Coronary artery bypass blood vessel patency OCT-FFR fusion image evaluation method and system

Abstract

The invention relates to the field of medical image processing, in particular to a coronary artery bypass blood vessel patency OCT-FFR fusion image evaluation method and system.The method comprises the steps that firstly, an OCT image and FFR data of a coronary artery bypass blood vessel of a patient are obtained, multi-scale wavelet transformation processing is conducted on the OCT image, and a feature image obtained after wavelet transformation is obtained; identifying a region of interest (OCT-ROI) in the OCT image based on the feature image; on the basis of FFR data, a functional abnormality region FFR-ROI is identified, and the limitation of single-modal evaluation is overcome by fusing OCT morphological information and FFR functional information; and an objective basis is provided for clinical treatment decisions based on a risk grading mechanism of multi-dimensional similarity.

Description

Technical Field

[0001] This invention relates to the field of medical image processing, specifically to a method and system for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images, and particularly to a method and system for evaluating the patency of coronary artery bypass grafts by combining optical coherence tomography (OCT) and fractional flow reserve (FFR) dual-modal data. Background Technology

[0002] Currently, assessing the patency of bypass vessels mainly relies on two techniques: optical coherence tomography (OCT) and fractional flow reserve (FFR) measurement.

[0003] OCT technology provides high-resolution vascular morphology information, clearly displaying vessel wall structure and plaque characteristics, but it cannot directly assess vascular function. FFR measurement, on the other hand, provides quantitative parameters of vascular function, reflecting hemodynamic characteristics, but lacks description of vascular microstructure. Both technologies have their advantages, but both have limitations when used alone, making it difficult to comprehensively assess the patency of bypass grafts.

[0004] In current technologies, OCT and FFR are usually used independently, requiring physicians to rely on experience to make a comprehensive judgment on the results of the two examinations, lacking an objective and quantitative method for fusion assessment. Furthermore, because OCT provides morphological information while FFR provides functional information, their information modalities differ, making direct fusion analysis difficult, which also limits the accuracy and reliability of the assessment.

[0005] Therefore, how to effectively integrate information from two different modalities, OCT and FFR, to achieve a comprehensive and accurate assessment of the patency of coronary artery bypass grafts has become a pressing technical problem that needs to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images. This method aims to overcome the limitations of single-modal evaluation in existing technologies and achieve a comprehensive and accurate evaluation of the patency of coronary artery bypass grafts by fusing OCT morphological information with FFR functional information.

[0007] This invention proposes a method for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images, including:

[0008] Acquire optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass grafts;

[0009] The OCT image is subjected to multi-scale wavelet transform processing to obtain the wavelet-transformed feature image;

[0010] Based on the feature image after wavelet transform, the region of interest (OCT-ROI) in the OCT image is identified;

[0011] Based on the FFR data, identify the functionally abnormal region FFR-ROI;

[0012] The identification of the region of interest (OCT-ROI) in the OCT image based on the feature image after wavelet transform includes:

[0013] Construct a multi-dimensional feature description system that includes morphological feature sets, texture feature sets, and color feature sets;

[0014] The color feature set is established based on the variance and average of the three components of hue, saturation and brightness in the HIS color space.

[0015] Based on the aforementioned multidimensional feature description system, the OCT image is segmented using the fuzzy C-means FCM clustering algorithm to obtain the region of interest (OCT-ROI) in the OCT image.

[0016] Preferably, the acquisition of optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass graft includes:

[0017] Multidimensional vascular images are simultaneously measured using an intravascular optical coherence tomography catheter, and the OCT images are continuously acquired from the proximal vascular inlet.

[0018] The FFR data of the coronary artery bypass graft were detected using a pressure measurement guidewire.

[0019] Preferably, the multi-scale wavelet transform processing of the OCT image includes:

[0020] The OCT image is subjected to wavelet transform using the coif5 wavelet basis to obtain wavelet coefficients at multiple decomposition levels.

[0021] The wavelet coefficients of the multiple decomposition levels are thresholded, with the threshold for low-frequency images set to 80% of the total number of pixels and the threshold for high-frequency images set to 90% of the total number of pixels.

[0022] The low-frequency image and the high-frequency image are divided into multiple segments, and the mean of each segment is calculated.

[0023] Based on the mean of each segment, the feature image after wavelet transform is obtained.

[0024] Preferably, a spatial mapping relationship is established between the OCT-ROI and the FFR-ROI;

[0025] Calculate the multidimensional similarity between the OCT-ROI and the FFR-ROI;

[0026] The patency of the patient's coronary artery bypass grafts was assessed based on the multidimensional similarity.

