Cardiovascular image-based myocardial death area image detection method and system

Abstract

The invention discloses a cardiac death area image detection method and system based on a cardiovascular image, and relates to the technical field of image detection, and the method comprises the steps: obtaining cardiovascular image data; in the current de-noising stage, de-noising processing is carried out on the cardiovascular image data by adopting an anisotropic diffusion filter; evaluating a noise suppression effect according to the denoised cardiovascular image data; according to the noise suppression effect, a de-noising processing strategy is adjusted, and the de-noising processing strategy is used for executing the next de-noising stage; after noise suppression processing of the cardiovascular image data is completed, a denoising quality grade is generated, and the noise suppression processing comprises a plurality of denoising stages; and according to the de-noising quality grade, carrying out myocardial death area image detection on the cardiovascular image data after noise suppression processing. According to the method, fine-grained de-noising processing can be carried out by dividing multiple de-noising stages, the de-noising effect is enhanced, image detection is achieved, and the accuracy and reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, and in particular to a method and system for detecting myocardial death regions based on cardiovascular imaging. Background Technology

[0002] In the diagnosis and treatment of cardiovascular diseases, accurate identification of myocardial death areas is crucial for disease assessment and treatment planning. Existing systems improve diagnostic efficiency and accuracy by processing image data such as magnetic resonance imaging (MRI) or computed tomography (CT) scans, performing denoising, segmentation, and quantitative analysis, thereby reducing the subjectivity of human interpretation. However, during system evolution, updates to the underlying computational library may introduce hidden performance degradation defects. This defect manifests when processing images with specific high-frequency noise features, leading to a significant decrease in the execution efficiency of key filtering algorithm functions. This results in insufficient iterations for denoising within a limited time, inadequate denoising effects, and more subtle artifacts and noise remaining in the image. When this noise enters the subsequent edge detection stage, it can easily introduce false intensity gradients or mask true gradients, causing deviations in boundary recognition, affecting the accurate segmentation of myocardial death areas, resulting in uneven boundaries or deviations from pathological boundaries, leading to low detection accuracy and reliability.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this invention is to propose a method and system for detecting myocardial death regions based on cardiovascular imaging. This method can enhance the denoising effect by dividing the image into multiple denoising stages and performing fine-grained denoising processing, thereby achieving image detection and improving accuracy and reliability.

[0005] On one hand, embodiments of the present invention provide an image detection method for myocardial death regions based on cardiovascular imaging, comprising the following steps: Acquire cardiovascular imaging data; In the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular image data; The noise suppression effect was evaluated based on the denoised cardiovascular imaging data. Based on the noise suppression effect, the denoising strategy is adjusted, which includes adjusting the number of iterations, switching the filtering method, and adjusting the lower limit of image quality, in order to execute the next denoising stage. After the cardiovascular imaging data has undergone noise suppression processing, a denoising quality level is generated. The noise suppression processing includes multiple denoising stages. Based on the denoising quality level, myocardial death region image detection is performed on the cardiovascular imaging data after noise suppression processing.

[0006] On the other hand, embodiments of the present invention provide an image detection system for myocardial death regions based on cardiovascular imaging, comprising: The data acquisition module is used to acquire cardiovascular imaging data; The denoising module is used to denoise the cardiovascular image data using an anisotropic diffusion filter in the current denoising stage. The effect evaluation module is used to evaluate the noise suppression effect based on the denoised cardiovascular imaging data. The strategy adjustment module is used to adjust the denoising processing strategy according to the noise suppression effect. The denoising processing strategy includes iteration number adjustment, filtering method switching and image quality lower limit adjustment, and is used to execute the next denoising stage. The noise reduction quality level generation module is used to generate a noise reduction quality level after the cardiovascular imaging data has undergone noise suppression processing. The noise suppression processing includes multiple noise reduction stages. The image detection module is used to perform myocardial death region image detection on the noise-suppressed cardiovascular imaging data according to the denoising quality level.

