Method and device for identifying diastolic period and systolic period of heart in CBCT image

Abstract

The invention provides a method and a device for identifying a diastolic period and a systolic period of a CBCT image, and the method comprises the steps: scanning a heart region through CBCT equipment, and collecting a plurality of two-dimensional projection images of the heart region; preprocessing the plurality of two-dimensional projection images so as to suppress scanning errors of CBCT equipment and enhance the distinction degree of heart motion signals and non-motion signals in the CBCT images; based on a PCA algorithm, extracting a feature value of each preprocessed two-dimensional projection image, the feature values being used for representing a heart motion state; fitting the characteristic values of all the two-dimensional projection images into a characteristic curve according to an acquisition time sequence; and identifying a peak value and a valley value of the characteristic curve, and determining that the two-dimensional projection image corresponding to the peak value is a diastolic period projection image and the two-dimensional projection image corresponding to the valley value is a systolic period projection image. According to the application, accurate and efficient identification of the key cardiac stage of the heart can be realized, and stable and reliable image support is provided for subsequent clinical operation.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to a method and apparatus for identifying the diastolic and systolic phases of the heart in CBCT images. Background Technology

[0002] In cardiac-related clinical procedures (such as cardiac angiography), it is necessary to determine the diastolic and systolic phases of the heart based on images acquired by C-arm CBCT (Cone Beam Computed Tomography) equipment to complete key steps such as cardiac contouring, providing a basis for surgical navigation and treatment planning. In the traditional process, doctors need to manually observe the two-dimensional projection image sequence generated by the CBCT equipment and judge the image frames corresponding to the diastolic and systolic phases based on experience. This process relies too heavily on the doctor's subjective experience, and the judgment results of different doctors or the same doctor at different times may differ, leading to inaccurate positioning of key nodes, which in turn affects the accuracy of subsequent contouring and the reliability of surgical plans. Moreover, this process requires doctors to review images frame by frame (usually covering multiple cardiac cycles), which is inefficient and seriously affects the progress of clinical procedures. Summary of the Invention

[0003] To overcome the problems existing in related technologies, this application provides a method and apparatus for identifying the diastolic and systolic phases of the heart in CBCT images.

[0004] According to a first aspect of the embodiments of this application, a method for identifying the diastolic and systolic phases of the heart in CBCT images is provided, the method comprising: The heart region is scanned using a CBCT scanner, and several two-dimensional projection images of the heart region are acquired. The aforementioned two-dimensional projection images are preprocessed to suppress scanning errors of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in the CBCT images. Based on the PCA algorithm, feature values ​​are extracted from each preprocessed two-dimensional projection image, and these feature values ​​are used to characterize the cardiac motion state. The feature values ​​of all two-dimensional projection images are fitted into feature curves according to the acquisition time sequence; Identify the peak and trough values ​​of the characteristic curve, determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the trough value is the systolic projection image of the heart.

[0005] According to a second aspect of the embodiments of this application, an apparatus for identifying the diastolic and systolic phases of the heart in CBCT images is provided, the apparatus comprising: The projection image acquisition module is used to scan the heart region using a CBCT device and acquire several two-dimensional projection images of the heart region. The preprocessing module is used to preprocess the plurality of two-dimensional projection images to suppress the scanning error of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in the CBCT images. The feature extraction module is used to extract feature values ​​from each preprocessed two-dimensional projection image based on the PCA algorithm. These feature values ​​are used to characterize the cardiac motion state. The feature curve fitting module is used to fit the feature values ​​of all two-dimensional projection images into feature curves according to the acquisition time sequence. The target projection image recognition module is used to identify the peak and trough values ​​of the feature curve, determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the trough value is the systolic projection image of the heart.

[0006] According to a third aspect of the embodiments of this application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect.

[0007] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0008] The technical solutions provided in this application embodiment may include the following beneficial effects: In this embodiment, an automated process replaces the traditional manual judgment mode. First, the two-dimensional projection image sequence acquired by the CBCT equipment is preprocessed in a targeted manner to optimize image quality and enhance cardiac motion signal characteristics. Then, based on the principal component analysis (PCA) algorithm, feature values ​​of each image are extracted and fitted into a time series curve. Finally, by identifying the peak and trough values ​​of the curve, the image frames corresponding to the diastolic and systolic phases of the heart are located, realizing accurate and efficient identification of key cardiac phases and providing stable and reliable image support for subsequent clinical operations.