[0027] Preferably, the identification of functionally abnormal regions FFR-ROI based on the FFR data includes:

[0028] The FFR data is preprocessed, including signal smoothing, baseline correction, outlier filtering, and data normalization.

[0029] Extract the pressure gradient features, time series features, waveform morphology features, and statistical features from the FFR data;

[0030] Based on the extracted features, the FFR data is segmented using the fuzzy C-means FCM clustering algorithm to obtain the functionally abnormal region FFR-ROI;

[0031] The blood flow reserve score corresponding to the functionally abnormal region FFR-ROI is less than 0.8.

[0032] Preferably, establishing the spatial mapping relationship between the OCT-ROI and the FFR-ROI includes:

[0033] Establish a unified reference coordinate system with the long axis of the blood vessel as the Z-axis;

[0034] Transform the OCT-ROI coordinates and FFR-ROI coordinates to the unified reference coordinate system respectively;

[0035] Identify anatomical landmarks in OCT-ROI and FFR-ROI, including vascular branches and calcifications;

[0036] Based on the anatomical landmarks, preliminary registration between OCT-ROI and FFR-ROI is performed;

[0037] A non-rigid deformation model was used to perform fine registration of local regions, resulting in the spatial mapping relationship between OCT-ROI and FFR-ROI.

[0038] Preferably, calculating the multidimensional similarity between the OCT-ROI and the FFR-ROI includes:

[0039] Calculate pixel-level similarity, including pixel value differences, region overlap, edge consistency, and intensity distribution similarity;

[0040] Calculate feature-level similarity, including shape similarity, texture similarity, orientation similarity, and scale similarity;

[0041] Calculate semantic-level similarity, including consistency of lesion type, correlation of severity, consistency of risk assessment, and correlation of treatment indications;

[0042] An adaptive weighting mechanism is used to weight and fuse the pixel-level similarity, feature-level similarity, and semantic-level similarity to obtain the multidimensional similarity between OCT-ROI and FFR-ROI.

[0043] Preferably, the adaptive weighting mechanism includes:

[0044] Initial weights are set based on expert experience;

[0045] The weights are dynamically adjusted based on image quality, feature saliency, and consistency assessment.

[0046] The weights are continuously optimized through historical data analysis, expert feedback, and pattern recognition.

[0047] Ensure that the weights satisfy the normalization constraint, that is, the sum of all weights equals 1.

[0048] Preferably, assessing the patency of the patient's coronary artery bypass graft based on the multidimensional similarity includes:

[0049] Set similarity thresholds, including low-risk thresholds and high-risk thresholds;

[0050] When the multidimensional similarity is lower than the low-risk threshold, the bypass vessel is deemed to have good patency.

[0051] When the multidimensional similarity is between the low-risk threshold and the high-risk threshold, it is determined that the bypass vessel has potential risks and requires regular follow-up.

[0052] When the multidimensional similarity is higher than the high-risk threshold and the OCT-ROI and FFR-ROI are spatially identical, it is determined that there is an occlusive lesion in the bypass vessel, and the antiplatelet drug regimen needs to be adjusted to strengthen antithrombotic treatment.

[0053] The OCT-FFR fusion image assessment system for coronary artery bypass graft patency includes:

[0054] The data acquisition module is used to acquire optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass grafts.

[0055] The wavelet transform processing module is used to perform multi-scale wavelet transform processing on the OCT image to obtain the wavelet-transformed feature image.

[0056] The OCT region identification module is used to identify the region of interest (OCT-ROI) in the OCT image based on the feature image after wavelet transform.

[0057] The FFR region identification module is used to identify functionally abnormal regions (FFR-ROIs) based on the FFR data.

[0058] The spatial mapping module is used to establish the spatial mapping relationship between the OCT-ROI and the FFR-ROI;

[0059] The similarity calculation module is used to calculate the multidimensional similarity between the OCT-ROI and the FFR-ROI;

[0060] A patency assessment module is used to assess the patency of the patient's coronary artery bypass graft based on the multidimensional similarity.

[0061] This invention uses multi-scale wavelet transform to process OCT images, accurately identifies regions of interest, and combines it with FFR data to identify regions with functional abnormalities. It establishes a spatial mapping relationship between the two and calculates multidimensional similarity, thereby achieving an objective assessment of the patency of bypass vessels.

[0062] The present invention has the following beneficial effects:

[0063] 1. It achieves effective integration of morphological and functional information, overcomes the limitations of single-modal assessment, and provides more comprehensive and accurate assessment results of bypass vessel patency.

[0064] 2. Using multi-scale wavelet transform technology to process OCT images can more accurately extract vascular structure features and improve the accuracy of region of interest identification.

[0065] 3. By establishing a spatial mapping relationship between OCT-ROI and FFR-ROI, the problem of image registration for different modalities was solved, laying the foundation for similarity calculation.

