Ultrasonic rapid imaging method combining difference compensator and adaptive denoising

Abstract

The invention discloses an ultrasonic rapid imaging method combining a difference compensator and adaptive denoising, and the method comprises the following steps: 1, building an ultrasonic imaging degradation model based on the spatial focusing and diffusion characteristics of ultrasonic sound field propagation, extracting the PSF information of local and out-of-focus imaging regions, converting the PSF information into a frequency domain through Fourier transform, and obtaining a frequency domain; based on ideal imaging conditions, analyzing frequency domain difference between local and overall PSF, designing a frequency domain difference compensator, expanding high-frequency information, and optimizing overall imaging performance; and 2, designing a signal-to-noise ratio self-estimation module, analyzing the signal-to-noise ratio condition of the input image, and constructing a self-adaptive parameterized denoising module in combination with the information of the decomposition level and the sub-band position, so as to effectively remove noise and enhance the image contrast, and obtain an ultrasonic reconstructed image with improved contrast. The method is efficient in calculation, does not need complex parameter adjustment, can quickly optimize the ultrasonic imaging quality, is suitable for traditional and portable ultrasonic equipment, and has wide application value in the field of medical imaging.

Description

TECHNICAL FIELD

[0001] The application relates to an ultrasonic imaging post-processing algorithm, in particular to an ultrasonic fast imaging method combining a difference compensator and adaptive denoising. BACKGROUND

[0002] Due to the advantages of low cost, portability, simple operation and non-invasiveness, ultrasonic imaging has become one of the most commonly used tools in medical diagnosis and is widely used in the fields of heart, abdomen, obstetrics, muscles and bones and blood vessels. However, the current ultrasonic equipment still has the limitation of relatively low resolution and contrast, which seriously restricts its application in fine diagnosis. In order to improve the imaging quality, researchers have explored image reconstruction technologies such as deep learning, beam forming and deconvolution, and have achieved certain results. However, even with the most advanced equipment, there is an inherent trade-off between the reconstruction speed, flexibility and imaging quality of ultrasonic imaging. High-resolution imaging often requires longer processing time, affecting real-time performance; and speeding up the processing speed may lead to a decrease in image quality. This challenge is particularly prominent in low-performance miniaturized devices, such as wearable and portable ultrasonic devices, which are limited by hardware resources and computing power, and there is an urgent need for an efficient method that can improve image quality without significantly increasing computational burden. Therefore, developing a practical technology that can quickly and automatically improve the quality of ultrasonic imaging is not only extremely challenging, but also has important clinical application value. SUMMARY

[0003] The application provides an ultrasonic fast imaging method combining a difference compensator and adaptive denoising to solve the mutual compromise of reconstruction time, imaging performance and flexibility in ultrasonic imaging. The method is computationally efficient and does not require complex parameter adjustment, can quickly optimize the quality of ultrasonic imaging, solves the limitations of ultrasonic equipment in resolution, signal-to-noise ratio and contrast, and is suitable for traditional and portable ultrasonic equipment, and has wide application value in the field of medical imaging.

[0004] The purpose of the application is achieved by the following technical solutions:

[0005] An ultrasonic fast imaging method combining a difference compensator and adaptive denoising, comprising the following steps:

[0006] Step one: based on the spatial focusing and diffusion characteristics of ultrasonic sound field propagation, an ultrasonic imaging degradation model is established, and the PSF information of the focusing and overall imaging area is extracted, the Fourier transform is converted to the frequency domain, and the local focusing area is taken as the target ideal imaging condition, the frequency domain difference between the local focusing area and the overall PSF is analyzed, a frequency domain difference compensator is designed, the high frequency information is expanded, the overall imaging performance is optimized, an image with enhanced resolution is obtained, and the specific steps are as follows:

[0007] Step one: under the assumption of linear propagation and weak scattering, the pressure field received by the ultrasonic system is modeled by using the first-order Born approximation, considering the influence of measurement noise, the ultrasonic imaging model represents the radio frequency (RF) image as the convolution result of the PSF and the tissue reflectance function (TRF), and superimposes a noise term, and the mathematical expression is as follows:

[0008]

