A blood vessel stenosis degree analysis method, device, equipment and readable storage medium

Abstract

An example of the present disclosure provides a method, device, equipment and readable storage medium for analyzing the degree of vascular stenosis, wherein the steps of the method comprise: dividing CT plain scan vascular medical images into a plurality of initial vascular segments for stenosis probability prediction, and generating the stenosis probability of each initial vascular segment; predicting the stenosis probability of a cross-segment vascular segment between adjacent initial vascular segments, and generating the stenosis probability of the cross-segment vascular segment; and determining the degree of vascular stenosis according to the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment. The example of the present disclosure predicts the stenosis probability of the initial vascular segment and the cross-segment vascular segment, so that the plaque across the border is not missed by segmentation, thereby improving the accuracy of the analysis of the degree of vascular stenosis.

Description

Technical Field

[0001] This solution relates to the field of medical data processing technology. More specifically, it relates to a method, analysis device, analysis equipment, and readable storage medium for analyzing the degree of vascular stenosis. Background Technology

[0002] Medical imaging uses signal acquisition and image processing technologies to achieve visual analysis of human tissues. Common medical imaging methods include ultrasound imaging, computed tomography (CT) imaging, and magnetic resonance imaging (MRI). These methods scan target areas of the human body to generate visual medical images. Doctors can use these visualized medical images to diagnose various conditions such as vascular plaques, aneurysms, pulmonary nodules, colon polyps, and coronary artery disease, and provide treatment recommendations.

[0003] Arteries, distributed throughout the body's organs, are the largest and most vital channels, responsible for delivering oxygen- and nutrient-rich blood to every part of the body. Arteries typically withstand significant blood flow and pressure. When arteries develop plaques, thrombosis, or other lesions, their pressure-bearing capacity weakens, significantly impacting blood flow and potentially threatening a patient's life. Vascular plaques are composed of lipids, connective tissue, and other solid components from the blood, adhering to the vessel walls. These plaques can lead to narrowing of the arteries, increasing the risk of heart disease and stroke.

[0004] Plain CT scans are a commonly used imaging method. Compared to contrast-enhanced CT and MRI, they offer advantages such as faster examination, lower radiation exposure, and lower cost. However, plain CT scans are not very effective at imaging certain vascular lesions, which can affect diagnostic accuracy. For example, the following problems exist: 1. Blurred Lumen Boundaries: Plain CT scans do not use contrast agents, resulting in minimal density difference between the vessel lumen and surrounding soft tissue. In plain CT images, the CT value of a normal vessel lumen is typically 30-50 HU (Huntersfield units), while the CT value of the vessel wall is approximately 40-60 HU. Their densities are similar, leading to indistinct boundaries. When atherosclerosis occurs in the vessel wall, forming plaques, the CT value of these plaques ranges from approximately 60-150 HU, again showing minimal density difference from the normal vessel wall. Therefore, in plain CT images, it is difficult to directly distinguish between the lumen boundary, the vessel wall boundary, and the plaque boundary, posing a fundamental challenge to the accurate measurement of the degree of stenosis.

[0005] 2. High-density masking effect of calcified plaques: Vascular plaques often contain calcified components, and the CT value of calcified tissue is extremely high, usually above 400-1000 HU. In plain CT images, the calcified area appears as a bright, dense shadow, and its high-density signal can mask the lumen area it encloses, producing a "bright masking dark" effect. This effect makes it impossible to directly observe the boundary between calcification and the lumen, and it is difficult to determine the degree to which the calcification protrudes into the lumen.

[0006] 3. Beamhardening artifacts and partial volume effect: High-density calcified plaques can also produce beamhardening artifacts, which appear as radial streaks or annular dark bands at the calcification edges. Simultaneously, due to the limited slice thickness of CT scans, the CT values ​​of adjacent pixels are the average of the tissue density within a certain volume, resulting in a partial volume effect. These artifacts and effects can lead to blurred calcification edges and distorted CT values ​​of surrounding tissues, further interfering with the interpretation of the degree of stenosis.

[0007] 4. Differences in different vascular locations: The diameter, course, and plaque morphology of different vascular locations vary greatly. For example, the diameter of coronary arteries is approximately 2-5 mm, while the diameter of cerebral blood vessels is approximately 1-4 mm. Even with the same stenosis rate, the imaging manifestations in different vessels can be drastically different. The same area of ​​calcification may only result in mild stenosis in large vessels, but may lead to severe stenosis in small vessels. This variability poses a challenge to standardizing analytical criteria.

[0008] 1. Limitations of Enhanced CT Enhanced CT scans fill the blood vessel lumen with a high-density contrast agent through injection, clearly displaying the morphology and degree of stenosis, effectively overcoming the technical limitations of plain CT scans. However, enhanced CT still has the following limitations: 1. Invasive: Enhanced CT scans require intravenous injection of contrast agents, making it an invasive procedure. Some patients may be allergic to the contrast agent (approximately 0.5% to 3% incidence), and in severe cases, anaphylactic shock may occur, endangering life. Furthermore, the puncture and injection itself carries risks such as hematoma and infection.

[0009] 2. High cost: Enhanced CT scans are significantly more expensive than plain CT scans, and require consumables such as high-pressure injectors and special contrast agents, resulting in higher overall costs.

