Image processing method and device based on interventional array ultrasonic transducer, and system
By using multi-level filtering based on interventional array ultrasound transducers and U-Net model processing, combined with region partitioning filtering algorithms, the problem of incomplete denoising in existing ultrasound images is solved, achieving high-precision image denoising and improving image quality and diagnostic accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HANGZHOU XINYING MEDICAL TECHNOLOGY CO LTD
- Filing Date
- 2026-01-26
- Publication Date
- 2026-05-29
AI Technical Summary
Existing ultrasound image denoising methods cannot completely remove noise, resulting in decreased image quality, increased risk of missed or misdiagnosed diagnoses, and the inability to preserve image details at the same time.
An image processing method based on interventional array ultrasound transducers is adopted, including noise assessment, multi-level filtering algorithm, U-Net model processing and region segmentation filtering. Through multi-level filtering and U-Net model, precise denoising is achieved. Combined with region filtering algorithm for edge and smooth images, accurate image denoising is achieved.
It improves the accuracy of ultrasound image denoising, avoids over-smoothing and edge loss, enhances image quality, and reduces the risk of missed diagnoses and misdiagnoses.
Smart Images
Figure CN122115262A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ultrasound image processing technology, and in particular to an image processing method, apparatus, and system based on an interventional array ultrasound transducer. Background Technology
[0002] In the field of clinical ultrasound diagnosis, array-type ultrasound transducers, with their advantage of multi-element collaborative imaging, can achieve high-resolution, high-frame-rate two-dimensional ultrasound image acquisition, making them a core hardware component for diagnosing areas such as the abdomen, heart, and superficial tissues. However, the imaging process inevitably introduces composite noise such as Gaussian noise, multiplicative speckle noise, and salt-and-pepper noise. This noise can obscure crucial diagnostic information such as lesion edges and vascular texture, reducing image quality and increasing the risk of missed or misdiagnosed diagnoses.
[0003] In recent years, deep learning has become a research hotspot in the field of medical imaging, and ultrasound image analysis is no exception. Thanks to its powerful feature learning capabilities, deep neural networks can automatically extract complex distribution features from large-scale data in ultrasound images, effectively modeling and suppressing noise, thereby improving image contrast and structural fidelity. However, existing ultrasound image denoising methods suffer from key problems such as incomplete noise reduction, insufficient detail preservation, poor real-time performance, and weak array adaptability, making it difficult to meet the image quality requirements of clinical ultrasound diagnosis.
[0004] It is evident that current ultrasound image denoising methods are incomplete and cannot simultaneously preserve image details, resulting in low accuracy in noise identification within ultrasound images. Therefore, a new ultrasound image processing method is needed to improve the accuracy of image denoising. Summary of the Invention
[0005] This invention provides an image processing method based on an interventional array ultrasonic transducer, which can improve the accuracy of image denoising.
[0006] The first aspect of this invention discloses an image processing method based on an interventional array ultrasound transducer, the method comprising: A two-dimensional ultrasound image to be processed is acquired; the two-dimensional ultrasound image is subjected to noise intensity assessment processing to obtain a noise assessment image; wherein, the two-dimensional ultrasound image to be processed is acquired through an interventional array ultrasound transducer; The noise evaluation image is processed by a first filtering algorithm to obtain a filtered and denoised image; the filtered and denoised image is then processed by a second filtering algorithm to obtain a preliminary denoised image. The pre-trained U-Net model is used to process the initial denoised image to obtain an accurate denoised image; The precisely denoised image is divided into regions to obtain an edge image and a smoothed image; a region filtering algorithm is applied to the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; and a denoised ultrasound image is obtained based on the edge-denoised image and the smoothed-denoised image.
[0007] As an optional implementation, in the first aspect of the present invention, the step of performing noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image includes: The gradient of all pixels in the two-dimensional ultrasound image is determined, and a first target pixel is selected from all pixels in the two-dimensional ultrasound image; a first noise dataset is obtained based on all the first target pixels; wherein the gradient of the first target pixel satisfies a preset first threshold range. The grayscale mean of all pixels in the two-dimensional ultrasound image is determined; the variance of all pixels is obtained based on the grayscale mean; a second target pixel is selected from all pixels; a second noise dataset is obtained based on all the second target pixels; wherein the variance of the second target pixel satisfies a preset second threshold range. Based on the mean and variance of all the pixels, the mean square value corresponding to all the pixels is obtained; a third target pixel is selected from all the pixels; a third noise dataset is obtained based on all the third target pixels; wherein the variance of the third target pixel satisfies a preset third threshold range; A noise assessment image is obtained based on the first noise dataset, the second noise dataset, and the third noise dataset.
[0008] As an optional implementation, in the first aspect of the present invention, the step of processing the noise evaluation image with a first filtering algorithm to obtain a filtered and denoised image includes: Determine the parameter values of the first filtering algorithm corresponding to the first noise dataset; obtain the corresponding first filtering algorithm based on the parameter values of the first filtering algorithm; process the noise evaluation image using the first filtering algorithm to obtain the filtered and denoised image; And, the filtered and denoised image is processed by a second filtering algorithm to obtain a preliminary denoised image, including: Determine the parameter values of the second filtering algorithm corresponding to different noise regions in the second noise dataset; obtain the second filtering algorithm corresponding to each noise region based on the parameter values of the second filtering algorithm; process the filtered and denoised image using the second filtering algorithm to obtain the preliminary denoised image; wherein, the noise region is used to select the distribution of different second noise intensities.
[0009] As an optional implementation, in the first aspect of the present invention, processing the preliminary denoised image using a pre-trained U-Net model to obtain an accurately denoised image includes: According to the preset labeling algorithm in the U-Net model, the first noise dataset and the second noise dataset in the preliminary denoised image are labeled to obtain the first labeled dataset; the third noise dataset in the preliminary denoised image is labeled to obtain the second labeled dataset. The first and second labeled datasets are processed using the pre-defined convolution algorithm in the U-Net model to obtain an accurately denoised image.
