessment. Among these applications, the quantitative measurement of blood flow velocity is a crucial research direction in photoacoustic microscopy, reflecting tissue perfusion status and metabolic levels, and is of great significance for early disease diagnosis and treatment evaluation.

[0004] Existing photoacoustic blood flow measurement methods mainly fall into three categories: Time autocorrelation method: Based on the fluctuation characteristics of a single-point photoacoustic signal over time, local blood flow velocity is estimated by calculating the decay time constant of the autocorrelation function. Photoacoustic Doppler method: Velocity components are calculated using the frequency shift of the photoacoustic signal caused by blood flow. Time drift method: Transverse flow velocity is obtained by analyzing the correlation time drift of adjacent B-scan frames.

[0005] Among these methods, the time autocorrelation method has become the most widely used photoacoustic blood flow quantification method due to its simplicity, lack of need for spectral separation, and good robustness to noise. This method involves continuously acquiring multiple A-line signals at the same location, calculating the characteristic time constant under different time lags, and fitting its exponential decay model. However, the traditional autocorrelation method has the following main problems: 1) Point-by-point acquisition, low efficiency: Each pixel needs to independently acquire a long-term sequence signal (usually thousands of pulses), and the calculation of the entire field of view takes a very long time, making it difficult to achieve real-time measurement; 2) Insufficient directionality: Since only the time fluctuations in the optical axis direction are analyzed, the autocorrelation method can only reflect the axial velocity component and cannot distinguish the blood flow direction; 3) Limited signal-to-noise ratio: In areas with low flow velocity or thin blood vessels, the autocorrelation function decays rapidly, resulting in poor fitting stability.

[0006] Therefore, existing photoacoustic blood flow measurement technologies generally suffer from problems such as long acquisition times and high computational loads. Specifically, the time autocorrelation method can only calculate the numerical value of blood flow velocity, but cannot determine the flow direction; while the photoacoustic Doppler method can obtain directional information, it requires extremely high signal spectral resolution and signal-to-noise ratio, making stable measurement difficult at the microvascular scale. Therefore, existing technologies struggle to achieve efficient, rapid, directional, and robust quantitative blood flow measurement while maintaining spatial resolution. Summary of the Invention

[0007] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a label-free photoacoustic microscopic method for joint measurement of blood flow direction and velocity based on spatiotemporal cross-correlation. This method can combine structural information for guidance, rapidly scan along the blood vessel direction, and achieve directional blood flow analysis through spatiotemporal cross-correlation, thereby significantly improving measurement speed and accuracy. At the same time, it achieves robust and efficient joint measurement of blood flow direction and velocity.

[0008] To achieve the above objectives, the present invention provides the following solution: a label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation, comprising: Two-dimensional structural images of the target tissue were acquired using a photoacoustic microscopy imaging platform, and the two-dimensional structural images were preprocessed and enhanced with blood vessels to obtain enhanced tubular regions. Based on the enhanced tubular region, blood vessel segmentation and morphological optimization are performed to obtain a binary blood vessel image. The centerline of the binary blood vessel image is extracted and the direction is calculated to obtain blood vessel direction information. Based on the blood vessel direction information, path planning is performed to obtain a scan point sequence. Then, the scan point sequence is optimized and time-synchronized to obtain a directional scan path. According to the directional scanning path, photoacoustic signal sequences are acquired along the blood vessel direction, and spatiotemporal cross-correlation analysis is performed on the photoacoustic signal sequences; The results of spatiotemporal cross-correlation analysis are used to quantitatively calculate blood flow direction and velocity, and to visualize the combined blood flow direction and velocity.

[0009] Optionally, a two-dimensional structural image of the target tissue is acquired using a photoacoustic microscopy imaging platform, and the two-dimensional structural image is preprocessed and enhanced with blood vessels to obtain an enhanced tubular region, including: The sample is irradiated with a pulsed laser to generate an ultrasonic signal. The ultrasonic signal is received by a spherical focusing ultrasonic transducer. The envelope information of the ultrasonic signal is then extracted by bandpass filtering and Hilbert transform to obtain a reconstructed high-contrast vascular network image. Based on the Frangi filter principle, the second derivative matrix of the high-contrast vascular network image is calculated, and eigenvalue analysis is performed on the second derivative matrix to obtain the first eigenvalue and the second eigenvalue. Determine whether the second feature value is negative and whether the absolute value of the second feature value is much greater than the absolute value of the first feature value. If so, it is identified as a tubular structure. Then, an enhancement function is used to detect vascular protrusion, resulting in an enhanced tubular region. The calculation expression of the enhancement function is as follows: ; in, For cyclic scaling parameters, The first eigenvalue, The second eigenvalue, To suppress noise, , To adjust the parameters.