[0066] 4. A multidimensional similarity assessment mechanism is adopted, which comprehensively considers pixel-level, feature-level and semantic-level information, thereby improving the reliability and clinical significance of the assessment results.

[0067] 5. A risk grading mechanism based on multidimensional similarity can provide objective evidence for clinical treatment decisions, guide the formulation of personalized antiplatelet therapy plans, and reduce the risk of postoperative complications. Attached Figure Description

[0068] Figure 1 This is a flowchart of the OCT-FFR fusion image assessment method for coronary artery bypass graft patency according to the present invention;

[0069] Figure 2 This is a structural framework diagram of the OCT-FFR fusion image assessment system for coronary artery bypass graft patency according to the present invention. Detailed Implementation

[0070] Please refer to Figures 1-2 The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0071] Reference Figure 1 The present invention provides a method for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images, comprising the following steps:

[0072] Step 1: Acquire OCT images and FFR data of the patient's coronary artery bypass graft. In one embodiment of the present invention, optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass graft are first acquired. Specifically, an OCT catheter and an FFR pressure guidewire (typically 0.014 inches in diameter) are inserted into the heart through the apex of the left ventricle to perform coronary artery bypass grafting. Multidimensional vascular images are simultaneously measured using an intravascular optical coherence tomography catheter (e.g., the OptoWire catheter from St. Jude Medical Systems, Inc.). Tomographic images of the vessels are continuously acquired from the proximal vessel inlet. Preferably, 300 frames of tomographic images are acquired continuously, with a sampling interval of 1 mm between each frame. Simultaneously, the fractional flow reserve data of the coronary artery bypass graft is detected using a pressure measurement guidewire.

[0073] This step provides foundational data for subsequent analysis. OCT images offer morphological information such as vessel wall structure and plaque characteristics, while FFR data provide quantitative parameters of vascular function. These two types of data complement each other, enabling a comprehensive assessment of bypass vessel patency.

[0074] Step 2: Perform multi-scale wavelet transform processing on the OCT image. After acquiring the OCT image, perform multi-scale wavelet transform processing on it to obtain the wavelet-transformed feature image. Specifically, select the coif5 wavelet basis to perform wavelet transform on the OCT image to obtain wavelet coefficients at multiple decomposition levels.

[0075] In a preferred embodiment of the present invention, the wavelet transform is set to 5 layers, which can fully extract multi-scale features while maintaining computational efficiency. After the wavelet transform, the obtained coefficients are thresholded, wherein the threshold for low-frequency images is... Set to 80% of the total number of pixels in the image, the threshold for high-frequency images. The threshold is set to 90% of the total number of pixels in the image. These threshold parameters have been clinically validated to effectively separate vascular structures from background noise.

[0076] The wavelet transform process can be represented by the following mathematical formula:

[0077] ,

[0078] in, These are the wavelet transform coefficients. The input is the OCT image signal. For wavelet basis functions, Indicates the scale parameter. Denotes the translation parameters. Wavelet basis functions. It is usually defined as:

[0079] ,

[0080] in, The coif5 wavelet basis selected in this invention has good smoothness and localization properties, serving as the mother wavelet function.

[0081] For two-dimensional images, wavelet transform can be performed using row transform and column transform respectively:

[0082] ,

[0083] in, These are the two-dimensional wavelet transform coefficients. It is a two-dimensional OCT image. and Let be the wavelet basis functions in the x and y directions, respectively. and These are the scale parameters in the x and y directions, respectively. and These are the translation parameters in the x and y directions, respectively.

[0084] Thresholding uses the following rules: for low-frequency coefficients ,if If the value is 0, retain it; otherwise, set it to 0. For high-frequency coefficients... ,if If the expression is true, keep it; otherwise, set it to 0. Represents the low-frequency coefficients of the j-th layer. Represents the high-frequency coefficients of the j-th layer. and These are the low-frequency and high-frequency thresholds, respectively.

[0085] Next, the low-frequency and high-frequency images are divided into multiple segments: 8 segments for the low-frequency image and 16 segments for the high-frequency image. The mean of each segment is calculated. Based on these means, the wavelet-transformed feature image is reconstructed using inverse wavelet transform.

[0086] The inverse wavelet transform can be expressed as:

[0087] ,

[0088] in, The reconstructed OCT feature image, For the wavelet coefficients after thresholding, the summation operation iterates through all combinations of scale and translation parameters.

[0089] This multi-scale wavelet transform process can effectively extract multi-level features from OCT images, providing a reliable foundation for subsequent region of interest identification.

[0090] Step 3: Identify the region of interest (OCT-ROI) in the OCT image. Based on the feature image after wavelet transform, this invention constructs a multi-dimensional feature description system that includes morphological feature sets, texture feature sets, and color feature sets, which is used to identify the region of interest (OCT-ROI) in the OCT image.

