Coronary artery angiography image-based collateral branch physiology intelligent evaluation system and method

Abstract

The invention discloses a collateral physiology intelligent evaluation system and method based on a coronary artery angiography image, and the system comprises an image collection module, a segmentation module, a reconstruction module, a modeling module, and an index calculation module. Wherein the image acquisition module is used for acquiring coronary artery angiography and screening minimum overlapped and shortened angiography projection images after CTO-PCI front and back side branch blood supply arteries and occluded blood vessels are opened; the segmentation module is used for carrying out frame-by-frame blood vessel segmentation on the obtained coronary artery angiography by adopting a deep learning algorithm; the reconstruction module is used for performing blood vessel three-dimensional volume reconstruction on the coronary artery binary image obtained through segmentation; the modeling module is used for modeling blood flow and calculating average blood flow; the index calculation module is used for calculating the collateral branch blood flow index according to the blood flow change before and after the CTO is opened. According to the method, the subjective defect of a traditional visual scoring system is eliminated, the operation process does not need to be interrupted, special consumables are not needed, and the medical cost and the occurrence risk of vascular injury related complications in the operation are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of medical information, and in particular to a system and method for intelligently evaluating collateral physiology based on coronary angiography images. Background Art

[0002] Chronic total occlusion (CTO) is one of the most complex and difficult clinical conditions to treat in the field of percutaneous coronary intervention (PCI). CTO vessels are often accompanied by a certain degree of collateral circulation development, which provides compensatory blood flow to the myocardium in the blood supply area of ​​the occluded vessels. Accurate quantitative assessment of the physiological function of collateral circulation is crucial for the adequate planning of CTO revascularization strategies and the prediction of postoperative efficacy. In the clinical management of CTO lesions, collateral circulation is of decisive significance for maintaining distal myocardial viability. However, the current functional assessment system is still limited to semi-quantitative, visual scoring systems derived from coronary angiography, such as the Rentrop grade and collateral connection (CC) grade, while detailed hemodynamic function assessment is limited in clinical practice due to its invasiveness. Such traditional methods have the following significant defects:

[0003] 1. Limited assessment accuracy: Relying on the operator's subjective experience, objective quantification of hemodynamic parameters cannot be achieved;

[0004] 2. Inefficient operation: The need to repeatedly switch interventional instruments and inject contrast agents prolongs the operation time.

[0005] 3. Increased medical costs: Dependence on consumables such as specialized pressure guidewires can increase the cost of a single test by 30%-50%;

[0006] 4. Uncontrollable risks: Complex operating steps can easily induce complications such as vascular spasm or perforation. Summary of the Invention

[0007] In response to the above-mentioned technical bottlenecks in the field of coronary collateral circulation function assessment, the present invention proposes a non-invasive, automated collateral circulation physiological function quantitative assessment scheme. Its core objectives are: to analyze standard coronary angiography image sequences through a deep learning model, eliminate the subjective defects of the traditional visual scoring system, and achieve objective quantification of the collateral flow index (CFI); automatically extract hemodynamic characteristics based on coronary angiography data, shorten the evaluation time without interrupting the surgical process; calculate entirely based on conventional angiography images, without the need for special consumables, effectively reducing medical costs and the risk of complications related to intraoperative vascular injury, so as to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above-mentioned object, one aspect of the present invention provides an intelligent assessment system for collateral physiology based on coronary angiography images, comprising an image acquisition module, a segmentation module, a reconstruction module, a modeling module, and an index calculation module, wherein:

[0009] The image acquisition module is used to acquire coronary angiography and screen angiographic projection images with minimal overlap and shortening of collateral blood supply arteries before and after CTO-PCI and after occluded blood vessels are opened;

[0010] The segmentation module is used to perform frame-by-frame vessel segmentation on the acquired coronary angiography using a deep learning algorithm model;

[0011] The reconstruction module is used to reconstruct the 3D volume of the coronary artery based on the segmented binary image.

[0012] The modeling module is used to model blood flow and calculate the average blood flow;

[0013] The index calculation module is used to calculate the collateral blood flow index based on the changes in blood flow before and after CTO opening.

