A thrombus assisted identification method and device based on blood flow monitoring

Abstract

This invention relates to a thrombus-assisted identification method based on blood flow monitoring. It constructs a complete ultrasound-assisted thrombus identification process adapted to the extracorporeal circulation arteriovenous chamber scenario in dialysis. Using non-invasive Doppler ultrasound as the detection medium, it achieves online thrombus monitoring without damaging existing blood tubing or contacting the patient's blood. Through dual-modal joint analysis of grayscale anatomical information and color Doppler flow information, it significantly improves the detection rate of early microthrombi and fibrin deposits compared to single pressure monitoring or manual visual observation. The fully automated data processing chain reduces the workload and subjective errors of manual interpretation by medical staff, enabling continuous thrombus early warning throughout dialysis, effectively reducing the risk of adverse events such as extracorporeal circulation coagulation and tubing blockage, and ensuring the safety and adequacy of dialysis treatment. Furthermore, the thrombus monitoring device is clip-shaped, allowing for easy clamping at any position in the venous or arterial chambers, facilitating data acquisition.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare technology, and in particular to a method and device for thrombus-assisted identification based on blood flow monitoring. Background Technology

[0002] Current clinical monitoring of coagulation in extracorporeal circulation loops still relies primarily on traditional methods, which have significant technical limitations. Firstly, there is indirect monitoring using pressure parameters, inferring the degree of blockage by collecting changes in arterial and venous pressure and transmembrane pressure. While this method requires no additional hardware, it only issues an alert after the thrombus has grown to the point of causing a significant increase in tubing resistance, exhibiting a clear lag. It completely fails to identify early microthrombi and thin-layer fibrin deposits, and the parameter values ​​are easily affected by factors such as pump speed, needle position, and patient position, resulting in a high false alarm rate. Secondly, there is manual visual inspection, where medical staff periodically observe blood clots and deposits in the arteriovenous chamber. This method is highly subjective, heavily reliant on the operator's clinical experience, and is based on intermittent sampling, making continuous monitoring impossible and failing to promptly detect rapidly progressing coagulation events. It also significantly increases the workload of clinical staff. Summary of the Invention

[0003] The purpose of this invention is to at least address one of the shortcomings of the prior art and provide a thrombus-assisted identification method based on blood flow monitoring.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: Specifically, a thrombus-assisted identification method based on blood flow monitoring is proposed, including the following: Acquire Doppler ultrasound data at the target location of the venous or arterial pot; The Doppler ultrasound data are preprocessed to obtain a two-dimensional RF data matrix; The two-dimensional RF data matrix is ​​converted into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image. The extent of the vascular lumen was located from the color blood flow overlay image, and regions of interest suspected of being thrombi were preliminarily screened. Extract three types of features from the region of interest: morphology, texture, and hemodynamics. Thrombus identification results are obtained by using three types of features: morphology, texture, and hemodynamics. The thrombus identification results are used to perform quantitative parameter calculations and generate a visualization report.

[0005] Furthermore, specifically, the method for acquiring Doppler ultrasound data at the target location of the venous or arterial pothole is as follows: The thrombosis monitoring device is a clip-on Doppler ultrasound probe that is clamped to the target location of the venous or arterial chamber.

[0006] Furthermore, specifically, the Doppler ultrasound data is preprocessed to obtain a two-dimensional RF data matrix, including: Step 1: The probe's piezoelectric array converts electrical signals into ultrasound waves and transmits them to the target user's associated arterial or venous tubules; the echoes are received by the probe array and converted back into analog electrical signals. Step 2: The analog electrical signal is amplified by a low-noise amplifier (LNA) to produce a weak echo; a time-gain compensated gas cascade (TGC) is used to increase the depth echo gain according to the depth, compensating for the propagation attenuation of ultrasound in tissue; a high-speed ADC samples and quantizes the analog signal, completing analog-to-digital conversion to obtain multi-channel digital radio frequency (RF) echo data. Step 3: Perform precise delay and weighted summation on the RF echo data of different array elements to achieve dynamic focusing and aperture control, complete spatial beamforming, and integrate multi-channel data into a line-by-line scanning two-dimensional RF data matrix.

[0007] Further, specifically, converting the two-dimensional RF data matrix into a grayscale anatomical image and a color flow overlay image includes, Step 4: Envelope detection is performed on the two-dimensional RF data matrix to extract the echo amplitude; the echo intensity with a wide dynamic range is mapped to 256 gray levels through logarithmic compression; image clarity is optimized by speckle noise suppression and tissue harmonic imaging to improve the boundary contrast between the blood vessel wall and the thrombus and obtain a gray-scale anatomical image. Step 5: Mix the two-dimensional RF data matrix with the orthogonal local oscillator signal, and down-convert to obtain two baseband complex signals, I in phase and Q orthogonal, namely the I / Q baseband orthogonal signal matrix, which completely preserves the phase information of the echo and provides a basis for blood flow velocity calculation; Step 6: Using a high-order high-pass wall filter, the strong clutter signals generated by the low-frequency motion of the blood vessel wall and valves in the I / Q baseband orthogonal signal matrix are filtered out, and only the high-frequency Doppler frequency shift signal generated by the motion of red blood cells is retained, so as to avoid tissue motion interfering with blood flow interpretation and obtain the filtered blood flow I / Q signal matrix. Step 7: Perform autocorrelation operation on the filtered blood flow I / Q signal matrix, and calculate the average blood flow velocity (i.e., blood flow velocity matrix), blood flow direction (i.e., blood flow direction matrix), and velocity variance (i.e., velocity variance matrix) of each sampling point by using the phase difference between adjacent echoes. Step 8: Map blood flow velocity to color encoding; based on priority determination, color is superimposed on effective blood flow pixels, while grayscale is retained in areas without blood flow; color Doppler fusion image is obtained by reducing image flicker noise through spatiotemporal smoothing filtering.