[0010] 3. High radiation dose: Enhanced CT requires multi-phase scanning (such as arterial phase, venous phase, delayed phase, etc.), and the total radiation dose is usually 2 to 4 times that of plain CT scan. Multiple examinations may accumulate a large amount of radiation exposure.

[0011] 4. Numerous contraindications: Iodine-containing contrast agents are contraindicated in patients with renal insufficiency; contrast agent injection may increase the burden on the heart in patients with heart failure; and pregnant women should avoid radiation examinations as much as possible. These contraindications limit the range of patients suitable for enhanced CT scans.

[0012] 5. Long examination time: Enhanced CT scans require multiple steps, including contrast agent injection, scanning, and delayed scanning, with a total examination time of approximately 15-30 minutes, significantly longer than the few minutes of a plain CT scan. For emergency patients or those unable to maintain a fixed position for extended periods, enhanced CT scans present significant challenges. Summary of the Invention

[0013] The purpose of this disclosure is to provide a method, apparatus, device, and readable storage medium for analyzing the degree of vascular stenosis, thereby improving the accuracy of vascular stenosis analysis.

[0014] To achieve the above objectives, the following technical solution is adopted: In a first aspect, the present disclosure provides a method for analyzing the degree of vascular stenosis, the steps of which include: The CT plain scan vascular medical image is divided into multiple initial vascular segments for stenosis probability prediction, generating the stenosis probability of each initial vascular segment; Stenosis probability is predicted for cross-segmental vascular segments between adjacent initial vascular segments, generating the stenosis probability of cross-segmental vascular segments; The degree of stenosis of a blood vessel is determined based on the stenosis probability of the initial segment and the stenosis probability of the segment spanning the vessel.

[0015] In some instances, the steps for predicting the stenosis probability by dividing a plain CT vascular medical image into multiple initial vascular segments and generating the stenosis probability for each initial vascular segment include: Based on the naming of vascular segments in CT plain vascular medical images, the CT plain vascular medical images are divided according to a predetermined distance to generate multiple initial vascular segment medical images. Using a stenosis probability prediction model, medical images of multiple initial vascular segments are used to predict the stenosis probability of each initial vascular segment.

[0016] In some instances, the step of predicting the stenosis probability of cross-segmental vascular segments between adjacent initial vascular segments to generate the stenosis probability of cross-segmental vascular segments includes: Extract the vascular segment between two adjacent initial vascular segments whose stenosis probability reaches the first threshold, and generate a cross-segment vascular segment. A stenosis probability prediction model is used to predict the stenosis probability of cross-segmental blood vessels.

[0017] In some instances, the step of extracting a vascular segment between two adjacent initial vascular segments whose stenosis probability reaches a first threshold, and generating a cross-segment vascular segment, includes: Using the junction of two adjacent vascular segments where the stenosis probability reaches the first threshold as the center position, a predetermined distance is extended to both sides of the center position along the vascular direction to form a cross-segment vascular segment.

[0018] In some instances, the step of extracting a vascular segment between two adjacent initial vascular segments whose stenosis probability reaches a first threshold, and generating a cross-segment vascular segment, includes: Using a preset step size and a preset stage length, a cross-segment vascular segment is formed between two adjacent initial vascular segments where the stenosis probability reaches the first threshold.

[0019] In some instances, the step of determining the degree of vascular stenosis based on the stenosis probability of the initial vascular segment and the stenosis probability of the trans-segment vascular segment includes: Based on the stenosis probability of the trans-segmental vessel segment and the stenosis probability of the initial vessel segment, the vessel segment in which the plaque is located is estimated. The degree of vascular stenosis is determined based on the probability of stenosis in the segment of the blood vessel where the plaque is located.

[0020] In some instances, the step of estimating the vascular segment in which the plaque is located based on the stenosis probability of the transsegmental vascular segment and the stenosis probability of the initial vascular segment includes: The stenosis probability of the trans-segmental vascular segment is compared with the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment; If the stenosis probability of a trans-segmental vascular segment is greater than or equal to the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the trans-segmental vascular segment. If the stenosis probability of the trans-segmental vascular segment is less than the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the initial vascular segment surrounding the trans-segmental vascular segment.

[0021] Secondly, an example of this disclosure provides a device for analyzing the degree of vascular stenosis, the device comprising: The initial prediction module divides the CT plain scan vascular medical image into multiple initial vascular segments to predict the stenosis probability and generates the stenosis probability of each initial vascular segment. The cross-segment prediction module predicts the stenosis probability of cross-segment blood vessels between adjacent initial blood vessel segments and generates the stenosis probability of the cross-segment blood vessel segments. The plaque severity analysis module determines the degree of vascular stenosis based on the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment.

[0022] Thirdly, an example of this disclosure provides a device for analyzing the degree of vascular stenosis, comprising: a memory, one or more processors; the memory and the processors are connected via a communication bus; the processors are configured to execute instructions in the memory; the memory stores instructions for performing the various steps of the method described above.

[0023] Fourthly, an example of this disclosure provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described above.

[0024] The beneficial effects of the examples disclosed herein 1) The examples disclosed herein, by adding the prediction of the stenosis probability of cross-segmental vascular segments between adjacent initial vascular segments based on the initial vascular segment analysis, can effectively solve the problem of low recognition caused by the plaque information being segmented into two segments when the plaque is located at the vascular segment boundary, and improve the accuracy of vascular stenosis degree analysis.

