Intelligent assessment method and system for coronary artery stenosis based on chest radiograph

Abstract

The application discloses a kind of based on chest radiograph coronary artery stenosis intelligent evaluation method and system, it is related to health risk assessment technical field, the method includes: obtaining the chest radiograph image data of target patient;Based on the time sequence fluctuation of calcification focus area in chest radiograph image data, the calcification activity index is calculated, determine the heart beat key point in chest radiograph image data, based on the displacement change between each image frame of heart beat key point, the collateral circulation compensation index of calcification focus area is calculated;Based on collateral circulation compensation index and calcification activity index, the dynamic risk factor corresponding to coronary stenosis is calculated;Dynamic risk factor is input to the preset neural network model and is iteratively trained, and based on the preset neural network model after training is completed to chest radiograph image data is predicted to process, obtains coronary stenosis risk assessment result.The application reaches the technical effect of reducing the clinical misjudgment risk of coronary artery stenosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health risk assessment, in particular to a chest radiograph coronary artery stenosis intelligent evaluation method and system. BACKGROUND

[0002] Coronary artery stenosis is an important problem in cardiovascular diseases, which is usually manifested as stenosis or obstruction of the coronary artery lumen, leading to insufficient blood supply to the myocardium, and further causing severe consequences such as angina pectoris and myocardial infarction.

[0003] Existing coronary artery stenosis intelligent evaluation methods are mostly based on extracting image features from single-frame static chest radiographs. However, chest radiography is limited by the two-dimensional projection principle and soft tissue resolution, and cannot directly visualize the fine anatomical structure of the coronary artery, resulting in a serious disconnection between the pathological anatomical information and the image features, and increasing the risk of clinical misjudgment. SUMMARY

[0004] The main purpose of the present application is to provide a chest radiograph coronary artery stenosis intelligent evaluation method and system, which aims to solve the technical problem in the related art that the extraction of image features from single-frame static chest radiographs for coronary artery stenosis evaluation leads to a serious disconnection between the pathological anatomical information and the image features, and increases the risk of clinical misjudgment.

[0005] To achieve the above-mentioned purpose, the embodiments of the present application provide a chest radiograph coronary artery stenosis intelligent evaluation method, which comprises:

[0006] Obtaining chest radiograph image data of a target patient from a preset database;

[0007] Based on the time series fluctuation of the calcification focus region in the chest radiograph image data, a calcification activity index is calculated, which is used to represent the activity of coronary stenosis progression;

[0008] Determining the heart beat key points in the chest radiograph image data, and based on the displacement change of the heart beat key points between each image frame, a collateral circulation compensation index of the calcification focus region is calculated;

[0009] Based on the collateral circulation compensation index and the calcification activity index, a dynamic risk factor corresponding to the coronary stenosis is calculated;

[0010] The dynamic risk factor is input into a preset neural network model for iterative training, and the chest radiograph image data is predicted based on the trained preset neural network model to obtain a coronary stenosis risk evaluation result.

[0011] In a possible implementation of the present application, based on the time series fluctuation of the calcification focus region in the chest radiograph image data, the calcification activity index is calculated, which comprises:

[0012] Image segmentation is performed on the chest radiograph image data to obtain a calcification mask in each image frame;

[0013] An average gray value of each calcification region in the calcification mask is calculated;

[0014] Based on the average gray value, a fluctuation intensity coefficient of each calcification region in the time sequence is calculated, and the fluctuation intensity coefficient is used to represent the time sequence fluctuation of the corresponding density of the calcification region.

[0015] An edge contour binary graph is obtained by detecting the boundary pixels of the calcification mask through an edge detection algorithm.

[0016] Based on the edge contour binary graph, an edge roughness index of each calcification region is calculated.

[0017] Based on the fluctuation intensity coefficient and the edge roughness index, a calcification activity index is calculated.

[0018] In a possible implementation of the present application, based on the edge contour binary graph, an edge roughness index of each calcification region is calculated, including:

[0019] Based on the edge contour binary graph, the contour perimeter and the equivalent circle area of the boundary pixels of the calcification region are determined.

[0020] Based on the equivalent circle area and the contour perimeter, an edge roughness index of each calcification region is calculated, and the edge roughness index is used to reflect the roughness of the edge of the calcification region.

[0021] In a possible implementation of the present application, based on the fluctuation intensity coefficient and the edge roughness index, a calcification activity index is calculated, including:

[0022] Based on the edge roughness index of each image frame, a roughness data sequence is constructed.

[0023] The roughness fluctuation intensity between the maximum sequence value and the minimum sequence value in the roughness data sequence is determined.

[0024] Based on the fluctuation intensity coefficient, the roughness fluctuation intensity, and the average value of the edge roughness index of each image frame, a calcification activity index is calculated.

[0025] In a possible implementation of the present application, based on the displacement change of the heart beat key points between each image frame, a collateral circulation compensation index of the calcification region is calculated, including:

[0026] Based on the Euclidean distance of the heart beat key points between each image frame, a displacement time sequence of each heart beat key point is determined.

[0027] Based on the displacement time sequence, the instantaneous phase time delay coefficient of each heart beat key point is calculated;

[0028] The calcification region is matched with the heart beat key point, and a plurality of key point calcification set is obtained;

[0029] The first Pearson correlation coefficient between the instantaneous phase time delay coefficient corresponding sequence and the roughness data sequence corresponding to the key point calcification set is calculated, and the first Pearson correlation coefficient is taken as the collateral circulation compensation index of the calcification region.