[0010] As an optional implementation, in the first aspect of the present invention, the step of performing region segmentation processing on the precisely denoised image to obtain an edge-denoised image and a smoothed denoised image includes: Gradient calculation is performed on all pixels in the precise denoised image to obtain the horizontal and vertical gradients of all pixels; based on the horizontal and vertical gradients of all pixels, the gradient change magnitude corresponding to all pixels is obtained. Filter all the pixels to obtain target edge pixels; obtain an edge image based on all the target edge pixels; Filter all the pixels to obtain target smoothed pixels; obtain a smoothed image based on all the target smoothed pixels; Wherein, the gradient change amplitude of the target edge pixel satisfies a preset first division condition, and the gradient change amplitude of the target smooth pixel satisfies a preset second division condition.
[0011] As an optional implementation, in the first aspect of the present invention, performing a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image includes: The target pixels in the edge image are processed by the region filtering algorithm to obtain a first template window; based on the first template window, a first candidate template is obtained; wherein, each of the first candidate templates is the same size as the first template window; Calculate the grayscale difference between the first template window and all the first candidate templates to obtain the first grayscale difference; based on the first grayscale difference, obtain the first weight distribution corresponding to all the first candidate templates; based on all the first weight distributions, obtain the edge-denoised image; The region filtering algorithm is applied to the target pixels in the smoothed image to obtain a second template window; based on the second template window, a second candidate template is obtained; wherein each of the second candidate templates is the same size as the second template window, and the size of the second template window is larger than the size of the first template window. Calculate the grayscale difference between the second template window and all the second candidate templates to obtain the second grayscale difference; based on the second grayscale difference, obtain the second weight distribution corresponding to all the second candidate templates; based on all the second weight distributions, obtain the smoothed and denoised image.
[0012] As an optional implementation, in the first aspect of the present invention, acquiring the two-dimensional ultrasound image to be processed includes: Based on a preset interventional array ultrasonic transducer, ultrasonic beams are emitted to the target area by activating different arrays; according to the interventional array ultrasonic transducer, echo data is obtained by receiving the echo signal corresponding to the ultrasonic beam. The echo data is used to obtain a beamforming line through beamforming; The two-dimensional ultrasound image is obtained by arranging all the beamforming lines.
[0013] A second aspect of the present invention discloses an image processing device based on an interventional array ultrasonic transducer, the device comprising: A noise assessment module is used to acquire a two-dimensional ultrasound image to be processed; to perform noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image; wherein, the two-dimensional ultrasound image to be processed is acquired through an interventional array ultrasound transducer; The preliminary processing module is used to process the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image; and to process the filtered and denoised image using a second filtering algorithm to obtain a preliminary denoised image. The precision processing module is used to process the initial denoised image using a pre-trained U-Net model to obtain a precision denoised image; An integrated processing module is used to perform region segmentation processing on the precisely denoised image to obtain an edge image and a smoothed image; to perform a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; and to obtain a denoised ultrasound image based on the edge-denoised image and the smoothed-denoised image.
[0014] As an optional implementation, in a second aspect of the invention, the step of performing noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image includes: The gradient of all pixels in the two-dimensional ultrasound image is determined, and a first target pixel is selected from all pixels in the two-dimensional ultrasound image; a first noise dataset is obtained based on all the first target pixels; wherein the gradient of the first target pixel satisfies a preset first threshold range. The grayscale mean of all pixels in the two-dimensional ultrasound image is determined; the variance of all pixels is obtained based on the grayscale mean; a second target pixel is selected from all pixels; a second noise dataset is obtained based on all the second target pixels; wherein the variance of the second target pixel satisfies a preset second threshold range. Based on the mean and variance of all the pixels, the mean square value corresponding to all the pixels is obtained; a third target pixel is selected from all the pixels; a third noise dataset is obtained based on all the third target pixels; wherein the variance of the third target pixel satisfies a preset third threshold range; A noise assessment image is obtained based on the first noise dataset, the second noise dataset, and the third noise dataset.
[0015] As an optional implementation, in a second aspect of the present invention, the step of processing the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image includes: Determine the parameter values of the first filtering algorithm corresponding to the first noise dataset; obtain the corresponding first filtering algorithm based on the parameter values of the first filtering algorithm; process the noise evaluation image using the first filtering algorithm to obtain the filtered and denoised image; And, the filtered and denoised image is processed by a second filtering algorithm to obtain a preliminary denoised image, including: Determine the parameter values of the second filtering algorithm corresponding to different noise regions in the second noise dataset; obtain the second filtering algorithm corresponding to each noise region based on the parameter values of the second filtering algorithm; process the filtered and denoised image using the second filtering algorithm to obtain the preliminary denoised image; wherein, the noise region is used to select the distribution of different second noise intensities.
[0016] As an optional implementation, in a second aspect of the invention, processing the initial denoised image using a pre-trained U-Net model to obtain an accurately denoised image includes: According to the preset labeling algorithm in the U-Net model, the first noise dataset and the second noise dataset in the preliminary denoised image are labeled to obtain the first labeled dataset; the third noise dataset in the preliminary denoised image is labeled to obtain the second labeled dataset. The first and second labeled datasets are processed using the pre-defined convolution algorithm in the U-Net model to obtain an accurately denoised image.
[0017] As an optional implementation, in a second aspect of the invention, the step of performing region segmentation processing on the precisely denoised image to obtain an edge-denoised image and a smoothed denoised image includes: Gradient calculation is performed on all pixels in the precise denoised image to obtain the horizontal and vertical gradients of all pixels; based on the horizontal and vertical gradients of all pixels, the gradient change magnitude corresponding to all pixels is obtained. Filter all the pixels to obtain target edge pixels; obtain an edge image based on all the target edge pixels; Filter all the pixels to obtain target smoothed pixels; obtain a smoothed image based on all the target smoothed pixels; Wherein, the gradient change amplitude of the target edge pixel satisfies a preset first division condition, and the gradient change amplitude of the target smooth pixel satisfies a preset second division condition.
