Variable-resolution ultrasonic imaging method based on bionic vision

Abstract

The invention discloses a variable-resolution ultrasonic imaging method based on bionic vision, which comprises the following steps: firstly, acquiring ultrasonic image data of a patient, then carrying out preprocessing and feature extraction on the acquired ultrasonic image data, then simulating human vision to carry out dynamic resolution adjustment on the preprocessed ultrasonic image data, and outputting. And then, controlling imaging parameters of ultrasonic imaging according to a dynamic resolution adjustment result to realize variable-resolution imaging, and generating a final ultrasonic imaging result. According to the variable-resolution ultrasonic imaging method based on bionic vision, by simulating the dynamic focusing capacity of a human vision system and combining the ultrasonic imaging technology, high-resolution imaging of a key area can be achieved, meanwhile, low-resolution imaging is adopted for a non-key area, and therefore the purposes of reducing the calculation complexity, improving the calculation efficiency and reducing the calculation cost are achieved. Therefore, the high-resolution reconstruction efficiency of the ultrasonic image is improved.

Description

Technical Field

[0001] The present invention relates to an ultrasonic imaging method, specifically a variable-resolution ultrasonic imaging method based on bionic vision, belonging to the technical fields of medical image processing and computer vision. Background Technique

[0002] Ultrasonic imaging uses an ultrasonic beam to scan the human body, and obtains images of internal organs by receiving and processing reflected signals. Ultrasonic imaging methods are often used to determine the position, size, and shape of organs, determine the scope and physical properties of lesions, provide anatomical diagrams of some glandular tissues, distinguish the normality and abnormality of fetuses, and are widely used in ophthalmology, obstetrics and gynecology, cardiovascular system, digestive system, and urinary system.

[0003] Common ultrasonic instruments for ultrasonic imaging include A-mode (amplitude modulation type), M-mode (bright spot scanning type), B-mode (brightness modulation type), color Doppler ultrasound (color ultrasound), etc. B-mode display is developed on the basis of A-mode and M-mode display technologies. It changes the amplitude modulation display of A-mode to brightness modulation display, and the brightness changes with the size of the echo signal, reflecting the two-dimensional sectional tomographic image of human tissues. It is a gray-scale black-and-white imaging with limited inspection content and relatively low inspection cost. Color Doppler ultrasound is an ultrasonic diagnostic technology that uses the Doppler principle to process the collected blood flow information with colors such as red, blue, and green and brightness and then superimposes it on the B-mode ultrasound image. The inspection scope of color ultrasound is wider than that of B-ultrasound, and the content is more detailed and clear. Therefore, the inspection cost is relatively expensive. According to the different display spaces, color Doppler ultrasound can be divided into two-dimensional, three-dimensional, and four-dimensional types. Two-dimensional color ultrasound only shows sectional and tomographic images and is the most widely used in clinical applications. Three-dimensional color ultrasound generally shows a three-dimensional structure stereogram, while four-dimensional color ultrasound can present real-time dynamic three-dimensional image information on the basis of three-dimensional color ultrasound. Three-dimensional and four-dimensional color ultrasounds are mainly used for obstetric examinations to exclude the situation of congenital dysplasia of fetuses.

[0004] Ultrasound images are different from medical images such as CT images and PET images. Since the propagation process of ultrasonic waves inside human tissues is quite complex, the reflection, refraction, scattering and other phenomena generated when ultrasonic signals pass through the surface of human tissues are the results of the combined influence of multiple mechanisms. Therefore, there are obvious differences in the contrast and clarity of images collected by different devices or different doctors. Moreover, low resolution and low contrast are the biggest drawbacks of ultrasound images, which bring many inconveniences to clinical diagnosis: on the one hand, the diagnostic results of ultrasound images are often affected by the doctor's clinical experience or their own professional level; on the other hand, ultrasound images with low resolution and low contrast are extremely prone to missed diagnosis. Take the obstetric examination of whether the fetus has congenital dysplasia problems as an example. For local deformity symptoms such as adhesion of the external ear of the fetus and adhesion of the fingers of the fetus, it is often difficult to directly identify them through ultrasound images with low resolution and low contrast. Therefore, how to use image super-resolution reconstruction technology to reconstruct and restore ultrasound images for clinical diagnosis has attracted more and more attention from scholars in the industry.