[0030] In a possible implementation of the present application, based on the Euclidean distance between the heart beat key points in each image frame, the displacement time sequence of each heart beat key point is determined, comprising:

[0031] For any heart beat key point, the first Euclidean distance of the key point displacement between adjacent image frames is calculated, and the absolute value of the first Euclidean distance is taken as the inter-frame displacement of each heart beat key point, the heart beat key point including the apex point, the left atrial appendage point and the ventricular diaphragm point;

[0032] Based on the inter-frame displacement, the displacement time sequence of each heart beat key point is determined.

[0033] In a possible implementation of the present application, based on the displacement time sequence, the instantaneous phase time delay coefficient of each heart beat key point is calculated, comprising:

[0034] The displacement time sequence is linearly fitted to obtain the displacement fitting curve of each heart beat key point;

[0035] The instantaneous phase information of each heart beat key point corresponding to the displacement fitting curve is determined;

[0036] For any heart beat key point, the instantaneous phase difference between the instantaneous phase information of the current heart beat key point and other different heart beat key points is calculated, and the absolute value mean of each instantaneous phase difference is taken as the instantaneous phase time delay coefficient.

[0037] In a possible implementation of the present application, the calcification region is matched with the heart beat key point to obtain a plurality of key point calcification set, comprising:

[0038] The mask gravity coordinates in the calcification mask are obtained, and the second Euclidean distance from the mask gravity coordinates to each heart beat key point is calculated;

[0039] Each calcification region is assigned to the region corresponding to the heart beat key point corresponding to the minimum second Euclidean distance, and a plurality of key point calcification set is generated.

[0040] In a possible implementation of the present application, a dynamic risk factor corresponding to coronary artery stenosis is calculated based on the collateral circulation compensation index and the calcification activity index, including:

[0041] A second Pearson correlation coefficient between the sequence corresponding to the calcification activity index and the sequence corresponding to the reciprocal of the collateral circulation compensation index is calculated, and the second Pearson correlation coefficient is taken as a correction weight.

[0042] Based on the calcification activity index, the collateral circulation compensation index and the correction weight, a dynamic risk factor corresponding to coronary artery stenosis is calculated.

[0043] To achieve the above-mentioned purpose, an intelligent evaluation system for chest radiograph coronary artery stenosis is also provided, which includes a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of the intelligent evaluation method for chest radiograph coronary artery stenosis as described above when executing the computer program.

[0044] The present application provides an intelligent evaluation method and system for chest radiograph coronary artery stenosis. In the related art, the image features are extracted from a single static chest radiograph to evaluate coronary artery stenosis, which leads to a serious disconnection between the pathological information and the image features, and increases the risk of clinical misjudgment. In the present application, the chest radiograph image data of a target patient is obtained from a preset database. First, the calcification activity index is calculated based on the time sequence fluctuation of the calcification focus region in the chest radiograph image data. The calcification activity index is used to represent the activity of coronary artery stenosis progression. Then, the heart beating key points in the chest radiograph image data are determined, and the collateral circulation compensation index of the calcification focus region is calculated through the displacement change of the heart beating key points between each image frame. Further, a dynamic risk factor corresponding to coronary artery stenosis is calculated based on the collateral circulation compensation index and the calcification activity index. The contribution of the synergistic effect of calcification activity and collateral circulation compensation capacity to the stenosis risk is quantified through the dynamic risk factor, so as to establish an accurate mapping relationship between the multi-dimensional dynamic pathological features and the stenosis degree. The preset neural network model is trained through the dynamic risk factor, and the chest radiograph image data is predicted through the trained preset neural network model to output the coronary artery stenosis risk evaluation result. The doctor is assisted in judgment through the coronary artery stenosis risk evaluation result, thereby reducing the risk of clinical misjudgment. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The flowchart of the first embodiment of the intelligent evaluation method for chest radiograph coronary artery stenosis of the present application is shown in the figure.

[0046] Figure 2 The flowchart of the second embodiment of the intelligent evaluation method for chest radiograph coronary artery stenosis of the present application is shown in the figure.

[0047] Figure 3A device structure schematic diagram of a hardware running environment involved in an embodiment of the present application. DETAILED DESCRIPTION

[0048] It should be understood that the specific embodiments described herein are merely intended to explain the present application and are not intended to limit the present application.

[0049] An intelligent evaluation method for coronary artery stenosis based on chest radiography is provided in the embodiments of the present application. In a first embodiment of the intelligent evaluation method for coronary artery stenosis based on chest radiography, referring to Figure 1 , the method comprises:

[0050] Step S10, acquiring chest radiography image data of a target patient from a preset database;

[0051] Step S20, calculating a calcification activity index based on time sequence fluctuations of a calcification focus region in the chest radiography image data, the calcification activity index being used to represent the activity of coronary artery stenosis progression;

[0052] Step S30, determining a heart beat key point in the chest radiography image data, and calculating a collateral circulation compensation index of the calcification focus region based on displacement changes of the heart beat key point between each image frame;

[0053] Step S40, calculating a dynamic risk factor corresponding to the coronary artery stenosis based on the collateral circulation compensation index and the calcification activity index;

[0054] Step S50, inputting the dynamic risk factor into a preset neural network model for iterative training, and performing prediction processing on the chest radiography image data based on the preset neural network model after the training is completed, to obtain a coronary artery stenosis risk evaluation result.