[0018] As an optional implementation, in a second aspect of the invention, performing a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image includes: The target pixels in the edge image are processed by the region filtering algorithm to obtain a first template window; based on the first template window, a first candidate template is obtained; wherein, each of the first candidate templates is the same size as the first template window; Calculate the grayscale difference between the first template window and all the first candidate templates to obtain the first grayscale difference; based on the first grayscale difference, obtain the first weight distribution corresponding to all the first candidate templates; based on all the first weight distributions, obtain the edge-denoised image; The target pixels in the smoothed image are processed by the region filtering algorithm to obtain a second template window; based on the second template window, a second candidate template is obtained; wherein, each of the second candidate templates is the same size as the second template window, and the size of the second template window is larger than the size of the first template window; Calculate the grayscale difference between the second template window and all the second candidate templates to obtain the second grayscale difference; based on the second grayscale difference, obtain the second weight distribution corresponding to all the second candidate templates; based on all the second weight distributions, obtain the smoothed and denoised image.
[0019] As an optional implementation, in a second aspect of the present invention, acquiring the two-dimensional ultrasound image to be processed includes: Based on a preset interventional array ultrasonic transducer, ultrasonic beams are emitted to the target area by activating different arrays; according to the interventional array ultrasonic transducer, echo data is obtained by receiving the echo signal corresponding to the ultrasonic beam. The echo data is used to obtain a beamforming line through beamforming; The two-dimensional ultrasound image is obtained by arranging all the beamforming lines.
[0020] A third aspect of the present invention discloses an apparatus comprising a memory and a processor, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the image processing method based on an interventional array ultrasonic transducer according to any of the first aspects of the present invention.
[0021] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the image processing method based on an interventional array ultrasonic transducer as described in any of the first aspects of the present invention.
[0022] Compared with the prior art, the present invention has the following beneficial effects: This invention acquires a two-dimensional ultrasound image to be processed, performs noise assessment on the two-dimensional ultrasound image to obtain a noise assessment image, and then uses the noise assessment image to accurately partition the image noise. A first filtering algorithm and a second filtering algorithm are then used to perform preliminary processing on the partitioned noise to obtain a preliminary denoised image. Because the preliminary processing can only remove low-order, high-intensity noise, it can improve the accuracy of subsequent processing. Then, the preliminary denoised image is precisely processed using the U-Net model to obtain a precisely denoised image. However, the obtained precisely denoised image suffers from over-smoothing and edge loss. Therefore, the precisely denoised image is further divided into regions to obtain an edge image and a smoothed image. Corresponding algorithms are then applied to the edge image and the smoothed image to obtain an edge-denoised image and a smoothed denoised image. Finally, the edge-denoised image and the smoothed denoised image are weighted and fused to obtain the denoised ultrasound image, thereby improving the accuracy of ultrasound image denoising. Attached Figure Description
[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 This is a schematic flowchart of an image processing method based on an interventional array ultrasonic transducer disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of a micro phased array disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an interventional array ultrasonic transducer disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of a composite shielding architecture for a multilayer flexible circuit disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an image processing device based on an interventional array ultrasonic transducer disclosed in an embodiment of the present invention; Figure 6 This is a schematic diagram of a device including a memory and a processor disclosed in an embodiment of the present invention. Detailed Implementation
[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0026] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.
[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0028] This invention discloses an image processing method based on an interventional array ultrasonic transducer, which can improve the accuracy of image denoising. The following sections provide detailed explanations.
[0029] Example 1 Please see Figure 1 , Figure 1 This invention discloses an image processing method based on an interventional array-type ultrasonic transducer. Figure 1 The described image processing method based on an interventional array ultrasonic transducer can be applied to an image processing device based on an interventional array ultrasonic transducer. This device can be integrated into a cloud server or a local server; the embodiments of this invention are not limited thereto. Figure 1 As shown, the image processing method based on an interventional array ultrasound transducer may include the following operations: Step 101: Obtain the two-dimensional ultrasound image to be processed; perform noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image.
[0030] In this embodiment of the invention, the two-dimensional ultrasound image to be processed is acquired by an interventional array ultrasound transducer.
[0031] In this embodiment of the invention, the two-dimensional ultrasound image to be processed can be a cardiac ultrasound image, a thyroid ultrasound image, or a liver ultrasound image; In the embodiments of the present invention, those skilled in the art will understand that the noise intensity assessment process is to summarize the different noise distribution intensities in the ultrasound image.
[0032] In this embodiment of the invention, the noise evaluation image may be, but is not limited to, the following noises: salt and pepper noise, speckle noise, and Gaussian noise.
[0033] Step 102: The first filtering algorithm is used to process the noise evaluation image to obtain a filtered and denoised image; the second filtering algorithm is used to process the filtered and denoised image to obtain a preliminary denoised image.
[0034] In this embodiment of the invention, the first filtering algorithm can be used to denoise one or more types of noise in the noise evaluation image; the filtered and denoised image can be a salt-and-pepper noise denoised image, a speckle noise denoised image, or a Gaussian noise denoised image; the second filtering algorithm can also be used to denoise one or more types of noise in the noise evaluation image; the preliminary denoised image can be an image after denoising one or more types of noise, such as a salt-and-pepper noise denoised image, a speckle noise denoised image, or a Gaussian noise denoised image. In this embodiment of the invention, the first filtering algorithm includes, but is not limited to, the following methods: median filtering, adaptive median filtering, switch median filtering, weighted median filtering, and iterative median filtering; the second filtering algorithm includes, but is not limited to, the following methods: nonlocal mean filtering and wavelet domain speckle denoising.
[0035] Step 103: Use the pre-trained U-Net model to process the initial denoised image to obtain the accurate denoised image.
[0036] In the embodiments of the present invention, those skilled in the art will understand that the pre-trained U-Net model is a model that has been trained with data before the model starts; the U-Net model can accurately denoise one or more types of noise in the initially denoised image.
[0037] Step 104: Perform region segmentation processing on the precisely denoised image to obtain an edge image and a smoothed image; perform a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; obtain the denoised ultrasound image based on the edge-denoised image and the edge-denoised image.