[0010] Optionally, based on the enhanced tubular region, blood vessel segmentation and morphological optimization are performed to obtain a binary blood vessel image. The centerline of the binary blood vessel image is extracted and its orientation is calculated to obtain blood vessel orientation information, including: Based on the enhanced tubular region, an adaptive threshold algorithm is used for binarization segmentation to extract the vascular region. According to the tubular characteristics of the blood vessels in the image at the local scale, a disk-shaped structural element with circular symmetry is selected to perform dilation and erosion operations on the vascular region to connect broken blood vessels and fill local holes without introducing directional bias, thereby completing the morphological closure operation and obtaining a binarized vascular image. Using median transformation or Lee's algorithm, the binarized blood vessel image is refined into a skeleton with a width of one pixel. Skeleton branches shorter than a preset threshold are removed to filter noise. Then, local principal component analysis is performed on each skeleton point to calculate the local orientation angle and obtain the blood vessel orientation information.

[0011] Optionally, based on the blood vessel orientation information, path planning is performed to obtain a scan point sequence, and then the scan point sequence is optimized and time-synchronized to obtain a directional scanning path, including: Based on the blood vessel orientation information, the skeleton points are sorted according to the blood vessel topology, and a main path distributed along the blood vessel centerline is generated using a B-spline curve fitting algorithm. Then, the main path is resampled at a fixed interval to generate a scan point sequence. Based on the scan point sequence, the scanning order of multiple blood vessel paths is calculated using an optimization algorithm, and the overlapping region detection algorithm is used to identify whether there are repeated samplings at blood vessel intersections or branch points, thus completing the scan optimization operation. The optimized scan point sequence is converted into a control signal. Based on the control signal, the FPGA is used to drive the scanning mechanism to perform directional scanning and timing synchronization, thereby obtaining the directional scanning path.

[0012] Optionally, according to the directional scanning path, a photoacoustic signal sequence is acquired along the blood vessel direction, and spatiotemporal cross-correlation analysis is performed on the photoacoustic signal sequence, including: The directional scanning path is executed, and photoacoustic signal sequences are acquired along the direction of the blood vessel to obtain a spatiotemporal data matrix. Based on the spatiotemporal data matrix, the cross-correlation function of adjacent point signals is calculated to obtain the similarity between the two signals under different time delays, thus completing the cross-correlation calculation. The peak position of the cross-correlation function curve is searched, and the peak shift time is extracted based on the peak position to obtain the peak time delay, thus completing the spatiotemporal cross-correlation analysis.

[0013] Optionally, the results of spatiotemporal cross-correlation analysis are used to quantitatively calculate blood flow direction and velocity, and to visualize the combined blood flow direction and velocity, including: The blood flow direction is automatically determined based on the sign of the peak time delay. When the peak time delay is greater than 0, it indicates that the downstream point signal lags behind the upstream point, and the blood flow is determined to be flowing in the forward direction along the scanning direction. When the peak time delay is less than 0, the blood flow is determined to be flowing in the reverse direction along the scanning direction, thus completing the blood flow direction determination. The blood flow velocity is calculated based on the fixed interval and the peak time delay, and the blood flow direction and blood flow velocity are mapped back to the original two-dimensional spatial coordinates to construct a complete two-dimensional blood flow vector field. The two-dimensional blood flow vector field is superimposed and fused with the high-contrast vascular network image to achieve a visual display that integrates structure and function.

[0014] This invention discloses the following technical advantages by providing a label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation: 1. Intelligent scanning guided by structural information: By extracting the direction of blood vessels from the photoacoustic image of the tissue structure in the pre-obtained scanning area, intelligent scanning of the beam along the direction of the blood vessels is realized, avoiding blind scanning and redundant sampling, and greatly improving the acquisition efficiency.

[0015] 2. Achieve blood flow directionality measurement: Through the spatiotemporal cross-correlation algorithm of adjacent points, the directionality (forward or reverse) of blood flow can be directly determined, making up for the deficiency of traditional autocorrelation method in that it cannot distinguish direction.

[0016] 3. Significantly accelerates measurement speed: The scanning method changes from area scanning to scanning along the blood vessel path, reducing the number of collection points by more than 60% while maintaining a high sampling rate for key blood flow areas, resulting in an overall measurement time reduction of 3 to 5 times.

[0017] 4. Enhanced noise resistance and robustness of results through signal cross-correlation analysis: This invention performs continuous sampling guided by the blood vessel direction, ensuring that the photoacoustic signals from adjacent sampling points remain highly correlated in time and space, forming a continuous spatiotemporal signal sequence along the flow direction. Since red blood cells or other absorbing particles exhibit a consistent motion trend within the blood vessel, the cross-correlation function of adjacent signals produces a sharp peak at the corresponding time delay, thereby accurately estimating blood flow velocity and direction. Simultaneously, background noise is spatially uncorrelated and cancels each other out in the cross-correlation calculation, increasing the proportion of effective signal, significantly enhancing the signal-to-noise ratio (SNR), and improving the stability of the results.

