Aortic valve automatic segmentation method based on multi-phase CTA images

Abstract

The present application relates to the technical field of medical image processing, and in particular to an aortic valve automatic segmentation method based on multi-phase CTA images. The method identifies the laminar flow dark area and the vortex light area on the cross section of the aortic sinus region; determines the leaflet-sinus wall distance according to the first characteristic size of the laminar flow dark area and the second characteristic size of the vortex light area; determines the first deviation index according to the change trend of the leaflet-sinus wall distance in the cardiac cycle; determines the second deviation index according to the difference characteristics between the CT value of the laminar flow dark area and the CT value of the vortex light area; if the first deviation index and the second deviation index both exceed the corresponding threshold, it is determined that the data is in the phase to be corrected; corrects the distance by using the correction weight calculated by the second deviation index; extracts the key parameters of each phase on the longitudinal section, associates the corrected leaflet-sinus wall distance to construct a three-dimensional dynamic model of the valve, and thus the three-dimensional dynamic model outputs more accurate valve motion data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing, and in particular to an aortic valve automatic segmentation method based on multi-phase CTA images. BACKGROUND

[0002] With the development of medical imaging technology, CTA (CT angiography) has become the key to accurate segmentation of aortic valves due to its high resolution and three-dimensional imaging advantages.

[0003] At present, in the prior art, valve segmentation mainly relies on single-phase CTA images, however, this method cannot capture the dynamic changes of aortic valve leaflet structure and boundary position in the cardiac cycle, resulting in difficulty in accurately identifying the running trajectory parameters of the valve; on the other hand, due to factors such as partial volume effect and uneven distribution of contrast agent in CT imaging, data noise is easily generated, which causes errors in the calculation of key parameters and affects the accuracy of the aortic valve motion trajectory parameters quantitatively segmented. SUMMARY

[0004] In order to solve the technical problem that the running trajectory parameters of the valve are difficult to accurately identify when the valve is segmented by relying on single-phase CTA images, the present application provides an aortic valve automatic segmentation method based on multi-phase CTA images, and the technical solution adopted is as follows:

[0005] The present application provides an aortic valve automatic segmentation method based on multi-phase CTA images, which comprises the following steps:

[0006] Collecting multi-phase CTA images of the aortic root covering the complete cardiac cycle; locating the aortic sinus region in each phase of the CTA images; identifying the laminar dark area and the vortex bright area on the transverse cross-sectional images of the aortic sinus region;

[0007] Determining the leaflet-sinus wall distance of each phase according to the first characteristic size of the laminar dark area and the second characteristic size of the vortex bright area in each phase; determining the first deviation index of each phase according to the change trend of the leaflet-sinus wall distance in the cardiac cycle; determining the second deviation index of each phase according to the difference characteristics between the CT value of the laminar dark area and the CT value of the vortex bright area in each phase;

[0008] If the first deviation index exceeds the preset first index threshold and the second deviation index exceeds the preset second index threshold, the corresponding phase is determined as a data to be corrected phase; determining the correction weight according to the second deviation index of the data to be corrected phase; correcting the leaflet-sinus wall distance of the data to be corrected phase using the correction weight;

[0009] Extracting morphological and functional dynamic parameters of the aortic valve at each phase on the longitudinal section image of the aortic sinus region, correlating the leaflet-sinus wall distance at each phase to construct a three-dimensional dynamic model of the valve, wherein the distance of the phase to be corrected is the value after correction.

[0010] Further, the first characteristic size of the laminar dark region and the second characteristic size of the vortex bright region are determined by the following process:

[0011] On the cross-sectional image of the aortic sinus region, the inner edge boundary contour of the laminar dark region and the outer edge boundary contour of the vortex bright region are extracted.

[0012] The polar coordinate system is established with the center point of the aortic root as the origin, and a plurality of preset sampling angles are set along the circumferential direction. At each sampling angle, the radial distance from the origin to the inner edge boundary contour of the laminar dark region is calculated as the first characteristic distance at the corresponding angle, and the radial distance from the origin to the outer edge boundary contour of the vortex bright region is calculated as the second characteristic distance at the corresponding angle.

[0013] The arithmetic mean of the first characteristic distances corresponding to all sampling angles in each phase is calculated as the first characteristic size of the laminar dark region in each phase.

[0014] The arithmetic mean of the second characteristic distances corresponding to all sampling angles in each phase is calculated as the second characteristic size of the vortex bright region in each phase.

[0015] Further, after extracting the inner edge boundary contour of the laminar dark region and the outer edge boundary contour of the vortex bright region, the contours are smoothed, and the smoothing process includes removing isolated noise points on the contours.

[0016] Further, the laminar dark region and the vortex bright region are identified on the cross-sectional image of the aortic sinus region by the following process:

[0017] Performing gray scale analysis on the cross-sectional image.

[0018] Based on a preset CT value threshold range, the high CT value region and the low CT value region in the image are identified. The CT value of the high CT value region is in a first preset range, the CT value of the low CT value region is in a second preset range, and the lower limit value of the first preset range is higher than the upper limit value of the second preset range.

[0019] The identified low CT value region is taken as the laminar dark region, and the identified high CT value region is taken as the vortex bright region.