[0038] In this embodiment of the invention, the region segmentation process can be to perform edge region detail preservation and smooth region strong noise reduction on the precisely denoised image, avoiding edge blurring or noise residue caused by single filtering. In this embodiment of the invention, the edge image can be a region divided into smaller pixel windows, while the smooth image can be a region divided into larger pixel windows. In this embodiment of the invention, the region filtering algorithm can be median filtering, adaptive median filtering, switch median filtering, weighted median filtering, or iterative median filtering; In this embodiment of the invention, the denoised ultrasound image can be an image that is a superposition of two edge-denoised images.
[0039] As can be seen, the embodiments of the present invention acquire a two-dimensional ultrasound image to be processed, and then perform noise evaluation processing on the two-dimensional ultrasound image to obtain a noise evaluation image. The noise evaluation image can be used to accurately partition the image noise. A first filtering algorithm and a second filtering algorithm can be used to perform preliminary processing on the partitioned noise to obtain a preliminary denoised image. Because the preliminary processing can only remove low-order, high-intensity noise, it can improve the accuracy of subsequent processing. Then, the preliminary denoised image is precisely processed using the U-Net model to obtain a precisely denoised image. However, the obtained precisely denoised image suffers from over-smoothing and edge loss. Therefore, the precisely denoised image is divided into regions to obtain an edge image and a smoothed image. Corresponding algorithms are executed on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image. Finally, the edge-denoised image and the smoothed-denoised image are weighted and fused to obtain a denoised ultrasound image, thereby improving the accuracy of ultrasound image denoising.
[0040] In an optional embodiment, performing noise intensity assessment processing on a two-dimensional ultrasound image to obtain a noise assessment image may include: Determine the gradient of all pixels in the two-dimensional ultrasound image, and select the first target pixel from all pixels in the two-dimensional ultrasound image; obtain the first noise dataset based on all the first target pixels; Determine the mean gray value of all pixels in the two-dimensional ultrasound image; based on the mean gray value, obtain the variance of all pixels, and select the second target pixels from all pixels; based on all the second target pixels, obtain the second noise dataset; Based on the mean and variance of all pixels, obtain the mean square value of all pixels; select the third target pixels from all pixels; obtain the third noise dataset based on all third target pixels; Based on the first noise dataset, the second noise dataset, and the third noise dataset, a noise assessment image is obtained.
[0041] In this optional embodiment, the gradient of the first target pixel satisfies a preset first threshold range, the variance of the second target pixel satisfies a preset second threshold range, and the variance of the third target pixel satisfies a preset third threshold range. The three threshold ranges can be set according to actual needs or experimental verification. For example, the three threshold ranges can be used to divide the pixels on the ultrasound image according to different rules, which is beneficial for subsequent accurate noise reduction.
[0042] In this optional embodiment, the first target pixel can be the grayscale change amplitude of each pixel in the image, that is, the gradient; and the pixels within a first threshold range are selected as the first target pixel; the first threshold range can be that the gradient change amplitude of the target pixel exceeds 3 times the overall gradient change amplitude of all pixels in the image. Optionally, the first noise dataset can be obtained by counting the number of all first target pixels, dividing by the total number of pixels in the entire image, and obtaining the proportion of first target pixels; when the proportion is less than 2%, it is judged as low-intensity first noise; when the proportion is between 2% and 5%, it is judged as medium-intensity first noise; when the proportion is greater than 5%, it is judged as high-intensity first noise; and the results of the judgment are recorded to form the first noise dataset. In this optional embodiment, the second target pixel can be obtained by calculating the mean gray value of all pixels in this area, then calculating the sum of squared differences between each pixel and the mean, and finally obtaining the variance; when the variance reaches a second threshold range, it is determined to be the second target pixel; the second threshold range can be set to a variance greater than or equal to 0.01; Optionally, the second noise dataset can be defined as follows: pixels with a variance greater than or equal to 0.01 and less than 0.03 are classified as low-intensity second noise pixels, and pixels with a variance greater than or equal to 0.03 are classified as high-intensity second noise pixels; by summing up all the results, the second noise dataset is obtained. In this optional embodiment, the third target pixel can be calculated by calculating the mean and variance of the pixel grayscale values in this area, and then calculating the ratio of the variance to the mean; when the ratio reaches a third threshold range, it is determined to be the third target pixel; the third threshold range can be set to a ratio of the variance to the mean of the pixel being greater than 0.2; Optionally, the third noise dataset can be defined as follows: pixels with a variance greater than or equal to 0.2 and less than 0.35 are classified as low-intensity third noise pixels, and pixels with a ratio greater than or equal to 0.35 are classified as high-intensity third noise pixels; by summing up all the results, the third noise dataset is obtained. In this optional embodiment, the noise evaluation image can be obtained by superimposing the first noise dataset, the first noise dataset, and the first noise dataset, and bringing together all different noise distributions of high intensity and low intensity.
[0043] As can be seen, this optional embodiment identifies and classifies different noises in the image, and also identifies the intensity of different noises, which facilitates subsequent accurate denoising; it gathers the characteristics of all noises, accurately assesses the type, intensity, and distribution of each type of noise, and finally provides a decision basis for regional and targeted denoising; thereby improving the accuracy of denoising.
[0044] In another optional embodiment, processing the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image may include: Determine the parameter values of the first filtering algorithm corresponding to the first noise dataset; obtain the corresponding first filtering algorithm based on the parameter values of the first filtering algorithm; process the noise evaluation image using the first filtering algorithm to obtain a filtered and denoised image; Furthermore, the filtered and denoised image is processed by a second filtering algorithm to obtain a preliminary denoised image, including: Determine the parameter values of the second filtering algorithm corresponding to different noise regions in the second noise dataset; based on the parameter values of the second filtering algorithm, obtain the second filtering algorithm corresponding to each noise region; process the filtered and denoised image using the second filtering algorithm to obtain a preliminary denoised image; In this alternative embodiment, the noise region is used to select different distributions of the second noise intensity.
