Renal artery image acquisition and processing method

Abstract

The invention discloses a renal artery image acquisition and processing method. The method comprises the following steps: acquiring high-resolution images of a renal artery and surrounding tissues thereof; applying an image denoising method to the high-resolution image to improve the image quality, enhance the image contrast and highlight the difference between the renal artery and its branches and surrounding tissues so as to obtain a preprocessed image; marking the acquired pre-processed image, and positioning and identifying a renal artery and a branch; segmenting the renal artery and the branches, and measuring key parameters of the segmented renal artery and the branches; and based on an AI algorithm, distinguishing the ablatable area, the to-be-ablated area and the non-ablatable area, and performing identification to obtain an identification image. According to the method, the high-resolution renal artery image is acquired, and the ablatable, non-ablatable and undetermined areas are identified and marked, so that the image precision and the ablation precision are improved, and the damage to normal tissues and the risk that the preset ablation degree is not reached are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of renal artery image processing, and in particular relates to a method for acquiring and processing a renal artery image. Background Art

[0002] Renal artery ultrasound denervation ablation technology is an emerging treatment method, which is gradually applied to the treatment of diseases such as hypertension due to its minimally invasive and real-time characteristics. However, the existing technology requires manual parameter setting based on the doctor's experience, and often faces the problem of inaccurate ablation range, which may lead to the following risks:

[0003] Ablation over a large area may cause damage to normal tissues and lead to postoperative complications, such as damage to renal function or adjacent tissues;

[0004] Insufficient ablation range will affect the treatment effect on the target tissue.

[0005] In order to improve the control over the ablation range in related technologies, image recognition is generally used to identify the renal artery to improve the accuracy of ablation. However, it is still impossible to effectively and accurately identify the renal artery, and the above-mentioned problems still exist. Summary of the invention

[0006] The purpose of the present invention is to provide a method for acquiring and processing a renal artery image to solve the problems in the prior art.

[0007] To this end, the present invention provides a method for acquiring and processing a renal artery image, comprising:

[0008] S100, obtains high-resolution images of the renal artery and surrounding tissues;

[0009] S200, applying an image denoising method to the high-resolution image to improve image quality, enhance image contrast, and highlight the difference between the renal artery and its branches and surrounding tissues to obtain a preprocessed image;

[0010] S300, marking the acquired preprocessed image to locate and identify the renal artery and its branches;

[0011] S400, segmenting the renal artery and its branches, and measuring key parameters of the segmented renal artery and its branches;

[0012] S500, based on the AI ​​algorithm, distinguishes the ablation area, the ablation area to be determined, and the non-ablation area, and marks them to obtain a marked image.

[0013] As a further description of the above technical solution, in step S300, the located and identified renal arteries and branches include at least the renal artery trunk, the primary branches of the renal arteries, the front end of the accessory renal artery, the middle section of the accessory renal artery and the proximal end of the accessory renal artery.

[0014] As a further description of the above technical solution, in step S400, the key parameters include the starting position of the renal artery and its branches, the length of the renal artery and its branches, and the diameter of the renal artery and its branches.

[0015] As a further description of the above technical solution, in step S500, the AI ​​algorithm includes:

[0016] S510, collecting and preprocessing kidney imaging videos to generate training images;

[0017] S520, based on a multi-stage cascaded convolutional neural network, performs model training based on a training image set to generate a feasible image recognition model.

[0018] As a further description of the above technical solution, the step of preprocessing the kidney visualization video in step S510 includes:

[0019] S511, extracting preset frames from the plurality of kidney development videos, and normalizing the preset frames to obtain a set of images to be processed;

[0020] S512, filtering all the image sets to be processed to reduce noise and enhance boundaries;

[0021] S513, classifying the images processed in step S512 into a training image set, a test image set, and a verification image set according to a preset ratio;

[0022] S514: semantically annotate the training image set to generate an image segmentation mask image.

[0023] As a further description of the above technical solution, the normalization process includes adjusting the multiple images to a uniform size and a uniform definition.

[0024] As a further description of the above technical solution, in step S513, the preset ratio is 5:2:1.

[0025] As a further description of the above technical solution, in S520, the image recognition model is tested using a test image set, and the correctness and accuracy of the image recognition model are tested based on the output results.

[0026] As a further description of the above technical solution, in S520, a verification image set is used to verify the feasibility of the image recognition model.

[0027] As a further description of the above technical solution, in step S500, the preceding steps of distinguishing the ablationable area, the to-be-determined ablation area, and the non-ablationable area include: identifying whether the image is a renal artery vessel;

[0028] Identify ablation-capable areas, pending ablation areas, and non-ablative areas, including malformed blood vessels, blood vessels with stents, parts within the renal parenchyma, smaller diameter branches, and areas adjacent to important tissues;

[0029] The ablation area includes the renal artery trunk, accessory renal arteries, and the extrarenal parenchymal part of the first-level branches;

[0030] The to-be-determined ablation area includes a portion between the non-ablative area and the ablative area.