[0008] Furthermore, specifically, the extent of the vascular lumen is located from the color blood flow overlay image, and regions of interest suspected of causing thrombosis are preliminarily screened, including... Step 9: Remove the device parameter bar and text labels from the image edges of the color Doppler fusion image; perform color space conversion from RGB to HSV, and separate the color blood flow component and grayscale background component; complete pixel spacing normalization and contrast adaptive enhancement to eliminate parameter differences between different devices and obtain a standardized dual-channel image; Step 10: Use a semantic segmentation network to perform pixel-level segmentation on the standardized dual-channel image, accurately distinguishing three types of regions: blood vessel wall, blood vessel lumen, and surrounding soft tissue, to obtain blood vessel wall boundary masks, blood vessel lumen region masks, and background tissue masks. Step 11: Based on the vessel wall boundary mask, vessel lumen region mask, and background tissue mask, mark the region with the preset condition of "no color blood flow signal filling and at the same time, abnormal grayscale echo"; filter out the normal anatomical structures of vascular valves and venous valves through morphological priors, and generate a set of suspected thrombus candidate region masks to obtain the thrombus candidate region mask.

[0009] Furthermore, specifically, three types of features—morphology, texture, and hemodynamics—are extracted from the region of interest, including: Step 12: Calculate the thrombus area, major axis / minor axis, sphericity, boundary smoothness, attachment base width to the vessel wall, and degree of lumen dilation for the grayscale image region corresponding to the candidate ROI; Echo characteristics: Statistical analysis of grayscale mean, standard deviation, and skewness within the region to determine low / equal / high echo type; Texture features: Parameters such as contrast, energy, homogeneity, and entropy are extracted using the Gray-Level Co-occurrence Matrix (GLCM) to reflect the internal homogeneity of the thrombus, ultimately yielding a grayscale morphological texture feature vector. Step 13: Calculate the proportion of the blood flow filling defect area to the lumen and the residual effective lumen area; extract the peak flow velocity at the stenosis, the flow velocity ratio before and after the stenosis, the resistance index RI, and the pulsatility index PI; identify indirect signs such as collateral circulation around the lesion to obtain the hemodynamic feature vector; Step 14: Calculate the rate of change of lumen cross-sectional area and the rate of change of blood flow filling before and after pressurization, and quantify the compressibility of the blood vessel to obtain the characteristic parameters of lumen compressibility.

[0010] Furthermore, specifically, thrombus identification results are obtained based on three types of features: morphology, texture, and hemodynamics. Step 15: Input the grayscale and blood flow fusion image into a multimodal CNN network, and output the pixel-level thrombus segmentation mask and thrombus staging probability, i.e., the thrombus discrimination result. The result includes normal, suspected thrombus, confirmed thrombus, confidence level, and accurate thrombus segmentation mask. Step 16: Based on echo intensity, motion patterns, and morphological characteristics, easily confused structures such as vascular valves, calcified plaques, acoustic artifacts, blood flow aliasing artifacts, and ultrasound side-lobe artifacts are excluded to obtain the final thrombus identification result that eliminates false evidence.

[0011] Furthermore, specifically, the thrombus identification results are used for quantitative parameter calculation and visualization report generation, including: Step 17: Calculate thrombus length, maximum cross-sectional area, and volume; calculate vessel diameter stenosis rate and area stenosis rate; assess occlusion grade; combine echo characteristics to assist in determining thrombus staging and obtain a structured set of quantitative thrombus parameters; Step 18: Overlay the thrombus boundary, measurement scale, and key numerical annotations onto the original image; summarize all parameters to generate a standardized diagnostic report.

[0012] The present invention also proposes a thrombus-assisted identification device based on blood flow monitoring, which applies the steps of the method described above. The device includes the following: The data acquisition module is used to acquire Doppler ultrasound data at the target location of the venous or arterial pothole; The first calculation module is used to preprocess the Doppler ultrasound data to obtain a two-dimensional RF data matrix; The second calculation module is used to convert the two-dimensional RF data matrix into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image. The ROI calculation module is used to locate the range of blood vessel lumen from color blood flow overlay images and initially screen out regions of interest suspected of thrombosis. The feature extraction module is used to extract three types of features from the region of interest: morphology, texture, and hemodynamics. The thrombus identification module is used to identify thrombi based on three types of features: morphology, texture, and hemodynamics. The visualization module is used to perform quantitative parameter calculations and generate visualization reports on the thrombus identification results.