[0009] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form part of this application, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0011] Figure 1 This is a flowchart illustrating a method for identifying the diastolic and systolic phases of the heart in a CBCT image according to an exemplary embodiment of this application.

[0012] Figure 2 This is a schematic diagram illustrating the structure of a hierarchical statistical model constructed based on the PCA algorithm according to an exemplary embodiment of this application.

[0013] Figure 3 This application illustrates an identified diastolic projection image of the heart according to an exemplary embodiment.

[0014] Figure 4 This application illustrates an identified projection image of the heart during systole according to an exemplary embodiment.

[0015] Figure 5 This is a schematic diagram of a device for identifying the diastolic and systolic phases of the heart in a CBCT image, according to an exemplary embodiment of this application.

[0016] Figure 6 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment of this application. Detailed Implementation

[0017] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0018] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0019] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0020] In cardiac-related clinical procedures (such as cardiac angiography), it is necessary to determine the diastolic and systolic phases of the heart based on images acquired by C-arm CBCT equipment to complete key steps such as cardiac contouring, providing a basis for surgical navigation and treatment planning. In the traditional process, doctors need to manually observe the two-dimensional projection image sequence generated by CBCT and judge the image frames corresponding to the diastolic and systolic phases based on experience. This process relies too heavily on the doctor's subjective experience, and the judgment results of different doctors or the same doctor at different times may differ, leading to inaccurate positioning of key nodes, which in turn affects the accuracy of subsequent contouring and the reliability of surgical plans. Moreover, this process requires doctors to review images frame by frame (usually covering multiple cardiac cycles), which is inefficient and seriously affects the progress of clinical procedures.

[0021] Based on this, and to address the problems existing in related technologies, this application provides a method for identifying the diastolic and systolic phases of the heart in CBCT images. This method replaces the traditional manual judgment mode with an automated process. First, it performs targeted preprocessing on the two-dimensional projection image sequence acquired by the CBCT device to optimize image quality and enhance cardiac motion signal characteristics. Then, it extracts feature values ​​for each image based on the PCA (Principal Component Analysis) algorithm and fits them into a time series curve. Finally, it locates the image frames corresponding to the diastolic and systolic phases of the heart by identifying the peaks and troughs of the curve, achieving accurate and efficient identification of key cardiac phases and providing stable and reliable image support for subsequent clinical operations.

[0022] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart illustrating a method for identifying the diastolic and systolic phases of the heart in a CBCT image according to an exemplary embodiment of this application. Figure 1 As shown, the method includes steps S101 to S105.

[0024] Step S101: Scan the heart region using a CBCT device to acquire several two-dimensional projection images of the heart region.

[0025] Specifically, in this embodiment, a commonly used C-arm CBCT device can be selected to perform dynamic scanning on the patient's heart region, acquiring several two-dimensional projection images of the heart region to generate a continuous two-dimensional projection image sequence.

[0026] Furthermore, to ensure the validity of subsequent cardiac cycle analysis data, the two-dimensional projection images acquired during the scanning process should meet the following two requirements: First, the two-dimensional projection images should cover at least two complete cardiac cycles to obtain sufficient cardiac motion data, avoiding deviations in subsequent motion curve fitting or missed identification of key cycle nodes due to insufficient data. Simultaneously, the reliability of the identification results can be further improved through repetitive comparison of multi-cycle data. Second, the acquisition time interval between adjacent two-dimensional projection images should remain fixed, preferably within the range of 10-50 milliseconds. The specific time interval can be set by comprehensively considering the heart rate (the normal adult resting heart rate is approximately 60-100 beats / minute, corresponding to a cardiac cycle of approximately 0.6-1 second) and the accuracy requirements of subsequent data processing; this embodiment does not impose a specific limitation. A fixed time interval ensures the temporal consistency of the image sequence, providing a stable time reference for time-based cardiac motion analysis and avoiding distortion of the motion curve due to fluctuations in time intervals.

[0027] Step S102: Preprocess several two-dimensional projection images to suppress scanning errors of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in the CBCT images.