[0025] 2) The examples disclosed herein establish a comparison and determination mechanism between the initial vascular segment and the cross-segment vascular segment: by comparing the stenosis probability of the cross-segment vascular segment with the stenosis probability of the surrounding initial vascular segment, it is possible to accurately locate whether the plaque is located in the cross-segment or the initial segment, thus avoiding missed detection or underestimation of risk of plaques in the cross-segment location.

[0026] 3) The examples disclosed herein utilize a stenosis probability prediction model trained with digital subtraction angiography data as the gold standard. It can accurately predict the probability of vascular stenosis based solely on CT plain scan images without the need for contrast agents, thus expanding the applicable population for vascular stenosis screening. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this disclosure, 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 in the examples of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A schematic diagram of the method for analyzing the degree of vascular stenosis described in this scheme is shown; Figure 2 This diagram illustrates an example of the initial vascular segment stenosis probability prediction process described in this scheme. Figure 3 A schematic diagram illustrating an example of the initial vascular segmentation method described in this scheme; Figure 4 This diagram illustrates an example of the cross-segmental vascular stenosis probability prediction process described in this scheme. Figure 5 A schematic diagram illustrating an example of the process for determining the degree of vascular stenosis as described in this scheme; Figure 6 A schematic diagram illustrating an example of the vascular stenosis analysis device described in this solution; Figure 7 This diagram illustrates an example of the vascular stenosis analysis device described in this solution. Detailed Implementation

[0029] To make the technical solutions and advantages in the embodiments of this disclosure clearer, the implementation methods in the embodiments of this disclosure will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0030] In clinical vascular imaging analysis, to achieve a precise assessment of vascular lesions in plain CT scans, it is usually necessary to divide the entire vessel into multiple segments for independent analysis. While this segment-based approach facilitates the organization and management of the analysis task, it also introduces a fundamental technical drawback: when a vascular plaque is located precisely at the boundary between two adjacent segments, the plaque information is mechanically divided into two segments, and each segment can only obtain partial, rather than complete, information about the plaque.

[0031] Suppose a plaque exists in a blood vessel at location X, with its physical boundary spanning the end point of segment A and the beginning point of segment B. In a traditional segment-independent analysis framework, the end point of segment A is defined as location X. Therefore, segment A can only identify a portion of the plaque signal at location X, which may only account for 30% to 70% of the plaque information. Similarly, the beginning point of segment B is defined as location X, so segment B can only identify another portion of the plaque signal at location X. As a result, the originally complete plaque information is rigidly segmented into two segments, each containing incomplete information. This fragmentation directly leads to a systematic underestimation of the degree of plaque stenosis.

[0032] Therefore, this disclosure aims to provide a method, apparatus, device, and readable storage medium for analyzing the degree of vascular stenosis, in order to address the problem of low recognition when plaques are located at the boundary of vascular segments and improve the accuracy of vascular stenosis analysis in CT plain scan images.

[0033] The following section will describe in detail a method for analyzing the degree of vascular stenosis proposed in this scheme, with reference to the accompanying drawings. Specifically, as shown in the figures... Figure 1 As shown, the method includes: S1. Divide the CT plain scan vascular medical image into multiple initial vascular segments for stenosis probability prediction and generate the stenosis probability of each initial vascular segment. S2. Predict the stenosis probability of cross-segmental vascular segments between adjacent initial vascular segments to generate the stenosis probability of cross-segmental vascular segments; S3. Determine the degree of stenosis of the blood vessel based on the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment.

[0034] In this disclosed example, a plain CT scanner can be used to scan a target physiological tissue area to obtain initial medical images.

[0035] Taking coronary arteries as an example, a plain CT scanner is used to scan the patient's chest area to obtain initial medical images. The scan range covers the entire heart region, and the scanning parameters are set to a tube voltage of 120kVp, an automatic adjustment of the tube current based on the patient's body size, a slice thickness of 0.5-1.0mm, and a reconstruction interval of 0.5mm. Because a plain scan is used, no contrast agent is required, making the examination convenient and particularly suitable for screening the degree of vascular stenosis in large populations.

[0036] Then, using a physiological tissue segmentation model, the initial medical image is segmented to obtain a plain CT medical image of the target physiological tissue region.

[0037] Physiological tissue segmentation models can be based on deep learning-trained neural network models, such as U-Net, ResNet, and nnU-Net architectures. These models can automatically identify and segment the heart region, removing irrelevant structures like the thoracic cavity and lung tissue, and outputting only plain CT images containing the heart region. The segmentation model, based on the nnU-Net architecture, is trained in a supervised manner on a large number of CT images with labeled heart boundaries, enabling accurate segmentation of the heart region.

[0038] Next, the vascular segmentation model was used to segment the CT plain scan medical images to obtain CT plain scan vascular medical images.

[0039] Vascular segmentation models are used to further segment vascular structures from plain CT images of the cardiac region. These models can be deep learning networks based on a 3D U-Net architecture, taking the plain CT image of the cardiac region as input and outputting the segmented results of the coronary arteries (i.e., plain CT vascular images). Because the contrast between blood vessels and surrounding tissues is low in plain CT images, vascular segmentation models require specially designed attention mechanisms and feature enhancement modules to improve segmentation accuracy. For example, vascular segmentation models can employ multi-scale feature fusion and channel attention mechanisms to enhance the segmentation ability of small vascular branches.