[0091] In one embodiment of the present invention, the color feature set is established based on the variance and average of the three components of hue (H), saturation (S), and lightness (I) in the HIS color space. For example, for each pixel in an OCT image, the RGB value is first converted to an HIS value, and then the variance of each component of H, S, and I within the region is calculated. , , and average , , , forming color feature vectors .in, , , Let H, S, and I represent the variances of the components, respectively. , , Let H, S, and I represent the average values ​​of the components, respectively. This represents the color feature vector.

[0092] The formula for converting RGB to HSI is:

[0093] ,

[0094] ,

[0095] ,

[0096] Where H represents the hue component (range from 0 to 360 degrees), S represents the saturation component (range from 0 to 1), I represents the luminance component (range from 0 to 255), and R, G, and B are the pixel values ​​of the red, green, and blue channels, respectively (range from 0 to 255). This represents the minimum value among R, G, and B. It's important to note that when B > G, .

[0097] The morphological feature set includes geometric parameters such as lumen diameter, area, and perimeter, while the texture feature set includes statistical features such as local ambiguity and homogeneity. Based on these features, the OCT image is segmented using the fuzzy C-means (FCM) clustering algorithm to obtain the region of interest (OCT-ROI) in the OCT image.

[0098] The objective function of the FCM clustering algorithm is:

[0099] ,

[0100] in, Describe the objective function. The fuzzy factor (usually taken as 2), For the sample The membership degree belonging to class j. Let the cluster center be the j-th class. For the sample size, The number of clusters, Indicates sample To the cluster center The Euclidean distance.

[0101] The FCM algorithm is optimized through iterative optimization. and , so that the objective function Minimize:

[0102] ,

[0103] ,

[0104] in, Indicates sample The membership degree belonging to class j. Denotes the cluster center of the j-th class. It is a clustering index, ranging from 1 to , This indicates summing over all clusters.

[0105] In this invention, the number of clusters C=3 is set, corresponding to the blood vessel lumen, blood vessel wall, and plaque region, respectively. The iteration termination condition is that the change in the objective function is less than a preset threshold (e.g., 0.001) or the maximum number of iterations (e.g., 100 times) is reached.

[0106] Through the above process, the present invention can accurately identify regions of interest in OCT images, especially patchy regions, providing a basis for subsequent analysis.

[0107] Step 4: Identify FFR-ROIs (Regions of Regions of Functional Abnormality) in FFR Data. Based on FFR data, this invention identifies FFR-ROIs. First, the FFR data is preprocessed, including signal smoothing, baseline correction, outlier filtering, and data standardization.

[0108] Signal smoothing employs a sliding window averaging method with a window size of 5 sampling points to filter out high-frequency noise. Baseline correction is achieved by subtracting the low-pass filtered signal, with a cutoff frequency set to 0.05Hz. Outlier filtering uses the 3σ criterion, classifying values ​​deviating from the mean by more than three times the standard deviation as outliers and replacing them with the local mean. Data standardization maps FFR values ​​to the [0,1] interval for easier subsequent processing.

[0109] Next, pressure gradient features, time series features, waveform morphology features, and statistical features of the FFR data are extracted. Pressure gradient features reflect pressure differences at different locations within the blood vessel and are an important indicator for assessing hemodynamic abnormalities. Time series features include the periodic changes in the pressure waveform, while waveform morphology features describe the shape of the pressure wave, such as peak value, trough value, and rise time. Statistical features include statistics such as mean, variance, skewness, and kurtosis.

[0110] Based on these characteristics, the FFR data is segmented using the FCM clustering algorithm to obtain functionally abnormal regions FFR-ROIs. In clinical practice, an FFR value less than 0.8 is generally considered functionally significant, indicating a risk of insufficient blood perfusion in that area. Therefore, this invention identifies regions with FFR values ​​less than 0.8 as functionally abnormal regions FFR-ROIs.

[0111] The formula for calculating the FFR value is:

[0112] ,

[0113] Here, FFR represents fractional flow reserve. Pressure at the distal end of the stenosis (unit: mmHg). The value is aortic pressure (unit: mmHg). When FFR < 0.8, it indicates a decrease in coronary blood flow reserve and a risk of functional ischemia.

[0114] By identifying regions with abnormal FFR function, this invention can locate areas in blood vessels with abnormal hemodynamics, preparing for subsequent fusion analysis with OCT regions of interest.

[0115] Step 5: Establish the spatial mapping relationship between OCT-ROI and FFR-ROI. In order to achieve the fusion of OCT morphological information and FFR functional information, this invention establishes the spatial mapping relationship between OCT-ROI and FFR-ROI.

