A method, device and medium for super-resolution imaging of microvessels based on point spread function and deconvolution

Through the microvascular super-resolution imaging method based on point diffusion function and deconvolution, the problems of insufficient accuracy and low image resolution of microvascular super-resolution imaging in the prior art are solved, and higher image resolution and noise suppression effects are achieved. Suitable for microvascular surgery and diagnosis of cardiovascular diseases.

CN114820360BActive Publication Date: 2024-11-01CENT SOUTH UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210420008.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-21
Publication Date
2024-11-01
Estimated Expiration
2042-04-21

AI Technical Summary

Technical Problem

The existing microvascular super-resolution imaging technology has problems such as insufficient accuracy, low image resolution and high professional level requirements for medical staff in domestic applications, and the image background noise suppression effect is poor.

Method used

Using the microvascular super-resolution imaging method based on point diffusion function and deconvolution, the ultrasonic echo signal is obtained and processed, signal preprocessing, beam synthesis, quadrature demodulation, and image postprocessing, including decoherence, interpolation, and microbubble Separation, point diffusion function deconvolution and positioning microbubble centroids, using Richardson-Lucy algorithm and two-dimensional normalized cross-correlation processing to improve image resolution and accuracy.

Benefits of technology

It significantly improves the accuracy and image resolution of microvascular imaging, reduces background noise, improves microbubble detection accuracy and efficiency, enhances the signal-to-noise ratio of the image, and improves the positioning accuracy and contrast of the microbubble centroid.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114820360B_ABST
    Figure CN114820360B_ABST
Patent Text Reader

Abstract

The present invention discloses a microvascular super-resolution imaging method, device and medium based on point spread function and deconvolution, the method: obtaining contrast-enhanced ultrasonic echo signals, performing signal pre-processing, beam synthesis and orthogonal demodulation to obtain ultrasonic images; performing image post-processing on the ultrasonic images, and superimposing to obtain microvascular super-resolution images; the image post-processing includes: decoherence, interpolation, microbubble separation, denoising, point spread function deconvolution, and positioning microbubble centroid; microbubble separation includes: using the imaging result of an ultrasonic microbubble model as the point spread function PSF1, performing two-dimensional normalized cross-correlation calculation on all sub-graphs of the image, and taking the sub-graph with the largest cross-correlation coefficient as the point spread function PSF2; performing two-dimensional normalized cross-correlation processing on the image using PSF2, retaining the sub-graph with a coefficient greater than a threshold value T; when the point spread function is deconvolved, PSF2 is used as the blur kernel of the deconvolution. The present invention can not only improve the accuracy and image resolution of imaging, but also effectively suppress the background noise of the image.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of ultrasonic imaging, and in particular relates to a microvascular super-resolution imaging method, equipment and medium based on point spread function and deconvolution. Background Art

[0002] The anatomical relationship of human blood vessels is crucial in clinical surgery. Preoperative exploration of microvessels and clarification of their anatomical structure are of great significance in many medical fields, such as preoperative flap design and postoperative care analysis in plastic surgery, preoperative evaluation of microvascular surgery, and prevention and diagnosis of cardiovascular diseases.

[0003] With the deepening of research and the further improvement of clinical needs, microbubble-enhanced Doppler ultrasound is a newly developed ultrasound imaging improvement technology in recent years. It is an enhanced mode of conventional color Doppler ultrasound imaging and its clinical application is gradually increasing. The imaging advantage of this technology is that it can show smaller blood vessels by enhancing echoes through microbubbles, while also combining the advantages of ultrasound such as real-time, no radiation, safety and environmental protection. Summary of the invention

[0004] In view of the limitations of the existing domestic super-resolution imaging technology for microvessels, such as insufficient reliability and accuracy, high professional requirements for medical staff, and low image resolution, the present invention proposes a microvessel super-resolution imaging method, equipment and medium based on point spread function and deconvolution, which can not only improve the imaging accuracy and image resolution, but also effectively suppress the background noise of the image.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A microvascular super-resolution imaging method based on point spread function and deconvolution, comprising:

[0007] S1: Acquire contrast-enhanced ultrasound echo signals, perform signal pre-processing, beamforming and orthogonal demodulation on them, and obtain N frames of ultrasound images;

[0008] S2: performing image post-processing on N frames of ultrasound images, and finally superimposing the processed N frames of ultrasound images to obtain a microvascular super-resolution image;

[0009] The image post-processing includes: decoherence, interpolation, microbubble separation, denoising, point spread function deconvolution, and positioning of microbubble centroid;

[0010] The microbubble separation comprises: for each frame of ultrasound image, firstly using the imaging result of an ultrasound microbubble model as the point spread function PSF1, performing two-dimensional normalized cross-correlation calculation on all sub-graphs in the ultrasound image, and taking the sub-graph with the largest cross-correlation coefficient as the point spread function PSF2 corresponding to the ultrasound image; then using the point spread function PSF2, further performing two-dimensional normalized cross-correlation processing on the ultrasound image, and retaining the sub-graph with the cross-correlation coefficient greater than a threshold value T in the ultrasound image;

[0011] During point spread function deconvolution, the point spread function PSF2 obtained by microbubble separation is used as the blur kernel for deconvolution of the corresponding ultrasound image.

[0012] Furthermore, a 10 um tungsten wire model was used to simulate the imaging result of the ultrasonic microbubble.

[0013] Furthermore, the calculation formula of the two-dimensional normalized cross-correlation is:

[0014]

[0015] Where (x, y) represents any subgraph S of the ultrasound image to be processed in two-dimensional normalized cross-correlation. x,y The coordinates of the upper left corner pixel of , (i, j) represents the coordinates of any pixel of the point spread function, and (x+i, y+j) represents the sub-image S x,y The coordinates of any pixel in S(x+i,y+j) represent the sub-graph S x,y h(i,j) represents the grayscale value of the pixel with coordinates (i,j) in the point spread function. represents the average gray value of the point spread function, m,n represents the pixel size of the point spread function, that is, the sub-image S x,y The pixel size in ;

[0016] Furthermore, the iterative Richardson-Lucy algorithm is used to perform point spread function deconvolution processing on the ultrasound image, which is expressed as:

[0017]

[0018] Among them, f k (x, y) represents the restored ultrasound image obtained by the k-th iterative deconvolution process, (x, y) represents the pixel coordinates; g(x, y) represents the denoised ultrasound image, and the initial condition is selected as f k=0 (x,y)=g(x,y) for iteration. When k increases, f k+1(x, y) gradually converges, and then the approximate value of the image can be obtained; h(x, y) represents the point spread function, and the negative sign of h(-x, -y) represents the flip of the point spread function h(x, y).

[0019] Furthermore, the signal pre-processing includes signal amplification, AD conversion, filtering and apodization; and the beamforming adopts coherent plane wave recombination and coherence factor weighting.

[0020] Furthermore, the image is decorrelated by singular value decomposition: first, the demodulated N frames of image (x, y, N) are recombined into a Casorati matrix (x×y, N); then, the matrix (x×y, N) is subjected to singular value decomposition to obtain a singular value matrix, and the middle part of the singular values ​​is extracted according to a preset boundary threshold to form a new singular value matrix. Finally, the new singular value matrix is ​​used to perform the inverse operation of the singular value decomposition to obtain a new N-frame image.

[0021] Furthermore, the de-coherent image is interpolated by bilinear interpolation, so that the pixel interval of the image is reduced to 10 um.

[0022] Furthermore, when locating the centroid of the microbubble, the intensity peaks of all isolated sub-images in the deconvoluted ultrasound image are used as the centroid position of one ultrasound microbubble, and only the pixel value of the microbubble centroid position is saved in the isolated sub-image.

[0023] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor implements beamforming and orthogonal demodulation and image post-processing in any one of the above methods.

[0024] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements beamforming and orthogonal demodulation and image post-processing in any of the above methods.