[0028] Different projection images used in clinical practice exhibit distinct characteristics in their presentation of motion signals due to differences in their imaging principles. For example, PET-CT (Positron Emission Tomography-Computed Tomography) projection images are based on tracer metabolic imaging, reflecting the metabolic state of tissue function. Their signals are metabolically related count signals, with significant differences between the signals of moving areas such as the heart and static tissues. Furthermore, the noise is photon counting noise, mainly concentrated in low-metabolic areas, while the signals in high-metabolic areas are relatively stable, and the signal boundaries of moving areas are clear. In contrast, the CBCT projection images used in this application are based on X-ray attenuation imaging, and the signals are represented by gray values ​​related to tissue density. The gray value differences of the soft tissues surrounding the heart are minimal, and the noise is mainly X-ray quantum noise, which is uniformly distributed throughout the image. At the same time, the dynamic signals of the heart are easily interfered with by the overlapping of tissue densities within the thoracic cavity, resulting in severe signal fluctuations and making it difficult to directly distinguish between moving and non-moving areas. Therefore, in order to suppress scanning errors of CBCT equipment and enhance the distinction between cardiac motion signals and non-motion signals in CBCT images, targeted preprocessing can be used to optimize the quality of CBCT projection images, laying the foundation for subsequent cardiac motion feature extraction. Specifically, the preprocessing process is as follows: First, air calibration is performed to eliminate system errors caused by equipment hardware factors. CBCT projection images are acquired collaboratively by multiple detector channels. Due to individual differences in detector parameters, afterglow effect, and fluctuations in X-ray tube exposure energy, the output values ​​of different channels may exhibit zero-point drift (some channels output zero, positive or negative deviations, etc.), leading to distortion of tissue attenuation information in the image. This embodiment uses pre-scanning of the air with the CBCT equipment to obtain air projection images as zero-point drift values ​​for each detector channel. These values ​​are then used as reference values ​​to correct the response errors of each detector in the two-dimensional projection image caused by zero-point drift. This correction follows the following X-ray attenuation formula: in, The grayscale values ​​of the air projection image. The tissue attenuation coefficient, For tissue thickness, To correct the grayscale values ​​of the original two-dimensional projected image, the following derivation is obtained: That is, after Corrected projection data.

[0029] To improve correction accuracy in dynamic scenes and obtain higher image quality, an optimized correction formula is further introduced: in This is the proportionality coefficient. For the latest static two-dimensional data acquisition at 0 degrees data, Two-dimensional data obtained during periodic historical correction The data is dynamically corrected through iterative benchmarking to ensure the accuracy of attenuation information for different scanning periods.

[0030] Next, noise suppression and smoothing are performed on the corrected two-dimensional projection image. In this embodiment, classic filter kernels such as Shepp-Logan and Hanning are used as a basis, and a smoothing filter and a Gaussian filter are superimposed to enhance the filtering of the two-dimensional projection image. The frequency domain formula for the smoothing filter can be: The formula for Gaussian filtering used for image noise removal can be: Where N is half the length of the truncation correction edge, and the second half is symmetrical to the first half.

[0031] Subsequently, the heart region is extracted as the region of interest (ROI) from the smoothed 2D projection image. In this embodiment, the heart ROI can be extracted from the 2D projection image using image segmentation algorithms (such as segmentation based on a preset grayscale threshold, basic edge detection algorithms, etc.), excluding areas that are not primarily affected by cardiac motion, such as lung tissue, chest wall, and blood vessels, thus reducing the interference of irrelevant data on subsequent PCA analysis. During extraction, the ROI must accurately encompass the entire ventricular region to ensure complete capture of ventricular motion changes during systole and diastole.

[0032] Finally, data normalization is performed, mapping the gray values ​​of the extracted pixels in the heart region to a preset value range (such as the [0,1] interval) through linear transformation and other methods. This eliminates the gray value differences between different images caused by slight fluctuations in scanning imaging parameters (such as slight changes in tube voltage and tube current), ensuring that all image data are consistent and meeting the input standardization requirements for subsequent feature extraction.

[0033] Step S103: Based on the PCA algorithm, extract the feature values ​​of each preprocessed two-dimensional projection image. These feature values ​​are used to characterize the cardiac motion state.

[0034] Specifically, considering the small grayscale difference between cardiac motion signals and surrounding tissue signals in CBCT projection images, and their susceptibility to noise and tissue overlap, simple linear feature extraction is insufficient to fully capture the dynamic characteristics of cardiac motion. To more accurately mine cardiac motion-related features, this embodiment can construct a feature extraction method based on the PCA algorithm, such as... Figure 2 The hierarchical statistical model shown includes L feature layers (L≥2). Each two-dimensional projection image, after preprocessing in step S102, is input. The first feature layer of the hierarchical statistical model extracts the linear features of the projection image. Subsequent feature layers of the hierarchical statistical model then use the output of the previous feature layer as input to progressively mine the nonlinear features of the projection image. The linear and nonlinear features are then weighted and fused according to preset weighting parameters, ultimately generating feature values ​​representing the cardiac motion state in each two-dimensional projection image. The specific process of extracting image feature values ​​through the hierarchical statistical model is as follows: First, the model input data is preprocessed. The two-dimensional projection image sequence preprocessed in step S102 is defined as the input data matrix. Where S represents the number of frames in the dynamic acquisition time (i.e., the total number of images acquired in step S101), and N represents the dimension of the probe data (i.e., the number of pixels in a single image). During the total acquisition time, the projection data can be divided into S frames according to a fixed unit duration. Data framing needs to balance the impact of statistical noise and the preservation of temporal information. To reduce noise interference and alleviate computational burden, the data matrix X can first be downsampled in the spatial domain (reducing the data volume while preserving key grayscale information); then, it can be centered using the formula... Eliminate data mean bias to ensure that data in each dimension meets the input requirements of the PCA algorithm.