[0013] The beneficial effects of this invention are as follows: This invention proposes a thrombus-assisted identification method and device based on blood flow monitoring. It constructs a complete ultrasound-assisted thrombus identification process adapted to the extracorporeal circulation arteriovenous chamber scenario in dialysis. Using non-invasive Doppler ultrasound as the detection medium, it achieves online thrombus monitoring without damaging existing blood tubing or contacting the patient's blood. Through dual-modal joint analysis of grayscale anatomical information and color Doppler blood flow information, it supports thrombus determination from both structural morphology and blood flow function dimensions. Compared to single pressure monitoring or manual visual observation, it significantly improves the detection rate of early microthrombi and fibrin deposits. The fully automated data processing chain reduces the workload and subjective errors of manual interpretation by medical staff, enabling continuous thrombus early warning throughout dialysis, effectively reducing the risk of adverse events such as extracorporeal circulation coagulation and tubing blockage, and ensuring the safety and adequacy of dialysis treatment. Furthermore, the thrombus monitoring device is clip-shaped, allowing for easy clamping at any position in the venous or arterial chambers, facilitating data acquisition. Attached Figure Description

[0014] The above and other features of this disclosure will become more apparent from the detailed description of the embodiments illustrated in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings: Figure 1 The diagram shows a flowchart of a thrombus-assisted identification method based on blood flow monitoring according to the present invention. Detailed Implementation

[0015] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The same reference numerals used throughout the accompanying drawings indicate the same or similar parts.

[0016] Example 1, referring to Figure 1 This invention proposes a thrombus-assisted identification method based on blood flow monitoring, comprising the following: Step 110: Acquire Doppler ultrasound data at the target location of the venous or arterial pothole; Step 120: Preprocess the Doppler ultrasound data to obtain a two-dimensional RF data matrix; Step 130: Convert the two-dimensional RF data matrix into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image; Step 140: Locate the vascular lumen from the color blood flow overlay image and preliminarily screen out regions of interest suspected of thrombosis; Step 150: Extract three types of features from the region of interest: morphology, texture, and hemodynamics. Step 160: Based on three types of features—morphology, texture, and hemodynamics—thrombus identification is performed to obtain thrombus identification results; Step 170: Calculate quantitative parameters and generate a visualization report based on the thrombus identification results.

[0017] In this embodiment 1, a complete ultrasound-assisted thrombus identification process adapted to the extracorporeal circulation arteriovenous chamber scenario in dialysis was constructed. Using non-invasive Doppler ultrasound as the detection medium, online thrombus monitoring can be achieved without damaging existing blood tubing or contacting the patient's blood. Through dual-modal joint analysis of grayscale anatomical information and color Doppler flow information, thrombus identification is supported from both structural morphology and blood flow function dimensions. Compared to single pressure monitoring or manual visual observation, this significantly improves the detection rate of early microthrombi and fibrin deposits. The fully automated data processing chain reduces the workload and subjective errors of manual interpretation by medical staff, enabling continuous thrombus early warning throughout dialysis, effectively reducing the risk of adverse events such as extracorporeal circulation coagulation and tubing blockage, and ensuring the safety and adequacy of dialysis treatment. Furthermore, the thrombus monitoring device is clip-shaped, allowing for easy clamping at any position in the venous or arterial chambers, facilitating data acquisition.

[0018] In a preferred embodiment of the present invention, the method for acquiring Doppler ultrasound data at the target location of the venous or arterial pothole is as follows: The thrombosis monitoring device is a clip-on Doppler ultrasound probe that is clamped to the target location of the venous or arterial chamber.

[0019] In this preferred embodiment, the thrombosis monitoring device is clip-shaped and used to monitor for thrombosis in the arterial or venous cistern of the blood vessels. The clip has a blood flow scanner to detect areas where blood flow is absent, suggesting the presence of a thrombosis. A small display screen at the end of the clip shows the percentage of thrombosis formation. The clip's monitoring location can be easily changed, for example, to the bottom or top of the cistern, allowing free switching between venous and arterial cisterns. The entire thrombosis monitoring device is rechargeable, and the clip also has a USB charging port at the end.

[0020] Using a clamp-on Doppler ultrasound probe as the data acquisition device, it can be directly clipped and fixed to the outer wall of the arterial and venous chambers of existing dialysis tubing without the need to modify standard dialysis tubing or puncture or contact with blood, completely avoiding the risk of cross-infection. Installation and operation are simple and quick, and it can be rapidly deployed when dialysis is started, adapting to the size of clinically standard arterial and venous chambers. At the same time, it can realize continuous online monitoring throughout the entire dialysis treatment, breaking through the limitations of traditional imaging examinations that can only screen once and cannot monitor throughout the entire process, providing a practical hardware acquisition solution for the early detection of extracorporeal circulation thrombosis.