[0025] Beneficial Effects

[0026] 1) In the present invention, a linear array ultrasonic transducer is used to transmit multi-angle, ultra-high pulse repetition rate planar ultrasonic waves to the target imaging area, and receive ultrasonic echo signals, which can effectively track ultrasonic microbubbles flowing along the microvessels, thereby improving the signal-to-noise ratio of the image;

[0027] 2) The introduction of microbubble point spread function PSF2 can greatly improve the detection accuracy and efficiency of microbubbles in two-dimensional normalized cross-correlation processing;

[0028] 3) Using the Richardson-Lucy deconvolution algorithm to restore the image can improve the positioning accuracy of the microbubble centroid, image resolution and contrast, and reduce background noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flowchart of the method according to an embodiment of the present invention;

[0030] Figure 2 This is a flowchart of an algorithm for image post-processing in the method according to an embodiment of the present invention;

[0031] Figure 3 Ultrasound B-mode imaging results of the melanoma area on the mouse leg.

[0032] Figure 4 This is the super-resolution result of microvessels in the melanoma area of ​​the mouse leg. The microbubble point spread function deconvolution calculation was not performed in the image post-processing in step 5.

[0033] Figure 5 This is the super-resolution result of microvessels in the melanoma area of ​​the mouse leg. The tungsten point spread function PSF1 was used to calculate the normalized cross-correlation coefficient in the image post-processing in step 5.

[0034] Figure 6 This is the super-resolution result of microvessels in the melanoma area of ​​the mouse leg. DETAILED DESCRIPTION

[0035] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0036] This embodiment provides a microvascular super-resolution imaging method based on point spread function and deconvolution. Figure 1 As shown, including:

[0037] Step 1, sulfur hexafluoride ultrasonic microbubbles, as a contrast agent that has been approved for clinical use, can enhance the ultrasonic echo signal of blood flow in blood vessels and improve the clarity and resolution of ultrasonic images. Therefore, a low concentration of sulfur hexafluoride ultrasonic microbubble suspension is first injected into the venous blood vessels of the imaging target, wherein the concentration of the ultrasonic microbubble suspension and the injection method should be different according to the imaging target and the imaging area. A specific implementation method of this embodiment for mice is: inject 5mL of 0.9% normal saline into a vial containing sulfur hexafluoride microbubble freeze-dried powder, shake vigorously to obtain a suspension, then extract 0.2mL of the suspension and dilute it 5 times again, and finally inject the low concentration of sulfur hexafluoride microbubble suspension into the mouse blood vessels at one time.

[0038] Step 2: Control the linear array ultrasonic transducer to transmit multi-angle, ultra-high pulse repetition rate planar ultrasonic waves to the target imaging area. The ultrasonic wave transmission angle range is -18° to +18°, the transmission angle interval is 6°, and the pulse repetition rate is 1.4KHZ.

[0039] As the ultrasound microbubbles flow in the blood vessels to the target imaging area, they resonate under the action of the high-frequency ultrasound waves emitted by the ultrasound transducer and emit ultrasound echoes.

[0040] Step 3: The integrated linear array ultrasonic transducer receives the ultrasonic echo signal of the target imaging area and transmits it to the Verasonics Vantage 256 (ultrasound multi-channel research platform) for signal pre-processing, where the signal pre-processing includes signal amplification, AD conversion, filtering, and apodization.

[0041] The aforementioned steps 1-3 are all prior art and are described in this application only to facilitate understanding of the contrast-enhanced ultrasound echo signal initially acquired by the present invention.

[0042] Step 4, Verasonics Vantage 256 transmits the processed signal via the PIC line to the computer host for signal beamforming and orthogonal demodulation, wherein beamforming uses coherent plane wave recombination and coherence factor weighting.

[0043] Step 5, further perform image post-processing on the orthogonal demodulated signal, wherein the image post-processing includes singular value decomposition (decoherence), interpolation, two-dimensional normalized cross-correlation processing (microbubble separation), denoising, point spread function deconvolution, and positioning of microbubble centroid. Figure 2 As shown, specifically including:

[0044] (1) First, the orthogonal demodulated IQ data (x, y, N) are reassembled into a Casorati matrix (x×y, N), and then the singular value decomposition is calculated to obtain a singular value matrix. Since in the singular value matrix, low-order singular values ​​represent static tissue signals and high-order singular values ​​represent noise, in order to extract effective microbubble signals, the singular values ​​in the middle part are selected. Finally, the new singular value matrix is ​​used to perform the inverse operation of the singular value decomposition and obtain a new IQ data set.