[0020] Further, the automatic aortic valve segmentation method based on multi-phase CTA images further includes:

[0021] The first area ratio of the laminar flow dark area and the second area ratio of the vortex flow bright area are calculated for multiple phases in a cardiac cycle; wherein the first area ratio is a ratio of a laminar flow dark area to a total area of a cross section of the aortic sinus region, and the second area ratio is a ratio of a vortex flow bright area to the total area of the cross section of the aortic sinus region;

[0022] Based on the time sequence variation characteristics of the leaflet-sinus wall distance, the first area ratio and the second area ratio in the cardiac cycle, each phase is classified into a first blood flow state period or a second blood flow state period.

[0023] Further, the first deviation index determination process comprises:

[0024] The leaflet-sinus wall distances of all phases in the same period are polynomially fitted to generate a distance change trend curve, and fitted distance values of each phase in the same period are obtained, wherein the same period represents the first blood flow state period or the second blood flow state period classified.

[0025] The difference between the leaflet-sinus wall distance of each phase and the fitted distance value of the same phase is calculated as a raw residual, and the standard deviation of all raw residuals in the same period is calculated.

[0026] The absolute value of the ratio of the raw residual of each phase to the standard deviation in the same period is taken as the first deviation index of the corresponding phase.

[0027] Further, the second deviation index determination process comprises:

[0028] Based on the CT value distribution of the pixel points in the laminar flow dark area, the first statistical characteristic value of the laminar flow dark area is calculated, and based on the CT value distribution of the pixel points in the vortex flow bright area, the second statistical characteristic value of the vortex flow bright area is calculated.

[0029] The difference between the first statistical characteristic value of the laminar flow dark area and the second statistical characteristic value of the vortex flow bright area of each phase is calculated to obtain a CT difference value of each phase.

[0030] The CT difference values of all phases in the same period are polynomially fitted to generate a CT difference value change trend curve, and fitted CT difference values of each phase are obtained, wherein the same period represents the first blood flow state period or the second blood flow state period classified.

[0031] The difference between the CT difference value of each phase and the corresponding fitted CT difference value is calculated as a CT raw residual of each phase, and the standard deviation of all CT raw residuals in the same period is calculated.

[0032] The absolute value of the ratio of the CT raw residual of each phase to the standard deviation in the same period is taken as the second deviation index of the corresponding phase.

[0033] Further, the calculation process of the first statistical characteristic value and the second statistical characteristic value comprises:

[0034] The arithmetic mean value of the CT values of all pixel points in the laminar dark region of each phase is calculated as the first statistical characteristic value of the laminar dark region.

[0035] The arithmetic mean value of the CT values of all pixel points in the vortex bright region of each phase is calculated as the second statistical characteristic value of the vortex bright region.

[0036] Further, the correction weight determination process comprises:

[0037] Exponential operation is performed with the natural constant as the base and the inverse of the second deviation index of each phase as the index to obtain the correction weight of each phase.

[0038] Further, the correction of the leaflet-sinus wall distance of the data to be corrected phase by using the correction weight comprises:

[0039] The fitting distance value corresponding to the data to be corrected phase is obtained.

[0040] The product of the correction weight and the leaflet-sinus wall distance of the data to be corrected phase is calculated as a first product, the difference between the numerical value 1 and the correction weight is calculated as a correction weight, and the product of the correction weight and the fitting distance value is calculated as a second product.

[0041] The sum of the first product and the second product is calculated as the corrected leaflet-sinus wall distance.

[0042] The present application has the following beneficial effects:

[0043] The present application determines the first deviation index by analyzing the change trend of the leaflet-sinus wall distance in the cardiac cycle, determines the second deviation index by analyzing the difference characteristics between the CT value of the laminar dark region and the CT value of the vortex bright region in each phase, and combines the first deviation index and the second deviation index to finely identify the phase of data anomaly in the cardiac cycle to eliminate accidental errors caused by single phase segmentation. The spatial scales of the laminar dark region and the vortex bright region are quantified by the first characteristic size and the second characteristic size to be compatible with their possible irregular shapes, so that the error of the determined leaflet-sinus wall distance is reduced. The correction weight calculated by the second deviation index is used to correct the problematic leaflet-sinus wall distance, and the leaflet-sinus wall distance is dynamically balanced by the correction weight, so that the change trend of the leaflet-sinus wall distance conforms to the hemodynamic characteristics of the blood flow passing through the aortic valve, and the subsequent acquisition of key data is avoided. Finally, the corrected distance, the leaflet attachment point, the opening angle and other key parameters are synchronously integrated to construct a three-dimensional dynamic model, so that more accurate valve movement data can be quickly output based on the three-dimensional dynamic model. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, below will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flow chart of an aortic valve automatic segmentation method based on multi-phase CTA images provided by an embodiment of the present application;

[0046] Figure 2 An example diagram of a laminar flow dark area and a vortex light area change process provided by an embodiment of the present application;

[0047] Figure 3 An example diagram of a first deviation index determination process provided by an embodiment of the present application;

[0048] Figure 4 An example diagram of a second deviation index determination process provided by an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined purposes, the specific embodiments, structures, features and effects of the aortic valve automatic segmentation method based on multi-phase CTA images according to the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0051] The specific scheme of the aortic valve automatic segmentation method based on multi-phase CTA images provided by the present application is described in detail below in combination with the drawings.