[0018] 5. Compatibility and scalability: The scanning range and sampling strategy can be flexibly adjusted according to different sample sizes and blood vessel densities.

[0019] 6. High-performance data processing and control architecture enables real-time directional blood flow imaging: FPGA real-time control and GPU parallel computing. The FPGA is responsible for laser triggering, MEMS for scan synchronization and data caching, while the GPU performs structural diagram analysis, vessel orientation extraction, and cross-correlation calculations, achieving millisecond-level visualization of directional blood flow velocity. This architecture significantly improves data processing speed and system responsiveness, meeting the experimental and clinical needs of real-time dynamic blood flow monitoring.

[0020] 7. Scalable Algorithm Framework and Adaptive Parameter Optimization Mechanism: This invention supports various signal analysis algorithms, including cross-correlation peak detection, sliding time window averaging, adaptive filtering, and signal enhancement methods based on structural similarity weighting. It can automatically adjust the relevant window length, step size, and threshold according to local signal-to-noise ratio or vascular morphology characteristics, thereby maintaining measurement accuracy and stability under different imaging depths and vascular types.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention; Figure 2 A flowchart of the spatiotemporal cross-correlation measurement method provided in an embodiment of the present invention; Figure 3 A schematic diagram of the spatiotemporal cross-correlation measurement method provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of blood flow direction recognition provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 , Figure 2 As shown, this invention provides a label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation, including: Step 1: Acquire two-dimensional structural images of the target tissue using a photoacoustic microscopy imaging platform, and perform image preprocessing and vascular enhancement on the two-dimensional structural images to obtain enhanced tubular regions; Step 1 includes: 1.1 The sample is irradiated with a pulsed laser to generate an ultrasonic signal. The ultrasonic signal is received by a spherical focusing ultrasonic transducer. The envelope information of the ultrasonic signal is then extracted by bandpass filtering and Hilbert transform to obtain a reconstructed high-contrast vascular network image.

[0027] Specifically, a pulsed laser with a wavelength of 532 / 560 nm, a pulse width of 5 ns, and a repetition frequency of 50 kHz is used to irradiate the sample surface. Hemoglobin and other light absorbers generate ultrasonic signals, which are received by a spherical focusing ultrasonic transducer (center frequency 50 MHz, bandwidth 60%). The signal is then processed by bandpass filtering (5-80 MHz) and Hilbert transform to extract envelope information and reconstruct a high-contrast vascular network image. The spatial resolution of the image needs to be better than 10 μm to ensure clear identification of microvessels. 1.2 Based on the Frangi filter principle, the second derivative matrix of the high-contrast vascular network image is calculated, and eigenvalue analysis is performed on the second derivative matrix to obtain the first eigenvalue and the second eigenvalue; 1.3 Determine whether the second feature value is negative and whether the absolute value of the second feature value is much greater than the absolute value of the first feature value. If so, identify it as a tubular structure, and then combine it with an enhancement function to show vascular protrusion, thus obtaining an enhanced tubular region; the calculation expression of the enhancement function is: ; in, For cyclic scaling parameters, The first eigenvalue, The second eigenvalue, To suppress noise, , To adjust the parameters.

[0028] Specifically, multi-scale Hessian matrix analysis is used for vessel enhancement. This method is based on the Frangi filter principle, calculating the second derivative matrix (Hessian matrix) of the image and analyzing its eigenvalues ​​to distinguish tubular structures. For each pixel, the eigenvalues ​​λ1 and λ2 of the Hessian matrix reflect the curvature characteristics of the tubular structure. When the absolute value of λ2 is much greater than λ1 and λ2 is negative, the region is identified as a vessel. The enhancement function is expressed as: in, Used to distinguish between tubular and plate-like structures Noise suppression is achieved by adjusting parameters β and c according to image characteristics (typical values ​​β=0.5, c=500). Multiscale analysis is performed using the cyclic scale parameter σ (range 5-15 μm), and the final vascular enhancement result is the maximum value at each scale to cover vessels of different diameters.

[0029] Step 2: Based on the enhanced tubular region, perform vessel segmentation and morphological optimization to obtain a binary vessel image. Then, extract the centerline and calculate the direction of the binary vessel image to obtain vessel direction information. Step 2 includes: 2.1 Based on the enhanced tubular region, an adaptive threshold algorithm is used for binarization segmentation to extract the vascular region. According to the fact that the blood vessels in the image exhibit approximately isotropic tubular characteristics at the local scale, a disk-shaped structural element with circular symmetry is selected to perform dilation and erosion operations on the vascular region to connect broken blood vessels and fill local holes without introducing directional bias, thereby completing the morphological closure operation and obtaining a binarized vascular image.

[0030] 2.2 Using median transformation or Lee algorithm, the binarized blood vessel image is thinned into a skeleton with a single pixel width. Skeleton branches shorter than a preset threshold are removed to filter noise. Local principal component analysis is then performed on each skeleton point to calculate the local orientation angle and obtain blood vessel orientation information.