[0014] Furthermore, in the image acquisition module, the right coronary artery angiography took LAO (left anterior oblique) 45°, anteroposterior position + head tilt 20°-25°; the left coronary artery angiography took RAO (right anterior oblique) 30° + foot tilt 20°, RAO 30° + head tilt 20°, LAO 45° + head tilt 25°, LAO 60° + foot tilt 30°.

[0015] Furthermore, the segmentation module uses CNN (Convolutional Neural Networks) as a feature extractor to capture local boundary feature maps and adopts a multi-head attention mechanism. The channel attention module enhances the feature extraction ability of the model by focusing on the importance of different feature channels, and the spatial attention module further refines the model's processing of spatial information by focusing on the importance of different spatial regions.

[0016] Furthermore, the thickness of each voxel in the angiographic projection direction is calculated to correct for the effects of two-dimensional image overlap and shortening. The vascular volume V is calculated as follows:

[0017] ,

[0018] Where x, y, and z are three-dimensional coordinates, A is the cross-sectional area of ​​the blood vessel, and l is the length of the blood vessel. is the thickness of blood vessels.

[0019] Furthermore, the blood flow modeling method in the modeling module is as follows: draw the time-volume curve for analysis, use a multi-stage filtering algorithm to perform signal enhancement processing on the curve, retain the trend characteristics of the contrast agent filling process, and suppress high-frequency noise interference, and then perform filling stage detection, calculate the first-order derivative dV / dt of the preprocessed curve to locate the rising edge, and identify the contrast agent inflow period through the threshold method.

[0020] Furthermore, the interval with the largest rise was selected as the fitting interval, the average blood flow parameters were estimated, and the hemodynamic model was established based on the optimized fitting interval. The blood flow calculation formula is as follows:

[0021] ,

[0022] in It's blood flow, is the cross-sectional area of ​​the blood vessel, V is the blood vessel volume, and t is the time;

[0023] The formula for calculating mean blood flow is as follows:

[0024] ,

[0025] where Q avg is the average blood flow, V peak is the peak vascular volume, V baseline is the baseline vascular volume, t peak is the time when the contrast agent reaches its peak, t onset is the arrival time of the contrast agent.

[0026] Furthermore, the collateral flow index (CFI) Auto The calculation formula is:

[0027] ;

[0028] wherein is the PCDA blood flow before CTO opening, is the PCDA blood flow after CTO opening, is the absolute change of collateral blood flow, which is simulated and calculated by the change of main collateral blood supply artery blood flow before and after CTO opening, or is the forward blood flow after the CTO vessel recanalization.

[0029] Another aspect of the present application provides a method for intelligent evaluation of collateral physiology based on coronary angiography images, comprising the following steps:

[0030] Step S1, collecting coronary angiography and screening collateral blood supply arteries before and after CTO-PCI and the smallest overlap and shortened angiography projection images after the occluded vessel is opened;

[0031] Step S2, performing frame-by-frame blood vessel segmentation on the obtained coronary angiography using a deep learning algorithm model;

[0032] Step S3, performing three-dimensional volume reconstruction on the segmented coronary artery binary image;

[0033] Step S4, modeling blood flow and calculating average blood flow;

[0034] Step S5, calculating the collateral blood flow index according to the blood flow change before and after CTO opening.

[0035] Compared with the prior art, the system and method have the following advantages:

[0036] 1. The intelligent evaluation system and method for collateral physiology based on coronary angiography images can effectively avoid the complications caused by traditional invasive measurement methods, does not need to change the existing PCI operation process, greatly shortens the intraoperative functional examination time, and has no related risks, and has good clinical application value;

[0037] 2. The present application captures the spatio-temporal correlation of the contrast agent filling process by combining the Transformer network and the multi-head attention mechanism, realizes millisecond-level dynamic analysis of blood vessel function parameters, obtains coronary artery segmentation and quickly and accurately calculates hemodynamic parameters, and has very strong operability and repeatability;

[0038] 3. The present application obtains dynamic blood flow parameters and blood vessel volume based on coronary angiography, and obtains coronary blood flow through hemodynamic modeling, avoiding the use of invasive and expensive catheter measurement equipment;

[0039] 4. The collateral blood flow index calculation model obtained in the present invention has strong clinical universality. It can adapt to the imaging characteristics of different angiography equipment through the transfer learning framework. By calculating the changes in blood flow in the collateral supply arteries before and after CTO opening, it is more in line with the actual changes in collateral blood flow and can accurately and quickly reflect the true and specific coronary physiological status of CTO patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flowchart of the intelligent assessment method for collateral physiology based on coronary angiography images.