[0035] Next, hierarchical feature extraction is initiated. For the first feature layer, the traditional PCA algorithm can be applied to extract linear features from the projected image. First, the feature space is searched. The projected dataset X is linearly projected onto this feature space, where In the feature space Projection weighting factor The basis functions or principal components of the first feature layer are represented by the superscript (1), which indicates the first layer of feature extraction. The PCA algorithm is based on the premise of minimal information loss. It transforms correlated variables into linear uncorrelated variables through orthogonal transformation. The basis vectors of the transformed feature space are called principal components. The eigenvectors corresponding to the minimum eigenvalues ​​are often related to noise. Discarding them can achieve the effect of noise reduction and signal extraction to a certain extent. Therefore, in order to solve the feature space, the optimization objective of this layer is "the covariance between the feature variables after space transformation is 0 and the internal variance is as large as possible", that is, its optimization formula is as follows: in Let be an S-dimensional identity matrix, and T denote the matrix transpose operation. Using the Lagrange multiplier method, construct the Lagrange function and... After differentiation, the above optimization problem is transformed into an eigenvalue problem: In the formula Let X be the diagonal matrix of eigenvalues ​​of the covariance matrix. For a single eigenvalue, the basis vector matrix of the eigenspace is obtained by solving. The corresponding eigenvalues ​​satisfy This leads to the first layer feature matrix. .

[0036] For the subsequent Layer feature layer (2≤ l ≤L), apply KPCA (Kernel Principal Component Analysis) algorithm to extract the first digit of the projected image. Nonlinear characteristics of layers Due to the input feature matrix In higher dimensions Linearly inseparable, can be addressed through nonlinear mapping functions. Implicit linear mapping is implemented, and the mapping relationship is as follows: It is the first The basis functions or principal components of the feature layer It is the characteristic matrix. In this case, for Applying the linear PCA algorithm, the following optimization task is obtained: From this optimization objective, we can see that the nonlinear mapping function Without explicit definition, the mapping relationship can be indirectly represented through the kernel matrix. Define the kernel matrix. , where the kernel matrix elements .

[0037] After substituting the optimization objective, the formula can be rewritten as: Applying the Lagrange multiplier method again, construct the Lagrange function and then... After differentiation, it transforms into an eigenvalue problem: The basis vector matrix of the eigenspace is obtained by solving. The corresponding eigenvalues ​​satisfy And thus obtain the first Layer feature matrix From the above eigenvalue formula, it can be seen that... The equations are linear in the kernel space, but the original data... It is non-linear.

[0038] The Gaussian kernel function is a classic robust radial basis function that effectively resists noise in the data. Its parameters determine the function's effective range; beyond this range, the data's influence is essentially negligible. Therefore, in this embodiment, the Gaussian kernel function is preferred. Its mathematical form is: in The parameters are those of the Gaussian kernel function. It should be noted that this application does not impose any restrictions on the selection of the kernel function; it can be adjusted according to actual needs. Other commonly used kernel functions, such as the polynomial kernel function, the Laplace kernel function, the Sigmoid kernel function, and the Cauchy kernel function, should also be within the scope of protection of this application.

[0039] After extracting features at each layer, weighted fusion of the multi-layer features can be performed. For a given data X, the first layer of linear features... Extractable via PCA, the first Layer (2≤ l ≤L) Nonlinear characteristics It can be extracted via KPCA. Signals are acquired externally. As a priori, the first l Layer weighting parameters The features of each layer can be determined using the least squares method, and the weighting parameters are determined by a prior strategy, as shown below: Taking the derivative of the above formula and setting it to zero, for the d-th (1≤d≤S) principal component of the k-th (1≤k≤L) feature layer, its weighting parameters... The solution is: Solving the above system of equations simultaneously, we get... The optimal solution can be found by iterating through all prior data. The optimal solution distribution can be obtained, and its average value can be used as the final weight optimization parameter. The features of each layer are weighted and summed according to the final weight to obtain the feature value corresponding to each two-dimensional projection image. This feature value can directly characterize the cardiac motion state.