[0009] Wherein, x and y represent the sampling directions in the transverse and longitudinal directions respectively, represents the observed RF image, represents the PSF, represents the measurement noise, represents the TRF to be solved, represents the convolution operation;

[0010] The ultrasonic imaging degradation model is reconstructed, the obtained radio frequency image is decomposed into the superposition of the TRF and the focusing area PSF and other out-of-focus area PSF convolution results, and the specific expression form is:

[0011]

[0012] Wherein, and are the PSF of the focusing area and the PSF in the out-of-focus range respectively, and represent the TRF information of the focusing area and the out-of-focus area respectively, is the obtained degradation image information;

[0013] Step two: under the focusing PSF and noise-free imaging conditions, the ultrasonic imaging result with higher imaging performance is obtained:

[0014]

[0015] Wherein, represents the imaging result of the focusing PSF and the noise-free imaging condition, and the Fourier transform is performed on the above formula, which is represented in the frequency domain as:

[0016]

[0017] where, , , are the Fourier transform results of , , , is the multiplication symbol, denote the frequency form of x, y, respectively;

[0018] In the out-of-focus region with poor imaging resolution, the imaging is represented as:

[0019]

[0020] where, denotes the imaging result of the out-of-focus region, and the above formula is represented in the spectral domain as:

[0021]

[0022] where, , , , are the Fourier transform forms of , , ,

[0023] Step III: According to Step I and Step II, the PSF and the noise-free imaging result under the ideal imaging condition are taken as the target to design the frequency domain difference compensator to expand the high-frequency content of the far-field region, and the calculation formula is:

[0024]

[0025] By minimizing the deviation between the ideal imaging result and the calibration output, the optimal solution of the difference compensator is obtained, as follows:

[0026]

[0027] where E denotes the expectation operation, is the expectation, and by simultaneously differentiating both sides of the above equation and letting the result be zero to obtain its optimal value, the explicit expression of is derived:

[0028]

[0029] where, is the transpose of ,​ is a regularization factor;

[0030] Finally, by integrating the above equations, the image with enhanced resolution is obtained as follows:

[0031]

[0032] wherein, denotes the inverse Fourier transform, denotes the resolution-enhanced reconstructed image;

[0033] Step two: based on the image with enhanced resolution obtained in step one, a signal-to-noise ratio self-estimation module is designed to analyze the signal-to-noise ratio of the input image, and an adaptive parameterized denoising module is constructed in combination with the decomposition level and sub-band position information to effectively remove noise and enhance image contrast, thereby obtaining an ultrasound reconstructed image with enhanced contrast, and the specific steps are as follows:

[0034] Step two one: a signal-to-noise ratio self-estimation module is designed to estimate the overall signal-to-noise ratio of the image, which takes a noisy image and a clean image as input, wherein: a wavelet denoising method is used to estimate the clean image, and then the signal-to-noise ratio is calculated as follows:

[0035]

[0036] wherein, SNR est is the signal-to-noise ratio estimate, is the noisy image, is the estimated clean image; then, the signal-to-noise ratio weight adjustment parameter is associated with SNR est , if SNR est is greater than 25, the image is considered relatively clean, is set to 1; for other cases, it is adjusted with 25 as the center, as follows:

[0037]

[0038] wherein, max denotes the maximum value operation;

[0039] Step two two: considering the decomposition level and its corresponding sub-band position, a weighting factor is designed: decomposition level , sub-band position , and combined with the signal-to-noise ratio weight adjustment parameter to adaptively adjust the threshold weight factor F of the noise, as follows:

[0040]

[0041] wherein, min denotes the minimum value operation, and:

[0042]

[0043]

[0044] Among them, k and L represent the current decomposition level and the maximum decomposition level respectively, s represents the decomposition subband, Represents the weight factor of the adjustment threshold, through The wavelet processing threshold is adaptively adjusted to filter out noise and obtain a guidance image. The guidance image is then used to guide dual-wavelet-based Wiener denoising, achieving a trade-off between edge protection and noise removal, ultimately obtaining a contrast-enhanced ultrasound reconstructed image.