[0041] Figure 2 Schematic diagram of the frame-by-frame segmentation principle of coronary artery lumen based on deep learning network architecture.

[0042] Figure 3 Vessel segmentation and CFI before and after PCI in case 1 Auto Computational graph.

[0043] Figure 4 For case 2, vascular segmentation and CFI before and after PCI Auto Computational graph. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0045] The collateral flow index is defined as the ratio of collateral blood flow to the forward blood flow after CTO revascularization. It is currently the most accurate diagnostic indicator of the physiological function of coronary collateral circulation. The present invention obtains a binary image of the coronary arteries through deep learning network segmentation, calculates the thickness value of each voxel point in the angiographic projection direction, corrects the effects of two-dimensional image overlap and foreshortening, obtains a time-varying PCDA and CTO vessel volume sequence and obtains the average blood flow. The collateral blood flow is simulated by the change in PCDA blood flow before and after CTO vessel patency. The CFI is calculated based on the absolute change in collateral blood flow and the ratio of the forward blood flow after the CTO vessel is restored to patency. The specific calculation steps are as follows:

[0046] Step S1: collecting and screening coronary angiography.

[0047] At least two orthogonal body position projections of left and right coronary arteries, right coronary angiography takes LAO 45°, positive position + head tilt position 20°-25°, if necessary, LAO 45°+ head tilt position 25°; Left coronary angiography takes RAO 30°+ foot tilt position 20°, RAO 30°+ head tilt position 20°, LAO 45°+ head tilt position 25°, LAO 60°+ foot tilt position 30° projection, if necessary, positive position + head tilt 20° and positive position + foot tilt 20°.

[0048] CTO is defined as complete occlusion of the original coronary artery forward blood flow, no lumen continuity, and TIMI (thrombolysis in myocardial infarction) grade 0 by angiographic evaluation, and the occlusion period is more than 3 months according to clinical events estimation or previous angiographic confirmation. CTO successful opening is defined as successful opening of completely occluded blood vessels and restoration of blood flow (residual stenosis <25% and TIMI 3 grade blood flow) after stent implantation. Before CTO vessel opening, collateral circulation is identified by bilateral angiography, that is, the same side or opposite artery (collateral artery) supplies blood to the distal occluded vessel through small arteries, and the collateral artery before and after CTO-PCI and the minimum overlap and shortened angiographic projection images after CTO vessel opening are screened.

[0049] In step S2, a deep learning algorithm model is used for frame-by-frame blood vessel segmentation of the obtained coronary angiography.

[0050] PCDA supplies blood to the distal occluded lesion of CTO vessel before CTO vessel opening to maintain the activity of distal myocardial cells. PCDA and CTO vessel segmentation can use traditional segmentation algorithm or deep learning algorithm model, which combines the advantages of CNN and Transformer network in local and global spatial feature extraction, uses CNN as a feature extractor in two-dimensional angiographic images to capture local boundary feature maps including blood vessel contour and gray scale. On this basis, a multi-head attention mechanism is used, in which the channel attention module enhances the feature extraction capability of the model by focusing on the importance of different feature channels, and the spatial attention module further refines the processing of spatial information by the model by focusing on the importance of different spatial regions. The multi-head attention mechanism can better distinguish between blood vessel structure and background, effectively improve the accuracy and details of blood vessel recognition, calculate the cross-spatial correlation between different angiographic frames, and thus improve the segmentation effect of the current angiography by using adjacent images as strong priori, to realize frame-by-frame segmentation of PCDA during contrast agent filling.

[0051] In step S3, the coronary artery binary image obtained by segmentation is subjected to blood vessel three-dimensional volume reconstruction.