[0005] Image super-resolution reconstruction technology aims to restore a noisy, blurred low-resolution image into a higher-resolution image that is clearer, has richer details, and less noise through a specific algorithm. Traditional super-resolution reconstruction algorithms include interpolation-based super-resolution algorithms and reconstruction-based super-resolution algorithms. The interpolation-based super-resolution algorithm obtains the gray value of the pixel at the interpolation point after magnification by calculating the gray value of the image, so as to realize the super-resolution reconstruction of the image. Classic interpolation algorithms include nearest neighbor interpolation, bilinear interpolation, bicubic interpolation, etc. The advantages of these algorithms are simple calculation and easy operation, but the effect is average and cannot meet the requirements of high-quality images. The reconstruction-based super-resolution algorithm predicts the high-resolution signal based on the low-resolution input signal according to the principles of uniform and non-uniform sampling, so as to realize the super-resolution reconstruction of the image. This type of algorithm is mainly divided into frequency domain method and spatial domain method. Compared with the interpolation method, these methods have better effects, but are complex in design and high in computational complexity, and have limited application ranges. Traditional super-resolution reconstruction algorithms, whether they are interpolation-based super-resolution algorithms or reconstruction-based super-resolution algorithms, perform super-resolution reconstruction on the entire image. However, the focus of the lesion is often within a local range. Take observing whether the external ear of the fetus is adhered as an example. It is only necessary to perform super-resolution reconstruction on the local range where the external ear of the fetus is located in the image, and there is no need to perform super-resolution reconstruction on other non-concerned areas in the image. Therefore, the traditional super-resolution reconstruction algorithm that performs super-resolution reconstruction on the entire image will not only complicate the calculation, but also cause waste of resources. Summary of the Invention

[0006] In view of the problems existing in the above-mentioned prior art, the present invention provides a variable-resolution ultrasonic imaging method based on bionic vision, which can realize local high-resolution reconstruction of different regions of ultrasonic images according to different focuses of attention, thereby reducing the computational complexity and improving the computational efficiency, so as to improve the efficiency of high-resolution reconstruction of ultrasonic images.

[0007] To achieve the above object, the variable-resolution ultrasonic imaging method based on bionic vision specifically includes the following steps:

[0008] Step1, data acquisition: Scan through an ultrasonic instrument to obtain ultrasonic image data;

[0009] Step2, data preprocessing: First, denoise and enhance the contrast of the obtained ultrasonic image to improve the image quality; then set the key area including the area of interest and optimize the image features to prepare for feature extraction;

[0010] Step3, feature extraction: Extract features from the preprocessed ultrasonic image to identify the feature of the area of interest in the key area;

[0011] Step4, dynamic resolution adjustment: According to the result of feature extraction, apply the principle of bionic vision to dynamically adjust the visual focus, specifically as follows:

[0012] Introduce integral and Gaussian weight functions to optimize the resolution adjustment. The mathematical expression for dynamically adjusting the imaging resolution is as follows:

[0013]

[0014] In the formula: R(x,y) represents the resolution at the position (x,y) in the image, where x and y represent the positions in the horizontal and vertical directions of the two-dimensional coordinate system respectively;

[0015] R max represents the maximum resolution that the system can provide. When a certain area in the image is identified as the key area, the resolution of this area is set to R max ;

[0016] Rkey represents a set containing all the images identified as key areas;

[0017] (x,y)∈Rkey conditional statement is used to judge whether the coordinate (x,y) belongs to the key area set Rkey. If the condition is true, the high resolution R max is applied. If the condition is false, the low resolution Rmin is applied;