[0040] After obtaining plain CT vascular images, the images can be divided into multiple initial vascular segments according to a predetermined division method. Then, the stenosis probability of each initial vascular segment is predicted, generating a stenosis probability for each segment. Specifically, for example... Figure 2 As shown, S101. Based on the naming of vascular segments in the CT plain scan vascular medical images, divide the CT plain scan vascular medical images according to a predetermined distance to generate multiple initial vascular segment medical images.

[0041] Taking the coronary arteries as an example, the coronary arteries have a standard segment naming system. For example, the modified AHA classification system divides the coronary arteries into segments such as the left main coronary artery (LM), the proximal segment of the left anterior descending artery (pLAD), the mid-segment of the left anterior descending artery (mLAD), the distal segment of the left anterior descending artery (dLAD), the diagonal branches (D1, D2), the proximal segment of the left circumflex artery (pLCX), the distal segment of the left circumflex artery (dLCX), the obtuse marginal branch (OM), the proximal segment of the right coronary artery (pRCA), the mid-segment of the right coronary artery (mRCA), the distal segment of the right coronary artery (dRCA), the posterior descending artery (PDA), and the posterior collateral branch (PLB).

[0042] like Figure 3 As shown, the coronary arteries are divided according to predetermined distances, that is, each vessel is divided into several initial vascular segments according to a certain length (such as 5mm, 10mm, etc.). The predetermined distance can be set according to the vessel diameter and lesion characteristics: for proximal vessels with larger diameters (such as the left main coronary artery and the proximal segment of the left anterior descending artery), the predetermined distance can be set to 10mm; for distal vessels with smaller diameters (such as the distal segment of the left anterior descending artery and the diagonal branch), the predetermined distance can be set to 5mm. The medical image of each initial vascular segment includes the morphology and density information of the vessel within that segment.

[0043] S102. Using a stenosis probability prediction model, predict the stenosis probability of each initial vascular segment based on medical images.

[0044] The stenosis probability prediction model is a model used to identify the probability of vascular stenosis by using digital subtraction angiography (DSA) data with vascular calcification as the gold standard and iteratively training it with a deep learning neural network.

[0045] Specifically, the training process for the narrow probability prediction model is as follows: i. Collect a large number of patients' CT plain scan images and corresponding DSA image data. DSA images are the gold standard for angiography and can clearly show the morphology and degree of stenosis of the vascular lumen. ii. Mark the location and degree of vascular stenosis in DSA images as the gold standard label; iii. Align the vascular segments in the CT plain scan images with the stenosis markers in the DSA images to construct training sample pairs from CT plain scan images to stenosis probabilities; iv. Using deep learning neural networks (such as 3D CNN, ResNet, etc.), with CT plain scan vascular segment images as input and the stenosis probability of the DSA gold standard as the supervision signal, end-to-end supervised training is performed. v. Through multiple rounds of iterative training, the network learns the mapping relationship between the characteristic manifestations of vascular stenosis in CT plain scan images and the stenosis probability of the DSA gold standard, and generates a stenosis probability prediction model.

[0046] After training, the stenosis probability prediction model can output a stenosis probability value (such as a continuous value between 0 and 1, or a discrete stenosis level) for a given CT scan image of a vascular segment. A higher stenosis probability indicates a greater likelihood that plaque is present in that vascular segment, leading to stenosis.

[0047] By inputting the medical images of each initial vascular segment into the stenosis probability prediction model, the stenosis probability of each initial vascular segment can be obtained. For example, the stenosis probability of the proximal segment of the left anterior descending artery is 0.72, the stenosis probability of the mid-segment of the left anterior descending artery is 0.15, and the stenosis probability of the proximal segment of the right coronary artery is 0.85, etc.

[0048] Step S2: Predict the stenosis probability of cross-segmental vascular segments between adjacent initial vascular segments to generate the stenosis probability of cross-segmental vascular segments.

[0049] Step S2 is the core step in this disclosure example, designed to address the issue of low recognition rates when plaques are located at the boundaries of vascular segments.

[0050] In step S2, each initial vascular segment is independently predicted for stenosis probability. When a plaque is located at the boundary between two adjacent initial vascular segments, the plaque is divided into two different segments. The plaque information contained in a single segment is incomplete, resulting in both segments having low predicted stenosis probability values, which in turn underestimates the degree of vascular stenosis at that location.

[0051] To address this issue, this step predicts the stenosis probability of cross-segmental vessels between adjacent initial vascular segments, such as... Figure 4 As shown.

[0052] S201. Extract the vascular segment between two adjacent initial vascular segments whose stenosis probability reaches the first threshold, and generate a cross-segment vascular segment.

[0053] The first threshold is a criterion used to screen initial vascular segments that require further cross-segment analysis. When the stenosis probability of an initial vascular segment reaches the first threshold, it indicates a high probability of plaque presence in that segment, necessitating further examination of the boundary region between that segment and adjacent segments. The first threshold can be set according to clinical needs, for example, to 0.3 or 0.5.

[0054] In the examples disclosed herein, trans-segmental vascular segments can be generated using the following two methods: The first method: taking the junction of two adjacent vascular segments with a stenosis probability reaching the first threshold as the center position, and extending a predetermined distance to both sides of the center position along the direction of the blood vessel to form a cross-segment vascular segment.