[0052] Please refer to Figure 1 which shows a flow chart of an aortic valve automatic segmentation method based on multi-phase CTA images provided by an embodiment of the present application. The method comprises:

[0053] S101: Collect multi-phase CTA images of the aortic root covering a complete cardiac cycle; locate the aortic sinus region in each phase of the CTA images; and identify the laminar flow dark area and the vortex light area on the cross-sectional images of the aortic sinus region.

[0054] Computed Tomography Angiography (CTA) is a medical imaging examination method combining Computed Tomography (CT) technology and angiography principles.

[0055] In order to accurately segment the aortic valve, a clear aortic root CTA image needs to be obtained, so some preliminary preparations need to be made before collecting multi-phase CTA images, for example, using a higher concentration of iodine contrast agent while ensuring that it does not cause significant adverse effects to the patient.

[0056] A complete cardiac cycle refers to the entire process of the heart from one contraction to the next contraction, including the systole (left ventricular ejection) and diastole (left ventricular filling) stages. In the systole stage, the left ventricle contracts, the aortic valve opens, and blood is ejected from the left ventricle into the aorta. In this period, the distance between the valve leaflet and the sinus wall gradually decreases to a minimum value. In the diastole stage, the left ventricle relaxes, the aortic valve closes, and blood flows back to form a vortex. In this period, the distance between the valve leaflet and the sinus wall gradually increases to a maximum value.

[0057] It should be noted that in the CTA image, the specific way of positioning the aortic sinus region and the specific way of identifying the laminar dark area and the vortex bright area are well-known techniques to those skilled in the art, and this embodiment will not be described again. For example, the existing nnU-Net model is used to locate the aortic sinus region and segment the cross-sectional image of the aortic sinus region. The Otsu ring section analysis algorithm is used to analyze the cross-sectional image of the aortic sinus region to determine the gray level histogram of the cross-sectional image, and then the cross-sectional image is divided into two parts: the laminar dark area and the vortex bright area.

[0058] It should be noted that during the systole period, due to the opening of the aortic valve leaflet, blood is rapidly ejected from the left ventricle into the main flow area of the aorta through the aortic valve orifice, and the blood flow is in a regular parallel flow (laminar flow state) with fast flow speed and rapid flushing of contrast agent. In addition, due to the expansion of the aortic sinus structure, the blood flow entering the sinus forms a rotating and slow vortex, and the contrast agent stays in the sinus for a longer time, resulting in a significantly higher contrast agent concentration than the laminar flow area in the center of the valve orifice. It can be seen that the CT value directly reflects the degree of X-ray absorption by blood, and the contrast agent is a strong X-ray absorbing material. The higher the concentration of contrast agent, the higher the corresponding CT value. Therefore, during the systole period, the blood flow channel in the aortic root CTA image, i.e., the laminar flow area, presents a low-CT-value "dark area" due to the low concentration of contrast agent. Correspondingly, the vortex area presents a high-CT-value "bright ring" in the CTA image, surrounding the periphery of the laminar flow area.

[0059] For example, in a complete cardiac cycle, the change process of the laminar dark area and the vortex bright area is shown in the following figure: Figure 2As shown, the change of CT value difference is synchronized with the opening and closing state of the aortic valve and the hemodynamic characteristics, for example, from the early systole to the mid-systole, generally showing that the left ventricular pressure rises, the valve leaflet gradually opens, the blood flow accelerates into the aorta, the flow velocity of the laminar dark area continuously rises, the contrast agent is quickly washed away, and the corresponding CT value gently decreases; the vortex area keeps a high CT value due to the retention of the contrast agent in the sinus, therefore, the CT value difference between the vortex bright area and the laminar dark area gradually increases with the decrease of the CT value of the laminar dark area, and the change curve is smooth (without sharp fluctuation). From the mid-systole to the early diastole, generally showing that the ventricular contractility decreases, the valve leaflet gradually closes, the flow velocity of the laminar dark area slows down, the contrast agent washing effect weakens, the CT value of the laminar dark area decreases slowly or even slightly rises, and the CT value of the vortex bright area is stable. Therefore, the increasing trend of the CT value difference slows down, and tends to be stable after the valve leaflet is completely closed.

[0060] It should be noted that, based on the change rule from systole to diastole, the change rule from the early diastole to the mid-diastole and then to the systole can be reversely analyzed. The CT value difference gradually decreases from the early diastole to the mid-diastole, and the change curve is smooth; the decreasing trend of the CT value difference slows down from the mid-diastole to the systole.

[0061] In this embodiment, the gray scale analysis is performed on the cross-sectional image; based on the preset CT value threshold range, the high-CT-value region and the low-CT-value region in the image are identified; the CT value of the high-CT-value region is in a first preset range, the CT value of the low-CT-value region is in a second preset range, and the lower limit value of the first preset range is higher than the upper limit value of the second preset range; the identified low-CT-value region is taken as the laminar dark area; and the identified high-CT-value region is taken as the vortex bright area.

[0062] The cross-sectional image refers to a two-dimensional image slice obtained by cutting perpendicular to the long axis of the aorta, which can clearly present the annular structure of the aortic sinus, the position of the valve leaflet attachment and the blood flow distribution state. The cross-sectional image usually shows an approximately circular aortic sinus cross-sectional profile.

[0063] It should be noted that the specific value range of the first preset range and the second preset range is adaptively adjusted according to the actual situation, which is not limited in this embodiment. For example, in order to avoid region overlap and ensure clear division of the high-CT-value region and the low-CT-value region, there is a clear interval between the first preset range and the second preset range, the first preset range (high-CT-value region) can be valued at 150HU-350HU, corresponding to the vortex bright area; and the second preset range (low-CT-value region) can be valued at 30HU-120HU, corresponding to the laminar dark area.