[0040] Step S104: Fit the feature values ​​of all two-dimensional projection images into feature curves according to the acquisition time sequence.

[0041] Considering that the feature values ​​extracted in step S103 are discrete data points, it is difficult to intuitively analyze the continuous periodic changes of cardiac motion based directly on discrete points, and it is easily affected by random noise, leading to misjudgment of extreme points. In order to obtain a smooth and continuous feature change curve that closely matches the real cardiac motion pattern, and at the same time accurately preserve the trend of feature value changes over time, this embodiment can use the weighted least squares method combined with cubic B-spline curves to complete the curve fitting of feature values. The specific process is as follows: First, data preparation is performed: a discrete data point set is constructed by using the acquisition time of each two-dimensional projection image as the horizontal axis and the corresponding feature values ​​extracted in step S103 as the vertical axis. The acquisition time can be calculated based on the fixed time interval set in step S101, assuming the initial acquisition time is... (The acquisition time of the first image is taken as 0ms), and the time interval between adjacent images is... The acquisition time for the i-th image (i=0,1,2,...,S-1, where S is the total number of frames) is... The corresponding eigenvalue is Ultimately, this forms a discrete data point set. .

[0042] Next, the B-spline curve is constructed: The first step is to define the node vector of the cubic B-spline curve, which can be achieved using a uniform node partitioning method. The number of nodes can be set to the number of discrete data points + 4 (i.e., S + 4 nodes). This setting satisfies the continuity condition of the second derivative of the cubic B-spline curve, avoiding abrupt changes at the boundaries and ensuring the overall smoothness of the fitted curve. The second step is to solve for the control vertices of the curve using the weighted least squares method: For each point in the discrete data set, a weight is assigned based on the standard deviation of its eigenvalues ​​(the smaller the standard deviation, the higher the data reliability, and the greater the weight). An optimization function is constructed with the objective of minimizing the weighted error between the fitted curve and the discrete points, and the control vertex sequence of the cubic B-spline curve is obtained. The third step is to verify the fitting accuracy by calculating the fitted curve f(t) at each discrete data point. Fitting error at (i.e.) It is necessary to ensure that the error does not exceed a preset threshold, such as 15% of the standard deviation of the eigenvalues. If the error exceeds the threshold, the node vector density or weight allocation rules should be readjusted until the accuracy requirements are met. The final continuous eigenvalue fitting curve f(t) can accurately reflect the trend of eigenvalue changes over time, providing a clear and reliable curve basis for subsequent identification of extreme points corresponding to the diastolic and systolic phases of the heart.

[0043] Step S105: Identify the peak and trough values ​​of the characteristic curve, and determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the trough value is the systolic projection image of the heart.

[0044] After obtaining the feature curve, to determine the diastolic and systolic projection images of the heart using the feature curve, it is necessary to first clarify the relationship between the feature curve and cardiac physiological motion: the feature value extracted in step S103 is essentially the core motion feature extracted from the grayscale data of the cardiac region using the PCA algorithm. The magnitude of this feature value is directly related to the variation amplitude of the grayscale distribution of tissues within the cardiac region. Based on the laws of cardiac physiological motion, ventricular motion is a periodic cycle of "diastole-contraction": when the ventricle enters diastole, the myocardium gradually relaxes, and the ventricular diameter increases accordingly. At this time, the spatial distribution of myocardium, blood, and other tissues within the cardiac region changes significantly, leading to a significant increase in the difference and fluctuation amplitude of grayscale values ​​within the region, thus causing the feature value to show an upward trend; when the ventricle enters systole, the myocardium actively contracts, the ventricular diameter decreases, the spatial distribution change of tissues within the cardiac region tends to be gentler, the difference and fluctuation amplitude of grayscale values ​​weaken, and the feature value subsequently shows a downward trend. It can be seen that the trend of the characteristic curve is positively correlated with the change of ventricular volume. The increase of the characteristic value corresponds to ventricular diastole and volume increase, and the decrease of the characteristic value corresponds to ventricular systole and volume decrease. Therefore, the two-dimensional projection image corresponding to the peak value of the characteristic curve is the diastolic state image when the ventricular volume is at its maximum, that is, the diastolic projection image of the heart; the two-dimensional projection image corresponding to the trough value is the systolic state image when the ventricular volume is at its minimum, that is, the systolic projection image of the heart.