[0018] C(x,y) represents the local contrast enhancement factor;

[0019] D(x, y) represents the depth dependence factor;

[0020] T(x, y) represents the time variation factor;

[0021] S(x, y) represents the spatial frequency modulation factor;

[0022] O(x, y) represents the tissue type identification factor;

[0023] B(x, y) represents the hemodynamic factor;

[0024] g(d, θ) represents a function that adjusts the resolution according to the distance and angle of the point (x', y') to the nearest key area, and is defined as:

[0025]

[0026] where β is a constant used to control the influence degree of the angle and distance on the resolution, θ is the angle of the point (x', y') to the nearest key area, and d threshold is a threshold used to define the influence range of the key area;

[0027] Z represents the normalization constant that ensures the sum of the integrals is 1;

[0028] λ represents the attenuation coefficient that controls the weight distribution of the integral terms;

[0029] represents the Gaussian weight function, which weights the contribution of each pixel point according to the distance of the point (x', y') to the point (x, y);

[0030] Step 5, Imaging control: Control the imaging parameters of ultrasonic imaging according to the results of the dynamic resolution adjustment, realize variable resolution imaging, and apply the ultrasonic image reconstruction algorithm and the dynamic resolution adjustment algorithm in Step 4 to control the imaging process to ensure high-resolution imaging of the key area;

[0031] Step 6, Result display: Display the final ultrasonic imaging result and highlight the concerned parts of the key area.

[0032] Furthermore, in Step 2, when optimizing the image features, a Gaussian filter is used for denoising, which can be specifically expressed as:

[0033]

[0034] In the formula: I(x ′ , y′) is the original graph; G is the Gaussian kernel;

[0035] Histogram equalization is used for contrast enhancement, which can be specifically expressed as:

[0036]

[0037] Where: L is the number of gray levels; n is the total number of pixels; CDF is the cumulative distribution function.

[0038] Furthermore, in Step 3, edge detection and morphological operation methods are used to identify the key regions, specifically as follows:

[0039] Use the Canny edge detection algorithm for edge detection, expressed as:

[0040] G(x,y) = max(|I x |,|I y |)

[0041] Where: I x and I y are the gradients of the image in the x and y directions respectively;

[0042] During morphological operation, erosion and dilation operations are used to highlight specific features, expressed as:

[0043]

[0044] Where: A is the image; B is the structuring element; Z 2 is the two-dimensional integer space, representing all possible pixel positions in the image and the structuring element; z is a point in the two-dimensional integer space Z 2 and represents a pixel position in the image; ω is a point in the structuring element B and represents a pixel position in the structuring element; B z is the translation of the structuring element B at point z, B z = {z + ω|ω ∈ B}.

[0045] Compared with the prior art, the variable-resolution ultrasonic imaging method based on bionic vision can achieve high-resolution imaging of key regions by simulating the dynamic focusing ability of the human visual system and combining ultrasonic imaging technology, while using low-resolution imaging for non-key regions, thereby reducing the computational complexity and improving the computational efficiency, and thus achieving the improvement of the high-resolution reconstruction efficiency of ultrasonic images. Brief Description of the Drawings

[0046] Figure 1 is the flowchart of the present invention;

[0047] Figure 2 is the comparison diagram before and after the high-resolution reconstruction of the ultrasonic image of the B-ultrasound image for examining whether there is congenital adhesion of the external ear of the fetus by using the method of the present invention, where (a) is the original B-ultrasound image and (b) is the processed B-ultrasound image. Detailed Embodiments

[0048] As Figure 1 shown, the variable-resolution ultrasonic imaging method based on bionic vision first collects ultrasonic image data of a patient, then preprocesses and extracts features from the collected ultrasonic image data, then simulates human vision to perform dynamic resolution adjustment on the preprocessed ultrasonic image data and output it, and then controls the imaging parameters of ultrasonic imaging according to the result of dynamic resolution adjustment to achieve variable-resolution imaging and generate the final ultrasonic imaging result.