[0021] In a preferred embodiment of the present invention, specifically, the Doppler ultrasound data is preprocessed to obtain a two-dimensional RF data matrix, including, Step 1: The probe's piezoelectric array converts electrical signals into ultrasound waves and transmits them to the target user's associated arterial or venous tubules; the echoes are received by the probe array and converted back into analog electrical signals. Step 2: The analog electrical signal is amplified by a low-noise amplifier (LNA) to produce a weak echo; a time-gain compensated gas cascade (TGC) is used to increase the depth echo gain according to the depth, compensating for the propagation attenuation of ultrasound in tissue; a high-speed ADC samples and quantizes the analog signal, completing analog-to-digital conversion to obtain multi-channel digital radio frequency (RF) echo data. Step 3: Perform precise delay and weighted summation on the RF echo data of different array elements to achieve dynamic focusing and aperture control, complete spatial beamforming, and integrate multi-channel data into a line-by-line scanning two-dimensional RF data matrix.

[0022] In this preferred embodiment, the three-stage preprocessing process of analog signal conditioning, analog-to-digital conversion, and digital beamforming can effectively amplify weak ultrasound echoes and compensate for the propagation attenuation of ultrasound waves in the arteriovenous ...

[0023] Among them, precise time delay and weighted summation are used to achieve dynamic focusing and aperture control, employing a time-delay superposition (DAS) beamforming algorithm commonly used in the ultrasonic field. Delay calculation: Based on the preset value of tissue / blood sound velocity (1540m / s), calculate the sound path difference from each array element to the current depth focus point, and divide the sound path difference by the sound velocity to obtain the corresponding reception delay. Dynamic focusing: The receiver adopts point-by-point dynamic focusing, updating the focusing delay once for each sampling point along the axial depth direction to achieve optimal focusing across the entire depth range; Dynamic aperture: The receiving aperture expands synchronously with the depth of focus, keeping the F-number (the ratio of focal length to aperture) constant and maintaining the consistency of lateral resolution.

[0024] In a preferred embodiment of the present invention, specifically, the two-dimensional RF data matrix is ​​converted into a grayscale anatomical image and a color blood flow overlay image, including, Step 4: Envelope detection is performed on the two-dimensional RF data matrix to extract the echo amplitude; the echo intensity with a wide dynamic range is mapped to 256 gray levels through logarithmic compression; image clarity is optimized by speckle noise suppression and tissue harmonic imaging to improve the boundary contrast between the blood vessel wall and the thrombus and obtain a gray-scale anatomical image. Step 5: Mix the two-dimensional RF data matrix with the orthogonal local oscillator signal, and down-convert to obtain two baseband complex signals, I in phase and Q orthogonal, namely the I / Q baseband orthogonal signal matrix, which completely preserves the phase information of the echo and provides a basis for blood flow velocity calculation; Step 6: Using a high-order high-pass wall filter, the strong clutter signals generated by the low-frequency motion of the blood vessel wall and valves in the I / Q baseband orthogonal signal matrix are filtered out, and only the high-frequency Doppler frequency shift signal generated by the motion of red blood cells is retained, so as to avoid tissue motion interfering with blood flow interpretation and obtain the filtered blood flow I / Q signal matrix. Step 7: Perform autocorrelation operation on the filtered blood flow I / Q signal matrix, and calculate the average blood flow velocity (i.e., blood flow velocity matrix), blood flow direction (i.e., blood flow direction matrix), and velocity variance (i.e., velocity variance matrix) of each sampling point by using the phase difference between adjacent echoes. Step 8: Map blood flow velocity to color encoding; based on priority determination, color is superimposed on effective blood flow pixels, while grayscale is retained in areas without blood flow; color Doppler fusion image is obtained by reducing image flicker noise through spatiotemporal smoothing filtering.

[0025] In a preferred embodiment of the present invention, specifically, the vascular lumen is located from the color blood flow overlay image, and regions of interest suspected of causing thrombosis are initially screened, including... Step 9: Remove the device parameter bar and text labels from the image edges of the color Doppler fusion image; perform color space conversion from RGB to HSV, and separate the color blood flow component and grayscale background component; complete pixel spacing normalization and contrast adaptive enhancement to eliminate parameter differences between different devices and obtain a standardized dual-channel image; Step 10: Use a semantic segmentation network to perform pixel-level segmentation on the standardized dual-channel image, accurately distinguishing three types of regions: blood vessel wall, blood vessel lumen, and surrounding soft tissue, to obtain blood vessel wall boundary masks, blood vessel lumen region masks, and background tissue masks. Step 11: Based on the vessel wall boundary mask, vessel lumen region mask, and background tissue mask, mark the region with the preset condition of "no color blood flow signal filling and at the same time, abnormal grayscale echo"; filter out the normal anatomical structures of vascular valves and venous valves through morphological priors, and generate a set of suspected thrombus candidate region masks to obtain the thrombus candidate region mask.

[0026] In this preferred embodiment, through the parallel processing link of grayscale imaging and color Doppler imaging, anatomical structure images and blood flow distribution images within the arteriovenous fossa can be output simultaneously: grayscale imaging, after noise reduction optimization, can clearly present the morphology and echo characteristics of thrombi and fibrin deposits within the fossa; the color Doppler link, through orthogonal demodulation, wall filtering, and autocorrelation calculation, can accurately extract red blood cell motion signals, accurately calculate blood flow velocity, direction, and turbulence level, and effectively filter out clutter interference caused by fossa wall vibration; the finally generated fused image can intuitively present the blood flow filling defect area within the lumen, while preserving tissue structure details, providing structural and functional data support for subsequent thrombus identification.