[0045] Based on this, in order to accurately identify the peaks and troughs of the feature curve, an adaptive threshold combined with a sliding window extreme value detection algorithm can be used to identify the peaks and troughs of the feature curve. The specific identification process is as follows: First, curve smoothing preprocessing is performed: Although the feature curve f(t) obtained by fitting in step S104 has eliminated some noise, it may still have slight fluctuations caused by accidental fluctuations in data acquisition. If it is directly used for extreme value detection, it is easy to produce misjudgments. Therefore, a Gaussian filter (Gaussian kernel standard deviation) can be applied to f(t) in advance. An optimal range of 0.5-1.0 can be selected (which effectively reduces noise without blurring the curve trend) to further eliminate residual noise and obtain a smooth curve. This ensures that the curve trend closely matches the actual heartbeat pattern.

[0046] Next, adaptive threshold calculation is performed: to avoid the difficulty of adapting a fixed threshold to curve fluctuations in different subjects (such as differences in heart rate or myocardial exercise intensity), a smoothed characteristic curve can be used. The overall characteristics are dynamically set with thresholds. First, the characteristic curve is calculated. Maximum value over the entire acquisition time range Minimum value The amplitude is obtained by the difference between the two. Furthermore, considering the characteristics of cardiac motion where "the peak value of diastolic characteristic is significantly higher than the baseline, and the trough value of systolic characteristic is significantly lower than the baseline," a peak detection threshold is set. (Ensure that only true peak values ​​significantly above the curve baseline are identified, excluding minor fluctuations), Valley detection threshold (Filter out the true valley values ​​that are significantly lower than the curve baseline). It is worth noting that the embodiments of this application do not impose specific restrictions on the peak detection threshold and the valley detection threshold, which can be dynamically adjusted according to the actual fluctuation of the curve to ensure that the true extreme values ​​can be effectively distinguished from noise interference.

[0047] Finally, extreme point determination is performed: a sliding window with a preset time unit length (the window size can be selected as 5-9 time units to cover the key phases of a single ventricular motion) can be used to move point by point along the time axis to traverse the smoothed characteristic curve. Calculate the local maxima and local minima of the feature curve within each sliding window. For a local maxima, if a local maxima exceeds the peak detection threshold... If the curve values ​​(i.e., the function values ​​of the characteristic curve within the corresponding window) in the adjacent windows before and after this point are all less than the local maximum value, then this local maximum value is determined as the peak point (denoted as ). (k=1,2,...); For local minima, if a local minimum is less than the valley detection threshold... If the curve values ​​in the adjacent windows before and after a point are all greater than the local minimum, then the local minimum is determined to be a valley point (denoted as ). (k=1,2,...) After identifying extreme points, the peak points can be used as a basis for analysis. Valley point Using the time coordinates corresponding to the feature curve, inversely match the original two-dimensional projection image: peak points. corresponding time The acquisition time in step S101 Matching yields the corresponding image frame number i, which is the diastolic projection image of the heart. In this frame, the ventricular diameter is at its maximum, and the myocardium is fully relaxed. Figure 3 (as shown); valley point corresponding time Similarly, by matching, image frame number j is obtained. This frame is the projected image of the heart during systole. The ventricular diameter is smallest in this frame, and the myocardium is fully contracted. Figure 4 As shown in the figure. Through this process, key frames of diastole and systole in each cycle can be automatically located in an image sequence covering multiple cardiac cycles, providing precise image support for subsequent clinical operations such as cardiac contouring and surgical navigation.

[0048] In addition, to ensure the accuracy of the identification results, after determining the diastolic and systolic projection images of the heart, the reliability of the results can be further verified through physiological characteristics. The specific process is as follows: First, image feature parameters are extracted from the identified diastolic and systolic projection images of the heart. Quantitative indicators directly related to cardiac motion, such as ventricular diameter and ventricular area, are selected: the ventricular diameter can be obtained by locating the ventricular wall boundary using an edge detection algorithm (such as the Canny operator) based on the heart region extracted in step S102, and then calculating it using the Euclidean distance between the two points; the ventricular area can be obtained by counting pixels in the region enclosed by the ventricular boundary.