[0055] The present embodiment aims to train the preset neural network model through the dynamic risk factor, and then perform prediction processing on the chest radiography image data through the preset neural network model after the training is completed, to output an accurate coronary artery stenosis risk evaluation result, thereby reducing the risk of clinical misjudgment.

[0056] The specific steps are as follows:

[0057] Step S10, acquiring chest radiography image data of a target patient from a preset database.

[0058] As an example, the intelligent evaluation method for coronary artery stenosis based on chest radiography can be applied to an intelligent evaluation device for coronary artery stenosis based on chest radiography, which belongs to an intelligent evaluation equipment for coronary artery stenosis based on chest radiography.

[0059] As an example, the chest radiograph-based coronary artery stenosis intelligent evaluation method can also be applied to a chest radiograph-based coronary artery stenosis intelligent evaluation system, which includes a data acquisition platform and an intelligent evaluation unit. The data acquisition platform is configured to acquire chest radiograph image data of a target patient, and the intelligent evaluation unit is configured to perform risk assessment based on the chest radiograph image data and output a risk assessment result.

[0060] As an example, the chest radiograph image data can be extracted from a preset database. The patient has undergone image acquisition in a historical time period. The image acquisition process can be as follows: a standard shooting position (posterior-anterior position and left lateral position) of a digital X-ray photography device is adopted, and the patient is in a straight standing position and is attached to a detector plane;

[0061] The photography device is set to a tube voltage of 100-150 kVp, a tube current of 2-10 mAs (adjusted according to the patient's body type), a focal-film distance of 180 cm, and continuous shooting at the end of deep inspiration and breath holding (to ensure that the lung field is fully expanded and the heart motion blur is reduced). A short-time sequence image sequence in the same breathing cycle (in the deep inspiration and breath holding state) is acquired. The acquired chest radiograph is preprocessed, including gray scale normalization, noise suppression, spatial coordinate calibration (aligning the heart region by using a deformable registration technology), and the like, to ensure that the input data meets the analysis requirements.

[0062] In step S20, a calcification activity index is calculated based on the time sequence fluctuation of the calcification focus region in the chest radiograph image data. The calcification activity index is used to represent the activity of the progression of coronary artery stenosis.

[0063] As an example, the chest radiograph image data is limited by its own resolution accuracy and cannot accurately display the coronary artery anatomical structure. The calcification of the hardening plaque around the coronary artery is a sign of stenosis progression. Therefore, the essence of the chest radiograph-based coronary artery stenosis intelligent evaluation is to quantify the mapping relationship between the chest radiograph calcification plaque and the real coronary artery stenosis pathology.

[0064] As an example, the time sequence fluctuation of the calcification focus region can be the change caused by the change of the calcification focus region over time. The progression of coronary artery stenosis is determined by the degree of change of the gray value of the coronary artery calcification focus region in the chest radiograph image data, that is, the calcification activity index is calculated.

[0065] In step S20 of the chest radiograph-based coronary artery stenosis intelligent evaluation, steps S21-S26 are further included, comprising:

[0066] In step S21, the chest radiograph image data is subjected to image segmentation processing to obtain a calcification focus mask in each image frame.

[0067] As an example, chest X-ray image data is not a single frame of image data, but a sequence of chest X-ray images including multiple image frames over a period of time, such as a chest X-ray sequence. The acquisition time period can be 5 seconds, 10 seconds, etc., and there is no specific limitation. The total number of frames is set to T, and each image frame is a static chest X-ray taken within the same respiratory cycle.

[0068] As an example, calcifications in the stenotic segment of the coronary artery are typically composed of calcium deposits, exhibiting a higher X-ray absorption density than the surrounding tissue. Therefore, in chest X-rays, calcified plaques in the stenotic segment of the coronary artery appear as high-brightness areas. An adaptive threshold segmentation algorithm (Otsu's method) is used for each frame of the chest X-ray to automatically calculate the optimal grayscale threshold, extract all high-brightness areas within the cardiac silhouette that meet the threshold condition, and mark them as calcification areas. Finally, a binary calcification mask for each image frame is output. The binary calcification mask can be obtained by performing frame-by-frame binary segmentation of cardiac image sequences (such as CT and ultrasound) to obtain a single-channel binary image (0=non-calcified, 1=calcified) containing only the "calcification" area in each frame.

[0069] Step S22: Calculate the average gray value of each calcification area in the calcification mask.

[0070] As an example, for each frame of chest X-ray image data, the spatial coordinates of all calcified lesion regions are located to ensure spatial consistency of the same calcified lesion region across different time series (achieved through coordinate registration). For each calcified lesion region, its average gray value in each frame of the chest X-ray image sequence is calculated. (This can be obtained by statistically analyzing the average grayscale value of all pixels within the calcification mask area).

[0071] Step S23: Based on the average gray value, calculate the fluctuation intensity coefficient of each calcification zone in the time series. The fluctuation intensity coefficient is used to characterize the time series fluctuation of the density corresponding to the calcification zone.

[0072] As an example, the fluctuation intensity coefficient of each calcification region over time is calculated. The method can be:

[0073]

[0074] Where T represents the number of image frames. and The fluctuation intensity coefficient represents the average gray value of the (t+1)th adjacent image frame and the average gray value of the tth image frame in the chest radiograph sequence. The value of reflects the degree of change in gray value of coronary calcifications in a short time-series chest radiograph, that is, the temporal fluctuation of the density corresponding to the calcification area. The larger the value, the faster the calcium salt deposition rate in the stenotic segment of the coronary artery and the acute progression of coronary stenosis.