[0045] Compared with the prior art, the present invention has the following advantages:

[0046] 1. Aiming at the spatial transformation characteristics in ultrasonic imaging, the present invention constructs an ultrasonic imaging degradation model based on the frequency domain information difference between the focal area and the global PSF, and designs a difference compensator to improve imaging resolution. At the same time, through in-depth analysis of noise distribution, a signal-to-noise ratio self-estimation module is constructed, and an adaptive denoising module is designed by combining hierarchical and positional information to effectively remove noise and enhance image contrast.

[0047] 2. The present invention proposes a solution that meets the needs of practical applications. It cleverly utilizes the difference in PSF frequency domain information, deeply considers the actual imaging capabilities, proposes a difference compensation method for quickly improving the resolution in the frequency domain, and designs an adaptive non-parametric denoising model in the wavelet domain, which can quickly and effectively improve the quality of ultrasonic imaging.

[0048] 3. While ensuring rapid reconstruction, the present invention achieves simultaneous optimization of ultrasound image resolution and contrast, overcoming the difficulty of complex parameter adjustment in traditional methods and improving the practicality and reliability of ultrasound imaging in medical diagnosis.

[0049] 4. The present invention is applicable to various medical imaging devices and complex imaging environments, providing a higher quality and more stable solution for ultrasound imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The flowchart of the ultrasound fast imaging method with combined difference compensator and adaptive denoising.

[0051] Figure 2 This is a block diagram of the execution structure of the ultrasonic fast imaging method with a combined difference compensator and adaptive denoising.

[0052] Figure 3 Designed for the SNRest module.

[0053] Figure 4 Phantom experiment results.

[0054] Figure 5 Human experiment results. DETAILED DESCRIPTION

[0055] The technical solutions of the present application are further described below in conjunction with the drawings, but are not limited thereto, and any modification or equivalent replacement to the technical solutions of the present application without departing from the spirit and scope of the technical solutions of the present application shall be encompassed in the protection scope of the present application.

[0056] Based on the focusing and diffusion characteristics of the ultrasonic sound field propagation, an ultrasonic imaging mathematical degradation model is established, on the basis of which, the information difference in the frequency domain between the local focusing area PSF and the overall imaging characteristics is fully considered, and the signal-to-noise ratio (SNR) estimation of the input data and the level and sub-band position information of the wavelet decomposition are combined to provide an ultrasonic fast imaging method combining difference compensator and adaptive denoising. Specifically, first, based on the focusing and diffusion characteristics of the ultrasonic sound field, a degradation model of the ultrasonic imaging system is constructed; then, the PSF of the local focusing area and the defocusing area is Fourier transformed to simplify the calculation process and improve the calculation efficiency, and on this basis, the noiseless ideal imaging in the focusing state is taken as the optimization target, a difference compensator is designed to compensate the high-frequency information of the defocusing area with poor imaging performance, so as to improve the resolution; further, the image with enhanced resolution is input into the signal-to-noise ratio self-estimation module to evaluate the image quality, and the level and sub-band position information of the wavelet decomposition are combined to analyze the noise distribution characteristics, so as to construct an adaptive denoising model without manual parameter adjustment, so as to improve the image contrast and optimize the imaging effect. As shown in Figure 1 and Figure 2 The specific steps are as follows:

[0057] Step one: based on the spatial focusing and diffusion characteristics of the ultrasonic sound field propagation, an ultrasonic imaging degradation model is established, and the PSF information of the focusing and overall imaging area is extracted, which is converted to the frequency domain through Fourier transform, and the local focusing area is taken as the target ideal imaging condition, the frequency domain difference between the local focusing area and the overall PSF is analyzed, a frequency domain difference compensator is designed to expand the high-frequency information and optimize the overall imaging performance, and an image with enhanced resolution is obtained, and the specific steps are as follows:

[0058] Step one: under the assumption of linear propagation and weak scattering, the pressure field received by the ultrasonic system can be modeled by using the first-order Born approximation, considering the influence of measurement noise, the ultrasonic imaging model represents the RF image as the convolution result of the PSF and the TRF, and superimposes a noise term, and the mathematical expression is as follows:

[0059]

[0060] where x and y represent the sampling direction in the lateral and longitudinal direction, respectively, denotes the observed RF image, denotes the PSF, denotes the measurement noise, denotes the TRF to be solved, denotes the convolution operation.