[0055] Specifically, for two adjacent initial vascular segments, such as the i-th segment and the (i+1)-th segment, where i is a positive integer, the cross-segment vascular segment extends a predetermined distance d to both sides along the vascular direction, centered on the boundary point, to form a cross-segment vascular segment of length 2d.

[0056] The selection of the predetermined distance d needs to take into account the following factors: ① Typical plaque size: The length of vascular plaque along the direction of the blood vessel is usually between 5 and 30 mm, with an average of about 15 mm. In order to cover most of the plaques that may be located near the boundary, the predetermined distance d should not be less than the average length of the plaque; ②CT scan slice thickness: In order to ensure the complete display of cross-segmental blood vessel segments in CT images, the predetermined distance d should be greater than 3 to 5 times the CT slice thickness; ③ Computational efficiency: The larger the predetermined distance d, the more blood vessel segments are spanned, and the greater the computational load. A balance needs to be struck between coverage and computational efficiency.

[0057] like Figure 3 As shown, assume that initial vascular segments A and B are two adjacent segments, and both have a stenosis probability reaching the first threshold. The junction of initial vascular segments A and B is point O. Extending a predetermined distance d (e.g., 5 mm) to both sides along the vessel's course, centered at point O, forms a vascular segment of length 2d, which is the transsegmental vascular segment. This transsegmental vascular segment covers the boundary region of initial vascular segments A and B, containing plaque information that may be segmented into the two segments. The predetermined distance d can be adjusted according to the vessel diameter and lesion characteristics: for larger vessels, d can be set to 5-10 mm; for smaller vessels, d can be set to 3-5 mm. The predetermined distance d should ensure that the transsegmental vascular segment can cover the possible plaque range in the boundary region, while not over-extending to introduce too much irrelevant information.

[0058] Taking the coronary arteries as an example, suppose the proximal left anterior descending artery (pLAD) and the mid-left anterior descending artery (mLAD) are two adjacent initial vascular segments with stenosis probabilities of 0.72 and 0.15, respectively. If the first threshold is set to 0.3, only pLAD reaches the first threshold, while mLAD does not, and in this case, there is no need to perform cross-segment analysis between these two segments. Suppose the stenosis probabilities of the proximal right coronary artery (pRCA) and the mid-right coronary artery (mRCA) are 0.85 and 0.45, respectively, and both reach the first threshold of 0.3. Then, taking the junction of pRCA and mRCA as the center, extend 5 mm to each side to form a cross-segment vascular segment with a length of 10 mm, and predict the stenosis probability of this cross-segment vascular segment.

[0059] The second method involves forming a cross-segment vascular segment between two adjacent initial vascular segments where the stenosis probability reaches the first threshold, using a preset step size and a preset stage length.

[0060] Specifically, assuming that initial vascular segments A and B are adjacent segments, and both have a stenosis probability reaching a first threshold, a sliding motion is performed between initial vascular segments A and B along the vessel's course at a preset step size s (e.g., 2 mm). Each time, a segment of a preset length L (e.g., 10 mm) is selected as a cross-segment vascular segment. This method allows for the acquisition of boundary regions between initial vascular segments with finer granularity, ensuring no possible plaque locations are missed. This is particularly beneficial for cerebral vessels, which have tortuous courses and significant diameter variations, ensuring comprehensive coverage of cross-segment detection.

[0061] S202. Use the stenosis probability prediction model to predict the stenosis probability of cross-segmental blood vessel segments and generate the stenosis probability of cross-segmental blood vessel segments.

[0062] Medical images of cross-segment vessels are input into a stenosis probability prediction model, which outputs the stenosis probability of the cross-segment vessel. Since the cross-segment vessel contains complete information about the boundary region of the initial vessel segment, if the plaque is indeed located at the boundary, the stenosis probability of the cross-segment vessel will be higher than the stenosis probability of the segmented individual initial vessel segment.

[0063] Specifically, for each cross-segment vascular segment, the same stenosis probability prediction model as in step S2 is used to predict and generate the stenosis probability P_cross_j of the cross-segment vascular segment, where j is the cross-segment number, j=i, and j is a positive integer.

[0064] The probability of stenosis across a segmental vessel reflects the information density of plaque crossing that boundary. A higher P_cross value indicates a plaque body located near the boundary, while a lower P_cross value indicates a plaque located within an initial segment far from the boundary.

[0065] In the examples disclosed herein, after predicting the probability of stenosis in the initial vascular segment and the probability of stenosis in the trans-segment vascular segment, the degree of stenosis can be further determined. Specifically, as... Figure 5 As shown, S301: Estimate the vascular segment in which the plaque is located based on the stenosis probability of the trans-segmental vascular segment and the stenosis probability of the initial vascular segment.

[0066] This step introduces the concept of plaque location localization. In traditional segmental analysis methods, due to boundary segmentation, plaque information is dispersed across multiple segments, with each segment only acquiring partial information about the plaque. Therefore, by comparing the stenosis probability of a cross-segmental vessel segment with the stenosis probability of the surrounding initial vessel segments, the optimal vessel segment containing the complete plaque or most of the plaque can be estimated, thereby determining the plaque's location information.