[0075] Step S24, detecting the boundary pixels of the calcification mask by an edge detection algorithm to obtain an edge contour binary image.

[0076] As an example, the fluctuation intensity coefficient can only reflect the time sequence fluctuation of the calcification density, but cannot distinguish the spatial pattern of the calcium deposition, such as uniform deposition and non-uniform dendritic deposition; in order to capture the morphological active features of the calcification, it is necessary to analyze the edge evolution rule, therefore, the Sobel edge detection algorithm (prior art) is performed on the calcification mask in each image frame of the chest image sequence to detect the boundary pixels of each calcification region and generate an edge contour binary image.

[0077] Step S25, calculating the edge roughness index of each calcification region based on the edge contour binary image.

[0078] As an example, in the stenosis progression period of the local coronary vessel segment, the calcium is rapidly deposited, the edge of the calcification is "spur-like", the edge contour is irregular, while in the chronic period of the stenosis, the calcification structure is mature, the edge is smooth, and the contour is close to a geometric circle (the calcium deposition tends to be uniform, and due to the change of the mechanical properties of the vessel wall, the calcification morphology tends to be stable), based on this, the edge roughness index of each calcification region is calculated by counting the contour perimeter of the boundary pixels in the edge contour binary image and the area of the calcification.

[0079] As an example, the edge roughness index is used to reflect the roughness of the edge of the calcification region, the larger the value, the more rough and irregular the edge of the calcification is, which indicates that the region is in the active calcification stage and the calcium deposition is not yet fully mature.

[0080] The step S25 of calculating the edge roughness index of each calcification region based on the edge contour binary image comprises:

[0081] Based on the edge contour binary image, the contour perimeter of the boundary pixels of the calcification region and the equivalent circle area are determined.

[0082] Based on the equivalent circle area and the contour perimeter, the edge roughness index of each calcification region is calculated, and the edge roughness index is used to reflect the roughness of the edge of the calcification region.

[0083] As an example, for each calcification region, the total length of the 8-neighbor connection of the boundary pixels is counted by the chain code method The contour perimeter is taken as the minimum circumscribed circle area of the calcification, and the equivalent circle area is obtained synchronously.

[0084] As an example, the calculation method of the edge roughness index R can be:

[0085]

[0086] wherein A represents the equivalent circle area, and P represents the contour perimeter.

[0087] In step S26, the calcification activity index is calculated based on the fluctuation intensity coefficient and the edge roughness index.

[0088] As an example, the calcification activity index is calculated by combining the fluctuation intensity coefficient and the edge roughness index to reflect the calcium deposition process in the calcification focus region. The calcium deposition accelerates the hardening process of blood vessels, and further aggravates the degree of stenosis of blood vessels, thereby laying a foundation for the subsequent dynamic risk factor.

[0089] In step S26, the calcification activity index is calculated based on the fluctuation intensity coefficient and the edge roughness index.

[0090] Based on the edge roughness index of each image frame, a roughness data sequence is constructed.

[0091] The roughness fluctuation intensity between the maximum sequence value and the minimum sequence value in the roughness data sequence is determined.

[0092] As an example, the roughness data sequence is constructed in units of image frames according to the edge roughness index of each image frame, wherein the roughness fluctuation intensity The calculation method of the roughness fluctuation intensity can be:

[0093]

[0094] represents the maximum sequence value, represents the minimum sequence value, and subscript t represents the time corresponding to the video frame, The larger the roughness fluctuation intensity is, the more the current calcification focus edge changes, and the more the acute phase calcification edge of the coronary stenosis segment changes.

[0095] The calcification activity index is calculated based on the fluctuation intensity coefficient, the roughness fluctuation intensity, and the average value of the edge roughness index of each image frame.

[0096] As an example, the calculation method of the calcification activity index may be:

[0097]

[0098] wherein represents the fluctuation intensity coefficient, represents the roughness fluctuation intensity, The average value of the edge roughness index is used to represent the edge roughness of the target calcification focus in the current cycle, and the value is used as a weight. When the average roughness of the calcification focus in the (respiratory) cycle is high, it indicates that the calcification focus itself is in an active state, and the pathological significance of the edge fluctuation is greater (the acute stenosis of the coronary artery is significantly increased).

[0099] Specifically, a higher value indicates that the blood vessel wall corresponding to the calcification focus region is undergoing rapid calcium deposition, accelerating the hardening process of the blood vessel, and further exacerbating the degree of stenosis of the blood vessel.

[0100] Step S30, determine the heart beat key points in the chest radiograph image data, and calculate the collateral circulation compensation index of the calcification focus region based on the displacement change of the heart beat key points between each image frame.

[0101] As an example, the selection of the heart beat key points is used to analyze the heart beat dynamics of the chest radiograph to capture the myocardial compensation reaction caused by coronary stenosis. Generally, coronary stenosis will cause myocardial compensation reaction, therefore, by detecting the change of the heart beat key points, the myocardial compensation reaction is determined.

[0102] Specifically, the myocardial compensation reaction caused by coronary stenosis is essentially that regional blood flow restriction leads to myocardial hypoxia, and the heart maintains blood circulation by increasing the frequency and intensity of beating, thereby producing a stronger motion or beating reaction near the stenosis segment. Therefore, by determining the displacement change of each heart beat point between each image frame, the beating reaction is determined, thereby calculating the collateral circulation compensation index of the calcification focus region. The collateral circulation compensation index is used to represent the correlation between the displacement change of the heart beat key points and the edge roughness of the calcification focus region. According to the correlation, it is determined whether the coronary stenosis region corresponding to the calcification focus has good collateral circulation compensation function.