[0064] S102: determine the leaflet-dilatory wall distance of each phase according to the first characteristic size of the laminar dark area and the second characteristic size of the vortex bright area in each phase; determine the first deviation index of each phase according to the change trend of the leaflet-dilatory wall distance in the cardiac cycle; and determine the second deviation index of each phase according to the difference between the CT value of the laminar dark area and the CT value of the vortex bright area in each phase.

[0065] The leaflet-dilatory wall distance is the average distance between the leaflet and the dilatory wall calculated based on the first characteristic size of the laminar dark area and the second characteristic size of the vortex bright area, which can be referred to as the distance in the subsequent description.

[0066] It should be noted that according to step S101, it can be understood that in the cardiac cycle, when the systole period is transitioning to the diastole period, the range of the laminar dark area gradually decreases, the range of the vortex bright area gradually increases, and the leaflet-dilatory wall distance shows an increasing trend; on the contrary, when the diastole period is transitioning to the systole period, the ranges of the laminar dark area and the vortex bright area and the leaflet-dilatory wall distance all show reverse corresponding change trends.

[0067] In this embodiment, the first area ratio of the laminar dark area and the second area ratio of the vortex bright area are calculated for a plurality of phases in the cardiac cycle; based on the time sequence change characteristics of the leaflet-dilatory wall distance, the first area ratio and the second area ratio in the cardiac cycle, each phase is classified into a first blood flow state period or a second blood flow state period.

[0068] The first area ratio reflects the proportion of laminar flow in local blood flow, wherein the first area ratio is the ratio of the area of the laminar dark area to the total area of the cross section of the aortic sinus region.

[0069] The second area ratio reflects the proportion of vortex flow in local blood flow, wherein the second area ratio is the ratio of the area of the vortex bright area to the total area of the cross section of the aortic sinus region.

[0070] It should be noted that in a non-extreme case, the aortic sinus region cannot be absent, and therefore the total area of the cross section of the aortic sinus region cannot be zero.

[0071] It needs to be understood that in combination with the physiological movement of the aortic valve and the hemodynamic characteristics, when the aortic valve is open (systole): the ventricular ejection pressure pushes the free edge of the valve leaflet outwardly to the aortic sinus wall, the valve leaflet gradually approaches the sinus wall, and thus the valve leaflet-sinus wall distance gradually decreases, until the valve leaflet is fully unfolded and attached to the sinus wall (the valve leaflet-sinus wall distance reaches the minimum value), at this time, the laminar flow dark area is enlarged due to the opening of the valve leaflet (the first area ratio is high), and the vortex flow area is reduced (the second area ratio is low); when the aortic valve is closed (diastole): the ventricle stops ejecting blood, the blood flow in the aorta reverses, pushing the free edge of the valve leaflet inwardly to contract, the valve leaflet gradually moves away from the sinus wall, and thus the valve leaflet-sinus wall distance gradually increases, until the free edge of the valve leaflet is attached to each other (the valve leaflet-sinus wall distance reaches the maximum value), at this time, the vortex flow area (the regurgitation area between the valve leaflet and the sinus wall) is enlarged (the second area ratio is high), and the laminar flow area disappears or is reduced (the first area ratio is low).

[0072] For example, if the valve leaflet-sinus wall distance of a certain phase is less than that of the previous phase and the first area ratio is greater than that of the previous phase, it is determined that it is the first blood flow state period, i.e. the systole; if the valve leaflet-sinus wall distance of a certain phase is greater than that of the previous phase and the first area ratio is less than that of the previous phase, it is determined that it is the second blood flow state period, i.e. the diastole.

[0073] The first blood flow state period corresponds to the systolic phase of the aortic valve. At this time, the ventricle contracts, the valve leaflet opens, the blood flow is ejected from the left ventricle to the aorta in the form of high-speed laminar flow, the laminar flow dominates the blood flow (the dark area is large), the vortex flow is weak (the bright area is small), and the valve leaflet and the sinus wall distance show a decreasing trend.

[0074] The second blood flow state period corresponds to the diastolic phase of the aortic valve. At this time, the ventricle relaxes, the valve leaflet closes, the blood flow slows down and forms a vortex flow in the aortic sinus (the bright area is large), the laminar flow disappears or weakens (the dark area is small), and the vortex flow can help buffer the impact of blood flow on the sinus wall and maintain the stability of the blood vessel.

[0075] It needs to be understood that in the transverse section image of the aortic sinus, the shape of the laminar flow dark area and the vortex flow bright area is usually approximately circular or circular-like (because the aortic sinus has a symmetrical cystic structure, and the hemodynamics is easy to form a circular vortex flow), however, during the systole, the laminar flow dark area is enlarged and axially extended, and the transverse section may be elliptical; during the diastole, when the valve leaflet is closed, the vortex flow bright area is squeezed by the valve leaflet and may be an irregular sector in the shape of "three-leafed embrace", therefore, when determining the valve leaflet-sinus wall distance, the influence of the irregular laminar flow dark area or vortex flow bright area on the distance also needs to be considered.

[0076] The first characteristic size is a quantitative value for characterizing the spatial scale of the laminar dark region, reflecting the average radial position of the laminar dark region relative to the center point of the aortic root, and is a quantitative description of the overall spatial distribution of the valve leaflet.