[0045] In this optional embodiment, the first filtering algorithm parameter value can be the core parameters of the first noise output in the generated noise evaluation image; for example, noise intensity level, noise density, etc.; then, the denoising priority and algorithm parameters of each region are determined according to these core parameters to obtain the first filtering algorithm parameter value; and the first filtering algorithm is specifically determined for the first filtering algorithm parameter value. In this optional embodiment, the noise region can be selected by dividing low-intensity second noise pixels and high-intensity second noise pixels generated in the noise evaluation image into different noise regions; the second filtering algorithm parameter value can be the core parameters for generating second noise from these noise regions, thereby determining the second filtering algorithm parameter value corresponding to different noise regions; then, the second filtering algorithm corresponding to pixels with different noise intensities is determined.
[0046] As can be seen, this optional embodiment uses targeted matching algorithms for different noise intensities. For low-intensity noise, targeted algorithms preserve details; for medium-intensity noise, targeted algorithms maintain a balanced effect; and for high-intensity noise, combined algorithms effectively suppress noise. This removes the first and second noises that are easy to process, outputs a preliminary denoised image, and clears obstacles for subsequent denoising of key noises.
[0047] In yet another optional embodiment, processing the initial denoised image using a pre-trained U-Net model to obtain a precisely denoised image may include: According to the pre-defined labeling algorithm in the U-Net model, the first noise dataset and the second noise dataset in the preliminary denoised image are labeled to obtain the first labeled dataset; the third noise dataset in the preliminary denoised image is labeled to obtain the second labeled dataset. The first and second labeled datasets are processed using the pre-defined convolution algorithm in the U-Net model to obtain accurately denoised images.
[0048] In this optional embodiment, those skilled in the art will understand that the labeling algorithm labels the noise data contained in the input image, which can distinguish different noise objects when used in the denoising algorithm; thus facilitating the algorithm to allocate resources and perform targeted denoising. In this optional embodiment, the first labeled dataset is the data after preliminary denoising; here, computational resources can be reduced to perform accurate denoising; the second labeled dataset is the data that has not yet been denoised, and the convolution algorithm in the U-Net model is used to increase computational resources on the second labeled dataset to ensure that accurate denoising is achieved. As can be seen, this optional embodiment can accurately distinguish between processed and unprocessed noise through the U-Net model, achieving precise focusing on the noise region and enabling the network to efficiently learn the characteristics of the noise; thus, it can remove noise more accurately.
[0049] In another optional embodiment, performing region segmentation processing on the precisely denoised image to obtain an edge-denoised image and a smoothed denoised image may include: Gradient calculation is performed on all pixels in the precisely denoised image to obtain the horizontal and vertical gradients of all pixels; based on the horizontal and vertical gradients of all pixels, the gradient change magnitude of each pixel is obtained. Filter all pixels to obtain the target edge pixels; obtain the edge image based on all target edge pixels; Filter all pixels to obtain the target smoothed pixels; obtain the smoothed image based on all target smoothed pixels.
[0050] In this optional embodiment, the gradient change magnitude of the target edge pixels satisfies a preset first division condition, and the gradient change magnitude of the target smooth pixels satisfies a preset second division condition; In this optional embodiment, the horizontal gradient can be the vertical edge in the detected image. The weight template assigns opposite weights to the left and right adjacent regions of the pixel, and no weight is assigned to the middle region. The gradient value in this direction is obtained by summing the products of adjacent pixels and weights. The greater the difference in weights, the more drastic the gray-scale change in the horizontal direction. Optionally, the vertical gradient can be the horizontal edge in the detected image. The weight template assigns opposite weights to the upper and lower adjacent regions of a pixel, while no weight is assigned to the middle region. Similarly, the gradient value in this direction is obtained by summing the products of adjacent pixels and their weights. The greater the difference in weights, the more drastic the change in grayscale in the vertical direction. Optionally, the gradient change magnitude can be obtained by first squaring the gradient values in both directions, then adding the two squared results, and finally taking the square root of the sum to obtain the final gradient change magnitude. In this optional embodiment, the first dividing condition can be a gradient change magnitude greater than 200; the second dividing condition can be a gradient change magnitude between 100 and 200; and a gradient change magnitude less than 100 is directly ignored. As can be seen, this optional embodiment divides the precise image into an edge image and a smooth image through post-processing; this facilitates subsequent filtering algorithms to preserve details in the edge region and reduce noise in the smooth region.
[0051] In yet another optional embodiment, performing a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image may include: The target pixels in the edge image are processed by a region filtering algorithm to obtain a first template window; based on the first template window, a first candidate template is obtained. Calculate the grayscale difference between the first template window and all first candidate templates to obtain the first grayscale difference; based on the first grayscale difference, obtain the first weight distribution corresponding to all first candidate templates; based on all first weight distributions, obtain the edge-denoised image; The target pixels in the smoothed image are processed by a region filtering algorithm to obtain a second template window; based on the second template window, a second candidate template is obtained. Calculate the grayscale difference between the second template window and all second candidate templates to obtain the second grayscale difference; based on the second grayscale difference, obtain the second weight distribution corresponding to all second candidate templates; based on all second weight distributions, obtain the smoothed and denoised image.