[0103] Step S40, based on the collateral circulation compensation index and the calcification activity index, a dynamic risk factor corresponding to the coronary stenosis is calculated.

[0104] The step S40 of calculating the dynamic risk factor corresponding to the coronary stenosis based on the collateral circulation compensation index and the calcification activity index includes:

[0105] Calculate the second Pearson correlation coefficient between the sequence corresponding to the calcification activity index and the sequence corresponding to the reciprocal of the collateral circulation compensation index, and use the second Pearson correlation coefficient as a correction weight.

[0106] ​As an example, the second Pearson correlation coefficient is the Pearson correlation coefficient between the sequence corresponding to the calcification activity index and the sequence corresponding to the reciprocal of the collateral circulation compensation index, which is used to represent the correlation between the two sequences. The reciprocal of the collateral circulation compensation index represents the degree of absence of collateral circulation compensation function, and the calcification activity index represents the activity of coronary stenosis progression. The collateral circulation compensation index represents the collateral circulation compensation function of the coronary stenosis region corresponding to the calcification focus. When the calcification activity index is larger, the collateral circulation compensation function is poorer, that is, the calcification activity index and the collateral circulation compensation index are negatively correlated, and the sequence corresponding to the calcification activity index and the sequence corresponding to the reciprocal of the collateral circulation compensation index are positively correlated. When the second Pearson correlation coefficient is larger, the positive correlation between the two sequences is stronger.

[0107] As an example, The collateral circulation compensation index represents the degree of absence of collateral circulation compensation function, and the reciprocal of the collateral circulation compensation index represents the degree of absence of collateral circulation compensation function. The second Pearson correlation coefficient is used as a correction weight. The larger the value is, the stronger the positive correlation between the calcification activity of the calcification focus and the regional collateral circulation compensation absence degree is, that is, the calcification activity of the calcification focus will significantly aggravate the regional collateral circulation compensation absence degree, and the calcification activity is the main factor causing the aggravation of the regional collateral circulation compensation absence degree.

[0108] Based on the calcification activity index, the collateral circulation compensation index, and the correction weight, a dynamic risk factor corresponding to the coronary stenosis is calculated.

[0109] As an example, the dynamic risk factor may be calculated in the following manner:

[0110]

[0111] In the formula, DRF represents the dynamic risk factor of the current chest radiograph calcification focus region corresponding to the coronary vessel segment stenosis, represents the calcification activity index, represents the collateral circulation compensation index, represents the collateral circulation compensation absence degree, represents the second Pearson correlation coefficient. A larger DRF value indicates that the calcification activity of the calcification focus is higher, and the collateral circulation compensation ability of the region is weaker, which means that the coronary vessel segment may have a more serious stenosis and have a greater impact on blood flow. The risk of stenosis of the coronary vessel in this region is higher.

[0112] In step S50, the dynamic risk factor is input into a preset neural network model for iterative training, and the chest radiograph image data is predicted based on the trained preset neural network model to obtain a coronary stenosis risk assessment result.

[0113] As an example, the dynamic risk factor As a key spatiotemporal dynamic feature input into a pre-trained preset neural network model (such as a gradient boosting decision tree, a temporal convolutional network, or a graph neural network), the model learns the complex relationship between the dynamic risk factor and the degree of coronary stenosis through offline training, and combines the patient's basic physiological parameters (such as immune markers, clinical history, etc.) as auxiliary features to enhance the prediction ability of the model. After the iterative training is completed, a trained preset neural network model is obtained.

[0114] As an example, based on the trained preset neural network model, the chest radiograph image data and the related physiological parameters of the target patient are predicted and processed to obtain a coronary stenosis risk assessment result, and the prediction probability or specific assessment level of the patient's coronary stenosis is provided to provide decision support for clinicians, helping to judge the patient's coronary health status and guide the subsequent treatment plan.

[0115] The present application provides a method for intelligent evaluation of coronary stenosis based on chest radiograph. In related art, the coronary stenosis is evaluated based on the image features extracted from a single static chest radiograph, which causes the disease anatomical information to be seriously disconnected from the image features, thereby increasing the risk of clinical misjudgment. In the present application, the chest radiograph image data of the target patient is obtained from a preset database. First, the calcification activity index is calculated based on the temporal fluctuations of the calcification region in the chest radiograph image data. The calcification activity index is used to represent the activity of coronary stenosis progression. Then, the heart beat key points in the chest radiograph image data are determined. The displacement changes of the heart beat key points between each image frame are calculated to obtain the collateral circulation compensation index of the calcification region. Then, based on the collateral circulation compensation index and the calcification activity index, the dynamic risk factor corresponding to the coronary stenosis is calculated. The dynamic risk factor quantifies the contribution of the synergistic effect of calcification activity and collateral circulation compensation capacity to the stenosis risk, thereby establishing an accurate mapping relationship between multi-dimensional dynamic pathological features and the degree of stenosis. The preset neural network model is trained by the dynamic risk factor, and the trained preset neural network model is used to predict and process the chest radiograph image data to output the coronary stenosis risk assessment result. The doctor is assisted in judgment by the coronary stenosis risk assessment result, thereby reducing the risk of clinical misjudgment.