[0031] Specifically, after vascular enhancement, binarization segmentation is performed to extract the vascular region. An adaptive thresholding algorithm (such as the Otsu method) is preferred, dynamically calculating the global threshold instead of a fixed threshold to improve adaptability to low-contrast images. The binarized image undergoes morphological closing operations, using disk-shaped structuring elements (radius 2μm) for dilation and erosion operations to connect broken vessels and fill holes, ensuring the continuity of the vascular network. Subsequently, the vascular centerline is extracted using a skeletonization algorithm. Using median transformation or the Lee algorithm, the binary image is thinned into a single-pixel-width skeleton, with skeleton points representing the center position of the vessel. To reduce noise interference, a short-branch filtering algorithm is applied: the length of each skeleton branch is calculated, and branches shorter than a preset threshold (e.g., 10 pixels) are removed. This threshold corresponds to an actual vessel length of approximately 100μm (based on image resolution). Finally, local principal component analysis (PCA) is performed based on the skeleton points to calculate the orientation angle θ(x,y) of each point. Specifically, a local window (e.g., 5×5 pixels) is extracted centered on the skeleton point. PCA is used to fit the distribution of points within the window, and the main direction vector is the local direction of the blood vessel. The direction angle θ is used for subsequent path generation.

[0032] Step 3: Based on the blood vessel direction information, perform path planning to obtain a scan point sequence, then optimize and time-synchronize the scan point sequence to obtain a directional scan path; Step 3 includes: 3.1 Based on the blood vessel orientation information, the skeleton points are sorted according to the blood vessel topology, and a main path distributed along the blood vessel centerline is generated using a B-spline curve fitting algorithm. Then, the main path is resampled at a fixed interval to generate a scan point sequence. 3.2 Based on the scan point sequence, the scanning order of multiple blood vessel paths is calculated using an optimization algorithm, and the overlapping region detection algorithm is used to identify whether there are repeated samplings at blood vessel intersections or branch points, thus completing the scan optimization operation; 3.3 The scan point sequence after scanning optimization is converted into a control signal. Based on the control signal, the FPGA is used to drive the scanning mechanism to perform directional scanning and timing synchronization to obtain the directional scanning path.

[0033] Step 3 in detail: Direction-guided scanning path generation utilizes extracted vascular orientation information to automatically plan and optimize the scanning trajectory, enabling rapid data acquisition along the vascular direction. This part includes path planning algorithms, scanning control mechanisms, and time synchronization to ensure the feasibility of spatiotemporal correlation analysis. Path planning is based on the orientation angle θ(x,y) of the vascular skeleton points. First, the skeleton points are sorted according to the vascular topology (e.g., from head to tail) to form an ordered point sequence. Subsequently, a B-spline curve fitting algorithm is used to generate a smooth path, avoiding mechanical vibration during the scanning process. The fitting process uses parametric interpolation: given a control point sequence, the B-spline function generates a continuous trajectory, the mathematical expression of which is: ; in, Let p be the B-spline basis function, and p be the order. The coordinates of the control points are used. After fitting, the path is resampled along the curve at a fixed interval of Δs = 10 μm to generate a sequence of scan points. For tortuous blood vessels, the path is adaptively segmented at the turning points, with the directional change within each segment limited to ±5° to ensure signal continuity.

[0034] For each blood vessel, the system first generates a main path along the centerline. The main path is generated by fitting a spline curve to the center point sequence, ensuring a smooth path. The system assigns scanning priorities based on the morphological characteristics of the blood vessels (diameter, length). For example, thicker and longer main vessels are scanned first to ensure rapid coverage of critical areas; thin capillaries are scanned subsequently to avoid omissions. The scanning order is optimized using a path planning algorithm to minimize the movement distance of the scanning mechanism (a simplified version of the Traveling Salesman Problem), reducing idle time. During the scanning path generation process, to minimize the movement distance of the scanning mechanism, the system employs an optimization algorithm similar to a simplified version of the Traveling Salesman Problem (TSP). This method treats multiple scanning paths as "city" nodes, calculates the Euclidean distance between path points, and uses heuristic strategies (such as nearest neighbor algorithms or genetic algorithms) to determine the optimal scanning order, thereby reducing idle time and improving measurement efficiency. This optimization ensures that the scanning mechanism has the shortest movement path when covering complex vascular networks, significantly reducing overall scanning time while maintaining the accuracy of directional scanning. The scanner (such as a MEMS galvanometer) executes each path sequentially in an optimized order. At vessel intersections or branch points, the paths may overlap. The system avoids duplicate sampling through an overlap region detection algorithm: for example, when multiple paths converge at an intersection, only the sampling point of one path is retained, and the other paths are skipped at that point. For tortuous vessels, the path is adaptively segmented at turning points to ensure that the directional change within each segment is gradual (limited to ±10°), avoiding signal distortion.