[0045] (2) Select bilinear interpolation to reduce the pixel interval of the ultrasound image to 10um.

[0046] (3) After obtaining the point spread function PSF1 of the 10 μm tungsten wire model in advance, calculate the maximum two-dimensional normalized cross-correlation coefficient between it and each frame of the two-dimensional ultrasound image after decoherence and interpolation, then find the corresponding area of ​​the maximum cross-correlation coefficient in the two-dimensional ultrasound image, and calibrate the area as the microbubble point spread function PSF2 of the current calculation frame of the two-dimensional ultrasound image.

[0047] The calculation formula of the two-dimensional normalized cross-correlation between each frame of the interpolated two-dimensional ultrasound image and the point spread function PSF1 is:

[0048]

[0049] Where (x, y) represents any subgraph S of the ultrasound image to be processed in two-dimensional normalized cross-correlation. x,y The coordinates of the upper left corner pixel of , (i, j) represents the coordinates of any pixel of the point spread function, and (x+i, y+j) represents the sub-image S x,y The coordinates of any pixel in S(x+i,y+j) represent the sub-graph S x,y h(i,j) represents the grayscale value of the pixel with coordinates (i,j) in the point spread function. represents the average gray value of the point spread function, m,n represents the pixel size of the point spread function, that is, the sub-image S x,y The pixel size in .

[0050] (4) Select the microbubble point spread function PSF2 as the template image for two-dimensional normalized cross-correlation processing, and calculate its cross-correlation coefficient with each frame of the ultrasound image, and then retain the sub-images in the ultrasound image whose cross-correlation coefficient is greater than the threshold T. The specific calculation method of the two-dimensional normalized cross-correlation coefficient is the same as the calculation formula in step (3), the only difference is that the point spread function is different:

[0051] (5) Select the microbubble point spread function PSF2 as the blur kernel of the Richardson-Lucy deconvolution algorithm. The number of iterations of the deconvolution algorithm is selected according to the imaging target and the imaging area. In this embodiment, 30 iterations are selected for the microvascular imaging of mouse leg melanoma. The specific calculation method of the Richardson-Lucy algorithm is as follows:

[0052]

[0053] Among them, f k (x, y) represents the restored ultrasound image obtained by the k-th iterative deconvolution process, (x, y) represents the pixel coordinates; g(x, y) represents the denoised ultrasound image, and the initial condition is generally selected as f k=0 (x,y)=g(x,y) for iteration. When k increases, f k+1 (x, y) gradually converges, and the approximate value of the image can be obtained. h(x, y) represents the point spread function, and the negative sign of h(-x, -y) represents the flip of the point spread function h(x, y).

[0054] (6) For the deconvolved ultrasound image, the intensity peaks of all isolated sub-images are used as the centroid position of an ultrasound microbubble, and only the pixel value of the microbubble centroid position is saved in the isolated sub-image.

[0055] Step 6, superimposing the N frames of two-dimensional ultrasound images obtained in step 5, and finally obtaining a microvascular super-resolution image.

[0056] from Figure 3-Figure 6 The results can clearly show the tissue structure and morphological characteristics of microvessels, such as origin, termination, diameter, position, course, variation, accompanying conditions, perforators, and relationship with skin flaps. Therefore, the method of the present invention has broad application prospects in the field of microvascular imaging, and can be actually applied to multiple medical fields such as preoperative evaluation of microvascular surgery, prevention and diagnosis of cardiovascular diseases, etc., thereby providing doctors with more reference and richer imaging data in the process of diagnosis and efficacy evaluation of patients, and using it as an auxiliary means for doctors to diagnose diseases, which can greatly improve the accuracy and scientificity of medical diagnosis.