[0077] The second characteristic size is a quantitative value for characterizing the spatial scale of the vortex bright region, reflecting the average radial position of the vortex bright region relative to the center point of the aortic root, and is a quantitative description of the overall spatial distribution of the sinus wall.

[0078] In this embodiment, on the cross-sectional image of the aortic sinus region, the inner edge boundary contour of the laminar dark region and the outer edge boundary contour of the vortex bright region are extracted; a polar coordinate system is established with the center point of the aortic root as the origin; a plurality of preset sampling angles are set along the circumferential direction; the radial distance from the origin to the inner edge boundary contour of the laminar dark region is calculated at each sampling angle direction as the first characteristic distance corresponding to the angle; the radial distance from the origin to the outer edge boundary contour of the vortex bright region is calculated as the second characteristic distance corresponding to the angle; the arithmetic mean of the first characteristic distances corresponding to all sampling angles in each phase is calculated as the first characteristic size of the laminar dark region in each phase; and the arithmetic mean of the second characteristic distances corresponding to all sampling angles in each phase is calculated as the second characteristic size of the vortex bright region in each phase.

[0079] It should be noted that before calculating the arithmetic mean of the first or second characteristic distance, the first or second characteristic distance of all sampling angles is subjected to outlier detection (such as Z-score method), and extreme values deviating from the mean by more than 3 times the standard deviation are removed to ensure the robustness of the arithmetic mean.

[0080] The inner edge boundary contour of the laminar dark region is a continuous contour line of the inner edge of the laminar dark region (low CT value region) in the cross-sectional image of the aortic sinus. In the heart, the inner edge boundary contour usually corresponds to the position of the free edge of the aortic valve leaflet. The laminar dark region is formed by high-speed laminar blood, and its inner side is adjacent to the valve leaflet. Therefore, the inner edge boundary contour can be regarded as the projection of the free edge of the valve leaflet on the cross-sectional image.

[0081] The outer edge boundary contour of the vortex bright region is a continuous contour line of the outer edge of the vortex bright region (high CT value region) in the cross-sectional image of the aortic sinus, which is the boundary between the vortex region and the surrounding tissue (such as the aortic sinus wall and myocardium). In the heart, the outer edge boundary contour usually corresponds to the inner side of the aortic sinus wall. The vortex bright region is formed by the retention of contrast agent, and its outer side is adjacent to the sinus wall. Therefore, the outer edge boundary contour can be regarded as the projection of the inner side of the sinus wall on the cross-sectional image.

[0082] It should be noted that after the inner edge boundary profile of the laminar flow dark area and the outer edge boundary profile of the vortex flow bright area are extracted, the profiles are smoothed, and the smoothing includes removing isolated noise points on the profiles, for example, the boundary profile is smoothed by using Gaussian filtering.

[0083] It should be noted that in order to ensure data consistency, the sampling angles can be uniformly arranged in the circumferential direction, and the difference values of adjacent sampling angles are equal, for example, one sampling angle is set every 30 degrees, and a total of 12 sampling angles are set. The specific way of arranging multiple sampling angles in the circumferential direction is a technology known to those skilled in the art, and will not be described in detail in the present embodiment.

[0084] In the present embodiment, the difference between the second characteristic size of the vortex flow bright area and the first characteristic size of the laminar flow dark area in each phase is calculated as the leaflet-sinus wall distance in each phase.

[0085] It can be understood that the leaflet-sinus wall distance = second characteristic size - first characteristic size.

[0086] The first deviation index determination process is as shown in Figure 3 , which includes:

[0087] S102-1: Perform polynomial fitting on the leaflet-sinus wall distances of all phases in the same period to generate a distance change trend curve and obtain the fitted distance values of each phase in the same period.

[0088] It should be noted that the same period refers to a period classified as the same blood flow state, for example, all phases classified as the first blood flow state period are referred to as phases in the same period.

[0089] It should be noted that the specific way of obtaining the fitted distance value by polynomial fitting is a technology known to those skilled in the art, and will not be described in detail in the present embodiment, for example, a polynomial function is used to fit the leaflet-sinus wall distances of all phases in the same period. In the fitting process, a suitable order (such as 2) is selected for the polynomial function, and then the coefficients are solved by the least squares method to obtain the final fitting function, which is the "distance change trend curve". For each phase in the same period, the corresponding fitted distance value is calculated by substituting the fitting function.

[0090] It should be noted that in the polynomial fitting process, different order polynomials are selected according to the data trend, for example, in the systolic period, the polynomial fitting is performed according to the trend of "the leaflet-sinus wall distance gradually decreasing", and a 2-order polynomial is used when the blood flow is stable; in the diastolic period, the polynomial fitting is performed according to the trend of "the distance gradually increasing" (such as the upward trend of the quadratic function), and a 3-order polynomial is used when the vortex flow changes complexly.

[0091] S102-2: Calculate the difference between the leaflet-cavity wall distance of each phase and the fitting distance value of the same phase as the original residual; calculate the standard deviation of all original residuals in the same period.

[0092] It should be noted that the method of calculating the standard deviation is a known technical means, and the present embodiment will not be described again.

[0093] S102-3: Take the absolute value of the ratio of the original residual of each phase in the same period to the standard deviation as the first deviation index of the corresponding phase.