[0067] In the process of plaque location estimation, it is necessary to compare the stenosis probability of the cross-segment vascular segment with the stenosis probability of the initial vascular segment surrounding the cross-segment vascular segment.

[0068] For example, for the j-th segment of a trans-segment vessel, its two sides are the i-th initial segment and the (i+1)-th initial segment, respectively. The three probability values ​​to be compared are: P_cross_j: The probability of stenosis in the j-th cross-segment of the blood vessel; P_initial_i: The stenosis probability of the i-th initial vascular segment; P_initial_i+1: The stenosis probability of the (i+1)th initial vascular segment.

[0069] The comparison can be performed by taking the maximum value. The larger value between P_initial_i and P_initial_i+1 is taken as the "probability of peripheral initial segment stenosis" P_initial_max_i, and then compared with P_cross_j: if P_cross_j ≥ P_initial_max_i, the plaque is located in the cross-segment vessel segment; otherwise, the plaque is located in the initial segment vessel segment.

[0070] If the stenosis probability of a trans-segmental vascular segment is greater than or equal to the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the trans-segmental vascular segment.

[0071] Judgment condition: P_cross_j ≥ P_initial_max_i; Determination: The optimal location of the intact plaque or most of the plaque is in the j-th cross-segment vascular segment.

[0072] If the stenosis probability of a trans-segmental vascular segment is less than the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the initial vascular segment.

[0073] Decision condition: P_cross_j < P_initial_max_i; Determination: The optimal location of a complete plaque or most of a plaque is at the i-th initial vascular segment or the (i+1)-th initial vascular segment. It is also possible that plaques of different sizes exist at the i-th initial vascular segment or the (i+1)-th initial vascular segment.

[0074] S302: Determine the degree of stenosis of the blood vessel based on the probability of stenosis in the segment of the blood vessel where the plaque is located.

[0075] Based on the above judgment results, the stenosis probability of the vascular segment where the plaque is located is taken as the assessment value of the degree of vascular stenosis.

[0076] For example, the degree of vascular stenosis can be classified into three levels based on the probability of stenosis: In the examples disclosed herein, the degree of narrowness can be represented in the following ways: (a) Single numerical representation: Take the representative value (such as the median) of the stenosis probability of the segment where the plaque is located as a quantitative indicator of the degree of stenosis; (b) Interval representation: At the same time, a narrow probability confidence interval (e.g., P25~P75) is given to reflect the uncertainty of the assessment; (c) Grade representation: Map the stenosis probability to a three-level classification result (mild, moderate, severe) to facilitate rapid clinical decision-making.

[0077] For example, the degree of vascular stenosis can be divided into 5 levels, such as: Grade 1: Stenosis probability 0-0.2%, essentially no stenosis; Grade 2: Stenosis probability 0.2-0.4, mild stenosis; Grade 3: Stenosis probability 0.4-0.6, moderate stenosis; Grade 4: Stenosis probability 0.6-0.8, severe stenosis; Grade 5: Stenosis probability 0.8-1.0, extremely severe stenosis.

[0078] Based on the stenosis probability corresponding to the determined plaque location (segmental or initial vascular segment), the degree of vascular stenosis in the current vascular region can be determined by referring to the above grading criteria.

[0079] By predicting the probability of stenosis across different segments of blood vessels, plaques located at segmental boundaries can be effectively detected. Through the comparison and judgment mechanism between the probability of stenosis across segments and the initial segment, the location of plaques can be accurately determined, avoiding missed detection or underestimation of risk of plaques in cross-segment locations. At the same time, the entire analysis process only requires plain CT images and does not require the injection of contrast agents, thus expanding the applicable population for vascular stenosis screening.

[0080] Based on the above-described method for analyzing the degree of vascular stenosis, this disclosure further provides a device 4 for analyzing the degree of vascular stenosis, such as... Figure 6 As shown, a vascular stenosis degree analysis device 4 includes an initial prediction module 401, a cross-segment prediction module 402, and a plaque degree analysis module 403.

[0081] The initial prediction module 401 is used to divide the CT plain scan vascular medical image into multiple initial vascular segments for stenosis probability prediction and generate the stenosis probability of each initial vascular segment.

[0082] Specifically, the initial prediction module 401 includes a segmentation unit and an initial prediction unit.

[0083] The segmentation unit is used to divide CT plain vascular images into multiple initial vascular segment images according to the naming of vascular segments and a predetermined distance. The segmentation unit incorporates a standard vascular segment naming system (such as the AHA classification standard for coronary arteries and the segment naming standard for cerebral vessels), which can automatically identify the segment boundaries in the vascular image and subdivide them according to the predetermined distance. The predetermined distance can be adaptively adjusted according to the vessel diameter and lesion characteristics: a larger predetermined distance is used for vascular segments with larger diameters, and a smaller predetermined distance is used for vascular segments with smaller diameters.

[0084] The initial prediction unit utilizes a stenosis probability prediction model to predict the stenosis probability of multiple initial vascular segments from medical images, generating a stenosis probability for each segment. The stenosis probability prediction model integrated within the initial prediction unit uses digital subtraction angiography (DSA) data of vascular calcification as the gold standard and is generated through iterative training using a deep learning neural network. The stenosis probability prediction model can output the stenosis probability value for a given CT plain scan vascular segment image based on the input image.