[0027] The process involves mixing with the quadrature local oscillator signal and down-converting to obtain the I / Q baseband complex signal, using the following scheme: Two digital local oscillator signals with a 90° phase difference are generated, with a frequency equal to the probe center frequency f0; The two-dimensional real RF matrix is ​​multiplied by the two local oscillator signals respectively to obtain the mixed signal; After filtering out the high-frequency components at twice the frequency by a low-pass filter, two baseband signals, one in phase (I) and one quadrature (Q), are output. These are combined into a complex signal matrix S(n) = I(n) + jQ(n), completely preserving the echo phase information. An equivalent alternative is to convert the real RF signal into an analytic complex signal using a Hilbert transform, directly extracting the I / Q components, achieving the same technical effect.

[0028] In a preferred embodiment of the present invention, specifically, three types of features—morphology, texture, and hemodynamics—are extracted from the region of interest, including: Step 12: Calculate the thrombus area, major axis / minor axis, sphericity, boundary smoothness, attachment base width to the vessel wall, and degree of lumen dilation for the grayscale image region corresponding to the candidate ROI; Echo characteristics: Statistical analysis of grayscale mean, standard deviation, and skewness within the region to determine low / equal / high echo type; Texture features: Parameters such as contrast, energy, homogeneity, and entropy are extracted using the Gray-Level Co-occurrence Matrix (GLCM) to reflect the internal homogeneity of the thrombus, ultimately yielding a grayscale morphological texture feature vector. Step 13: Calculate the proportion of the blood flow filling defect area to the lumen and the residual effective lumen area; extract the peak flow velocity at the stenosis, the flow velocity ratio before and after the stenosis, the resistance index RI, and the pulsatility index PI; identify indirect signs such as collateral circulation around the lesion to obtain the hemodynamic feature vector; Step 14: Calculate the rate of change of lumen cross-sectional area and the rate of change of blood flow filling before and after pressurization, and quantify the compressibility of the blood vessel to obtain the characteristic parameters of lumen compressibility.

[0029] In a preferred embodiment of the present invention, specifically, thrombus identification results are obtained based on three types of features: morphology, texture, and hemodynamics. Step 15: Input the grayscale and blood flow fusion image into a multimodal CNN network, and output the pixel-level thrombus segmentation mask and thrombus staging probability, i.e., the thrombus discrimination result. The result includes normal, suspected thrombus, confirmed thrombus, confidence level, and accurate thrombus segmentation mask. Step 16: Based on echo intensity, motion patterns, and morphological characteristics, easily confused structures such as vascular valves, calcified plaques, acoustic artifacts, blood flow aliasing artifacts, and ultrasound side-lobe artifacts are excluded to obtain the final thrombus identification result that eliminates false evidence.

[0030] In this preferred embodiment, image standardization preprocessing can eliminate image differences caused by different ultrasound equipment and different acquisition parameters, improving the algorithm's adaptability to different equipment; pixel-level lumen segmentation based on semantic segmentation network has higher segmentation accuracy than traditional thresholding and edge detection methods, and can accurately adapt to the chamber morphology of arteriovenous sacs; automatic screening of suspected thrombus regions of interest can significantly reduce the calculation scope of subsequent feature extraction and recognition, improving the overall processing speed; combined with morphological prior screening to remove normal structures within the sac, false positives can be reduced in the initial screening stage, locking in effective target areas for subsequent accurate identification.

[0031] In addition, the high-order high-pass wall filter adopts the wall filtering scheme commonly used in ultrasound blood flow detection: Filter type: A linear phase finite-length unit impulse response (FIR) high-pass filter or a polynomial regression wall filter is used, typically of order 4 to 8. Cutoff frequency setting: Set to 50~200Hz according to the detection scenario to filter out strong clutter signals generated by low-frequency movements of structures such as the pot wall and valves, and retain high-frequency Doppler shift signals generated by red blood cell movement; Design rationale: The movement velocity of tissue / vessel walls is much lower than that of blood flow, corresponding to a Doppler frequency offset lower than the cutoff frequency, thereby achieving separation of clutter and blood flow signals.

[0032] In addition, tissue harmonic imaging optimizes image clarity using pulse inversion harmonic imaging (PIHI) technology, which is commonly used in clinical ultrasound equipment. The transmitting end emits two fundamental pulses with the same amplitude but opposite phase to the same scan line in succession; The RF signals from the two echoes are received and superimposed. The fundamental components of linear propagation cancel each other out, while the second harmonic components generated by nonlinear propagation are superimposed and enhanced. The second harmonic component is extracted for subsequent envelope detection and grayscale imaging to suppress fundamental noise and artifacts and improve the contrast between the pot wall and the thrombus boundary.

[0033] In a preferred embodiment of the present invention, specifically, the thrombus identification results are subjected to quantitative parameter calculation and visualization report generation, including: Step 17: Calculate thrombus length, maximum cross-sectional area, and volume; calculate vessel diameter stenosis rate and area stenosis rate; assess occlusion grade; combine echo characteristics to assist in determining thrombus staging and obtain a structured set of quantitative thrombus parameters; Step 18: Overlay the thrombus boundary, measurement scale, and key numerical annotations onto the original image; summarize all parameters to generate a standardized diagnostic report.