[0049] The following takes the obstetric examination for excluding the presence of congenital adhesion of the external ear of a fetus as an example to further illustrate the present invention.

[0050] Step1, data acquisition: Scan the abdomen of a pregnant woman through an ultrasonic instrument to obtain ultrasonic image data of the fetus.

[0051] Step2, data preprocessing: First, denoise and enhance the contrast of the obtained ultrasonic image of the fetus to improve the image quality; then use the external ear as the key area to optimize the image features to prepare for feature extraction.

[0052] When optimizing the image features, use a Gaussian filter for denoising, which can be specifically expressed as:

[0053]

[0054] In the formula: I(x ′ , y′) is the original graph; G is the Gaussian kernel.

[0055] Use histogram equalization for contrast enhancement, which can be specifically expressed as:

[0056]

[0057] In the formula: L is the number of gray levels; n is the total number of pixels; CDF is the cumulative distribution function.

[0058] Step3, feature extraction: Extract features from the preprocessed ultrasonic image of the fetus to identify the morphology and the position features of the external ear of the fetus.

[0059] Use methods such as edge detection and morphological operations to identify the key area, specifically as follows:

[0060] Use the Canny edge detection algorithm for edge detection, which can be expressed as:

[0061] G(x, y) = max(|i x |, |i y |)

[0062] In the formula: I x , I y are the gradients of the image in the x and y directions respectively.

[0063] During morphological operations, erosion and dilation operations are used to highlight specific features, which can be expressed as:

[0064]

[0065] where: A is the image; B is the structuring element; Z 2 is a two-dimensional integer space representing all possible pixel positions in the image and the structuring element; z is a point in the two-dimensional integer space Z 2 and represents a pixel position in the image; ω is a point in the structuring element B and represents a pixel position in the structuring element; B z is the translation of the structuring element B at point z, that is, B z = {z + ω | ω ∈ B}.

[0066] Step4, Dynamic resolution adjustment: According to the results of feature extraction, identify the key area of the fetal external ear. Apply the principle of bionic vision to dynamically adjust the visual focus, extract features from the key area of the external ear, and at the same time reduce the processing load of non-key areas, and then output.

[0067] Introduce integral and Gaussian weight functions to optimize the resolution adjustment, dynamically adjust the imaging resolution, use high-resolution imaging for the key area of the external ear, and use low-resolution imaging for other areas. The mathematical expression for dynamically adjusting the imaging resolution is as follows:

[0068]

[0069] where: R(x, y) represents the resolution at position (x, y) in the image, which is a two-dimensional coordinate system, where x and y represent positions in the horizontal and vertical directions respectively;

[0070] R max represents the maximum resolution that the system can provide. When a certain area in the image is identified as a key area, the resolution of this area will be set to R max to ensure the clarity and details of key information;

[0071] Rkey represents a set containing all the identified key image areas, which usually contain information crucial for diagnosis or analysis, such as the state of the fetal external ear, etc.;

[0072] (x, y) ∈ Rkey conditional statement is used to determine whether the coordinate (x, y) belongs to the key area set Rkey. If the condition is true, that is, this position belongs to the key area, then the high resolution R max is applied. If the condition is false, that is, this position does not belong to the key area, then the low resolution Rmin is applied;

[0073] C(x, y) represents the local contrast enhancement factor, which can be adjusted according to the local contrast of each point in the image to enhance the details in the image;

[0074] D(x, y) represents the depth-dependent factor, which can adjust the resolution according to the depth of each point in the image considering the attenuation of ultrasonic waves at different depths;

[0075] T(x, y) represents the time-varying factor, which can be adjusted according to the change of the image over time to meet the requirements of dynamic imaging;

[0076] S(x, y) represents the spatial frequency modulation factor, which can be adjusted according to the spatial frequency of each point in the image to enhance the details and contrast in the image;