[0052] For the subtracted coronary angiography image, the relationship between the image gray value and the contrast agent concentration and the blood vessel thickness is:

[0053] ,

[0054] in represents the grayscale value, represents the contrast agent concentration, Indicates the thickness of blood vessels. is a constant, is the mass attenuation coefficient of the contrast agent, which is proportional to the X-ray energy. is a constant that takes into account the acquisition and amplification of the X-ray imaging system.

[0055] Based on the deep learning network segmentation, the coronary artery binary image is obtained, and the thickness value of each voxel in the angiographic projection direction is calculated. The overlap and shortening effects of the two-dimensional image are corrected. The vascular volume V is expressed as the product of the two-dimensional vascular pixels and the corresponding vascular thickness. The formula is as follows:

[0056] ,

[0057] Where x, y, and z are three-dimensional coordinates, A is the cross-sectional area of ​​the blood vessel, and l is the length of the blood vessel. is the thickness of blood vessels.

[0058] Step S4: Model the blood flow and calculate the average blood flow.

[0059] First, a time-volume curve is plotted for analysis. A multi-stage filtering algorithm is used to enhance the curve's signal, preserving the trend characteristics of the contrast agent filling process while suppressing high-frequency noise. Next, the filling phase is detected. The first-order derivative dV / dt of the preprocessed curve is calculated to locate the rising edge, and a threshold method is used to identify the contrast agent inflow phase.

[0060] The interval with the largest rise was selected as the fitting interval to estimate the mean blood flow parameters. A hemodynamic model was established based on the optimized fitting interval, where blood flow is equal to the product of the mean blood flow velocity and the cross-sectional area.

[0061] ,

[0062] in It's blood flow, is the cross-sectional area of ​​the blood vessel, and V is the blood vessel volume.

[0063] The formula for calculating mean blood flow is as follows:

[0064]

[0065] where Q avg is the average blood flow, V peak is the peak vascular volume, V baseline is the baseline vascular volume, t peak is the time when the contrast agent reaches its peak, t onsetis the arrival time of the contrast agent (detected by the threshold method).

[0066] Step S5: Calculate the collateral blood flow index based on the blood flow changes before and after the CTO is opened.

[0067] CTO before and after interventional treatment, CFI Auto It can be approximately expressed as the absolute change in collateral blood flow ( ) and forward blood flow after CTO vessel restoration ( or ) ratio. The collateral flow index (CFI) can be calculated by simulating the changes in blood flow in the main collateral artery before and after CTO opening. Auto .

[0068] Collateral flow index (CFI) Auto The calculation formula is:

[0069] ;

[0070] in It is the PCDA blood flow before CTO opening, It is the blood flow of PCDA after CTO opening, is the absolute change in collateral blood flow, which is calculated by simulating the change in blood flow in the main collateral artery before and after CTO opening. or To restore the patency of CTO vessels and restore the forward blood flow.

[0071] like Figure 3 Shown are the vessel segmentation and CFI before and after PCI in case 1. Auto Calculation diagram. Case 1: Male, 42 years old, height 172cm, weight 80.5kg, BMI: 27.21kg / m 2 , Rentrop Level 3, CC Level 2, CFI ref = 0.37. CFI is calculated according to the method of the present invention Auto It is 0.41.

[0072] like Figure 4 Shown are the vessel segmentation and CFI before and after PCI in case 2. Auto Calculation diagram. Case 2: Male, 67 years old, height 164cm, weight 63.5kg, BMI: 23.61kg / m 2 , Rentrop Level 3, CC Level 2, CFI ref = 0.50. CFI is calculated according to the method of the present invention Auto It is 0.42.

[0073] The differences between the calculated results and the reference values ​​for the two cases above are both within an acceptable range. This demonstrates that the present invention uses a deep learning model to analyze standard coronary angiography image sequences, eliminating the subjective flaws of traditional visual scoring systems and achieving objective quantification of the collateral flow index. It also automatically extracts hemodynamic characteristics based on coronary angiography data, shortening evaluation time without interrupting the surgical process. Calculations are performed entirely based on conventional angiography images, eliminating the need for specialized consumables and effectively reducing medical costs and the risk of complications associated with intraoperative vascular injury.