[0007] The embodiments of this application include at least the following beneficial effects: First, cardiovascular imaging data is acquired. Then, in the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular imaging data. Next, the noise suppression effect is evaluated based on the denoised cardiovascular imaging data, and the denoising strategy is adjusted according to the noise suppression effect to execute the next denoising stage. After the cardiovascular imaging data completes the noise suppression process, a denoising quality level is generated. Finally, based on the denoising quality level, myocardial death region image detection is performed on the noise-suppressed cardiovascular imaging data. Thus, by dividing the data into multiple denoising stages, fine-grained denoising processing can be performed, enhancing the denoising effect and achieving image detection, thereby improving accuracy and reliability.

[0008] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description and the drawings. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0010] Figure 1 This is a flowchart of an image detection method for myocardial death regions based on cardiovascular imaging, according to an embodiment of the present invention. Figure 2This is a schematic diagram of the structure of a myocardial death region image detection system based on cardiovascular imaging, according to an embodiment of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0012] In cardiovascular disease diagnosis and treatment, accurate identification of myocardial death areas is crucial for disease assessment and treatment planning. Existing systems improve diagnostic efficiency and accuracy by processing image data such as magnetic resonance imaging (MRI) or computed tomography (CT) scans, performing denoising, segmentation, and quantitative analysis, while reducing the subjectivity of human interpretation. However, during system evolution, updates to the underlying computational library may introduce hidden performance degradation defects. This defect manifests when processing images with specific high-frequency noise features, leading to a significant decrease in the execution efficiency of key filtering algorithm functions. This results in insufficient iterations for denoising within a limited time, inadequate denoising effects, and more subtle artifacts and noise remaining in the image. When this noise enters the subsequent edge detection stage, it can easily introduce false intensity gradients or mask true gradients, causing deviations in boundary recognition, affecting the accurate segmentation of myocardial death areas, resulting in uneven boundaries or deviations from pathological boundaries, low detection accuracy, and low reliability.

[0013] Specifically, the existing system aims to provide accurate assessments of myocardial viability by processing various cardiovascular imaging modalities, such as cardiac magnetic resonance imaging (MRI) or computed tomography (CT). Its typical workflow includes image data acquisition, preprocessing (e.g., noise suppression), accurate segmentation of cardiac structures (particularly the endocardium and endoderm), and quantitative calculation of myocardial death regions based on this. However, the system underwent a standard software library upgrade, which included a third-party low-level computational library specifically for image processing. This library may contain an undiscovered performance degradation defect. This defect may only become apparent when processing image data with specific high-frequency noise characteristics. In such cases, a critical filtering algorithm function in the updated library may execute significantly less efficiently than the older version. This defect does not cause system crashes or obvious error messages but rather subtly affects system performance in a data-characteristic-dependent manner, making it extremely difficult to detect during routine regression testing or quality assurance checks. For example, while the library's new memory access pattern may improve speed in most cases, it could actually lead to decreased memory access efficiency and become a performance bottleneck when encountering highly localized and rapidly fluctuating pixel intensity values ​​common in high-frequency noisy image data.

[0014] Therefore, due to the unexpected decrease in the execution efficiency of the aforementioned specific algorithm functions, when the system processes image data that happens to have such high-frequency noise characteristics, the calculation process of the smoothing filter parameters used for image denoising cannot complete enough iterations to reach the optimal convergence state within the limited processing time set by the system to ensure real-time performance. Modern image analysis systems typically need to operate under strict real-time or near-real-time constraints to support rapid clinical decision-making. These constraints set a fixed time budget for each processing step. If an iterative optimization process, such as the calculation of the optimal parameters for anisotropic diffusion filters or nonlocal mean filters, is forced to terminate prematurely due to the slowdown of underlying calculation steps, then the filter parameters will not converge sufficiently to their ideal state. This means that the filter has failed to be optimally adjusted for the specific noise characteristics of the image data. For this particular portion of image data, the final denoising effect will be slightly insufficient, and the processed image will retain more subtle artifacts and noise than after normal processing. This slightly deficiently denoised image data will cause the subsequent edge detection logic for identifying the boundaries of the myocardium and endocardium to have a slight deviation in judging those small, irregular boundary features in the image when entering the subsequent edge detection logic for identifying the boundaries of the myocardium and endocardium. This subtle deviation directly affects the accurate segmentation of the myocardial death area, resulting in an uneven segmentation boundary or a slight offset from the actual pathological boundary, leading to low detection accuracy and reliability.