[0116] Further, with reference to Figure 2 Based on the first embodiment of the present application, another embodiment of the present application is provided. In this embodiment, the step S30 of calculating the collateral circulation compensation index of the calcification region based on the displacement changes of the heart beat key points between each image frame comprises:

[0117] Step S31, based on the Euclidean distance between the heart beat key points between each image frame, determine the displacement time sequence of each heart beat key point.

[0118] As an example, different displacement changes of the heart beat key points between different image frames are generated, the displacement changes are represented by calculating the Euclidean distances of the heart beat key points between different image frames, and then the displacement values of each heart beat key point are determined, and the displacement time sequence of each heart beat key point is constructed according to the displacement values.

[0119] The step S31 further includes:

[0120] For any heart beat key point, the first Euclidean distance of the key point displacement between adjacent image frames is calculated, and the absolute value of the first Euclidean distance is taken as the inter-frame displacement of each heart beat key point. The heart beat key points include the apex point, the left atrial appendage point and the ventricular diaphragmatic point.

[0121] Based on the inter-frame displacement, the displacement time sequence of each heart beat key point is determined.

[0122] As an example, in each image frame of the short-time sequence chest radiograph image data, the following heart beat key points are automatically located by using an anatomical topology constraint algorithm, the heart beat key points include the apex point, the left atrial appendage point and the ventricular diaphragmatic point. The apex point can be the intersection point of the left ventricular farthest segment and the diaphragm projection, the left atrial appendage point can be the turning point of the pulmonary artery segment and the left ventricular segment, and the ventricular diaphragmatic point can be the highest point where the right ventricle meets the diaphragm.

[0123] As an example, the first Euclidean distance represents the Euclidean distance of any heart beat key point between adjacent image frames, the adjacent inter-frame displacement of each heart beat key point is tracked by using the pyramid optical flow method, and the spatial coordinate Euclidean distance of the displacement between adjacent image frames, i.e. the first Euclidean distance, is calculated, and the absolute value of the first Euclidean distance is taken as the inter-frame displacement of each heart beat key point.

[0124] As an example, after obtaining the inter-frame displacement of each type of heart beat key point, the displacement time sequence of each heart beat key point is constructed: The total number of image frames corresponding to the chest radiograph image data).

[0125] In step S32, the instantaneous phase time delay coefficient of each heart beat key point is calculated based on the displacement time sequence.

[0126] ​As an example, the instantaneous phase time delay coefficient is used to characterize the pathological asynchrony degree in the process of the heart structure beating. The greater the value is, the more the heart structure part corresponding to the heart beat key point loses synchronization with other parts. The blood flow reduction in the coronary artery stenosis area leads to the compensatory contraction of the myocardium in this area, which further affects the coordination between it and other heart regions. The instantaneous phase time delay coefficient is calculated by calculating the instantaneous phase change of each sequence value in the displacement time sequence.

[0127] The step S32 of calculating the instantaneous phase time delay coefficient of each heart beat key point based on the displacement time sequence comprises:

[0128] The displacement time sequence is linearly fitted to obtain the displacement fitting curve of each heart beat key point.

[0129] The instantaneous phase information of the displacement fitting curve corresponding to each heart beat key point is determined.

[0130] As an example, the displacement time sequence corresponding to each heart beat key point is mapped in a two-dimensional space coordinate system, and the displacement fitting curve of the displacement time sequence of each heart beat key point is obtained by linear fitting (polynomial fitting method): wherein subscript r represents the heart beat key point.

[0131] As an example, after generating the displacement fitting curves of all heart beat key points , the fast Fourier transform (FFT) is performed to convert them into frequency domain data, and then the instantaneous phase information of the fitting curve corresponding to each heart beat key point is obtained (the phase information is converted into data by using the trigonometric function cos, which is convenient for subsequent data calculation).

[0132] For any heart beat key point, the instantaneous phase difference between the instantaneous phase information between the current heart beat key point and other different heart beat key points is calculated, and the absolute value mean of each instantaneous phase difference is taken as the instantaneous phase time delay coefficient.

[0133] As an example, for any heart beat key point, the instantaneous phase difference between the instantaneous phase information between the current heart beat key point and other different heart beat key points is calculated wherein, represents the instantaneous phase information of the current heart beat key point, represents the instantaneous phase information of the other heart beat key point, and the absolute value of the instantaneous phase difference is determined Since there are multiple heart beat key points, the instantaneous phase difference of each is integrated to obtain the mean value, which is the instantaneous phase time delay coefficient of the current heart beat key point.

[0134] In step S33, the calcification region is matched with the heart beat key point to obtain a plurality of key point calcification sets.

[0135] As an example, when performing correlation calculation of the calcification region and the heart beat key point, all calcification regions in each image frame are grouped according to spatial attribution, the calcification region is matched with the heart beat key point to obtain a calcification set corresponding to the assigned heart beat key point, that is, a key point calcification set, in which the calcification region and the corresponding heart beat key point are assigned to the same spatial attribution, and the key point calcification set is represented as: wherein represents the heart beat key point, represents the calcification region assigned to the corresponding heart beat key point.

[0136] In step S33, the calcification region is matched with the heart beat key point to obtain a plurality of key point calcification sets, including:

[0137] The mask barycenter coordinates in the calcification mask are obtained, and the second Euclidean distance from the mask barycenter coordinates to each heart beat key point is calculated.

[0138] As an example, the mask barycenter coordinates can be the geometric centroid of all valid pixels (usually regions with a value of 1 or 255) in the mask, and the second Euclidean distance represents the Euclidean distance from the mask barycenter coordinates to each heart beat key point.