[0031] Beneficial effects:

[0032] The present invention provides a method for acquiring and processing renal artery images, which acquires high-resolution images of the renal artery and surrounding tissues, and then identifies, distinguishes and marks ablative areas, non-ablative areas and pending ablation areas around the renal artery tissue in the image. This can not only effectively improve the accuracy of the image, but also effectively identify and mark different areas in the renal artery image, effectively improve the accuracy of subsequent ablation, and effectively reduce damage to normal tissues caused by ablation over a large range and the problem of failing to reach a preset ablation degree due to ablation over a smaller range. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 A schematic flow chart of the method for acquiring and processing renal artery images provided by the present invention.

[0035] Figure 2 A schematic diagram of the flow of the AI ​​algorithm in the renal artery image acquisition and processing method provided by the present invention.

[0036] Figure 3 The present invention provides a flowchart of the steps of preprocessing kidney visualization video in the method for acquiring and processing renal artery images provided by the present invention.

[0037] Figure 4 A schematic diagram of segmenting and marking a renal artery image according to the renal artery image acquisition and processing method provided by the present invention. DETAILED DESCRIPTION

[0038] The content of the present invention can be more easily understood by selecting the following detailed description of the preferred implementation method of the present invention and the embodiments included. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. When there is a conflict, the definition in this specification shall prevail.

[0039] The embodiment of the present application provides a method for acquiring and processing a renal artery image, which solves the problem that the image recognition of the renal artery in the prior art is not accurate enough, which affects the difficulty of medical staff in accurately locating the ablation position during subsequent ablation, and affects the accuracy of the tissue to be ablated and the ablation effect. The present application acquires a high-resolution image of the renal artery and surrounding tissues, and then identifies, distinguishes and marks the ablative area, non-ablative area and pending ablation area around the renal artery tissue in the image, which can not only effectively improve the accuracy of the image, but also effectively identify and mark different areas in the renal artery image, which can effectively improve the accuracy of subsequent ablation, and can effectively reduce the damage to normal tissue caused by ablation over a large range and the problem of failing to reach the preset ablation degree due to ablation over a small range.

[0040] like Figure 1-4 As shown, a method for acquiring and processing a renal artery image comprises:

[0041] S100, obtains high-resolution images of the renal artery and its surrounding tissues, specifically, can obtain high-resolution medical images of the renal artery and its surrounding structures through digital subtraction angiography, computed tomography and magnetic resonance angiography technology;

[0042] S200, applying an image denoising method to the high-resolution image to improve image quality, enhance image contrast, and highlight the difference between the renal artery and its branches and the surrounding tissues to obtain a preprocessed image. Specifically, the image quality can be improved by Wiener filtering, smoothing filtering, etc., and the influence of artifacts on subsequent analysis can be reduced to facilitate subsequent segmentation.

[0043] S300, marking the acquired preprocessed image to locate and identify the renal artery and its branches;

[0044] S400, segmenting the renal artery and its branches, and measuring key parameters of the segmented renal artery and its branches;

[0045] Specifically, medical image processing algorithms can be used. In some implementations, region growing methods, graph cut algorithms, etc. can be used to automatically identify and segment the acquired images to ensure that the renal artery and its main branches can be accurately located. By using the VTK (Visualization Toolkit) library, the two-dimensional contours of the renal artery and its main branches are extracted, and the contours are smoothed to reduce interference factors and improve recognition accuracy.

[0046] S500, based on the AI ​​algorithm, distinguishes between ablative areas, pending ablation areas, and non-ablative areas, and marks them to obtain a marked image. Specifically, through the AI ​​algorithm, after the processed image is input into the system relying on the algorithm, the ablative areas, pending ablation areas, and non-ablative areas in the image can be automatically identified, and each of the above areas can be marked at the same time, and finally a marked image can be obtained, which is convenient for medical staff to formulate ablation strategies according to the ablation location during subsequent ablation, which is more accurate and targeted, and can effectively reduce the damage to normal tissues caused by ablation over a large range and the problem of failing to reach the preset ablation level due to ablation over a small range.

[0047] Optionally, in some embodiments, the ablation-capable area, the ablation-to-be-determined area, and the non-ablation-capable area may be marked with boxes or dyes, so that medical staff can more intuitively identify each area. Figure 4 .

[0048] Optionally, in step S300, the located and identified renal arteries and branches include at least the renal artery trunk, the primary branches of the renal arteries, the front end of the accessory renal artery, the middle section of the accessory renal artery, and the proximal end of the accessory renal artery.

[0049] Optionally, in step S400, the key parameters include the starting position of the renal artery and its branches, the length of the renal artery and its branches, and the diameter of the renal artery and its branches. After obtaining the diameter of the renal artery and its branches, appropriate ablation power energy can be adapted according to the blood vessel diameter.

[0050] Optionally, in step S500, the AI ​​algorithm includes:

[0051] S510, collecting and preprocessing kidney imaging videos to generate training images, wherein the collection of kidney imaging videos includes kidney imaging videos of multiple patients, and at the same time, collecting videos of different ages, genders, normal growth, placement of vascular stents or vascular malformations, etc., and collecting kidney vascular imaging videos under different fields of view, clarity, angles and proportions to improve the accuracy of subsequent image recognition model recognition. The specific steps for image preprocessing include:

[0052] S511, extracting preset frames from the plurality of kidney development videos, and performing normalization processing on the preset frames to obtain a set of images to be processed, wherein the normalization processing includes processing the images into images of uniform size and clarity to form the set of images to be processed.