[0015] The embodiments of this application will be explained in detail below with reference to the accompanying drawings: Figure 1 This is an optional flowchart of an image detection method for myocardial death regions based on cardiovascular imaging provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S101 to S106.

[0016] Step S101: Acquire cardiovascular imaging data; Step S102: In the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular image data; Step S103: Evaluate the noise suppression effect based on the denoised cardiovascular imaging data; Step S104: Adjust the denoising strategy according to the noise suppression effect. The denoising strategy includes adjusting the number of iterations, switching the filtering method, and adjusting the lower limit of image quality, which is used to execute the next denoising stage. Step S105: After the cardiovascular imaging data has been processed for noise suppression, a denoising quality level is generated. The noise suppression process includes multiple denoising stages. Step S106: Based on the denoising quality level, perform myocardial death region image detection on the cardiovascular imaging data after noise suppression processing.

[0017] Steps S101 to S106 as shown in the embodiments of this application can perform fine-grained denoising processing by dividing the process into multiple denoising stages, thereby enhancing the denoising effect, achieving image detection, and improving accuracy and reliability.

[0018] In some embodiments, steps S101-S106 may involve acquiring cardiovascular imaging data first. Patient MRI or CT image data can be retrieved from a hospital's PACS (Picture Archiving and Communication System), or data can be received in real time via an interface directly connected to the imaging equipment. Alternatively, image data can be manually imported into the system via external storage media (such as a USB drive or network shared folder). It is understood that cardiovascular imaging data refers to image information acquired through various medical imaging techniques (such as MRI, CT, ultrasound, etc.) that reflects the structure and function of the cardiovascular system. This data is typically stored in digital format and contains rich pixel intensity, texture, and morphological information.

[0019] In the current denoising stage, an anisotropic diffusion filter is used to denoise cardiovascular image data. The anisotropic diffusion filter can be configured to perform strong smoothing in flat areas of the image, while performing weaker smoothing in edge areas such as myocardial boundaries, thus preserving important anatomical details. Alternatively, a fixed-parameter anisotropic diffusion filter can be used for initial denoising; these parameters can be set through experimental analysis of a large number of cardiovascular images. It is understood that the anisotropic diffusion filter is a nonlinear smoothing filter, characterized by its ability to preserve edge information while smoothing the image. By adjusting the diffusion coefficient based on local gradient information, it achieves strong smoothing in flat areas and weaker smoothing in edge areas, thereby effectively removing noise and preserving image details.

[0020] The noise suppression effect is then evaluated based on the denoised cardiovascular image data. The change in signal-to-noise ratio (SNR) or peak signal-to-noise ratio (PSNR) before and after denoising can be calculated to quantify the degree of noise reduction. Image smoothness and detail preservation can also be measured by calculating the local variance or structural similarity index (SSIM). The denoising strategy is adjusted based on the noise suppression effect, including adjusting the number of iterations, switching filtering methods, and adjusting the lower limit of image quality for the next denoising stage. For example, if the noise suppression effect is poor, the number of iterations of the anisotropic diffusion filter can be increased to enhance the denoising intensity. Alternatively, if the current filter is inefficient in handling a specific type of noise, it can be switched to another filtering method more suitable for that noise characteristic, such as non-local mean (NLM) filtering or wavelet thresholding denoising. Furthermore, the lower limit of image quality can be adjusted based on the evaluation results; for example, when the noise suppression effect is poor, the lower limit of image quality can be appropriately relaxed to avoid over-processing that leads to the loss of important details. It is understandable that iteration number adjustment refers to dynamically changing the number of filter iterations based on the current denoising effect; filtering method switching refers to selecting between different filtering algorithms based on image characteristics or noise type; and image quality lower limit adjustment refers to setting an acceptable minimum image quality standard to balance denoising effect and processing time.