[0061] In the above method, the PSF is assumed to be spatially invariant, however, this assumption does not hold in practical applications, resulting in distortion in the recovered image. Due to the focusing and diffusing characteristics of the ultrasound sound field propagation, the imaging resolution is highest in the focal region, while the imaging performance is poorer in other regions. Based on this, an ultrasound imaging degradation model is reconstructed, and the obtained RF image is decomposed into the superposition of the TRF and the convolution result of the focal region PSF and the out-of-focus region PSF. The specific expression form is:

[0062]

[0063] where, and are the PSF of the focal region and the PSF in the out-of-focus range, respectively, and represent the TRF information of the focal region and the out-of-focus region, respectively, is the obtained degraded image information.

[0064] Step two: if the imaging result with higher imaging performance can be obtained under the condition of the focal PSF and noise-free imaging:

[0065]

[0066] where, denotes the imaging result of the focal PSF and the noise-free imaging condition, and the Fourier transform of the above formula can be expressed in the frequency domain as:

[0067]

[0068] where, , , are the Fourier transform results of , , respectively, is a multiplication symbol, denote the frequency form of x, y, respectively;

[0069] In the out-of-focus region with poor imaging resolution, the imaging is expressed as:

[0070]

[0071] in, represents the imaging result of the defocused area. The above formula is expressed in the spectrum domain as:

[0072]

[0073] in, 、 、 、 They are 、 、 , The Fourier transform form of .

[0074] Step 13: Based on steps 11 and 12, design a frequency domain differential compensator with the PSF of ideal imaging conditions and noise-free imaging results as the goal. To expand the high-frequency content in the far field, the calculation formula is:

[0075]

[0076] By minimizing the deviation between the ideal imaging result and the calibrated output, a differential compensator is obtained. The best solution is as follows:

[0077]

[0078] Where E represents the expected operation, To obtain the optimal value of the above equation, we can deduce that The explicit expression for :

[0079]

[0080] in, for The transpose of is the regularization factor;

[0081] Finally, by integrating the above equations, we can obtain an image with enhanced resolution as follows:

[0082]

[0083] in, represents the inverse Fourier transform, Represents the reconstructed image with improved resolution.

[0084] Step 2: Resolution and contrast are two key indicators in ultrasound imaging, and both are indispensable. To further suppress noise such as clutter and artifacts to improve contrast, based on the image with enhanced resolution obtained in Step 1, a signal-to-noise ratio self-estimation module is designed. This module analyzes the signal-to-noise ratio of the input image and constructs an adaptive parametric denoising module based on the decomposition level and subband position information. This module effectively removes noise and enhances image contrast, resulting in a contrast-enhanced ultrasound reconstructed image. The specific steps are as follows:

[0085] Step 21: If Figure 3 As shown in FIG, a signal-to-noise ratio self-estimation module is designed to estimate the overall signal-to-noise ratio of the image. The module requires two inputs: a noisy image and a clean image. However, for ultrasound images, it is impractical to obtain a noise-free signal in the real world. To solve this problem, the present invention adopts a wavelet denoising method with general parameter settings to estimate the clean image, and then performs signal-to-noise ratio calculation as follows:

[0086]

[0087] Among them, SNR est is the signal-to-noise ratio estimate, is a noise image, is the estimated clean image. Then, the present invention adjusts the signal-to-noise ratio weight parameter and SNR est Related, specifically: If SNR est If it is greater than 25, the image is considered relatively clean. Set it to 1. For other cases, adjust it with 25 as the base center, as shown below:

[0088]

[0089] Among them, max means taking the maximum value operation.