[0035] Scanning control is implemented using an FPGA (Field-Programmable Gate Array) module. Path point coordinates are converted into voltage signals, driving the scanning mechanism (such as a MEMS micro-scanning mirror) to deflect. The MEMS mirror has a scanning frequency range of 10–50 kHz, enabling millisecond-level directional scanning. A timing synchronization mechanism ensures strict alignment of laser triggering, scanning position, and signal acquisition: the FPGA generates a synchronization clock, laser pulses trigger at each scanning point, and ultrasound signals are acquired by a high-speed ADC (sampling rate 250 MS / s), with a synchronization error not exceeding 10 ns. This accuracy corresponds to a flow velocity resolution of 0.05 mm / s, meeting the requirements for microvascular measurement.

[0036] The generation of the scanning path is tightly coupled with the vascular structure. For example, for complex vascular networks, the system automatically generates multiple parallel paths (spaced 10 μm apart) covering the entire vascular region. The path length can be extended according to the target vessel or manually limited by the user. During the scanning process, the photoacoustic signal sequence P(s,t) is recorded in real time, where s is the spatial position along the path and t is the time coordinate. The signal delay between adjacent sampling points is used for spatiotemporal cross-correlation analysis.

[0037] Step 4: Acquire photoacoustic signal sequences along the blood vessel direction according to the directional scanning path, and perform spatiotemporal cross-correlation analysis on the photoacoustic signal sequences; Step 4 includes: 4.1 Execute the directional scanning path and acquire photoacoustic signal sequences along the blood vessel direction to obtain a spatiotemporal data matrix. Based on the spatiotemporal data matrix, calculate the cross-correlation function of adjacent point signals to obtain the similarity between the two signals under different time delays and complete the cross-correlation calculation. 4.2 Search for the peak position of the cross-correlation function curve, extract the correlation peak shift time based on the peak position, obtain the peak time delay, and complete the spatiotemporal cross-correlation analysis.

[0038] Step 4 in detail: Signal preprocessing and feature extraction: To enhance the dynamic components of blood flow signals, suppress interference from static tissue structures, and improve the signal-to-noise ratio of cross-correlation analysis, the specific processing method is as follows: The envelope of the filtered signal is extracted using the Hilbert Transform, followed by high-pass filtering and differential processing to further enhance the dynamic components of blood flow and eliminate the influence of static tissue signals on blood flow calculation.

[0039] Spatiotemporal correlation function calculation and delay extraction: like Figure 3 As shown, blood flow causes a time delay in the signals from adjacent sampling points. Figure 3 The "step-like delay" phenomenon in [the text] can be observed. This delay can be accurately extracted through cross-correlation analysis. The peak position (Δs, Δtpeak) is extracted by analyzing the cross-correlation function along the time and space dimensions.

[0040] (1) Calculation of cross-correlation function For the preprocessed signals of adjacent sampling points si and si+1, calculate the spatiotemporal cross-correlation function: R(Δs,Δt)=∫P(si,t) P(si+1,t+Δt) dt; in: Δs is the fixed spatial distance between adjacent sampling points (a known quantity, such as 10 μm); Δt is a time delay variable (scan parameter), representing the time delay variable. relatively The time required for translation; t is the time coordinate (integral variable) in the depth direction; R(Δs,Δt) is the cross-correlation function value, which reflects the similarity between two signals at a delay of Δt. The scan range of Δt is set according to the expected flow rate, typically [-500 ns, +500 ns]; Physical meaning: The peak value of R(Δs,Δt) corresponds to the optimal time matching point between the two signals. For example... Figure 3 As shown, the cross-correlation function exhibits a Gaussian distribution.

[0041] (2) Extraction of cross-correlation peak positions The "peak value" here refers to the peak value of the cross-correlation function R(Δs,Δt) curve, not the peak value of the A-line signal P(si,t) itself.

[0042] The relevant peak shift time is extracted by searching for the location of the maximum value of the R(Δs,Δt) curve.

[0043] Δtpeak = argmax R(Δs, Δt) Δt∈[-T,T] Where Δtpeak is the time delay corresponding to the cross-correlation function reaching its maximum value.

[0044] argmax represents the value of Δt that makes R(Δs,Δt) reach its maximum value.

[0045] T is the upper limit of the search range (e.g., 100 ns, which is related to the laser repetition rate).

[0046] Step 5: Quantitatively calculate blood flow direction and velocity using the results of spatiotemporal cross-correlation analysis, and visualize the combined blood flow direction and velocity. Step 5 includes: 5.1 The blood flow direction is automatically determined based on the sign of the peak time delay. When the peak time delay is greater than 0, it indicates that the downstream point signal lags behind the upstream point, and the blood flow is determined to be in the forward direction along the scanning direction. When the peak time delay is less than 0, the blood flow is determined to be in the reverse direction along the scanning direction, thus completing the blood flow direction determination. 5.2 Calculate the blood flow velocity based on the fixed interval and the peak time delay, and map the blood flow direction and blood flow velocity back to the original two-dimensional spatial coordinates to construct a complete two-dimensional blood flow vector field; 5.3 The two-dimensional blood flow vector field is superimposed and fused with the high-contrast vascular network image to achieve a visual display that integrates structure and function.