[0116] First, a unified reference coordinate system is established with the long axis of the blood vessel as the Z-axis. OCT images are typically represented in cylindrical coordinates, while FFR data is one-dimensional data along the long axis of the blood vessel. Transforming both to a unified reference coordinate system is the first step in registration.

[0117] Let the coordinates of the point in the OCT image be... ,in Radial distance (unit: mm). Angle (unit: radians) The coordinates of the FFR data point are axial positions (unit: mm). (Unit: mm). Coordinate transformation can be expressed as:

[0118] ,

[0119] ,

[0120] in, This represents the coordinates of the OCT point in a unified coordinate system. This represents the z-coordinate in the OCT coordinate system. and These represent the minimum and maximum values ​​of the OCT coordinates, respectively. To standardize the coordinate system, the length (unit: mm) is usually taken as the actual length of the blood vessel segment. Similarly, This represents the coordinates of the FFR point in a unified coordinate system. Indicates the original coordinates of FFR. and These represent the minimum and maximum values ​​of the FFR coordinates, respectively.

[0121] Next, anatomical landmarks, including vascular branches and calcifications, were identified in the OCT-ROI and FFR-ROI. These landmarks are key reference points for registration. In OCT images, vascular branches appear as abrupt changes in the luminal contour, while calcifications appear as highly reflective areas. In FFR data, vascular branches typically show significant changes in pressure gradients.

[0122] Based on these anatomical landmarks, preliminary registration is performed between OCT-ROI and FFR-ROI. The registration process can be represented as finding the optimal transformation parameter T that minimizes the registration error:

[0123] ,

[0124] in, Indicates the transformation parameters. This represents the i-th marker point in OCT. This represents the i-th marker point in the FFR. Indicates the number of markers. This represents the square of the Euclidean distance between the transformed OCT marker and the corresponding FFR marker. This indicates the search for the value of T that minimizes the following expression.

[0125] Because blood vessels may deform during different examinations, simple rigid transformations may not meet the requirements for accurate registration. Therefore, this invention employs a non-rigid deformation model for fine registration of local areas. Non-rigid deformation can be achieved through a free deformation model or B-spline transformation.

[0126] Taking B-spline transformation as an example, the deformation field can be represented as a linear combination of control points:

[0127] ,

[0128] in, Point The deformation vector, , , Let these represent the B-spline basis functions in the three directions, , , For index variables (range 0 to 3), Indicates control points, yes Position in the local coordinate system. Represents a triple sum, respectively for... From 0 to 3, From 0 to 3 and Accumulate from 0 to 3.

[0129] Through the above registration process, this invention establishes a precise spatial mapping relationship between OCT-ROI and FFR-ROI, laying the foundation for subsequent similarity calculation.

[0130] Step Six: Calculate the multidimensional similarity between OCT-ROI and FFR-ROI. Based on the established spatial mapping relationship, this invention calculates the multidimensional similarity between OCT-ROI and FFR-ROI, including pixel-level similarity, feature-level similarity and semantic-level similarity.

[0131] Pixel-level similarity primarily examines the degree of difference in pixel values ​​within corresponding regions, regional overlap, edge consistency, and intensity distribution similarity. Feature-level similarity focuses on the matching degree of high-level features such as shape similarity, texture similarity, orientation similarity, and scale similarity. Semantic-level similarity is assessed from the perspective of clinical significance, including consistency of lesion type, correlation of severity, consistency of risk assessment, and correlation of treatment indications.

[0132] These similarity metrics are calculated using different methods. For example, for pixel value differences, mean squared error (MSE) or structural similarity index (SSIM) can be used:

[0133] ,

[0134] ,

[0135] Where MSE represents the mean squared error, and N represents the number of pixels. This represents the value of the i-th pixel in the OCT image. This represents the value of the i-th pixel in the FFR image. This represents summing over all pixels. SSIM stands for Structural Similarity Index. and These represent the registered OCT and FFR images, respectively. and These represent the mean values ​​of the OCT and FFR images, respectively. and These represent the standard deviations of the OCT and FFR images, respectively. This represents the covariance of the OCT and FFR images. and The stability constant is usually set to 0. and , where L is the dynamic range of pixel values.

[0136] Shape similarity can be calculated using the Dice coefficient or Hausdorff distance:

[0137] ,

[0138] ,

[0139] Where Dice represents the Dice coefficient. This represents the intersection area of ​​the OCT-ROI and FFR-ROI. and represent the areas of the OCT-ROI and FFR-ROI, respectively. Hausdorff represents the Hausdorff distance. This represents the one-way Hausdorff distance from set A to set B, where max represents the maximum value and min represents the minimum value. This represents the Euclidean distance between points a and b.