[0021] After noise suppression processing is performed on cardiovascular imaging data, a denoising quality level is generated. This noise suppression process includes multiple stages. For example, the cardiovascular imaging data can be classified into different quality levels such as "Excellent," "Good," "Medium," and "Poor" based on the residual noise level, edge sharpness, and overall visual clarity after noise suppression. Alternatively, a scoring system can be used to represent the denoising quality level. This system comprehensively considers multiple denoising indicators, such as signal-to-noise ratio, structural similarity index, and edge preservation.

[0022] Finally, based on the denoising quality level, myocardial death region detection is performed on the noise-suppressed cardiovascular imaging data. For example, if the denoising quality level is excellent, a high-sensitivity edge detection algorithm and a fine segmentation model can be used to detect even small lesion areas. If the denoising quality level is low, a more robust segmentation algorithm can be used, combined with morphological operations for post-processing to reduce the impact of noise on the detection results.

[0023] This embodiment first acquires cardiovascular imaging data, which may contain various types of noise. In the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular imaging data. This filter can adaptively smooth the image based on local features, preserving important details such as myocardial boundaries as much as possible while removing noise. After denoising, the system evaluates the noise suppression effect based on the processed cardiovascular imaging data to quantify the performance of the current denoising stage. Based on the evaluation results, the system adjusts the denoising strategy for the next denoising stage. For example, if the noise suppression effect is not ideal, the system may increase the number of filter iterations to enhance the denoising intensity, or switch to a filtering method more suitable for the current noise characteristics. This adaptive adjustment mechanism ensures that the denoising process can be optimized according to the actual data conditions, avoiding the limitations of fixed parameters. After all denoising stages of the cardiovascular imaging data have been completed, the system generates a denoising quality level. This level is a comprehensive evaluation of the final denoised image quality. Finally, based on the generated denoising quality level, myocardial death region image detection is performed on the noise-suppressed cardiovascular imaging data. For example, high-quality denoised images can employ more refined and sensitive detection algorithms, while lower-quality images may require more robust algorithms to reduce noise interference with the detection results. Through the above steps, this embodiment generates high-quality denoised images and performs accurate detection of myocardial death regions, from initial denoising to effect evaluation and strategy adjustment. This dynamic adaptive strategy adjustment mechanism enables the system to effectively address potential performance degradation defects in the underlying computing library, ensuring reliable denoising results in various complex noise environments, thereby significantly improving the accuracy of myocardial death region detection and the reliability of clinical diagnosis.

[0024] Through the above technical solution, this embodiment enables the system to monitor the performance of the denoising process in real time and dynamically adjust denoising parameters and methods based on the actual effect by evaluating the noise suppression effect and adjusting the denoising processing strategy. This adaptive adjustment capability can effectively address potential hidden performance defects in the underlying computing library, avoiding subsequent decreases in detection accuracy due to insufficient denoising. Furthermore, this embodiment evaluates the quality of cardiovascular image data after noise suppression by generating a denoising quality level. This level provides important quality criteria for subsequent myocardial death region image detection, allowing the detection algorithm to adaptively adjust based on image quality, further improving detection accuracy and robustness. This embodiment overcomes the problems of unstable denoising effects and decreased detection accuracy in complex noise environments, especially when there are performance defects in the underlying software library. It can continuously provide high-quality denoised images, ensuring accurate identification of myocardial death regions, thereby significantly improving the reliability of the intelligent detection system and enhancing the data reference value.

[0025] In some embodiments, in step S103, evaluating the noise suppression effect based on the denoised cardiovascular imaging data may include, but is not limited to, the following steps: Obtain the local variance and peak signal-to-noise ratio of the previous denoising stage; Based on the preset sliding window size and the denoised cardiovascular image data, calculate the local variance of the current denoising stage; The degree of smoothness optimization is calculated based on the local variance of the previous denoising stage and the local variance of the current denoising stage. Calculate the peak signal-to-noise ratio for the current denoising stage based on the denoised cardiovascular imaging data; The degree of signal-to-noise ratio optimization is calculated based on the peak signal-to-noise ratio of the previous denoising stage and the peak signal-to-noise ratio of the current denoising stage. The noise suppression effect is evaluated based on the degree of smoothness optimization and signal-to-noise ratio optimization.