[0034] In this preferred embodiment, key indicators such as thrombus size, degree of stenosis, and occlusion grade can be quantitatively presented through structured thrombus quantitative parameter calculation, assisting clinical staff in accurately assessing the severity of thrombus and determining the timing of intervention and treatment plan; visual annotation can intuitively mark the location and extent of thrombus on the original image, facilitating rapid reading of results by medical staff; standardized diagnostic reports can be directly linked to the clinical information system for archiving, supporting follow-up comparisons of multiple dialysis sessions, achieving standardized output of thrombus identification results, and effectively improving clinical diagnosis and treatment efficiency and medical record management quality.

[0035] Specifically, the semantic segmentation network used for pixel-level segmentation employs the U-Net series network architecture, which is commonly used in the field of medical image segmentation. Network input: Standardized dual-channel image, the first channel is a grayscale anatomical image, and the second channel is a color blood flow component image; Network architecture: It adopts the classic U-Net or U-Net++ encoder-decoder structure, and fuses deep and shallow features through skip connections to output a three-class probability map with the same size as the input. Post-processing: The probability map is converted into three types of pixel masks: pot wall, lumen, and surrounding structures through the argmax operation; the network is trained using an annotated clinical / extracorporeal tubing ultrasound dataset, and the loss function is a combination of cross-entropy loss and Dice loss.

[0036] In addition, step 11 uses morphological priors to screen out vascular valves and venous valves based on their normal anatomical structure, and the screening is performed using anatomical and ultrasound imaging priors. Location prior: Valves are mostly symmetrically attached to the sinus sites on the inner wall of the ampulla, and are distributed in pairs along the long axis of the fluid. Morphological prior: The valve has a slender crescent-shaped structure with uniform echo, smooth borders, and is continuous with the cup wall, without any space-occupying bulges; Motion prior: The valve pulsates with blood flow, exhibiting a regular opening-closing motion, which can be verified by the motion characteristics of adjacent Doppler data frames; Structures that simultaneously meet the above characteristics are judged as normal valves and are removed from the candidate thrombus region.

[0037] In step 14, the rate of change in the cross-sectional area of ​​the lumen before and after compression is obtained using existing techniques, such as the compression ultrasound, commonly used in the diagnosis of venous thrombosis. 1. Supplementary data acquisition phase: Apply standardized stepped pressure using a clamp-on probe, or in conjunction with an external miniature pressurization mechanism, to simultaneously acquire two sets of Doppler ultrasound data in the resting and pressurized states; the pressurization standard is based on the complete closure of the lumen without thrombus, which is a generally accepted clinical operating procedure; 2. Parameter Calculation: Calculate the cross-sectional area of ​​the lumen mask for both states, and calculate the rate of change using the formula: ; The thrombus area cannot be compressed due to its solid mass, and the corresponding change rate is significantly lower than that of the normal lumen segment, which is one of the core signs for differential diagnosis of thrombosis.

[0038] In step 15, the grayscale and blood flow fused image is input into the multimodal CNN network, which adopts the dual-stream feature fusion CNN architecture commonly used in the field of medical image analysis. Dual-stream branch design: Two parallel feature extraction branches are set up, namely grayscale image branch and color blood flow image branch. The backbone network can adopt general convolutional networks such as ResNet and VGG. Feature fusion: In the deep layers of the network, the two feature streams are fused by channel concat or attention-weighted fusion, integrating morphological texture information and hemodynamic information. Dual-task output: After feature fusion, two output heads are connected. One outputs a pixel-level thrombus segmentation mask, and the other outputs the thrombus staging / confidence classification probability, realizing multi-task joint output of segmentation and classification.

[0039] In step 16, exclusions are made according to the general identification criteria of ultrasound diagnostics, which are mature existing technologies, including: Calcified plaques: appear as focal hyperechoic areas with clear posterior acoustic shadowing, tightly attached to the vessel / pot wall, without any corresponding mass effect of blood flow filling defects; Acoustic artifacts: located directly behind strong echo structures (such as calcification, pipe joints), appearing as strips of low / no echo along the direction of the sound beam, without clear solid boundaries, and shifting with the probe angle; Blood flow aliasing artifact: appears in high-velocity blood flow regions, manifested as color reversal of colored blood flow, with high-velocity frequency shifts exceeding the Nyquist limit visible in the spectral Doppler, and no real echoes in the corresponding grayscale image; Side lobe artifacts: These are distributed symmetrically in mirror image to the strong echoes of the main lobe, with blurred shapes, and significantly shift with changes in probe scanning angle, without corresponding solid structures. By comparing these characteristics with the core features of solid thrombi and blood flow filling defects, false positive areas can be excluded.

[0040] In step 17, echogenicity features are used to assist in determining the thrombus stage. The staging criteria and corresponding echogenicity features are as follows: The diagnosis is based on the clinically accepted ultrasound staging criteria for thrombosis, which is a mature and existing diagnostic consensus. Acute phase (within 1-2 weeks of onset): The thrombus is hypoechoic or anechoic, homogeneous internally, with clear boundaries, and the lumen is often dilated. The thrombus can drift with the blood flow. Subacute phase (2 weeks to 3 months after onset): The thrombus has a medium echogenicity, the internal echoes gradually become heterogeneous, the boundaries become clearer, and the luminal dilation is reduced; Chronic phase (more than 3 months after onset): The thrombus is hyperechoic or heterogeneous, and may be accompanied by calcification, thickening of the vessel wall / catheter wall, and signs of recanalization may appear in the lumen.