[0140] This invention employs an adaptive weighting mechanism to weight and fuse these similarity metrics, obtaining a multidimensional similarity between OCT-ROI and FFR-ROI. The adaptive weighting mechanism first sets initial weights based on expert experience; for example, the initial weights for pixel-level, feature-level, and semantic-level similarity are 0.3, 0.4, and 0.3, respectively. Then, the weights are dynamically adjusted based on image quality, feature saliency, and consistency assessment. For example, when the OCT image quality is poor, the weight of pixel-level similarity is reduced; when the feature extraction results are significant, the weight of feature-level similarity is increased.

[0141] The weight adjustment formula can be expressed as:

[0142] ,

[0143] ,

[0144] in, This indicates the adjusted weights. This represents the initial weight of the i-th similarity index. This represents the adjustment factor (usually 0.2). This represents the quality factor (ranging from -1 to 1). This represents the final weights after normalization. This represents the sum of all adjusted weights.

[0145] Through historical data analysis, expert feedback, and pattern recognition, the weights can be continuously optimized to adapt to the characteristics of different patients and different disease types. The weights satisfy the normalization constraint, that is, the sum of all weights equals 1.

[0146] The final formula for calculating multidimensional similarity is:

[0147] ,

[0148] in, This represents the final multidimensional similarity. This represents the weight of the i-th similarity index. This represents the value of the i-th similarity index. This indicates that all similarity metrics are summed.

[0149] By calculating the multidimensional similarity between OCT-ROI and FFR-ROI, this invention can quantitatively assess the spatial consistency between morphological and functional abnormalities, providing an objective basis for assessing bypass vessel patency.

[0150] Step Seven: Assess the patency of the patient's coronary artery bypass grafts. Based on multidimensional similarity, this invention assesses the patency of the patient's coronary artery bypass grafts. First, similarity thresholds are set, including low-risk, intermediate-risk, and high-risk thresholds. In clinical practice, these thresholds are typically determined based on extensive clinical data and expert consensus; for example, a low-risk threshold is set to 0.3, an intermediate-risk threshold to 0.6, and a high-risk threshold to 0.8.

[0151] When the multidimensional similarity is below the low-risk threshold (0.3), the bypass graft is considered to have good patency. This indicates that the spatial distribution of morphological abnormalities detected by OCT is inconsistent with that of functional abnormalities detected by FFR, possibly because the two tests detect different types of problems, or one of the tests is a false positive. In this case, the bypass graft is considered to have good patency, and the patient can continue treatment according to the standard antiplatelet therapy regimen.

[0152] When the multidimensional similarity falls between the low-risk threshold (0.3) and the high-risk threshold (0.8), the bypass graft is considered to pose a potential risk and requires regular follow-up. This indicates that the morphological abnormalities detected by OCT and the functional abnormalities detected by FFR have some spatial overlap, but are not entirely consistent. In such cases, although there may be no obvious clinical symptoms at present, there is a risk of developing serious problems, requiring enhanced monitoring.

[0153] When the multidimensional similarity is higher than the high-risk threshold (0.8) and the OCT-ROI and FFR-ROI are spatially perfectly matched, it is determined that there is an occlusive lesion in the bypass vessel, and the antiplatelet drug regimen needs to be adjusted to strengthen antithrombotic therapy. This indicates that the morphological abnormalities are highly consistent with the functional abnormalities, and there is a high probability of serious vascular lesions, such as unstable plaques or thrombosis, requiring active intervention and treatment.

[0154] In such cases, it is generally recommended to increase the dose of antiplatelet drugs or use multiple antiplatelet drugs in combination. For example, clopidogrel (75 mg / day) or ticagrelor (90 mg twice daily) can be added to the regular aspirin (100 mg / day) regimen. For high-risk patients, short-term use of low-dose anticoagulants, such as rivaroxaban (2.5 mg twice daily), can also be considered.

[0155] Through this risk grading mechanism based on multidimensional similarity, the present invention can provide clinicians with objective and quantitative assessment results and treatment suggestions, help develop personalized treatment plans, improve treatment effects, and reduce the risk of complications.

[0156] Reference Figure 2The present invention also provides a coronary artery bypass graft patency OCT-FFR fusion image evaluation system, including: a data acquisition module 1, a wavelet transform processing module 2, an OCT region recognition module 3, an FFR region recognition module 4, a spatial mapping module 5, a similarity calculation module 6, and a patency evaluation module 7.

[0157] Data acquisition module 1 is used to acquire optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass graft. This module includes an OCT imaging device, an FFR measurement device, and a data transmission and storage unit. The OCT imaging device typically uses an intravascular optical coherence tomography catheter (such as the OptoWire catheter from St. Jude Medical Systems, Inc.), with an axial resolution of 10-15 μm and a retraction speed of 20-40 mm / s. The FFR measurement device uses a pressure guidewire, typically 0.014 inches in diameter, capable of accurately measuring pressure values ​​at different locations within the vessel.