[0049] The extracted image feature parameters were then compared with the physiological characteristics of the heart during diastole and systole. For example, according to clinical physiological data, there are significant differences in the characteristics of the normal adult heart during diastole and systole: the short-axis diameter of the left ventricle during diastole is typically 45-55 mm, and the cavity area of ​​the left ventricle is 80-12 mm. The short-axis diameter of the left ventricle during systole is 25-35 mm, and the cavity area of ​​the left ventricle is 20-40 mm. Furthermore, the diastolic ventricular diameter and ventricular area of ​​the same subject must be greater than those of the systolic ventricular diameter (the difference should be no less than 15 mm and 40 mm, respectively). If the extracted parameters meet the above range and size relationship (e.g., diastolic left ventricular short axis diameter 50mm, systolic 30mm, difference 20mm; diastolic area 100), then... 30 minutes of systole The difference is 70. If the image feature parameters match the physiological characteristics, then the currently identified diastolic projection image and systolic projection image of the heart are determined as the target projection images.

[0050] If the image feature parameters do not conform to physiological characteristics (e.g., diastolic ventricular diameter is smaller than systolic diameter, or the difference is less than 10 mm), the recognition process needs to be readjusted. For example, the peak detection threshold can be adjusted first. Valley detection threshold (such as) Downgraded to , Upgraded to If the extreme point determination is still not satisfied after adjusting the threshold, the process is backtracked to the preprocessing step, such as enhancing the noise suppression strength (e.g., increasing the Gaussian filter kernel). The process involves optimizing the extraction range of the cardiac region or adjusting the number of feature layers in the hierarchical statistical model (e.g., increasing the number of feature layers to improve feature extraction accuracy) until the image feature parameters such as ventricular diameter and ventricular area of ​​the re-identified projection image conform to the physiological characteristics of diastole and systole, ultimately determining the target projection image. This verification and feedback mechanism effectively avoids misjudgments caused by noise interference and individual differences, ensuring the clinical reliability of the results.

[0051] Corresponding to the aforementioned embodiments of the method for identifying the diastolic and systolic phases of the heart in CBCT images, this application also provides an apparatus for identifying the diastolic and systolic phases of the heart in CBCT images. Figure 5 This is a schematic diagram illustrating the structure of a device for identifying the diastolic and systolic phases of the heart in CBCT images, according to an exemplary embodiment of this application. Figure 5 As shown, the device includes: The projection image acquisition module 501 is used to scan the heart region using a CBCT device and acquire several two-dimensional projection images of the heart region. The preprocessing module 502 is used to preprocess several two-dimensional projection images to suppress scanning errors of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in CBCT images. The feature extraction module 503 is used to extract the feature values ​​of each preprocessed two-dimensional projection image based on the PCA algorithm. These feature values ​​are used to characterize the cardiac motion state. The feature curve fitting module 504 is used to fit the feature values ​​of all two-dimensional projection images into feature curves according to the acquisition time sequence. The target projection image recognition module 505 is used to identify the peak and valley values ​​of the feature curve, and determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the valley value is the systolic projection image of the heart.

[0052] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0053] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0054] Corresponding to the embodiments of the foregoing methods, this application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein, when the processor executes the computer program, it implements the steps of the method for identifying the diastolic and systolic phases of the heart in CBCT images as described in any of the above embodiments.

[0055] For example, processors include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), or field-programmable gate arrays (FPGAs).

[0056] For example, the memory may include at least one type of storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc.

[0057] Figure 6 This is a structural block diagram of a computer device according to an exemplary embodiment of this application. Figure 6 As shown, at the hardware level, the computer device includes a processor 601, an internal bus 602, a network interface 603, memory 604, and non-volatile memory 605, and may also include other hardware required for business operations. One or more embodiments of this application can be implemented in software, for example, the processor 601 reads the corresponding computer program from the non-volatile memory 605 into the memory 604 and then runs it. Of course, in addition to software implementation, one or more embodiments of this application do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the above processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0058] Corresponding to the embodiments of the foregoing methods, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for identifying the diastolic and systolic phases of the heart in CBCT images as described in any of the above embodiments.

[0059] Corresponding to the embodiments of the foregoing methods, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method for identifying the diastolic and systolic phases of the heart in CBCT images as described in any of the above embodiments.

[0060] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0061] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention filed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the foregoing claims.

[0062] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0063] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for identifying the diastolic and systolic phases of the heart in CBCT images, characterized in that, include: The heart region is scanned using a CBCT scanner, and several two-dimensional projection images of the heart region are acquired. The aforementioned two-dimensional projection images are preprocessed to suppress scanning errors of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in the CBCT images. Based on the PCA algorithm, feature values ​​are extracted from each preprocessed two-dimensional projection image, and these feature values ​​are used to characterize the cardiac motion state. The feature values ​​of all two-dimensional projection images are fitted into feature curves according to the acquisition time sequence; Identify the peak and trough values ​​of the characteristic curve, determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the trough value is the systolic projection image of the heart.