[0094] The first deviation index is used to quantify the overall deviation between the leaflet-cavity wall distance of each phase and the fitting distance value of the corresponding phase in the same blood flow state period.

[0095] It should be noted that the greater the absolute value of the ratio of the original residual of a certain phase to the standard deviation, the greater the deviation degree of the leaflet-cavity wall distance obtained from the phase as an outlier on the distance change trend curve, i.e. the greater the deviation degree.

[0096] It should be noted that even in the same period, in actual situations, the leaflet-cavity wall distance is constantly changing over time. Due to some objective factors such as voxel mixing, the leaflet-cavity wall distance and the corresponding fitting distance value are usually not exactly the same, so the original residual cannot all be zero, and the standard deviation cannot be zero.

[0097] The second deviation index determination process is shown in Figure 4 , which includes:

[0098] S102-4: Calculate the first statistical characteristic value of the laminar dark area based on the CT value distribution of the pixel points in the laminar dark area, and calculate the second statistical characteristic value of the vortex bright area based on the CT value distribution of the pixel points in the vortex bright area.

[0099] For example, when the CT value of the laminar dark area or the vortex bright area meets the following conditions, the arithmetic mean is used to calculate the statistical characteristic value: 1. The CT value distribution in the region presents a unimodal and approximately normal distribution; 2. The pixel CT values in the region are uniformly distributed in space (such as homogeneous blood flow, no artifact soft tissue); 3. The signal-to-noise ratio of the laminar dark area or the vortex bright area .

[0100] It should be noted that the specific method of judging the CT value distribution of the laminar dark area or the vortex bright area and calculating the signal-to-noise ratio is a known technical means for those skilled in the art, and the present embodiment will not be described again.

[0101] ​S102-5: Calculate the difference between the first statistical feature value of the laminar flow dark area and the second statistical feature value of the vortex flow bright area in each phase, and obtain the CT difference value of each phase.

[0102] In this embodiment, the arithmetic mean of the CT values of all pixel points in the laminar flow dark area in each phase is calculated as the first statistical feature value of the laminar flow dark area; and the arithmetic mean of the CT values of all pixel points in the vortex flow bright area in each phase is calculated as the second statistical feature value of the vortex flow bright area.

[0103] S102-6: Perform polynomial fitting on the CT difference values of all phases in the same period to generate a CT difference value trend curve, and obtain the fitted CT difference value of each phase.

[0104] S102-7: Calculate the difference between the CT difference value of each phase and the corresponding fitted CT difference value as the CT original residual of each phase; and calculate the standard deviation of all CT original residuals in the same period.

[0105] S102-8: Take the absolute value of the ratio of the CT original residual of each phase to the standard deviation in the same period as the second deviation index of the corresponding phase.

[0106] The second deviation index is used to quantify the overall deviation between the CT difference value of each phase and the fitted CT difference value of the corresponding phase in the same blood flow state period.

[0107] It should be understood that the greater the absolute value of the ratio of the CT original residual of a certain phase to the standard deviation, the greater the deviation, i.e., the CT difference value obtained from the phase is an outlier on the CT difference value trend curve.

[0108] S103: If the first deviation index exceeds the preset first index threshold and the second deviation index exceeds the preset second index threshold, determine that the corresponding phase is a data correction phase; determine the correction weight according to the second deviation index of the data correction phase; and correct the leaf-sinus wall distance of the data correction phase using the correction weight.

[0109] It should be understood that in order to avoid the possibility of a single index exceeding the standard being an accidental error (such as noise at the imaging moment), but when both indexes exceed the standard, the probability of accidental error is extremely low, and it is more likely to be a systematic problem (such as device parameter error, patient position change during data acquisition, abnormal valve movement causing measurement reference disorder, etc.), therefore, the first deviation index and the second deviation index are combined to determine whether the data extracted in the phase needs to be corrected.

[0110] It should be noted that the specific values of the preset first index threshold and the preset second index threshold are determined according to actual conditions, and the present embodiment is not limited to specific values. For example, when the preset first index threshold is 3, it can be found that the measured value of the leaflet-sinus wall distance deviates greatly from the normal trend in the same period. When the preset second index threshold is 2, it can be found that the CT difference value deviates greatly from the normal change trend of the CT difference value in the same period.

[0111] In the present embodiment, the exponential function with the natural constant as the base and the inverse of the second deviation index of each phase as the index is calculated to obtain the correction weight of each phase.

[0112] It should be noted that when the leaflet-sinus wall distance of the data to be corrected is corrected by using the correction weight, only the phases that meet the first and second deviation index thresholds in the same blood flow state period are corrected, and the correction across periods (such as the systolic and diastolic periods) is not performed.

[0113] Since the greater the second deviation index of a phase, the greater the overall deviation between the leaflet-sinus wall distance of the phase and the fitting distance value of the phase, at this time, the leaflet-sinus wall distance of the phase should be corrected to be close to the fitting distance value, and in the correction process, the weight of the leaflet-sinus wall distance determined by the phase is smaller, that is, the correction weight is smaller, and therefore the weight of the fitting distance value is greater, therefore, the correction weight of each phase can be represented by the following formula:

[0114]

[0115] wherein, the correction weight is represented by W; the exponential function with the natural constant as the base is represented by exp; the second deviation index is represented by D.