[0085] The cross-segment prediction module 402 is used to predict the stenosis probability of cross-segment vascular segments between adjacent initial vascular segments and generate the stenosis probability of cross-segment vascular segments.

[0086] Specifically, the cross-segment prediction module 402 includes a threshold filtering unit, a cross-segment construction unit, and a cross-segment prediction unit.

[0087] The threshold screening unit is used to screen initial vascular segments whose stenosis probability reaches a first threshold. When the stenosis probability of two adjacent initial vascular segments both reach the first threshold, a cross-segment analysis process is triggered. The threshold screening unit supports automatic setting and manual adjustment of the first threshold, and different thresholds can be set according to different vascular regions and clinical needs.

[0088] The cross-segment construction unit is used to extract vascular segments between two adjacent initial vascular segments whose stenosis probability reaches a first threshold, generating cross-segment vascular segments. The cross-segment construction unit supports two construction methods: Method 1: Using the junction of two adjacent vascular segments with a stenosis probability reaching the first threshold as the center position, extend a predetermined distance to both sides of the center position along the vascular course to form a cross-segment vascular segment. The cross-segment vascular segment constructed in this way covers the area on both sides of the segment boundary with the segment boundary as the center, which is suitable for the directional detection of plaques at the boundary position.

[0089] Method 2: Using a preset step size and preset stage length, a cross-segment vascular segment is formed between two adjacent initial vascular segments where the stenosis probability reaches the first threshold. This method covers the boundary area with fine granularity and is suitable for blood vessels with tortuous course and large diameter variations.

[0090] The cross-segment prediction unit is used to predict the stenosis probability of cross-segment vessels using a stenosis probability prediction model, generating the stenosis probability of the cross-segment vessel. For multiple cross-segment vessels generated using a sliding window method, the cross-segment prediction unit predicts the stenosis probability of each cross-segment vessel separately, and takes the maximum value as the representative stenosis probability of the cross-segment vessel in that connecting region.

[0091] The plaque severity analysis module 403 is used to determine the degree of stenosis of a blood vessel based on the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment.

[0092] Specifically, the patch severity analysis module 403 includes a probability comparison unit, a patch location unit, and a risk classification unit.

[0093] The probability comparison unit is used to compare the stenosis probability of a trans-segmental vascular segment with the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment.

[0094] The plaque localization unit is used to determine the plaque location based on the comparison results: if the stenosis probability of the cross-segment vessel segment is greater than or equal to the stenosis probability of the surrounding initial vessel segment, the plaque is located in the cross-segment vessel segment; if the stenosis probability of the cross-segment vessel segment is less than the stenosis probability of the surrounding initial vessel segment, the plaque is located in the initial vessel segment.

[0095] The risk grading unit determines the degree of vascular stenosis in the current vascular region based on the probability of stenosis corresponding to the location of plaques. Vascular stenosis is classified into 5 levels, and the risk grading unit has a built-in grading standard that can automatically output the degree of stenosis based on the probability of stenosis. The risk grading unit also supports outputting plaque location annotation maps, marking the specific location of plaques (segmental location or initial segment location) and the corresponding degree of stenosis on the vascular image, allowing doctors to intuitively understand the plaque distribution.

[0096] Furthermore, in the examples disclosed herein, an image acquisition module can be further configured in the vascular stenosis analysis device to acquire CT plain scan vascular medical images. This image acquisition module includes an image receiving subunit, a tissue segmentation subunit, and a vascular segmentation subunit. The image receiving subunit receives initial medical images from the CT equipment, supports reading multiple medical image standard formats such as DICOM, and can interface with a hospital image archiving and communication system (PACS) to achieve automatic acquisition and transmission of image data. The tissue segmentation subunit integrates a physiological tissue segmentation model, capable of automatically segmenting the initial medical images and outputting CT plain scan medical images of the target physiological tissue region. The tissue segmentation subunit incorporates multiple physiological tissue segmentation models, automatically selecting the corresponding segmentation model based on the type of the target region: the heart segmentation model is trained based on the nnU-Net architecture, the head and neck segmentation model is trained based on the 3D U-Net architecture, etc. The vascular segmentation subunit integrates a vascular segmentation model, capable of further segmenting vascular structures from the CT plain scan medical images of the target physiological tissue region and outputting CT plain scan vascular medical images. The vascular segmentation model employs multi-scale feature fusion and channel attention mechanisms to enhance the segmentation capability of small vascular branches.

[0097] Based on the above-described embodiments of the method for analyzing the degree of vascular stenosis, this solution further provides a computer-readable storage medium. This computer-readable storage medium is used to implement the program product of the above-described method for analyzing the degree of vascular stenosis. It may employ a portable compact disk read-only memory (CD-ROM) and include program code, and can run on a device, such as a personal computer. However, the program product of this solution is not limited to this. In this document, the readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0098] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0099] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0100] Program code contained on a computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0101] Program code for performing the operations of this solution can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0102] Based on the above-described method for analyzing the degree of vascular stenosis, this solution further provides an electronic device. For example... Figure 7 The electronic device shown is merely an example and should not impose any limitations on the functionality and scope of use of this embodiment.

[0103] like Figure 7As shown, electronic device 5 is represented in the form of a general-purpose computing device. The components of electronic device 5 may include, but are not limited to: at least one storage unit 501, at least one processing unit 502, a display unit 505, and a bus 503 for connecting different system components.