2. The method according to claim 1, characterized in that, Preprocessing of the plurality of two-dimensional projection images includes: Based on the air projection images acquired by the CBCT device, the response error of the detector in the plurality of two-dimensional projection images is corrected to obtain the plurality of corrected two-dimensional projection images. A preset filtering algorithm is used to perform noise suppression and smoothing on several corrected two-dimensional projection images; The heart region is extracted from several smoothed two-dimensional projection images using a preset image segmentation algorithm. The gray values ​​of the extracted pixels within the heart region are mapped to a preset range to eliminate gray value differences caused by fluctuations in imaging parameters between different images.

3. The method according to claim 1, characterized in that, Based on the PCA algorithm, feature values ​​are extracted from each preprocessed 2D projection image, including: Based on the PCA algorithm, construct a hierarchical statistical model that includes at least two feature layers; Each preprocessed two-dimensional projection image is used as input data and fed into the first feature layer of the hierarchical statistical model to extract the linear features of each preprocessed two-dimensional projection image. The subsequent feature layers of the hierarchical statistical model take the output of the previous feature layer as input and extract the nonlinear features of each preprocessed two-dimensional projection image layer by layer. The linear and nonlinear features are weighted and fused based on preset weighting parameters to generate feature values ​​for each preprocessed two-dimensional projection image.

4. The method according to claim 1, characterized in that, The feature values ​​of all two-dimensional projected images are fitted into feature curves in chronological order of acquisition time, including: A discrete data point set is constructed by using the acquisition time of each two-dimensional projection image as the horizontal axis and the feature value of each two-dimensional projection image as the vertical axis. The discrete data point set is fitted with a weighted least squares method and a cubic B-spline curve to obtain the characteristic curve.

5. The method according to claim 1, characterized in that, Identifying the peaks and valleys of the characteristic curve includes: Determine the maximum value, minimum value, and amplitude of the characteristic curve throughout the entire acquisition time range; Based on the maximum value, minimum value, and amplitude, set the peak detection threshold and the valley detection threshold; By traversing the feature curves through a sliding window, the local maximum and local minimum values ​​of the feature curves within each sliding window are calculated. Local maximum values ​​exceeding the peak detection threshold are determined as peak values, and local minimum values ​​below the valley detection threshold are determined as valley values.

6. The method according to claim 1, characterized in that, The plurality of two-dimensional projection images cover at least two complete cardiac cycles, and / or the acquisition time interval between two adjacent two-dimensional projection images in the plurality of two-dimensional projection images is fixed.

7. The method according to claim 1, characterized in that, After determining the diastolic and systolic projection images of the heart, the following steps are also included: Image feature parameters are extracted from the diastolic projection image and the systolic projection image of the heart, and the image feature parameters include at least the ventricular diameter and / or ventricular area; If the image feature parameters conform to the physiological characteristics of cardiac diastole and cardiac systole, then the diastolic projection image and the systolic projection image are determined as target projection images; If the image feature parameters do not conform to the physiological characteristics of cardiac diastole and cardiac systole, then the peak detection threshold, trough detection threshold and / or preprocessing steps are readjusted until the image feature parameters of the adjusted diastolic projection image and cardiac systolic projection image conform to the physiological characteristics.

8. A device for identifying the diastolic and systolic phases of the heart in CBCT images, characterized in that, include: The projection image acquisition module is used to scan the heart region using a CBCT device and acquire several two-dimensional projection images of the heart region. The preprocessing module is used to preprocess the plurality of two-dimensional projection images to suppress the scanning error of the CBCT device and enhance the distinction between cardiac motion signals and non-motion signals in the CBCT images. The feature extraction module is used to extract feature values ​​from each preprocessed two-dimensional projection image based on the PCA algorithm. These feature values ​​are used to characterize the cardiac motion state. The feature curve fitting module is used to fit the feature values ​​of all two-dimensional projection images into feature curves according to the acquisition time sequence. The target projection image recognition module is used to identify the peak and trough values ​​of the feature curve, determine that the two-dimensional projection image corresponding to the peak value is the diastolic projection image of the heart, and the two-dimensional projection image corresponding to the trough value is the systolic projection image of the heart.

9. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method according to any one of claims 1 to 7.

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

Images