[0090] Step 22: The distribution of noise is also closely related to the level of wavelet decomposition and the subband position. The low-frequency subband contains more image information, while the high-frequency component is more susceptible to noise. Therefore, applying a larger threshold in the high-frequency subband can more effectively suppress noise. On the other hand, as the decomposition level increases, the generated subband frequency bandwidth gradually narrows, and the corresponding threshold should be appropriately reduced. Based on this analysis, the present invention comprehensively considers the decomposition level and its corresponding subband position, and designs the weighting factor: decomposition level , subband position , and adjust the parameters in combination with the signal-to-noise ratio weight The noise threshold weight factor F is adaptively adjusted as follows:

[0091]

[0092] wherein min denotes a min operation, and

[0093]

[0094]

[0095] wherein k and L denote the current decomposition level and the maximum decomposition level, respectively, and s denotes a decomposition subband, denotes a weight factor that adjusts the threshold, by The wavelet processing threshold is adaptively adjusted to filter out noise in a more robust and flexible manner, and a guided image is obtained. The guided image is then used to guide the Wiener denoising framework of the dual wavelet basis to perform a trade-off between edge protection and noise removal, and finally a contrast-enhanced ultrasound reconstruction image is obtained.

[0096] Embodiment:

[0097] Step one: Based on the linear propagation theory and the weak scattering assumption, the first-order Born approximation is used to model the signal received by the ultrasound system, and the influence of measurement noise is fully considered. In order to address the influence of spatial transformation characteristics on imaging quality in traditional ultrasound imaging, the RF image is decomposed into a focused region PSF and an out-of-focus region PSF, which are respectively convolved with their corresponding TRF, and are superimposed to more accurately depict the spatial transformation effect in the imaging process. On this basis, the PSF of the focused region and the noise-free condition are used as the optimization target, and the optimization modeling is performed in the frequency domain to focus on solving the problem of decreased imaging quality in the out-of-focus region. For this purpose, a frequency domain difference compensator is designed to optimize the compensation strategy by minimizing the deviation between the ideal imaging result and the calibration output, so as to adaptively restore the high-frequency information of the out-of-focus region. Finally, this method effectively expands the frequency bandwidth of the out-of-focus region and improves the spatial resolution of ultrasound imaging, so that the overall imaging quality is closer to the ideal imaging effect, and more accurate tissue structure information is provided for ultrasound diagnosis.

[0098] Step two: Based on the resolution-enhanced image obtained in step one, an adaptive denoising module is further designed to improve the contrast. Specifically, the SNRest module is designed to estimate the overall signal-to-noise ratio of the image. Since the noise-free reference image cannot be directly obtained from the ultrasound image, the wavelet denoising method is used to preprocess the input image to estimate the clean image and calculate its signal-to-noise ratio. Subsequently, the signal-to-noise ratio estimate is associated with the weight adjustment parameter When the estimated signal-to-noise ratio is higher than the set threshold (such as 25), it indicates that the image is relatively clean, and the weight is set to 1 at this time; otherwise, the set threshold is dynamically adjusted to enhance the denoising processing capability for low signal-to-noise ratio images. Meanwhile, the influence level information and position weight factor of the decomposition level and subband position on noise are also considered and Specifically, the distribution characteristics of noise in different frequency subbands are different, the low-frequency subband contains more image information, and the high-frequency subband is more susceptible to noise pollution, therefore, a larger denoising threshold is applied in the high-frequency subband, and the threshold of the low-frequency subband is relatively small, in addition, with the increase of the wavelet decomposition level, the frequency bandwidth of the subband becomes narrower, and the corresponding denoising threshold needs to be gradually reduced.

[0099] Finally, the method is verified, and the results are shown in Figs. Figure 4 and Figure 5 It can be seen from the results that the method can meet the real-time requirement, does not need to rely on any parameter adjustment, and has good comprehensive performance in image resolution and contrast.

Claims

1. A rapid ultrasound imaging method combining a difference compensator and adaptive denoising, characterized in that The method comprises the following steps: Step 1: Based on the spatial focusing and diffusion characteristics of ultrasonic sound field propagation, an ultrasonic imaging degradation model is established. The PSF information of the focused and overall imaging areas is extracted and converted to the frequency domain through Fourier transform. The local focused area is used as the target ideal imaging condition. The frequency domain difference between the local focused area and the overall PSF is analyzed. A frequency domain difference compensator is designed to expand the high-frequency information, optimize the overall imaging performance, and obtain an image with enhanced resolution. Step 2: Based on the image with enhanced resolution obtained in step 1, a signal-to-noise ratio self-estimation module is designed to analyze the signal-to-noise ratio of the input image. An adaptive parametric denoising module is constructed by combining the information of decomposition level and subband position to effectively remove noise and enhance image contrast, thereby obtaining an ultrasound reconstructed image with enhanced contrast.