[0158] Wavelet transform processing module 2 is used to perform multi-scale wavelet transform processing on OCT images to obtain wavelet-transformed feature images. This module implements the wavelet transform process, including selecting the coif5 wavelet basis, setting threshold parameters, segmented processing, and feature image reconstruction. Utilizing GPU acceleration technology, this module can complete wavelet transform processing in milliseconds, meeting the needs of real-time clinical processing.

[0159] OCT Region Recognition Module 3 is used to identify regions of interest (OCT-ROIs) in OCT images based on feature images after wavelet transform. This module implements the OCT-ROI recognition process, including multidimensional feature extraction and FCM clustering segmentation. This module can accurately identify plaque regions in OCT images, especially high-risk structures such as lipid cores.

[0160] The FFR Region Identification Module 4 is used to identify functionally abnormal regions (FFR-ROIs) based on FFR data. This module implements the FFR-ROI identification process, including FFR data preprocessing, feature extraction, and clustering segmentation. This module can accurately identify functionally abnormal regions with FFR values ​​less than 0.8, providing a foundation for subsequent analysis.

[0161] The spatial mapping module 5 is used to establish the spatial mapping relationship between OCT-ROI and FFR-ROI. This module implements the spatial mapping establishment process, including coordinate system transformation, marker point recognition, preliminary registration, and non-rigid deformation correction. This module solves the problem of registration between different modal images and lays the foundation for similarity calculation.

[0162] The similarity calculation module 6 is used to calculate the multidimensional similarity between OCT-ROI and FFR-ROI. This module implements the similarity calculation process, including pixel-level, feature-level, and semantic-level similarity calculations, as well as adaptive weight optimization. This module can comprehensively evaluate the similarity between OCT-ROI and FFR-ROI, providing objective and quantitative evaluation results.

[0163] The patency assessment module 7 is used to evaluate the patency of coronary artery bypass grafts based on multidimensional similarity. This module implements the patency assessment process, including risk stratification and treatment recommendation generation. It provides clinicians with clear assessment results and specific treatment recommendations, supporting the development of personalized treatment plans.

[0164] The system of this invention exchanges data between its various modules through standardized interfaces, forming a complete processing flow. The system adopts a modular design, allowing each module to be independently upgraded and replaced, exhibiting good scalability and maintainability. The system supports integration with Hospital Information Systems (HIS) and Picture Archiving Systems (PACS), facilitating clinical applications.

[0165] The system's hardware configuration includes a high-performance medical imaging workstation, large-capacity storage devices, and a high-resolution display. The software employs a multi-layered architecture, comprising a data layer, algorithm layer, business layer, presentation layer, and interface layer. The system supports concurrent access by multiple users and features robust data security and privacy protection mechanisms.

[0166] Through the coordinated operation of the above modules, the system of the present invention can achieve a comprehensive and accurate assessment of the patency of coronary artery bypass grafts, providing clinicians with a powerful decision support tool, improving treatment outcomes, and reducing the risk of complications.

[0167] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images, characterized in that... include: Acquire optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass grafts; The OCT image is subjected to multi-scale wavelet transform processing to obtain the wavelet-transformed feature image; Based on the feature image after wavelet transform, the region of interest (OCT-ROI) in the OCT image is identified; Based on the FFR data, identify the functionally abnormal region FFR-ROI; Establish a spatial mapping relationship between the OCT-ROI and the FFR-ROI; Calculate the multidimensional similarity between the OCT-ROI and the FFR-ROI; The patency of the patient's coronary artery bypass grafts was assessed based on the multidimensional similarity. The identification of the region of interest (OCT-ROI) in the OCT image based on the feature image after wavelet transform includes: Construct a multi-dimensional feature description system that includes morphological feature sets, texture feature sets, and color feature sets; The color feature set is established based on the variance and average value of the three components of hue, saturation and brightness in the HSI color space. Based on the aforementioned multidimensional feature description system, the OCT image is segmented using the fuzzy C-means FCM clustering algorithm to obtain the region of interest (OCT-ROI) in the OCT image.

2. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 1, characterized in that, The acquisition of optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass graft includes: Multidimensional vascular images are simultaneously measured using an intravascular optical coherence tomography catheter, and the OCT images are continuously acquired from the proximal vascular inlet. The FFR data of the coronary artery bypass graft were detected using a pressure measurement guidewire.

3. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 1, characterized in that, The multi-scale wavelet transform processing of the OCT image includes: The OCT image is subjected to wavelet transform using the coif5 wavelet basis to obtain wavelet coefficients at multiple decomposition levels. The wavelet coefficients of the multiple decomposition levels are thresholded, with the threshold for low-frequency images set to 80% of the total number of pixels and the threshold for high-frequency images set to 90% of the total number of pixels. The low-frequency image and the high-frequency image are divided into multiple segments, and the mean of each segment is calculated. Based on the mean of each segment, the feature image after wavelet transform is obtained.

4. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 1, characterized in that, Establish a spatial mapping relationship between the OCT-ROI and the FFR-ROI; Calculate the multidimensional similarity between the OCT-ROI and the FFR-ROI; The patency of the patient's coronary artery bypass grafts was assessed based on the multidimensional similarity.

5. The method for evaluating the patency of coronary artery bypass grafts using OCT-FFR fusion images according to claim 1, characterized in that, The identification of functionally abnormal regions (FFR-ROIs) based on the FFR data includes: The FFR data is preprocessed, including signal smoothing, baseline correction, outlier filtering, and data normalization. Extract the pressure gradient features, time series features, waveform morphology features, and statistical features from the FFR data; Based on the extracted features, the FFR data is segmented using the fuzzy C-means FCM clustering algorithm to obtain the functionally abnormal region FFR-ROI; The blood flow reserve score corresponding to the functionally abnormal region FFR-ROI is less than 0.

8.

6. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 4, characterized in that, The process of establishing the spatial mapping relationship between the OCT-ROI and the FFR-ROI includes: Establish a unified reference coordinate system with the long axis of the blood vessel as the Z-axis; Transform the OCT-ROI coordinates and FFR-ROI coordinates to the unified reference coordinate system respectively; Identify anatomical landmarks in OCT-ROI and FFR-ROI, including vascular branches and calcifications; Based on the anatomical landmarks, preliminary registration between OCT-ROI and FFR-ROI is performed; A non-rigid deformation model was used to perform fine registration of local regions, resulting in the spatial mapping relationship between OCT-ROI and FFR-ROI.

7. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 4, characterized in that, The calculation of the multidimensional similarity between the OCT-ROI and the FFR-ROI includes: Calculate pixel-level similarity, including pixel value differences, region overlap, edge consistency, and intensity distribution similarity; Calculate feature-level similarity, including shape similarity, texture similarity, orientation similarity, and scale similarity; Calculate semantic-level similarity, including consistency of lesion type, correlation of severity, consistency of risk assessment, and correlation of treatment indications; An adaptive weighting mechanism is used to weight and fuse the pixel-level similarity, feature-level similarity, and semantic-level similarity to obtain the multidimensional similarity between OCT-ROI and FFR-ROI.

8. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 7, characterized in that, The adaptive weighting mechanism includes: Initial weights are set based on expert experience; The weights are dynamically adjusted based on image quality, feature saliency, and consistency assessment. The weights are continuously optimized through historical data analysis, expert feedback, and pattern recognition. Ensure that the weights satisfy the normalization constraint, that is, the sum of all weights equals 1.

9. The method for evaluating coronary artery bypass graft patency using OCT-FFR fusion images according to claim 8, characterized in that, The assessment of the patency of the patient's coronary artery bypass graft based on the multidimensional similarity includes: Set similarity thresholds, including low-risk thresholds and high-risk thresholds; When the multidimensional similarity is lower than the low-risk threshold, the bypass vessel is deemed to have good patency. When the multidimensional similarity is between the low-risk threshold and the high-risk threshold, it is determined that the bypass vessel has potential risks and requires regular follow-up. When the multidimensional similarity is higher than the high-risk threshold and the OCT-ROI and FFR-ROI are spatially identical, it is determined that there is an occlusive lesion in the bypass vessel, and the antiplatelet drug regimen needs to be adjusted to strengthen antithrombotic treatment.

10. A coronary artery bypass graft patency assessment system based on OCT-FFR fusion images, implementing the method described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire optical coherence tomography (OCT) images and fractional flow reserve (FFR) data of the patient's coronary artery bypass grafts. The wavelet transform processing module is used to perform multi-scale wavelet transform processing on the OCT image to obtain the wavelet-transformed feature image. The OCT region identification module is used to identify the region of interest (OCT-ROI) in the OCT image based on the feature image after wavelet transform. The FFR region identification module is used to identify functionally abnormal regions (FFR-ROIs) based on the FFR data. The spatial mapping module is used to establish the spatial mapping relationship between the OCT-ROI and the FFR-ROI; The similarity calculation module is used to calculate the multidimensional similarity between the OCT-ROI and the FFR-ROI; A patency assessment module is used to assess the patency of the patient's coronary artery bypass graft based on the multidimensional similarity.

Images