[0074] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent assessment system for collateral physiology based on coronary angiography images, characterized by: It includes image acquisition module, segmentation module, reconstruction module, modeling module, and index calculation module, among which: The image acquisition module is used to acquire coronary angiography and screen angiographic projection images with minimal overlap and shortening of collateral blood supply arteries before and after CTO-PCI and after occluded blood vessels are opened; The segmentation module is used to perform frame-by-frame vessel segmentation on the acquired coronary angiography using a deep learning algorithm; The reconstruction module is used to reconstruct the 3D volume of the coronary artery based on the segmented binary image. The modeling module is used to model blood flow and calculate the average blood flow; The index calculation module is used to calculate the collateral blood flow index based on the changes in blood flow before and after CTO opening.

2. The intelligent assessment system for collateral physiology based on coronary angiography images according to claim 1, characterized in that: In the image acquisition module, the right coronary artery angiography took LAO45°, anteroposterior position + head tilt 20°-25°; the left coronary artery angiography took RAO30° + foot tilt 20°, RAO30° + head tilt 20°, LAO45° + head tilt 25°, LAO60° + foot tilt 30°.

3. The intelligent collateral physiology assessment system based on coronary angiography images according to claim 1, characterized in that: In the segmentation module, CNN is used as a feature extractor to capture local boundary feature maps and a multi-head attention mechanism is used. The channel attention module enhances the feature extraction ability of the model by focusing on the importance of different feature channels, and the spatial attention module further refines the model's processing of spatial information by focusing on the importance of different spatial regions.

4. The intelligent assessment system for collateral physiology based on coronary angiography images according to claim 1, characterized in that: The thickness of each voxel in the angiographic projection direction was calculated, and the effects of 2D image overlap and shortening were corrected. The vascular volume V was calculated using the following formula: ， Where x, y, and z are three-dimensional coordinates, A is the cross-sectional area of ​​the blood vessel, and l is the length of the blood vessel. is the thickness of blood vessels.

5. The intelligent assessment system for collateral physiology based on coronary angiography images according to claim 1, characterized in that: The blood flow modeling method in the modeling module is as follows: draw the time-volume curve for analysis, use a multi-stage filtering algorithm to perform signal enhancement processing on the curve, retain the trend characteristics of the contrast agent filling process, and suppress high-frequency noise interference. Then, the filling stage is detected, the first-order derivative dV / dt of the preprocessed curve is calculated to locate the rising edge, and the contrast agent inflow period is identified by the threshold method.

6. The intelligent assessment system for collateral physiology based on coronary angiography images according to claim 5, characterized in that: The interval with the largest rise was selected as the fitting interval, and the average blood flow parameters were estimated. The hemodynamic model was established based on the optimized fitting interval. The blood flow calculation formula is as follows: ， in It's blood flow, is the cross-sectional area of ​​the blood vessel, V is the blood vessel volume, and t is the time; The formula for calculating mean blood flow is as follows: ， where Q avg is the average blood flow, V peak is the peak vascular volume, V baseline is the baseline vascular volume, t peak is the time when the contrast agent reaches its peak, t onset is the arrival time of the contrast agent.

7. The intelligent assessment system for collateral physiology based on coronary angiography images according to claim 1, characterized in that: Collateral flow index (CFI) Auto The calculation formula is: ； in It is the PCDA blood flow before CTO opening, It is the blood flow of PCDA after CTO opening, is the absolute change in collateral blood flow, which is calculated by simulating the change in blood flow in the main collateral artery before and after CTO opening. or To restore the patency of CTO vessels and restore the forward blood flow.

8. An intelligent assessment method for collateral physiology based on coronary angiography images, characterized in that: The following steps are involved: Step S1, acquiring coronary angiography and screening angiographic projection images with minimal overlap and shortening of collateral feeding arteries before and after CTO-PCI and after occluded vessels are opened; Step S2, performing frame-by-frame vessel segmentation on the acquired coronary angiography using a deep learning algorithm model; Step S3, performing 3D volume reconstruction on the segmented coronary artery binary image; Step S4, modeling the blood flow and calculating the average blood flow; Step S5: Calculate the collateral blood flow index based on the blood flow changes before and after the CTO is opened.

Images