2. The ultrasonic rapid imaging method with a combined difference compensator and adaptive denoising according to claim 1 is characterized in that The specific steps of step one are as follows: Step 1: Under the assumptions of linear propagation and weak scattering, the pressure field received by the ultrasound system is modeled using the first-order Born approximation. Taking into account the influence of measurement noise, the ultrasound imaging model represents the RF image as the convolution result of the PSF and the tissue reflectance function TRF, and superimposes the noise term. Its mathematical expression is as follows: Among them, x and y represent the horizontal and vertical sampling directions respectively. Indicates that the RF image is observed, represents PSF, represents the measurement noise, represents the TRF to be sought, Represents the convolution operation; The ultrasound imaging degradation model is reconstructed to decompose the acquired RF image into the superposition of the convolution results of TRF and PSF of the focused area and PSF of other defocused areas. The specific expression is: in, and They are the PSF of the focused area and the PSF of the out-of-focus area, and Represents the TRF information of its focused area and out-of-focus area respectively, is the obtained degraded image information; Step 1 and 2: Obtain ultrasound imaging results with higher imaging performance under focused PSF and noise-free imaging conditions: in, The imaging result of the focused PSF and noise-free imaging conditions is expressed by Fourier transforming the above equation in the spectrum domain: in, 、 、 They are 、 、 The Fourier transform result of is the multiplication symbol, Represent the frequency forms of x and y respectively; In the out-of-focus area where the imaging resolution is poor, the imaging is represented as: in, represents the imaging result of the defocused area. The above formula is expressed in the spectrum domain as: in, 、 、 、 They are 、 、 , The Fourier transform form of Step 13: Based on steps 11 and 12, design a frequency domain differential compensator with the PSF of ideal imaging conditions and noise-free imaging results as the goal. The high-frequency content in the extended far-field region is calculated as: By minimizing the deviation between the ideal imaging result and the calibrated output, a differential compensator is obtained. The best solution is as follows: Where E represents the expected operation, To obtain the optimal value of the above equation, we can derive The explicit expression for : in, for The transpose of is the regularization factor; Finally, by integrating the above equations, we obtain an image with enhanced resolution as follows: in, represents the inverse Fourier transform, Represents the reconstructed image with improved resolution.

3. The ultrasonic rapid imaging method with a combined difference compensator and adaptive denoising according to claim 1 is characterized in that The specific steps of step 2 are as follows: Step 21: Design a signal-to-noise ratio self-estimation module to estimate the overall signal-to-noise ratio of the image. This module takes the noisy image and the clean image as input. Step 2: Consider the decomposition level and its corresponding sub-band position, and design the weighting factor: Decomposition level , subband position , and adjust the parameters in combination with the signal-to-noise ratio weight The threshold weight factor of the noise is adaptively adjusted as follows: Among them, min means taking the minimum value operation, Represents the weight factor for adjusting the noise threshold, through The wavelet processing threshold is adaptively adjusted to filter out noise and obtain a guidance image. The guidance image is then used to guide dual-wavelet-based Wiener denoising, achieving a trade-off between edge protection and noise removal, ultimately obtaining a contrast-enhanced ultrasound reconstructed image.

4. The ultrasonic rapid imaging method with a combined difference compensator and adaptive denoising according to claim 3 is characterized in that In step 21, a wavelet denoising method is used to estimate a clean image, and then the signal-to-noise ratio is calculated as follows: Among them, SNR est is the signal-to-noise ratio estimate, is a noise image, is the estimated clean image; then, the signal-to-noise ratio weight adjustment parameter and SNR est If the SNR est If it is greater than 25, the image is considered relatively clean. Set to 1; for other cases, adjust it with 25 as the base center, as shown below: Among them, max means taking the maximum value operation.

5. The ultrasonic rapid imaging method with a combined difference compensator and adaptive denoising according to claim 3 is characterized in that In step 22, the decomposition level , sub-band position The following conditions are met: Where k and L represent the current decomposition level and the maximum decomposition level, respectively, and s represents the decomposition subband.

Images