[0104] The storage unit 501 stores program code that can be executed by the processing unit 502, causing the processing unit 502 to perform the steps of the various exemplary embodiments described in the above-described method for analyzing the degree of vascular stenosis. For example, the processing unit 502 can perform actions such as... Figure 1 The steps are shown in the figure.

[0105] Storage unit 501 may include volatile storage units, such as random access memory (RAM) and / or cache storage units, and may further include read-only memory (ROM).

[0106] Storage unit 501 may also include programs / utilities with program modules, such program modules including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0107] Bus 503 may include a data bus, an address bus, and a control bus.

[0108] Electronic device 5 can also communicate with one or more external devices 506 (e.g., keyboard, pointing device, Bluetooth device, etc.), which can be done through input / output (I / O) interface 504. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 5, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0109] Obviously, the above embodiments in the examples of this disclosure are merely examples for clearly illustrating the examples of this disclosure, and are not intended to limit the implementation of the examples of this disclosure. For those skilled in the art, other variations or modifications can be made based on the above description. It is impossible to exhaustively list all implementation methods here. Any obvious variations or modifications derived from the technical solutions in the examples of this disclosure are still within the protection scope of the examples of this disclosure.

Claims

1. A method for analyzing the degree of vascular stenosis, characterized in that, The steps of this method include: The CT plain scan vascular medical image is divided into multiple initial vascular segments for stenosis probability prediction, generating the stenosis probability of each initial vascular segment; Stenosis probability is predicted for cross-segmental vascular segments between adjacent initial vascular segments, generating the stenosis probability of cross-segmental vascular segments; The degree of stenosis of a blood vessel is determined based on the stenosis probability of the initial segment and the stenosis probability of the segment spanning the vessel.

2. The method for analyzing the degree of vascular stenosis according to claim 1, characterized in that, The steps for dividing a plain CT vascular image into multiple initial vascular segments and predicting the stenosis probability of each initial vascular segment include: Based on the naming of vascular segments in CT plain vascular medical images, the CT plain vascular medical images are divided according to a predetermined distance to generate multiple initial vascular segment medical images. Using a stenosis probability prediction model, medical images of multiple initial vascular segments are used to predict the stenosis probability of each initial vascular segment.

3. The method for analyzing the degree of vascular stenosis according to claim 1, characterized in that, The step of predicting the stenosis probability of cross-segment vascular segments between adjacent initial vascular segments and generating the stenosis probability of cross-segment vascular segments includes: Extract the vascular segment between two adjacent initial vascular segments whose stenosis probability reaches the first threshold, and generate a cross-segment vascular segment. A stenosis probability prediction model is used to predict the stenosis probability of cross-segmental blood vessels.

4. The method for analyzing the degree of vascular stenosis according to claim 3, characterized in that, The step of extracting a vascular segment between two adjacent initial vascular segments whose stenosis probability reaches a first threshold, and generating a cross-segment vascular segment, includes: Using the junction of two adjacent vascular segments where the stenosis probability reaches the first threshold as the center position, a predetermined distance is extended to both sides of the center position along the vascular direction to form a cross-segment vascular segment.

5. The method for analyzing the degree of vascular stenosis according to claim 3, characterized in that, The step of extracting a vascular segment between two adjacent initial vascular segments whose stenosis probability reaches a first threshold, and generating a cross-segment vascular segment, includes: Using a preset step size and a preset stage length, a cross-segment vascular segment is formed between two adjacent initial vascular segments where the stenosis probability reaches the first threshold.

6. The method for analyzing the degree of vascular stenosis according to claim 1, characterized in that, The step of determining the degree of vascular stenosis based on the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment includes: Based on the stenosis probability of the trans-segmental vessel segment and the stenosis probability of the initial vessel segment, the vessel segment in which the plaque is located is estimated. The degree of vascular stenosis is determined based on the probability of stenosis in the segment of the blood vessel where the plaque is located.

7. The method for analyzing the degree of vascular stenosis according to claim 6, characterized in that, The step of estimating the vascular segment in which the plaque is located based on the stenosis probability of the trans-segmental vascular segment and the stenosis probability of the initial vascular segment includes: The stenosis probability of the trans-segmental vascular segment is compared with the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment; If the stenosis probability of a trans-segmental vascular segment is greater than or equal to the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the trans-segmental vascular segment. If the stenosis probability of the trans-segmental vascular segment is less than the stenosis probability of the initial vascular segment surrounding the trans-segmental vascular segment, then the plaque is located in the initial vascular segment surrounding the trans-segmental vascular segment.

8. A device for analyzing the degree of vascular stenosis, characterized in that, The device includes: The initial prediction module divides the CT plain scan vascular medical image into multiple initial vascular segments to predict the stenosis probability and generates the stenosis probability of each initial vascular segment. The cross-segment prediction module predicts the stenosis probability of cross-segment blood vessels between adjacent initial blood vessel segments and generates the stenosis probability of the cross-segment blood vessel segments. The plaque severity analysis module determines the degree of vascular stenosis based on the stenosis probability of the initial vascular segment and the stenosis probability of the cross-segment vascular segment.

9. A device for analyzing the degree of vascular stenosis, characterized in that, include: Memory, one or more processors; The memory and processor are connected via a communication bus; the processor is configured to execute instructions from the memory. The memory stores instructions for performing the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images