[0052] In this optional embodiment, each first candidate template is the same size as the first template window; each second candidate template is the same size as the second template window, and the size of the second template window is larger than the size of the first template window; In this optional embodiment, the region filtering algorithm may set a template window and search range for the current image to avoid confusion between the search and the background texture; Optionally, the first candidate template can be an image template with the same size as the first template window selected one by one within the search range. The gray-level similarity between each first candidate template and the template window is compared. If the texture difference in the edge area is large, only candidate templates with highly consistent gray-level distribution are retained. Optionally, the first weight distribution can be based on the grayscale difference between the first candidate template and the first template window. The smaller the difference, the greater the weight, indicating that the candidate template contributes more to suppressing the artifacts of the current pixel. At the same time, because the weight decays quickly, only the first candidate template with extremely high similarity will be assigned an effective weight. Optionally, the edge denoising image can be obtained by multiplying the corresponding pixel values of all first candidate templates with their respective weights, summing all the product results, and finally normalizing with the total weight; the weights of the edge regions are concentrated on a few first candidate templates to ensure that the edge details are not blurred after filtering. In this optional embodiment, the second candidate template can be an image template with the same size as the second template window selected one by one within the search range, and the grayscale similarity between each second candidate template and the template window is compared. The texture of the smooth area is uniform, the similarity requirement is relatively relaxed, and more second candidate templates are allowed to enter the subsequent calculation. Optionally, the second weight distribution can be based on the grayscale difference between the second candidate template and the second template window. The smaller the difference, the greater the weight, indicating that the candidate template contributes more to suppressing the artifacts of the current pixel. At the same time, the weight decay rate is slow, so even candidate templates with moderate similarity will be assigned a certain weight, thus expanding the effective weighting range. Optionally, the smoothed and denoised image can be obtained by multiplying the corresponding pixel values of all second candidate templates with their respective weights, summing all the product results, and finally normalizing with the total weight. The smoothed image has a wider weight distribution, and by averaging a large number of similar pixels, it effectively cancels the random fluctuations of residual noise, making the image more uniform and natural. As can be seen, this optional embodiment can prioritize the preservation of edge details and eliminate only minor artifacts by processing the edge region; processing the smooth image can further smooth residual noise and improve the overall naturalness of the image; thus achieving a balance between thorough noise reduction and detail preservation of the image.
[0053] In yet another alternative embodiment, acquiring the two-dimensional ultrasound image to be processed may include: Based on a preset interventional array ultrasonic transducer, ultrasonic beams are emitted to the target area by activating different arrays; according to the interventional array ultrasonic transducer, the echo signal corresponding to the ultrasonic beam is received to obtain echo data. The echo data is beamformed to obtain a beamforming line. All beamforming lines are arranged to obtain a two-dimensional ultrasound image.
[0054] In this optional embodiment, the interventional array ultrasound transducer employs composite processing techniques and materials to achieve miniaturization of the transducer's pitch direction and the gap between adjacent array elements. This reduces the pitch dimension of the intracardiac phased array transducer to approximately 60% of that of traditional transducers while maintaining its original performance, making it suitable for intracardiac ultrasound imaging in any group, including children. The transducer's structure is as follows: Figure 2 As shown; Optional, such as Figure 3As shown, the composite process refers to the piezoelectric ceramic being self-connected in the Z direction, responsible for generating and receiving ultrasonic waves; while the polymer is connected in all three directions, playing a supporting role, reducing acoustic impedance, and improving flexibility. In this optional embodiment, such as Figure 4 As shown, by integrating the signal layer and the back copper layer into the same flexible circuit, and superimposing double-sided copper shielding on the outside of the folded multi-channel FPC, an embedded + fully enclosed composite electromagnetic shielding system is constructed; this ensures that each signal is completely surrounded by double copper layers, significantly reducing crosstalk between adjacent signal lines and external electromagnetic interference; this architecture greatly optimizes the clarity and signal-to-noise ratio of intracardiac ultrasound imaging, providing greater operational flexibility and reliability for minimally invasive surgery.
[0055] As can be seen, this optional embodiment, by utilizing a novel interventional array ultrasound transducer to acquire the image to be processed, can significantly reduce crosstalk between adjacent signal lines and external electromagnetic interference, optimize the clarity and signal-to-noise ratio of intracardiac ultrasound imaging, and improve the accuracy of ultrasound image denoising.
[0056] Example 2 Please see Figure 5 , Figure 5 This invention discloses an image processing device based on an interventional array ultrasonic transducer, which may include: The noise assessment module 201 is used to acquire the two-dimensional ultrasound image to be processed; and to perform noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image. The preliminary processing module 202 is used to process the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image; and to process the filtered and denoised image using a second filtering algorithm to obtain a preliminary denoised image. The precision processing module 203 is used to process the preliminary denoised image using a pre-trained U-Net model to obtain a precision denoised image; The integrated processing module 204 is used to perform region segmentation processing on the precisely denoised image to obtain an edge image and a smoothed image; to perform a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; and to obtain a denoised ultrasound image based on the edge-denoised image and the smoothed-denoised image.
[0057] As can be seen, the embodiments of the present invention acquire a two-dimensional ultrasound image to be processed, and then perform noise evaluation processing on the two-dimensional ultrasound image to obtain a noise evaluation image. The noise evaluation image can be used to accurately partition the image noise. A first filtering algorithm and a second filtering algorithm can be used to perform preliminary processing on the partitioned noise to obtain a preliminary denoised image. Because the preliminary processing can only remove low-order, high-intensity noise, it can improve the accuracy of subsequent processing. Then, the preliminary denoised image is precisely processed using the U-Net model to obtain a precisely denoised image. However, the obtained precisely denoised image suffers from over-smoothing and edge loss. Therefore, the precisely denoised image is divided into regions to obtain an edge image and a smoothed image. Corresponding algorithms are executed on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image. Finally, the edge-denoised image and the smoothed-denoised image are weighted and fused to obtain a denoised ultrasound image, thereby improving the accuracy of ultrasound image denoising.
[0058] In an optional embodiment, the noise assessment module 201 performs noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image, which may include: Determine the gradient of all pixels in the two-dimensional ultrasound image, and select the first target pixel from all pixels in the two-dimensional ultrasound image; obtain the first noise dataset based on all the first target pixels; Determine the mean gray value of all pixels in the two-dimensional ultrasound image; based on the mean gray value, obtain the variance of all pixels, and select the second target pixels from all pixels; based on all the second target pixels, obtain the second noise dataset; Based on the mean and variance of all pixels, obtain the mean square value of all pixels; select the third target pixels from all pixels; obtain the third noise dataset based on all third target pixels; Based on the first noise dataset, the second noise dataset, and the third noise dataset, a noise assessment image is obtained.
[0059] It is evident that by identifying and classifying different types of noise in an image, and also recognizing the intensity of different noises, it is easier to perform accurate denoising subsequently. By concentrating the characteristics of all noises, the type, intensity, and distribution of each type of noise can be accurately assessed, ultimately providing a basis for decision-making regarding regional and targeted denoising, thereby improving the accuracy of denoising.