[0053] S512, filtering all the image sets to be processed to reduce noise and filter to enhance boundaries, so as to facilitate subsequent AI image recognition training;

[0054] S513, classifying the images processed in step S512 into a training image set, a test image set, and a verification image set according to a preset ratio, wherein the preset ratio is 5:2:1;

[0055] S514, semantically annotating the training image set to generate an image segmentation mask image, and the preprocessing step of the data set is now completed.

[0056] S520, based on a multi-level cascaded convolutional neural network, performs model training based on a training image set to generate a feasible image recognition model. Specifically, the currently popular medical imaging artificial intelligence Unet model can be selected. This model is essentially a multi-level cascaded convolutional neural network (CNN), which can fully extract features at different levels of the image and has sufficient learning and identification capabilities. Different model parameters can be selected when defining the Unet training model. During the model training process, the output loss and accuracy data are plotted as function curves to facilitate the adjustment of training parameters and model parameters.

[0057] Then use the test image set to predict the model and test the correctness and accuracy of the model. At the same time, use the validation image set to verify the feasibility of the generated model.

[0058] After verifying that the model is feasible, the generated image recognition model is deployed on the target platform. After the corresponding test image is input into the platform, the ablation area, the pending ablation area and the non-ablative area can be identified by the impact recognition model and marked with boxes or dyes, so that medical staff can distinguish each area more intuitively. Alternatively, a corresponding ablation system can be set up to implement different ablation strategies according to different areas to improve the accuracy and specificity of ablation, improve the ablation effect and ensure safety.

[0059] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for acquiring and processing a renal artery image, characterized in that: include: S100, obtains high-resolution images of the renal artery and surrounding tissues; S200, applying an image denoising method to the high-resolution image to improve image quality, enhance image contrast, and highlight the difference between the renal artery and its branches and surrounding tissues to obtain a preprocessed image; S300, marking the acquired preprocessed image to locate and identify the renal artery and its branches; S400, segmenting the renal artery and its branches, and measuring key parameters of the segmented renal artery and its branches; S500, based on the AI ​​algorithm, distinguishes the ablation area, the ablation area to be determined, and the non-ablation area, and marks them to obtain a marked image.

2. A method for acquiring and processing a renal artery image according to claim 1, characterized in that: In step S300, the renal arteries and branches located and identified include at least the main renal artery, the primary branches of the renal arteries, the front end of the accessory renal arteries, the middle section of the accessory renal arteries, and the proximal end of the accessory renal arteries.

3. A method for acquiring and processing a renal artery image according to claim 1, characterized in that: In step S400 , key parameters include the starting position of the renal artery and its branches, the length of the renal artery and its branches, and the diameter of the renal artery and its branches.

4. The method for acquiring and processing a renal artery image according to claim 1, characterized in that: In step S500, the AI ​​algorithm includes: S510, collecting and preprocessing kidney imaging videos to generate training images; S520, based on a multi-stage cascaded convolutional neural network, performs model training based on a training image set to generate a feasible image recognition model.

5. A method for acquiring and processing a renal artery image according to claim 4, characterized in that: In step S510, the steps of preprocessing the kidney visualization video include: S511, extracting preset frames from the plurality of kidney development videos, and normalizing the preset frames to obtain a set of images to be processed; S512, filtering all the image sets to be processed to reduce noise and enhance boundaries; S513, classifying the images processed in step S512 into a training image set, a test image set, and a verification image set according to a preset ratio; S514: semantically annotate the training image set to generate an image segmentation mask image.

6. A method for acquiring and processing a renal artery image according to claim 5, characterized in that: The normalization process includes adjusting the multiple images to a uniform size and a uniform definition.

7. A method for acquiring and processing a renal artery image according to claim 5, characterized in that: In step S513, the preset ratio is 5:2:

1.

8. The method for acquiring and processing a renal artery image according to claim 5, characterized in that: In S520, the image recognition model is tested using the test image set, and the correctness and accuracy of the image recognition model are tested based on the output results.

9. The method for acquiring and processing a renal artery image according to claim 5, characterized in that: In S520, the validation image set is used to validate the feasibility of the image recognition model.

10. The method for acquiring and processing a renal artery image according to claim 1, characterized in that: In step S500, the preceding steps of distinguishing the ablationable area, the to-be-determined ablation area, and the non-ablative area include identifying whether the image is a renal artery vessel; Identify ablation-capable areas, pending ablation areas, and non-ablative areas, including malformed blood vessels, blood vessels with stents, parts within the renal parenchyma, smaller diameter branches, and areas adjacent to important tissues; The ablation area includes the renal artery trunk, accessory renal arteries, and the extrarenal parenchymal part of the first-level branches; The to-be-determined ablation area includes a portion between the non-ablative area and the ablative area.

Images