[0139] Each calcification region is assigned to the corresponding region of the heart beat key point corresponding to the minimum second Euclidean distance to generate a plurality of key point calcification sets.

[0140] In step S34, the first Pearson correlation coefficient between the instantaneous phase time delay coefficient corresponding sequence and the roughness data sequence corresponding to the key point calcification set is calculated, and the first Pearson correlation coefficient is taken as the collateral circulation compensation index of the calcification region.

[0141] As an example, the roughness data sequence is represented as wherein is the calcification region number in the image, is the image frame number in the sequence, and R represents the edge roughness index.

[0142] As an example, the sequence composed of the instantaneous phase time delay coefficient U between each heart beat key point and other heart beat key points is determined Data sequence: ), and the temporal sequence of the edge roughness of coronary artery calcifications within the allocation region where the key point of cardiac pulsation is located ( ).

[0143] As an example, calculate a data sequence With data sequence The first Pearson correlation coefficient between them, which ranges from -1 to 1, is defined as the collateral circulation compensation index of the corresponding coronary artery calcification mask region. ).

[0144] Specifically, when When the value is greater than 0, the closer it is to 1, the stronger the positive correlation between the B sequence (instantaneous phase time delay coefficient) and the C sequence (calcification edge roughness). In this case, the asynchronous beating in certain areas of the heart is positively correlated with the edge roughness of the coronary calcification in those areas, meaning that coronary artery stenosis leads to reduced local blood flow, and the myocardium compensates for the insufficient blood flow by increasing its beating. At this time, the stenotic segment of the coronary artery corresponding to the calcification may have good collateral circulation compensation function, thereby reducing the risk of coronary artery stenosis.

[0145] when When the value of is (0,1), the two are negatively correlated, which means that the local blood flow reduction caused by coronary artery stenosis is insufficient, and the compensatory pulsation of the myocardium is insufficient, which cannot effectively compensate for the loss of blood flow. At this time, the stenotic segment of the coronary artery corresponding to the calcification lacks effective collateral circulation compensation function, and the risk caused by coronary artery stenosis will be further aggravated.

[0146] In this embodiment, the collateral circulation compensation index is calculated based on the instantaneous phase time delay coefficient between each key point of cardiac pulsation and the edge roughness time sequence. This is used to determine whether the coronary artery stenosis segment corresponding to the calcification area has effective collateral circulation compensation function, thereby accurately determining the degree of risk caused by coronary artery stenosis.

[0147] Specifically, this application also provides a coronary artery stenosis intelligent assessment system based on chest X-ray. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the coronary artery stenosis intelligent assessment method based on chest X-ray as described in the above embodiments.

[0148] Reference Figure 3 , Figure 3 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application.

[0149] like Figure 3As shown, the chest radiograph-based coronary stenosis intelligent evaluation device can include a processor 1001, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection communication between the processor 1001 and the memory 1005.

[0150] Optionally, the chest radiograph-based coronary stenosis intelligent evaluation device can further include a user interface, a network interface, a camera, an RF (Radio Frequency) circuit, a sensor, a WiFi module, and the like. The user interface can include a display screen (Display), an input sub-module such as a keyboard (Keyboard), and the optional user interface can further include a standard wired interface, a wireless interface. The network interface can include a standard wired interface, a wireless interface (such as a WI-FI interface).

[0151] Those skilled in the art can understand that, Figure 3 The structure of the chest radiograph-based coronary stenosis intelligent evaluation device shown in the above embodiments does not constitute a limitation on the chest radiograph-based coronary stenosis intelligent evaluation device, and can include more or fewer components than those shown, or combine certain components, or different component arrangements.

[0152] As Figure 3 As shown, the memory 1005 as a storage medium can include an operating system, a network communication module, and a chest radiograph-based coronary stenosis intelligent evaluation program. The operating system is a program that manages and controls the hardware and software resources of the chest radiograph-based coronary stenosis intelligent evaluation device, supports the running of the chest radiograph-based coronary stenosis intelligent evaluation program and other software and / or programs. The network communication module is used to realize the communication between the components in the memory 1005, and the communication with other hardware and software in the chest radiograph-based coronary stenosis intelligent evaluation system.

[0153] In Figure 3 In the chest radiograph-based coronary stenosis intelligent evaluation device shown in the above embodiments, the processor 1001 is used to execute the chest radiograph-based coronary stenosis intelligent evaluation program stored in the memory 1005, and realize the steps of any one of the chest radiograph-based coronary stenosis intelligent evaluation methods described above.

[0154] The specific embodiments of the chest radiograph-based coronary stenosis intelligent evaluation device of the present application are basically the same as the above-mentioned embodiments of the chest radiograph-based coronary stenosis intelligent evaluation method, and will not be repeated here.

[0155] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0156] The above-mentioned sequence numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0157] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and necessary general hardware platforms, and of course, they can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) as described above, and includes a number of instructions for making a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device) execute the methods described in various embodiments of the present application.

[0158] The above is only the preferred embodiment of the present application, and does not limit the application range of the present application. Any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or directly or indirectly applied in other related technical fields, is also included in the application protection range of the present application.