[0077] O(x, y) represents the tissue type recognition factor, which can be adjusted according to the tissue type of each point in the image to meet the imaging requirements of different tissue types;

[0078] B(x, y) represents the hemodynamic factor, which can be adjusted according to the hemodynamic characteristics of each point in the image to enhance the detection and analysis of blood flow signals;

[0079] g(d, θ) represents a function that adjusts the resolution according to the distance and angle of the point (x', y') to the nearest key region. This function can be defined as:

[0080]

[0081] where β is a constant used to control the influence degree of the angle and distance on the resolution, θ is the angle of the point (x', y') to the nearest key region, and d threshold is a threshold used to define the influence range of the key region;

[0082] Z represents the normalization constant to ensure that the sum of the integrals is 1;

[0083] λ represents the attenuation coefficient to control the weight distribution of the integral term;

[0084] represents the Gaussian weight function, which weights the contribution of each pixel point according to the distance of the point (x', y') to the point (x, y).

[0085] By introducing variable-resolution imaging with integrals and Gaussian weight functions, the key region can be more accurately located while maintaining the overall quality and details of the image.

[0086] Step 5, Imaging Control: Control the imaging parameters of ultrasonic imaging according to the results of dynamic resolution adjustment to achieve variable-resolution imaging. Apply ultrasonic image reconstruction algorithms and the dynamic resolution adjustment algorithm in Step 4 to control the imaging process and ensure high-resolution imaging of key areas.

[0087] The ultrasonic image reconstruction algorithm is a process of converting the signals collected by ultrasonic sensors into images. Commonly used ultrasonic image reconstruction algorithms include delay-and-sum algorithm, deconvolution algorithm, algebraic reconstruction technique, and deep learning-based reconstruction algorithms, etc. In this embodiment, the deconvolution algorithm is used for ultrasonic image reconstruction to improve the resolution of ultrasonic images, and the original signal is restored by removing the convolution effect in the signal.

[0088] Assume that the collected signal s(n) is the convolution result of the target signal o(n) and the convolution kernel h(n), plus the noise n(n), then we have:

[0089] s(n) = o(n) * h(n) + n(n)

[0090] Using the known s(n) and h(n), restore o(n) by deconvolution operation.

[0091] Step 6, Result Display: Display the final ultrasonic imaging result and highlight the key area of the outer ear.

[0092] Regarding the situation of whether the fetus has congenital adhesion of the outer ear, the comparison charts before and after high-resolution reconstruction of ultrasonic images using this bionic vision-based variable-resolution ultrasonic imaging method are as Figure 2 shown. As Figure 2 can be seen, after being processed by this bionic vision-based variable-resolution ultrasonic imaging method, the image quality is significantly improved. The clarity of the image increases, the details and edges are sharper, the gray-scale restoration is more real and natural, the saturation and contrast are optimized, the noise is effectively reduced, and the image is cleaner; in addition, the dynamic range expansion enables the details of both the high-light and shadow parts to be retained, and the resolution is also improved. In particular, the details of the outer ear in the center of the image can be clearly distinguished.

[0093] This bionic vision-based variable-resolution ultrasonic imaging method can achieve high-resolution imaging of key areas by simulating the dynamic focusing ability of the human visual system and combining ultrasonic imaging technology. At the same time, low-resolution imaging is adopted for non-key areas, thereby reducing the computational complexity and improving the computational efficiency, so as to improve the efficiency of high-resolution reconstruction of ultrasonic images.