[0047] Step 5 in detail: like Figure 4 As shown, forward flow identification: When the blood flows along the scanning direction (from s1 to s4), the signal at the downstream sampling point si+1 has a positive delay relative to the upstream sampling point si. The peak of the cross-correlation function R(Δs,Δt) occurs at Δtpeak > 0 (marked by a solid circle in the figure, located to the right of the reference line Δt=0); Physical meaning: The signal of si needs to be shifted to the right (positively delayed) by Δtpeak time to achieve the best match with the signal of si+1, indicating that the signal of si+1 is indeed lagging behind si; Judgment result: Δtpeak > 0 → Blood flow is along the scanning direction (positive direction) like Figure 4 As shown, reverse flow identification: When the blood flow is reversed (opposite to the scanning direction), the signal from the upstream sampling point lags behind that of the downstream sampling point. The peak of the cross-correlation function R(Δs,Δt) occurs when Δtpeak < 0 (marked by a hollow circle in the figure, located to the left of the reference line Δt=0); Physical meaning: The signal of si+1 needs to be shifted to the left (negative delay) to match si, indicating that the signal of si+1 is actually earlier than si; Judgment result: Δtpeak < 0 → Blood flow is opposite to the scanning direction (reverse flow).

[0048] Adaptive weighted smoothing and vector field generation: The velocity field is subjected to weighted smoothing based on the local signal-to-noise ratio to construct a continuous blood flow direction vector map.

[0049] Results fusion and display: The structural image is superimposed with the blood flow vector field to achieve integrated visualization of structure and function.

[0050] This invention may include: (a) Scan path generation: Several scan paths can be automatically generated based on the direction of the vessel centerline. The preferred path spacing is 10μm to achieve full vessel coverage while balancing spatial sampling density and scanning speed. The scan length can be automatically extended according to the length of the target vessel, or the area can be manually defined by the user.

[0051] (b) The scanning actuator may be one or more of the following combinations: MEMS galvanometer, acousto-optic deflector (AOD) or dual-axis electric displacement stage.

[0052] The preferred method is to use a galvanometer and a MEMS micro-scanner, with a single-axis scanning frequency range of 10 kHz-40 kHz and an angular accuracy better than 0.05°, which can achieve fast and highly repeatable directional scanning. For large field-of-view imaging, a two-dimensional electric platform can be used to drive the transducer to move along the blood vessel axis for scanning and acquisition.

[0053] (c) The ultrasonic detection transducer is preferably an ultra-wideband hollow spherical focused transducer with a center frequency of 30 MHz - 50 MHz, a bandwidth ≥ 60%, and a focal length of 5 mm - 10 mm to balance spatial resolution and signal penetration depth.

[0054] (d) The photoacoustic signal acquisition module adopts a synchronous trigger structure, and the synchronization error between the laser trigger signal and the sampling clock does not exceed 10 ns. The data acquisition rate is not less than 100 MS / s to ensure the accuracy of spatio-temporal cross-correlation.

[0055] (e) The spatio-temporal cross-correlation analysis adopts a sliding window acceleration algorithm. The window length is preferably 100 - 500 ns and can be automatically adjusted according to the signal bandwidth; the cross-correlation peak detection uses a sub-pixel fitting method (such as parabolic or Gaussian fitting), and the peak time delay resolution can be better than 5 ns, corresponding to a flow velocity resolution better than 0.05 mm / s.

[0056] (f) The rule for determining the blood flow direction is as follows: If the peak time offset Δtpeak of the cross-correlation function of adjacent sampling point signals > 0, it is determined that the blood flow is in the scanning direction; if Δtpeak < 0, it is determined that the blood flow is in the reverse direction; when |Δtpeak| < tth (threshold), the local flow velocity is considered approximately zero. The threshold tth is preferably set to 0.1Δs / c, where c is the speed of sound and Δs is the sampling point spacing.

[0057] (g) The fusion display of the blood vessel structure and the flow field can adopt a vector superposition mode or a pseudo-color coding method, where the blood flow direction is represented by color (such as red / blue representing forward / backward), and the blood flow velocity is represented by brightness or arrow length to achieve the visualization of structure - function integration.

[0058] (h) The control and data processing unit preferably includes a heterogeneous architecture of FPGA + GPU. The FPGA is responsible for laser triggering, scanning control, and signal caching, and the GPU is responsible for spatio-temporal cross-correlation calculation, blood flow direction reconstruction, and real-time visualization; The system can achieve a single-frame scanning period ≤ 100 ms, which can meet the requirements of quasi-real-time dynamic blood flow monitoring.