[0159] It should be noted that: the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0160] Each of the embodiments in the present specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A method for intelligent assessment of coronary artery stenosis based on chest X-ray, characterized in that, The method includes: Retrieve chest X-ray image data of the target patient from a pre-set database; Based on the temporal fluctuations of calcification areas in the chest X-ray image data, a calcification activity index is calculated, specifically including: The chest X-ray image data is processed by image segmentation to obtain the calcification mask in each image frame; Calculate the average gray value of each calcification zone in the calcification zone mask; Based on the average gray value, the fluctuation intensity coefficient of each calcification zone in the time series is calculated, and the fluctuation intensity coefficient is used to characterize the time series fluctuation of the density corresponding to the calcification zone. The boundary pixels of the calcification mask are detected by an edge detection algorithm to obtain a binary image of the edge contour. Based on the binary image of the edge contour, the edge roughness index of each calcification zone is calculated; Based on the fluctuation intensity coefficient and the edge roughness index, the calcification activity index is calculated, which is used to characterize the activity of coronary artery stenosis progression. Identify key cardiac pulsation points in the chest X-ray image data, and calculate the collateral circulation compensation index of the calcified lesion region based on the displacement changes of these key cardiac pulsation points between each image frame. Specifically, this includes: Based on the Euclidean distance between the key points of heart pulsation in each image frame, the displacement time sequence of each key point of heart pulsation is determined; Based on the displacement time sequence, the instantaneous phase time delay coefficient of each key point of heartbeat is calculated; By matching the calcification area with the key points of cardiac pulsation, a set of multiple key point calcifications is obtained. Calculate the first Pearson correlation coefficient between the sequence corresponding to the instantaneous phase time delay coefficient and the roughness data sequence corresponding to the set of key point calcifications, and use the first Pearson correlation coefficient as the collateral circulation compensation index of the calcification region. Based on the collateral circulation compensation index and the calcification activity index, the dynamic risk factors corresponding to coronary artery stenosis are calculated, specifically including: Calculate the second Pearson correlation coefficient between the sequence corresponding to the calcification activity index and the inverse sequence corresponding to the collateral circulation compensation index, and use the second Pearson correlation coefficient as the correction weight; Based on the calcification activity index, the collateral circulation compensation index, and the adjusted weight, the dynamic risk factor corresponding to coronary artery stenosis is calculated, wherein the calculation formula for the dynamic risk factor is as follows: in, This indicates the dynamic risk factor for coronary artery stenosis corresponding to the calcification area on the current chest X-ray. Indicates the calcification activity index, Indicates the collateral circulation compensation index. This represents the adjusted weights, where exp is the natural constant; The dynamic risk factors are input into a preset neural network model for iterative training, and the chest X-ray image data is then used for prediction based on the trained preset neural network model to obtain the coronary artery stenosis risk assessment result.

2. The intelligent assessment method for coronary artery stenosis based on chest X-ray as described in claim 1, characterized in that, The calculation of the edge roughness index of each calcification region based on the edge contour binary map includes: Based on the edge contour binary map, the contour perimeter and equivalent circle area of ​​the boundary pixels of the calcification foci region are determined. Based on the equivalent circle area and the contour perimeter, the edge roughness index of each calcification zone is calculated. The edge roughness index is used to reflect the roughness of the edge of the calcification zone.

3. The intelligent assessment method for coronary artery stenosis based on chest X-ray as described in claim 1, characterized in that, The calculation of the calcification activity index based on the fluctuation intensity coefficient and the edge roughness index includes: Based on the edge roughness index of each image frame, a roughness data sequence is constructed; Determine the roughness fluctuation intensity between the maximum and minimum sequence values ​​in the roughness data sequence; The calcification activity index is calculated based on the fluctuation intensity coefficient, the roughness fluctuation intensity, and the average value of the edge roughness index of each image frame.

4. The intelligent assessment method for coronary artery stenosis based on chest X-ray as described in claim 1, characterized in that, The determination of the displacement temporal sequence of each cardiac pulsation key point based on the Euclidean distance between each image frame includes: For any key point of heart pulsation, calculate the first Euclidean distance between adjacent image frames of the key point displacement, and take the absolute value of the first Euclidean distance as the inter-frame displacement of each key point of heart pulsation. The key points of heart pulsation include the apex of the heart, the left atrial appendage, and the ventricular diaphragmatic surface. Based on the inter-frame displacement, the displacement time sequence of each key point of heartbeat is determined.

5. The intelligent assessment method for coronary artery stenosis based on chest X-ray as described in claim 1, characterized in that, The instantaneous phase time delay coefficient for each key point of cardiac pulsation, calculated based on the displacement time sequence, includes: The displacement time sequence is linearly fitted to obtain the displacement fitting curve for each of the key points of heart pulsation. Determine the instantaneous phase information of the displacement fitting curve corresponding to each key point of heart pulsation; For any key point of heart pulsation, calculate the instantaneous phase difference between the current key point of heart pulsation and other different key points of heart pulsation, and use the average of the absolute values ​​of the instantaneous phase differences as the instantaneous phase time delay coefficient.

6. The intelligent assessment method for coronary artery stenosis based on chest X-ray as described in claim 1, characterized in that, The process of matching the calcified lesion region with the key points of cardiac pulsation yields a set of multiple key point calcified lesions, including: Obtain the centroid coordinates of the calcification mask, and calculate the second Euclidean distance from the centroid coordinates of the mask to each of the key points of cardiac pulsation; Each calcification region is assigned to the region corresponding to the cardiac pulsation key point with the minimum second Euclidean distance, generating multiple key point calcification sets.

7. A smart assessment system for coronary artery stenosis based on chest X-ray, characterized in that, The system includes a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 6.

Images