[0041] The present invention also proposes a thrombus-assisted identification device based on blood flow monitoring, which applies the steps of the method described above. The device includes the following: The data acquisition module is used to acquire Doppler ultrasound data at the target location of the venous or arterial pothole; The first calculation module is used to preprocess the Doppler ultrasound data to obtain a two-dimensional RF data matrix; The second calculation module is used to convert the two-dimensional RF data matrix into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image. The ROI calculation module is used to locate the range of blood vessel lumen from color blood flow overlay images and initially screen out regions of interest suspected of thrombosis. The feature extraction module is used to extract three types of features from the region of interest: morphology, texture, and hemodynamics. The thrombus identification module is used to identify thrombi based on three types of features: morphology, texture, and hemodynamics. The visualization module is used to perform quantitative parameter calculations and generate visualization reports on the thrombus identification results.

[0042] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0043] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0044] Although the description of the invention has been quite detailed and particularly of several described embodiments, it is not intended to limit it to any of these details or embodiments or any particular embodiment, but should be considered as providing a broad possible interpretation of the claims by referring to the appended claims and taking into account the prior art, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors in order to provide a useful description, and non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications.

[0045] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any embodiment that achieves the technical effects of the present invention using the same means should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations can be made to the technical solutions and / or implementation methods.

Claims

1. A thrombus-assisted identification method based on blood flow monitoring, characterized in that, Including the following: Acquire Doppler ultrasound data at the target location of the venous or arterial pot; The Doppler ultrasound data are preprocessed to obtain a two-dimensional RF data matrix; The two-dimensional RF data matrix is ​​converted into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image. The extent of the vascular lumen was located from the color blood flow overlay image, and regions of interest suspected of being thrombi were preliminarily screened. Extract three types of features from the region of interest: morphology, texture, and hemodynamics. Thrombus identification results are obtained by using three types of features: morphology, texture, and hemodynamics. The thrombus identification results are used to perform quantitative parameter calculations and generate a visualization report.

2. The thrombus-assisted identification method based on blood flow monitoring according to claim 1, characterized in that, Specifically, the method for acquiring Doppler ultrasound data at the target location of the venous or arterial pothole is as follows: The thrombosis monitoring device is a clip-on Doppler ultrasound probe that is clamped to the target location of the venous or arterial chamber.

3. The thrombus-assisted identification method based on blood flow monitoring according to claim 1, characterized in that, Specifically, the Doppler ultrasound data is preprocessed to obtain a two-dimensional RF data matrix, including: Step 1: The probe's piezoelectric array converts electrical signals into ultrasound waves and transmits them to the target user's associated arterial or venous tubules; the echoes are received by the probe array and converted back into analog electrical signals. Step 2: The analog electrical signal is amplified by a low-noise amplifier (LNA) to produce a weak echo; the deep echo gain is increased by a time-gain compensated GC according to depth to compensate for the propagation attenuation of ultrasound in tissue; the high-speed ADC samples and quantizes the analog signal to complete the analog-to-digital conversion and obtain multi-channel digital radio frequency (RF) echo data. Step 3: Perform precise delay and weighted summation on the RF echo data of different array elements to achieve dynamic focusing and aperture control, complete spatial beamforming, and integrate multi-channel data into a line-by-line scanning two-dimensional RF data matrix.

4. The thrombus-assisted identification method based on blood flow monitoring according to claim 3, characterized in that, Specifically, the two-dimensional RF data matrix is ​​converted into a grayscale anatomical image and a color flow overlay image, including: Step 4: Perform envelope detection on the two-dimensional RF data matrix to extract the echo amplitude; Logarithmic compression maps the echo intensity with a wide dynamic range to 256 gray levels; By suppressing speckle noise and optimizing image clarity through tissue harmonic imaging, grayscale anatomical images are obtained by improving the boundary contrast between the blood vessel wall and the thrombus. Step 5: Mix the two-dimensional RF data matrix with the orthogonal local oscillator signal, and down-convert to obtain two baseband complex signals, I in phase and Q orthogonal, namely the I / Q baseband orthogonal signal matrix, which completely preserves the phase information of the echo and provides a basis for blood flow velocity calculation; Step 6: Using a high-order high-pass wall filter, filter out the clutter signals generated by the low-frequency motion of the blood vessel wall and valves in the I / Q baseband orthogonal signal matrix, and retain only the high-frequency Doppler frequency shift signal generated by the motion of red blood cells, so as to avoid tissue motion interfering with blood flow interpretation and obtain the filtered blood flow I / Q signal matrix. Step 7: Perform autocorrelation operation on the filtered blood flow I / Q signal matrix, and calculate the average blood flow velocity (i.e., blood flow velocity matrix), blood flow direction (i.e., blood flow direction matrix), and velocity variance (i.e., velocity variance matrix) of each sampling point by using the phase difference between adjacent echoes. Step 8: Map blood flow velocity to color coding; Based on priority determination, effective blood flow pixels are overlaid with color, while grayscale is preserved in areas without blood flow. A color Doppler fused image is obtained by reducing image flicker noise through spatiotemporal smoothing filtering.