[0059] (i) For the signal processing algorithm, it supports multiple cross-correlation acceleration implementation methods, including fast Fourier transform (FFT) cross-correlation, phase-spectrum-based delay estimation, and fast matching algorithm based on sparse templates; the system can adaptively select the optimal strategy according to the signal-to-noise ratio.

[0060] Example 2 Hardware configuration: Laser source: wavelength 532 nm, pulse width 5 ns, repetition frequency 50 kHz, single pulse energy density less than 20 mJ / cm² 2 (Complies with ANSI Z136.1 laser safety standard); MEMS scanning mirror: dual-axis deflection range ±5°, scanning frequency 2000Hz, driving voltage ±10V, angular resolution better than 0.05°; Ultrasonic transducer: spherical focusing, center frequency 50 MHz, bandwidth 60%, focal length 6.5 mm; Acquisition system: FPGA + high-speed ADC (200 MS / s, 12-bit), with a synchronization error of less than 10 ns with laser triggering; Data processing platform: GPU (NVIDIA RTX 3050Ti) performs real-time reconstruction and cross-correlation calculations; Samples: Live / stereoscopic mouse ear, brain, or skin vascular models (flow rate range 0.1-2 mm / s).

[0061] Experimental procedure: (1) Structural diagram acquisition: Photoacoustic structural images I(x,y) were obtained in MEMS full-field scanning mode. Envelope reconstruction was performed using bandpass filtering (5-80 MHz) and Hilbert transform to obtain a clear vascular network map.

[0062] (2) Vessel extraction and orientation identification: Frangi filtering was applied to enhance the vascular signal, followed by the extraction of the vascular centerline using an adaptive threshold segmentation and skeletonization algorithm; then, the orientation angle of each point was extracted using local principal component analysis (PCA).

[0063] (3) Directional guidance scan path generation: The system automatically generates scanning trajectories along the centerline of the blood vessel, with a path spacing of approximately 10 μm.

[0064] MEMS galvanometers perform high-speed scanning along this path to obtain time-continuous signals.

[0065] (4) Spatiotemporal signal acquisition and cross-correlation calculation: Acquire the photoacoustic signal matrix P(s,t). Calculate the cross-correlation function for the signals at adjacent positions s and s+Δs: R(τ)=∫P(s,t) P(s+Δs,t+τ) dt Blood flow velocity is calculated by detecting the peak time delay τpeak of R(τ): ν = Δsτpeak If the main peak appears when τ>0, it is determined that the blood flow is along the scanning direction; if τ<0, it is the reverse flow.

[0066] (5) Blood flow vector generation: The flow velocity and direction of each scan path are mapped onto the structural diagram to obtain a directional blood flow velocity distribution map.

[0067] In pseudo-color displays, color indicates direction (red / blue represents forward / backward), and brightness indicates speed.

[0068] Results analysis: Experiments show that, compared with the traditional autocorrelation method, the signal-to-noise ratio of this method is improved by about 40%, while the flow direction can be determined and the average measurement time is reduced by about 7 times.

[0069] The present invention can also be modified as follows according to specific needs: The light source wavelength can be selected as 532 nm or 560 nm to suit different situations; The transducer type can be a spherical focusing transducer, a flat field transducer, or a hydrophone; The scanning mechanism can be replaced by an acousto-optic deflector, a galvanometer assembly, or a two-dimensional displacement motor; Cross-correlation calculations can be implemented on FPGAs, GPUs, or embedded processing units.

[0070] Therefore, this invention provides a label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation. It can combine structural information guidance to quickly scan along the blood vessel direction and realize directional blood flow analysis through spatiotemporal cross-correlation, thereby significantly improving the measurement speed and accuracy. At the same time, it realizes robust and efficient joint measurement of blood flow direction and velocity.

[0071] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0072] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A label-free photoacoustic microscopic blood flow direction and velocity combined measurement method based on spatiotemporal cross-correlation, characterized in that, include: Two-dimensional structural images of the target tissue were acquired using a photoacoustic microscopy imaging platform, and the two-dimensional structural images were preprocessed and enhanced with blood vessels to obtain enhanced tubular regions. Based on the enhanced tubular region, blood vessel segmentation and morphological optimization are performed to obtain a binary blood vessel image. The centerline of the binary blood vessel image is extracted and the direction is calculated to obtain blood vessel direction information. Based on the blood vessel direction information, path planning is performed to obtain a scan point sequence. Then, the scan point sequence is optimized and time-synchronized to obtain a directional scan path. According to the directional scanning path, photoacoustic signal sequences are acquired along the blood vessel direction, and spatiotemporal cross-correlation analysis is performed on the photoacoustic signal sequences; The results of spatiotemporal cross-correlation analysis are used to quantitatively calculate blood flow direction and velocity, and to visualize the combined blood flow direction and velocity.