[0060] In another optional embodiment, the preliminary processing module 202 performs a first filtering algorithm on the noise evaluation image to obtain a filtered and denoised image, which may include: Determine the parameter values of the first filtering algorithm corresponding to the first noise dataset; obtain the corresponding first filtering algorithm based on the parameter values of the first filtering algorithm; process the noise evaluation image using the first filtering algorithm to obtain a filtered and denoised image; Furthermore, the filtered and denoised image is processed by a second filtering algorithm to obtain a preliminary denoised image, including: Determine the parameter values of the second filtering algorithm corresponding to different noise regions in the second noise dataset; based on the parameter values of the second filtering algorithm, obtain the second filtering algorithm corresponding to each noise region; process the filtered and denoised image using the second filtering algorithm to obtain a preliminary denoised image; As can be seen, by applying targeted matching algorithms to different noise intensities, low-intensity noise is preserved by specific algorithms, medium-intensity noise is maintained by specific algorithms, and high-intensity noise is effectively suppressed by combined algorithms. This process removes the first and second noises that are easy to handle, outputting a preliminary denoised image and clearing obstacles for subsequent denoising of key noises.
[0061] In another optional embodiment, the precision processing module 203 processes the initial denoised image using a pre-trained U-Net model to obtain a precisely denoised image, which may include: According to the pre-defined labeling algorithm in the U-Net model, the first noise dataset and the second noise dataset in the preliminary denoised image are labeled to obtain the first labeled dataset; the third noise dataset in the preliminary denoised image is labeled to obtain the second labeled dataset. The first and second labeled datasets are processed using the pre-defined convolution algorithm in the U-Net model to obtain accurately denoised images. As can be seen, the U-Net model can accurately distinguish between processed and unprocessed noise, achieve precise focusing on noisy regions, and enable the network to efficiently learn the characteristics of noise; thus, it can remove noise more accurately.
[0062] In another optional embodiment, the integration processing module 204 performs region segmentation processing on the precisely denoised image to obtain an edge-denoised image and a smoothed denoised image, which may include: Gradient calculation is performed on all pixels in the precisely denoised image to obtain the horizontal and vertical gradients of all pixels; based on the horizontal and vertical gradients of all pixels, the gradient change magnitude of each pixel is obtained. Filter all pixels to obtain the target edge pixels; obtain the edge image based on all target edge pixels; Filter all pixels to obtain the target smoothed pixels; obtain the smoothed image based on all target smoothed pixels.
[0063] As can be seen, the final post-processing of the precise image divides it into an edge image and a smooth image; this facilitates subsequent filtering algorithms in preserving details in the edge region and reducing noise in the smooth region.
[0064] In another optional embodiment, the integration processing module 204 performs a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image, which may include: The target pixels in the edge image are processed by a region filtering algorithm to obtain a first template window; based on the first template window, a first candidate template is obtained. Calculate the grayscale difference between the first template window and all first candidate templates to obtain the first grayscale difference; based on the first grayscale difference, obtain the first weight distribution corresponding to all first candidate templates; based on all first weight distributions, obtain the edge-denoised image; The target pixels in the smoothed image are processed by a region filtering algorithm to obtain a second template window; based on the second template window, a second candidate template is obtained. Calculate the grayscale difference between the second template window and all second candidate templates to obtain the second grayscale difference; based on the second grayscale difference, obtain the second weight distribution corresponding to all second candidate templates; based on all second weight distributions, obtain the smoothed and denoised image.
[0065] As can be seen, processing the edge region can prioritize the preservation of edge details and eliminate only minor artifacts; processing the smooth image can further smooth residual noise and improve the overall naturalness of the image; thus achieving a balance between thorough noise reduction and detail preservation.
[0066] In yet another alternative embodiment, the noise assessment module 201, in acquiring the two-dimensional ultrasound image to be processed, may include: Based on a preset interventional array ultrasonic transducer, ultrasonic beams are emitted to the target area by activating different arrays; according to the interventional array ultrasonic transducer, the echo signal corresponding to the ultrasonic beam is received to obtain echo data. The echo data is beamformed to obtain a beamforming line. All beamforming lines are arranged to obtain a two-dimensional ultrasound image.
[0067] As can be seen, this optional embodiment, by utilizing a novel interventional array ultrasound transducer to acquire the image to be processed, can significantly reduce crosstalk between adjacent signal lines and external electromagnetic interference, optimize the clarity and signal-to-noise ratio of intracardiac ultrasound imaging, and improve the accuracy of ultrasound image denoising.
[0068] Example 3 Please see Figure 6 , Figure 6 This is a schematic diagram of a structure including a memory and a processor device disclosed in an embodiment of the present invention. Figure 6 As shown, the image processing device based on the interventional array ultrasonic transducer may include: Memory 301 storing executable program code; Processor 302 coupled to memory 301; The processor 302 calls the executable program code stored in the memory 301 to execute some or all of the steps in any of the image processing methods based on interventional array ultrasonic transducers in Embodiment 1 of the present invention.
[0069] Example 4 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the image processing methods based on interventional array ultrasonic transducers disclosed in Embodiment 1 of this invention.
[0070] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0071] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.
[0072] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. An image processing method based on an interventional array ultrasonic transducer, characterized in that, The method includes: A two-dimensional ultrasound image to be processed is acquired; the two-dimensional ultrasound image is subjected to noise intensity assessment processing to obtain a noise assessment image; wherein, the two-dimensional ultrasound image to be processed is acquired through an interventional array ultrasound transducer; The noise evaluation image is processed by a first filtering algorithm to obtain a filtered and denoised image; the filtered and denoised image is then processed by a second filtering algorithm to obtain a preliminary denoised image. The pre-trained U-Net model is used to process the initial denoised image to obtain an accurate denoised image; The precisely denoised image is divided into regions to obtain an edge image and a smoothed image; a region filtering algorithm is applied to the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; and a denoised ultrasound image is obtained based on the edge-denoised image and the smoothed-denoised image.