Claims

1. A variable-resolution ultrasonic imaging method based on bionic vision, characterized in that, Specifically, it includes the following steps: Step1, Data acquisition: Scan through an ultrasonic instrument to obtain ultrasonic image data; Step2, Data preprocessing: First, denoise and enhance the contrast of the obtained ultrasonic image to improve the image quality; then set the key area including the area of interest, and optimize the image features to prepare for feature extraction; Step3, Feature extraction: Extract features from the preprocessed ultrasonic image to identify the features of the area of interest in the key area; Step4, Dynamic resolution adjustment: According to the results of feature extraction, apply the principle of bionic vision to dynamically adjust the visual focus, specifically as follows: Introduce the integral and Gaussian weight functions to optimize the resolution adjustment. The mathematical expression for dynamically adjusting the imaging resolution is as follows: In the formula: R(x, y) represents the resolution at the position (x, y) in the image, where x and y represent the positions in the horizontal and vertical directions of the two-dimensional coordinate system respectively; R max represents the maximum resolution that the system can provide. When a certain area in the image is recognized as a key area, the resolution of this area is set to R max ; Rkey represents a set containing all the images identified as key areas; (x,y) ∈ Rkey conditional statement is used to determine whether the coordinates (x,y) belong to the key region set Rkey. If the condition is true, high-resolution R is applied max , if the condition is false, low-resolution Rmin is applied; C(x, y) represents the local contrast enhancement factor; D(x, y) represents the depth-dependent factor; T(x, y) represents the time-varying factor; S(x, y) represents the spatial frequency modulation factor; O(x, y) represents the tissue type recognition factor; B(x, y) represents the hemodynamic factor; g(d, θ) represents a function that adjusts the resolution according to the distance and angle of the point (x', y') to the nearest key area, defined as: where β is a constant for controlling the influence degree of the angle and distance on the resolution, θ is the angle from the point (x', y') to the nearest key area, and d threshold is a threshold for defining the influence range of the key area; Z represents the normalization constant that ensures the sum of the integrals is 1; λ represents the attenuation coefficient that controls the weight distribution of the integral term; represents a Gaussian weight function that weights the contribution of each pixel point according to the distance from the point (x', y') to the point (x, y); Step5, Imaging control: Control the imaging parameters of ultrasonic imaging according to the results of dynamic resolution adjustment to achieve variable-resolution imaging. Apply the ultrasonic image reconstruction algorithm and the dynamic resolution adjustment algorithm in Step4 to control the imaging process and ensure high-resolution imaging of the key area; Step6, Result display: Display the final ultrasonic imaging result and highlight the area of interest in the key area.

2. The method for variable-resolution ultrasonic imaging based on bionic vision according to claim 1, wherein In Step2, when optimizing the image features, a Gaussian filter is used for denoising, which can be specifically expressed as: where: I(x ′ , y′) is the original graph; G is the Gaussian kernel; Histogram equalization is used for contrast enhancement, which can be specifically expressed as: In the formula: L is the number of gray levels; n is the total number of pixels; CDF is the cumulative distribution function.

3. The variable-resolution ultrasonic imaging method based on bionic vision according to claim 1, characterized in that In Step3, edge detection and morphological operation methods are used to identify the key area, specifically as follows: The Canny edge detection algorithm is used for edge detection, expressed as: G(x,y) = max(|I x |,|I y |) Where: I x and I y are the gradients of the image in the x and y directions, respectively; During morphological operations, erosion and dilation operations are used to highlight specific features, expressed as: where: A is an image; B is a structuring element; Z 2 is a two-dimensional integer space representing all possible pixel positions in the image and the structuring element; z is a point in the two-dimensional integer space Z 2 and represents a pixel position in the image; ω is a point in the structuring element B, representing a pixel position in the structuring element; B z is the translation of the structuring element B at the point z, B z ={z + ω|ω ∈ B}.

4. The method for variable-resolution ultrasonic imaging based on bionic vision according to claim 1, wherein In Step5, the ultrasonic image reconstruction algorithm applied is the deconvolution algorithm, which restores the original signal by removing the convolution effect in the signal, specifically as follows: Assume that the collected signal s(n) is the convolution result of the target signal o(n) and the convolution kernel h(n), plus the noise n(n), then there is: s(n) = o(n) * h(n) + n(n) Based on the known s(n) and h(n), use deconvolution operation to restore o(n).

Images