5. The thrombus-assisted identification method based on blood flow monitoring according to claim 4, characterized in that, Specifically, the extent of the vascular lumen is located from color flow overlay images, and regions of interest suspected of causing thrombosis are preliminarily screened, including... Step 9: Crop the device parameter bar and text annotations from the image edges of the color Doppler fused image; Perform color space conversion from RGB to HSV, separating the color blood flow component from the grayscale background component; Pixel spacing normalization and contrast adaptive enhancement are completed to eliminate parameter differences between different devices and obtain a standardized dual-channel image; Step 10: Use a semantic segmentation network to perform pixel-level segmentation on the standardized dual-channel image, accurately distinguishing three types of regions: blood vessel wall, blood vessel lumen, and surrounding soft tissue, to obtain blood vessel wall boundary masks, blood vessel lumen region masks, and background tissue masks. Step 11: Based on the vessel wall boundary mask, vessel lumen region mask, and background tissue mask, mark the region with the preset condition of "no color blood flow signal filling and the presence of grayscale abnormal echo"; filter out the normal anatomical structures of vascular valves and venous valves through morphological priors, and generate a set of suspected thrombus candidate region masks to obtain the thrombus candidate region mask.

6. The thrombus-assisted identification method based on blood flow monitoring according to claim 5, characterized in that, Specifically, three types of features are extracted from the region of interest: morphology, texture, and hemodynamics. Step 12: Calculate the thrombus area, major axis / minor axis, sphericity, boundary smoothness, attachment base width to the vessel wall, and degree of lumen dilation for the grayscale image region corresponding to the candidate ROI; Echo characteristics: Statistical analysis of grayscale mean, standard deviation, and skewness within the region to determine low / equal / high echo type; Texture features: Parameters such as contrast, energy, homogeneity, and entropy are extracted using the Gray-Level Co-occurrence Matrix (GLCM) to reflect the internal homogeneity of the thrombus, ultimately yielding a grayscale morphological texture feature vector. Step 13: Calculate the proportion of the blood flow filling defect area to the lumen and the residual effective lumen area; extract the peak flow velocity at the stenosis, the flow velocity ratio before and after the stenosis, the resistance index RI, and the pulsatility index PI; identify indirect signs such as collateral circulation around the lesion to obtain the hemodynamic feature vector; Step 14: Calculate the rate of change of lumen cross-sectional area and the rate of change of blood flow filling before and after pressurization, and quantify the compressibility of the blood vessel to obtain the characteristic parameters of lumen compressibility.

7. A thrombus-assisted identification method based on blood flow monitoring according to claim 6, characterized in that, Specifically, thrombus identification results are obtained based on three types of features: morphology, texture, and hemodynamics. Step 15: Input the grayscale and blood flow fusion image into a multimodal CNN network, and output the pixel-level thrombus segmentation mask and thrombus staging probability, i.e., the thrombus discrimination result. The result includes normal, suspected thrombus, confirmed thrombus, confidence level, and accurate thrombus segmentation mask. Step 16: Based on echo intensity, motion patterns, and morphological characteristics, easily confused structures such as vascular valves, calcified plaques, acoustic artifacts, blood flow aliasing artifacts, and ultrasound side-lobe artifacts are excluded to obtain the final thrombus identification result that eliminates false evidence.

8. The thrombus-assisted identification method based on blood flow monitoring according to claim 7, characterized in that, Specifically, the thrombus identification results are used for quantitative parameter calculation and visualization report generation, including: Step 17: Calculate the thrombus length, maximum cross-sectional area, and volume; Calculate the stenosis rate of vessel diameter and the stenosis rate of vessel area; assess the occlusion grade; By combining echo characteristics to assist in the determination of thrombus staging, a set of structured quantitative thrombus parameters is obtained; Step 18: Overlay the thrombus boundary, measurement scale, and key numerical annotations onto the original image; summarize all parameters to generate a standardized diagnostic report.

9. A thrombus-assisted identification device based on blood flow monitoring, characterized in that, The apparatus comprising the steps of the method according to any one of claims 1-8, wherein the apparatus includes the following: The data acquisition module is used to acquire Doppler ultrasound data at the target location of the venous or arterial pothole; The first calculation module is used to preprocess the Doppler ultrasound data to obtain a two-dimensional RF data matrix; The second calculation module is used to convert the two-dimensional RF data matrix into a grayscale anatomical image and a color blood flow superimposed image, i.e., a color Doppler fused image. The ROI calculation module is used to locate the range of blood vessel lumen from color blood flow overlay images and initially screen out regions of interest suspected of thrombosis. The feature extraction module is used to extract three types of features from the region of interest: morphology, texture, and hemodynamics. The thrombus identification module is used to identify thrombi based on three types of features: morphology, texture, and hemodynamics. The visualization module is used to perform quantitative parameter calculations and generate visualization reports on the thrombus identification results.

Images