2. The image processing method based on an interventional array ultrasonic transducer according to claim 1, characterized in that, The noise intensity assessment processing of the two-dimensional ultrasound image to obtain a noise assessment image includes: The gradient of all pixels in the two-dimensional ultrasound image is determined, and a first target pixel is selected from all pixels in the two-dimensional ultrasound image; a first noise dataset is obtained based on all the first target pixels; wherein the gradient of the first target pixel satisfies a preset first threshold range. The grayscale mean of all pixels in the two-dimensional ultrasound image is determined; the variance of all pixels is obtained based on the grayscale mean; a second target pixel is selected from all pixels; a second noise dataset is obtained based on all the second target pixels; wherein the variance of the second target pixel satisfies a preset second threshold range. Based on the mean and variance of all the pixels, the mean square value corresponding to all the pixels is obtained; a third target pixel is selected from all the pixels; a third noise dataset is obtained based on all the third target pixels; wherein the variance of the third target pixel satisfies a preset third threshold range; A noise assessment image is obtained based on the first noise dataset, the second noise dataset, and the third noise dataset.
3. The image processing method based on an interventional array ultrasonic transducer according to claim 2, characterized in that, The step of processing the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image includes: Determine the parameter values of the first filtering algorithm corresponding to the first noise dataset; obtain the corresponding first filtering algorithm based on the parameter values of the first filtering algorithm; process the noise evaluation image using the first filtering algorithm to obtain the filtered and denoised image; And, the filtered and denoised image is processed by a second filtering algorithm to obtain a preliminary denoised image, including: Determine the parameter values of the second filtering algorithm corresponding to different noise regions in the second noise dataset; obtain the second filtering algorithm corresponding to each noise region based on the parameter values of the second filtering algorithm; process the filtered and denoised image using the second filtering algorithm to obtain the preliminary denoised image; wherein, the noise region is used to select the distribution of different second noise intensities.
4. The image processing method based on an interventional array ultrasonic transducer according to claim 3, characterized in that, The process of using a pre-trained U-Net model to process the initial denoised image to obtain a precisely denoised image includes: According to the preset labeling algorithm in the U-Net model, the first noise dataset and the second noise dataset in the preliminary denoised image are labeled to obtain the first labeled dataset; the third noise dataset in the preliminary denoised image is labeled to obtain the second labeled dataset. The first and second labeled datasets are processed using the pre-defined convolution algorithm in the U-Net model to obtain an accurately denoised image.
5. The image processing method based on an interventional array ultrasonic transducer according to claim 4, characterized in that, The process of dividing the precisely denoised image into regions to obtain an edge-denoised image and a smoothed denoised image includes: Gradient calculation is performed on all pixels in the precise denoised image to obtain the horizontal and vertical gradients of all pixels; based on the horizontal and vertical gradients of all pixels, the gradient change magnitude corresponding to all pixels is obtained. Filter all the pixels to obtain target edge pixels; obtain an edge image based on all the target edge pixels; Filter all the pixels to obtain target smoothed pixels; obtain a smoothed image based on all the target smoothed pixels; Wherein, the gradient change amplitude of the target edge pixel satisfies a preset first division condition, and the gradient change amplitude of the target smooth pixel satisfies a preset second division condition.
6. The image processing method based on an interventional array ultrasonic transducer according to claim 5, characterized in that, The step of performing a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image includes: The target pixels in the edge image are processed by the region filtering algorithm to obtain a first template window; based on the first template window, a first candidate template is obtained; wherein, each of the first candidate templates is the same size as the first template window; Calculate the grayscale difference between the first template window and all the first candidate templates to obtain the first grayscale difference; based on the first grayscale difference, obtain the first weight distribution corresponding to all the first candidate templates; based on all the first weight distributions, obtain the edge-denoised image; The target pixels in the smoothed image are processed by the region filtering algorithm to obtain a second template window; based on the second template window, a second candidate template is obtained; wherein, each of the second candidate templates is the same size as the second template window, and the size of the second template window is larger than the size of the first template window; Calculate the grayscale difference between the second template window and all the second candidate templates to obtain the second grayscale difference; based on the second grayscale difference, obtain the second weight distribution corresponding to all the second candidate templates; based on all the second weight distributions, obtain the smoothed and denoised image.
7. The image processing method based on an interventional array ultrasonic transducer according to claim 6, characterized in that, The acquisition of the two-dimensional ultrasound image to be processed includes: Based on a preset interventional array ultrasonic transducer, ultrasonic beams are emitted to the target area by activating different arrays; according to the interventional array ultrasonic transducer, echo data is obtained by receiving the echo signal corresponding to the ultrasonic beam. The echo data is used to obtain a beamforming line through beamforming; The two-dimensional ultrasound image is obtained by arranging all the beamforming lines.
8. An image processing device based on an interventional array ultrasonic transducer, characterized in that, The device includes: A noise assessment module is used to acquire a two-dimensional ultrasound image to be processed; to perform noise intensity assessment processing on the two-dimensional ultrasound image to obtain a noise assessment image; wherein, the two-dimensional ultrasound image to be processed is acquired through an interventional array ultrasound transducer; The preliminary processing module is used to process the noise evaluation image using a first filtering algorithm to obtain a filtered and denoised image; and to process the filtered and denoised image using a second filtering algorithm to obtain a preliminary denoised image. The precision processing module is used to process the initial denoised image using a pre-trained U-Net model to obtain a precision denoised image; An integrated processing module is used to perform region segmentation processing on the precisely denoised image to obtain an edge image and a smoothed image; to perform a region filtering algorithm on the edge image and the smoothed image to obtain an edge-denoised image and a smoothed-denoised image; and to obtain a denoised ultrasound image based on the edge-denoised image and the smoothed-denoised image.
9. An apparatus comprising a memory and a processor, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the image processing method based on the interventional array ultrasonic transducer as described in any one of claims 1-7.
10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by a processor, are used to execute the image processing method based on an interventional array ultrasonic transducer as described in any one of claims 1-7.