[0057] The present invention also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and wherein when the computer program is executed by the processor, the processor implements beamforming and orthogonal demodulation and image post-processing in the method described in the above embodiment.

[0058] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements beamforming, orthogonal demodulation, and image post-processing in the method described in the above embodiment.

[0059] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A microvascular super-resolution imaging method based on point spread function and deconvolution, characterized in that: include: S1: Acquire contrast-enhanced ultrasound echo signals, perform signal pre-processing, beamforming and orthogonal demodulation on them, and obtain N frames of ultrasound images; S2: performing image post-processing on N frames of ultrasound images, and finally superimposing the processed N frames of ultrasound images to obtain a microvascular super-resolution image; The image post-processing includes: decoherence, interpolation, microbubble separation, denoising, point spread function deconvolution, and positioning of microbubble centroid; The microbubble separation comprises: for each frame of ultrasound image, firstly using the imaging result of an ultrasound microbubble model as the point spread function PSF1, performing two-dimensional normalized cross-correlation calculation on all sub-graphs in the ultrasound image, and taking the sub-graph with the largest cross-correlation coefficient as the point spread function PSF2 corresponding to the ultrasound image; then using the point spread function PSF2, further performing two-dimensional normalized cross-correlation processing on the ultrasound image, and retaining the sub-graph with the cross-correlation coefficient greater than a threshold value T in the ultrasound image; During point spread function deconvolution, the point spread function PSF2 obtained by microbubble separation is used as the blur kernel for deconvolution of the corresponding ultrasound image; The calculation formula of the two-dimensional normalized cross-correlation is: Where (x, y) represents any subgraph S of the ultrasound image to be processed in two-dimensional normalized cross-correlation. x,y The coordinates of the upper left corner pixel of , (i, j) represents the coordinates of any pixel of the point spread function, and (x+i, y+j) represents the sub-image S x,y The coordinates of any pixel in S(x+i,y+j) represent the sub-graph S x,y h(i,j) represents the grayscale value of the pixel with coordinates (i,j) in the point spread function. represents the average gray value of the point spread function, m,n represents the pixel size of the point spread function, that is, the sub-image S x,y The pixel size in ; The iterative Richardson-Lucy algorithm is used to perform point spread function deconvolution processing on the ultrasound image, which is expressed as: Among them, f k (x, y) represents the restored ultrasound image obtained by the k-th iterative deconvolution process, (x, y) represents the pixel coordinates; g(x, y) represents the denoised ultrasound image, and the initial condition is selected as f k=0 (x, y) = g(x, y) is iterated, h(x, y) represents the point spread function, and h(-x, -y) takes a negative sign to represent the flipping of the point spread function h(x, y).

2. The method according to claim 1, characterized in that A 10 um tungsten wire model was used to simulate the imaging results of the ultrasound microbubble model.

3. The method according to claim 1, characterized in that The signal pre-processing includes signal amplification, AD conversion, filtering and apodization; the beam synthesis adopts coherent plane wave compounding and coherence factor weighting.

4. The method according to claim 1, characterized in that: The image is decorrelated by singular value decomposition: first, the demodulated N frames of image are reassembled into a Casorati matrix; then, the matrix (x×y, N) is subjected to singular value decomposition to obtain a singular value matrix, and the middle singular values ​​are extracted according to the preset boundary threshold to form a new singular value matrix; finally, the new singular value matrix is ​​used to perform the inverse operation of singular value decomposition to obtain a new N-frame image.

5. The method according to claim 1, characterized in that The de-coherent image is interpolated using bilinear interpolation to reduce the pixel interval of the image to 10um.

6. The method according to claim 1, characterized in that When locating the centroid of microbubbles, the intensity peaks of all isolated sub-images in the deconvoluted ultrasound image are used as the centroid position of one ultrasound microbubble, and only the pixel value of the microbubble centroid position is saved in the isolated sub-image.

7. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor implements beamforming and orthogonal demodulation and image post-processing in the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the beamforming, orthogonal demodulation and image post-processing in the method according to any one of claims 1 to 6 are implemented.