[0026] In some embodiments, the local variance and peak signal-to-noise ratio (PSNR) of the previous denoising stage can be obtained first. Local variance refers to the dispersion of pixel grayscale values ​​within a local region of the image; a smaller value generally indicates a smoother image with less noise. Peak signal-to-noise ratio (PSNR) is an objective standard for measuring image quality, used to quantify the relative intensity of signal and noise; a higher value generally indicates better image quality and more significant noise suppression. This historical data can be obtained from a historical database, providing a benchmark for evaluating the effectiveness of the current denoising stage.

[0027] Then, based on the preset sliding window size and the denoised cardiovascular image data, the local variance of the current denoising stage is calculated. Introducing a sliding window allows for fine-grained analysis of local image regions, avoiding the possibility of local noise or detail information being obscured by global variance. For example, a 3x3 sliding window size can be set to calculate the variance of the cardiovascular image data, yielding the local variance. The degree of smoothness optimization is then calculated based on the local variance of the previous and current denoising stages. This metric directly reflects the improvement in image smoothness achieved by the denoising process.

[0028] Next, based on the denoised cardiovascular image data, the peak signal-to-noise ratio (PSNR) of the current denoising stage is calculated. The degree of SNR optimization is then calculated by comparing the PSNR of the previous and current denoising stages. This metric quantifies the improvement in image SNR achieved by the denoising process.

[0029] Finally, the noise suppression effect is evaluated based on the degree of smoothness optimization and signal-to-noise ratio optimization. A comprehensive and objective assessment of the current noise suppression effect can be achieved by considering both the calculated smoothness optimization and signal-to-noise ratio optimization.

[0030] This embodiment achieves dynamic and quantitative evaluation of noise suppression effectiveness by introducing the acquisition of local variance and peak signal-to-noise ratio from the previous denoising stage and combining them with the corresponding calculation results of the current denoising stage. Specifically, changes in local variance reflect improvements in image smoothness, while changes in peak signal-to-noise ratio indicate improvements in image signal-to-noise ratio. By calculating the degree of smoothness optimization and signal-to-noise ratio optimization, the contribution of the current denoising process to image quality can be accurately measured. This evaluation mechanism based on stage-by-stage comparison allows the denoising process to be fed back and adjusted according to actual results.

[0031] Through the above technical solution, this embodiment can provide a more refined and objective evaluation of noise suppression effect. By comprehensively considering the degree of smoothness optimization and signal-to-noise ratio optimization, this embodiment provides a multi-dimensional and quantitative evaluation standard, which can more accurately determine the effectiveness of denoising processing, provide a reliable basis for subsequent denoising processing strategy adjustments, and ultimately help obtain higher quality cardiovascular imaging data, laying a solid foundation for the image detection of myocardial death areas.

[0032] In some embodiments, in step S104, adjusting the denoising processing strategy according to the noise suppression effect may include, but is not limited to, the following steps: Obtain the computation time and computation resource usage pattern of the current denoising stage. The computation resource usage pattern includes the instantaneous utilization of the central processing unit core, memory access locality, and data cache hit rate. The computation time of the current denoising stage is matched with the normal operating time baseline table to obtain the expected suppression effect; Calculate the noise suppression effect deviation based on the noise suppression effect and the expected suppression effect; By matching computing resource usage patterns with characteristic fluctuation patterns, target fluctuation patterns are obtained. Characteristic fluctuation patterns include normal high load patterns and efficiency degradation patterns. The degree of change in computational efficiency is assessed based on the suppression effect deviation and the target fluctuation pattern; Adjust the noise reduction strategy based on the degree of change in computational efficiency.

[0033] In some embodiments, adjusting strategies solely based on noise suppression effectiveness may not fully balance efficient utilization of computational resources and processing efficiency. For example, while some denoising strategies may achieve good noise suppression, they may be accompanied by excessive computation time or unreasonable resource consumption, thereby affecting the real-time performance of cardiovascular imaging data processing and the overall system performance. Failure to address these issues may lead to inefficiency when processing large amounts of image data, or even resource bottlenecks, ultimately impacting the timeliness and accuracy of myocardial death region image detection.