[0116] In the present embodiment, the fitting distance value corresponding to the data to be corrected is obtained, the product of the correction weight and the leaflet-sinus wall distance of the data to be corrected is calculated as a first product, the difference between the value 1 and the correction weight is calculated as a correction weight, the product of the correction weight and the fitting distance value is calculated as a second product, and the sum of the first product and the second product is calculated as the corrected leaflet-sinus wall distance.

[0117] S104: Extract the morphological and functional dynamic parameters of the aortic valve of each phase on the longitudinal cross-sectional image of the aortic sinus region, correlate the leaflet-sinus wall distance of each phase to construct a three-dimensional dynamic model of the valve, and output the trajectory parameters representing the valve motion from the model, wherein the distance of the data to be corrected is the corrected value.

[0118] The longitudinal sectional image of the aortic sinus region, i.e. the sectional imaging along the long axis direction of the aorta (i.e. the main flow direction of blood from the left ventricle into the aorta), can directly show the up-down diameter of the aortic sinus, the longitudinal movement trajectory of the leaflet and the longitudinal adjacent relationship with the surrounding structure.

[0119] The morphological and functional dynamic parameters of the aortic valve: the morphological parameters of the aortic valve mainly describe the anatomical structure characteristics of the aortic valve; the functional dynamic parameters of the aortic valve mainly reflect the hemodynamic state of the aortic valve in the cardiac cycle; for example, the morphological and functional dynamic parameters of the aortic valve at least include the coordinates of the leaflet attachment point, the position of the leaflet apex and the opening angle of the leaflet.

[0120] The coordinates of the leaflet attachment point represent the position of the fixed point where the leaflet is connected to the wall of the aortic sinus in the coordinate system, for example, in the three-dimensional coordinates with the center of the aortic root as the origin, the coordinates of the fixed point where the leaflet is connected to the wall of the aortic sinus.

[0121] The position of the leaflet apex refers to the coordinates of the highest point (in the systolic phase) or the closing point (in the diastolic phase) of the free edge of the leaflet, for example, in the three-dimensional coordinates with the center of the aortic root as the origin, the coordinates of the highest point of the free edge of the leaflet in the systolic phase.

[0122] The opening angle of the leaflet, with the leaflet attachment point as the vertex, the line segment connecting the attachment point and the leaflet apex, forms an angle with the long axis of the aorta, wherein in the systolic phase, the opening angle of the leaflet increases (such as from 30° to 90°, reflecting the opening degree of the leaflet); in the diastolic phase, the opening angle of the leaflet decreases (such as from 90° to 10°, reflecting the closing degree of the leaflet).

[0123] It should be noted that the trajectory parameters of the valve movement at least include the maximum opening angle, the distance change rate and the attachment point-apex distance change rate.

[0124] The maximum opening angle is the maximum opening angle of the leaflet in the systolic phase; the distance change rate is the change rate of the corrected leaflet-sinus wall distance with time, which is used to reflect the speed of the opening of the leaflet; the attachment point-apex distance change rate is the fluctuation range of the distance from the leaflet attachment point to the apex in the same period, which is used to reflect the stretching and contraction of the leaflet itself.

[0125] It should be noted that the specific way of constructing the three-dimensional dynamic model of the valve is a well-known technical means to those skilled in the art, and the present embodiment will not be described again. For example, first, the leaflet attachment point coordinates, vertex positions, opening and closing angles extracted from the longitudinal sectional images are matched in space and time with the leaflet-canal wall distance after the correction of the transverse sectional images. The parameters of the same phase correspond to the same cardiac cycle time, forming a "static snapshot" of the valve at that moment. Then, all the "static snapshots" of all phases in the cardiac cycle are connected in time sequence, and the transition state between adjacent phases is supplemented by an interpolation algorithm to form a continuous three-dimensional dynamic model. The model can intuitively show the complete movement process of the leaflet from opening to closing and then to opening.

[0126] It should be noted that the specific way of obtaining the maximum opening and closing angle, the rate of change of the distance, and the rate of change of the distance between the attachment point and the vertex is an existing technical means, and the present embodiment will not be described again.

[0127] 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.

[0128] Each of the embodiments in the specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. An automatic aortic valve segmentation method based on multi-temporal CTA images, characterized in that, The method includes: Acquire multi-phase CTA images of the aortic root covering the entire cardiac cycle; locate the aortic sinus region in the CTA images of each phase; identify laminar flow dark areas and eddy flow bright areas on the cross-sectional images of the aortic sinus region; Based on the first characteristic size of the laminar flow dark area and the second characteristic size of the vortex bright area in each phase, the leaflet-sinus wall distance of each phase is determined; based on the changing trend of the leaflet-sinus wall distance of each phase in the cardiac cycle, the first deviation index of each phase is determined; based on the difference characteristics between the CT values ​​of the laminar flow dark area and the CT values ​​of the vortex bright area in each phase, the second deviation index of each phase is determined. If the first deviation index exceeds the preset first index threshold and the second deviation index exceeds the preset second index threshold, the corresponding time phase is determined to be the data phase to be corrected; the correction weight is determined according to the second deviation index of the data phase to be corrected; the leaflet-sinus wall distance of the data phase to be corrected is corrected using the correction weight. Morphological and functional dynamic parameters of the aortic valve at each time phase are extracted from longitudinal cross-sectional images of the aortic sinus region. A three-dimensional dynamic model of the valve is constructed by associating the leaflet-sinus wall distance at each time phase. The model outputs trajectory parameters characterizing the valve motion, wherein the distance of the time phase to be corrected is the corrected value.

2. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 1, characterized in that, The process of determining the first characteristic dimension of the laminar dark region and the second characteristic dimension of the eddy bright region includes: On the cross-sectional image of the aortic sinus region, the inner edge boundary contour of the laminar flow dark area and the outer edge boundary contour of the vortex bright area are extracted; A polar coordinate system is established with the center point of the aortic root as the origin; multiple preset sampling angles are set along the circumference; the radial distance from the origin to the inner edge boundary contour of the laminar dark area is calculated in each sampling angle direction, which is used as the first feature distance of the corresponding angle; the radial distance from the origin to the outer edge boundary contour of the vortex bright area is calculated, which is used as the second feature distance of the corresponding angle. Calculate the arithmetic mean of the first characteristic distances corresponding to all sampling angles in each time phase, and use it as the first characteristic size of the laminar dark region in each time phase; Calculate the arithmetic mean of the second feature distances corresponding to all sampling angles in each time phase, and use it as the second feature size of the eddy bright region in each time phase.

3. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 2, characterized in that, After extracting the inner edge boundary contour of the laminar dark region and the outer edge boundary contour of the eddy bright region, the contours are smoothed, and the smoothing process includes removing isolated noise points on the contours.

4. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 1, characterized in that, The identification of laminar flow dark areas and eddy flow bright areas on cross-sectional images of the aortic sinus region includes: Perform grayscale analysis on the cross-sectional image; Based on a preset CT value threshold range, high CT value regions and low CT value regions in the image are identified; wherein, the CT value of the high CT value region is within a first preset range, and the CT value of the low CT value region is within a second preset range, and the lower limit of the first preset range is higher than the upper limit of the second preset range. The identified low CT value regions are designated as laminar flow dark regions; the identified high CT value regions are designated as eddy flow bright regions.

5. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 1, characterized in that, The method further includes: For multiple phases within the cardiac cycle, calculate the first area ratio of the laminar flow dark region and the second area ratio of the eddy flow bright region; where the first area ratio is the ratio of the laminar flow dark region area to the total cross-sectional area of ​​the aortic sinus region, and the second area ratio is the ratio of the eddy flow bright region area to the total cross-sectional area of ​​the aortic sinus region. Based on the temporal variation characteristics of the leaflet-sinus wall distance, the first area proportion, and the second area proportion during the cardiac cycle, each phase is classified into the first blood flow state period or the second blood flow state period.

6. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 5, characterized in that, The process for determining the first deviation index includes: Polynomial fitting was performed on the leaflet-sinus wall distance of all phases within the same period to generate a distance change trend curve, and the fitted distance value of each phase within the same period was obtained. The same period represents the first blood flow state period or the second blood flow state period obtained by classification. Calculate the difference between the leaflet-sinus wall distance at each time phase and the fitted distance value at the same time phase, and use it as the original residual; calculate the standard deviation of all original residuals within the same period. The absolute value of the ratio of the original residual to the standard deviation for each time phase within the same period is used as the first deviation index for the corresponding time phase.

7. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 5, characterized in that, The second deviation index determination process includes: Based on the CT value distribution of pixels in the laminar flow dark area, the first statistical characteristic value of the laminar flow dark area is calculated, and based on the CT value distribution of pixels in the eddy flow bright area, the second statistical characteristic value of the eddy flow bright area is calculated. The difference between the first statistical characteristic value of the laminar dark region and the second statistical characteristic value of the eddy bright region in each time phase is calculated to obtain the CT difference value of each time phase. Polynomial fitting was performed on the CT difference values ​​of all time phases within the same period to generate the CT difference value change trend curve, and the fitted CT difference value of each time phase was obtained. Here, the same period represents the first blood flow state period or the second blood flow state period obtained by classification. Calculate the difference between the CT difference value and the corresponding fitted CT difference value at each time phase, and use it as the original CT residual for each time phase; calculate the standard deviation of all original CT residuals within the same period. The absolute value of the ratio of the raw CT residuals to the standard deviation of each time phase within the same period is used as the second deviation index for the corresponding time phase.

8. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 7, characterized in that, The calculation process of the first statistical characteristic value and the second statistical characteristic value includes: Calculate the arithmetic mean of the CT values ​​of all pixels in the laminar flow dark area at each time phase, and use it as the first statistical feature value of the laminar flow dark area; The arithmetic mean of the CT values ​​of all pixels within the eddy current bright region at each time phase is calculated and used as the second statistical characteristic value of the eddy current bright region.

9. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 6, characterized in that, The process of determining the correction weights includes: Using the natural constant as the base and the opposite of the second deviation index of each time phase as the exponent, an exponential calculation is performed to obtain the correction weight of each time phase.

10. The method for automatic aortic valve segmentation based on multi-temporal CTA images according to claim 9, characterized in that, The step of correcting the leaflet-sinus wall distance of the phase in which the data is to be corrected using correction weights includes: Obtain the corresponding fitting interval value when the data is to be corrected; The product of the correction weight and the leaflet-sinus wall distance at the time phase to be corrected is calculated as the first product; the difference between the value 1 and the correction weight is calculated as the correction weight; and the product of the correction weight and the fitted distance value is calculated as the second product. Calculate the sum of the first and second products to obtain the corrected leaflet-sinus wall distance.

Images