2. The label-free photoacoustic microscopic blood flow direction and velocity combined measurement method based on spatiotemporal cross-correlation according to claim 1, characterized in that, Two-dimensional structural images of the target tissue were acquired using a photoacoustic microscopy imaging platform. These images were then preprocessed and enhanced with vascular enhancement to obtain enhanced tubular regions, including: The sample is irradiated with a pulsed laser to generate an ultrasonic signal. The ultrasonic signal is received by a spherical focusing ultrasonic transducer. The envelope information of the ultrasonic signal is then extracted by bandpass filtering and Hilbert transform to obtain a reconstructed high-contrast vascular network image. Based on the Frangi filter principle, the second derivative matrix of the high-contrast vascular network image is calculated, and eigenvalue analysis is performed on the second derivative matrix to obtain the first eigenvalue and the second eigenvalue. Determine whether the second feature value is negative and whether the absolute value of the second feature value is much greater than the absolute value of the first feature value. If so, it is identified as a tubular structure. Then, an enhancement function is used to detect vascular protrusion, resulting in an enhanced tubular region. The calculation expression of the enhancement function is as follows: ； wherein is a cycle scale parameter, is a first eigenvalue, is a second eigenvalue, is a suppression noise, , is an adjustment parameter.

3. The label-free photoacoustic microscopic blood flow direction and velocity combined measurement method based on spatiotemporal cross-correlation according to claim 2, characterized in that, Based on the enhanced tubular region, vessel segmentation and morphological optimization are performed to obtain a binary vessel image. The centerline of the binary vessel image is extracted and its direction is calculated to obtain vessel orientation information, including: Based on the enhanced tubular region, an adaptive threshold algorithm is used for binarization segmentation to extract the vascular region. According to the tubular characteristics of the blood vessels in the image at the local scale, a disk-shaped structural element with circular symmetry is selected to perform dilation and erosion operations on the vascular region to connect broken blood vessels and fill local holes without introducing directional bias, thereby completing the morphological closure operation and obtaining a binarized vascular image. Using median transformation or Lee's algorithm, the binarized blood vessel image is refined into a skeleton with a width of one pixel. Skeleton branches shorter than a preset threshold are removed to filter noise. Then, local principal component analysis is performed on each skeleton point to calculate the local orientation angle and obtain the blood vessel orientation information.

4. The label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation according to claim 3, characterized in that, Based on the blood vessel orientation information, path planning is performed to obtain a scan point sequence. This scan point sequence is then optimized and time-synchronized to obtain a directional scanning path, including: Based on the blood vessel orientation information, the skeleton points are sorted according to the blood vessel topology, and a main path distributed along the blood vessel centerline is generated using a B-spline curve fitting algorithm. Then, the main path is resampled at a fixed interval to generate a scan point sequence. Based on the scan point sequence, the scanning order of multiple blood vessel paths is calculated using an optimization algorithm, and the overlapping region detection algorithm is used to identify whether there are repeated samplings at blood vessel intersections or branch points, thus completing the scan optimization operation. The optimized scan point sequence is converted into a control signal. Based on the control signal, the FPGA is used to drive the scanning mechanism to perform directional scanning and timing synchronization, thereby obtaining the directional scanning path.

5. The label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation according to claim 4, characterized in that, According to the directional scanning path, photoacoustic signal sequences are acquired along the blood vessel direction, and spatiotemporal cross-correlation analysis is performed on the photoacoustic signal sequences, including: The directional scanning path is executed, and photoacoustic signal sequences are acquired along the direction of the blood vessel to obtain a spatiotemporal data matrix. Based on the spatiotemporal data matrix, the cross-correlation function of adjacent point signals is calculated to obtain the similarity between the two signals under different time delays, thus completing the cross-correlation calculation. The peak position of the cross-correlation function curve is searched, and the peak shift time is extracted based on the peak position to obtain the peak time delay, thus completing the spatiotemporal cross-correlation analysis.

6. The label-free photoacoustic microscopic blood flow direction and velocity joint measurement method based on spatiotemporal cross-correlation according to claim 5, characterized in that, The results of spatiotemporal cross-correlation analysis are used to quantitatively calculate blood flow direction and velocity, and to visualize the combined blood flow direction and velocity, including: The blood flow direction is automatically determined based on the sign of the peak time delay. When the peak time delay is greater than 0, it indicates that the downstream point signal lags behind the upstream point, and the blood flow is determined to be flowing in the forward direction along the scanning direction. When the peak time delay is less than 0, the blood flow is determined to be flowing in the reverse direction along the scanning direction, thus completing the blood flow direction determination. The blood flow velocity is calculated based on the fixed interval and the peak time delay, and the blood flow direction and blood flow velocity are mapped back to the original two-dimensional spatial coordinates to construct a complete two-dimensional blood flow vector field. The two-dimensional blood flow vector field is superimposed and fused with the high-contrast vascular network image to achieve a visual display that integrates structure and function.

Images