[0034] To achieve this, we can first obtain the computation time and computational resource usage patterns of the current denoising stage. Computation time refers to the amount of time required to complete the current denoising stage; for example, we can record the precise timestamps from the start to the end of the denoising process. Computational resource usage patterns include CPU core utilization, memory locality of access, and cache hit rate. CPU core utilization reflects the processor's workload at a given moment; memory locality of access refers to the locality characteristics exhibited by a program when accessing memory, including temporal and spatial locality, and its level directly affects memory access efficiency; cache hit rate represents the proportion of data found in the cache and is an important indicator of cache efficiency. These indicators can be obtained through system-level performance monitoring tools or performance counters embedded in the processing flow, with the aim of comprehensively understanding the resource consumption status of the current denoising stage.

[0035] The computation time of the current denoising stage is then matched against an uptime baseline table to obtain the expected suppression effect. The uptime baseline table is an empirical or pre-experimental database that records the standard computation time required to achieve a specific noise suppression effect under different computational environments and denoising strategies. By comparing the actual computation time with the baseline table, the theoretically achievable noise suppression level can be inferred. The suppression effect deviation is then calculated based on the noise suppression effect and the expected suppression effect. The suppression effect deviation is the difference between the actual noise suppression effect and the expected suppression effect inferred from the baseline table; this deviation quantifies the degree to which the actual performance of the current denoising strategy deviates from its theoretical performance for a given computation time.

[0036] Next, the computing resource usage pattern is matched with the characteristic fluctuation pattern to obtain the target fluctuation pattern. The characteristic fluctuation pattern is a number of typical resource usage patterns summarized based on historical data or expert experience, which can include normal high-load mode and efficiency degradation mode. Normal high-load mode indicates that system resources are fully utilized and processing efficiency is normal, while efficiency degradation mode indicates that resources are occupied but processing efficiency is low, such as due to memory jitter or cache failure. By matching, it is possible to identify which typical mode the current resource usage belongs to. And based on the suppression effect deviation and the target fluctuation pattern, the degree of change in computing efficiency is evaluated. The degree of change in computing efficiency is a comprehensive indicator that combines the gap between the actual performance of the denoising effect and the expected performance, as well as the actual usage of computing resources, to determine whether the computing efficiency of the current denoising process has improved, remained stable, or decreased.

[0037] Finally, the denoising strategy is adjusted based on the degree of change in computational efficiency. This includes adjusting the number of iterations, switching filtering methods, and adjusting the lower limit of image quality. For example, if computational efficiency drops significantly, even if noise suppression is still acceptable, it may be necessary to reduce the number of iterations, switch to a filtering method with lower computational complexity, or appropriately relax the lower limit of image quality, sacrificing some denoising effect for higher processing efficiency. Conversely, if computational efficiency is high and there is still a margin, increasing the number of iterations or using a more refined filtering method can be considered to further improve denoising quality.

[0038] This embodiment addresses the issue of potentially overlooking resource consumption and processing efficiency when adjusting denoising strategies by introducing an evaluation of computational efficiency. Specifically, by acquiring the computation time and resource usage patterns of the current denoising stage, the system can comprehensively perceive the resource consumption of the processing. Matching the computation time with an uptime baseline table quantifies the gap between the actual and theoretical performance of the current strategy, i.e., the suppression effect deviation. Simultaneously, by analyzing resource usage patterns and matching them with characteristic fluctuation patterns, potential efficiency bottlenecks or resource waste patterns can be identified. Therefore, by comprehensively evaluating the degree of change in computational efficiency based on the suppression effect deviation and the target fluctuation pattern, strategy adjustments can dynamically balance denoising quality and computational efficiency. For example, when good denoising results are found but computational efficiency is low, the system can proactively adjust the strategy, such as reducing the number of iterations or switching to a lighter filtering method, thereby avoiding unnecessary resource waste and ensuring efficient system operation while meeting denoising requirements.

[0039] To illustrate this technical solution more clearly, a specific example is used below. Suppose that in denoising cardiovascular imaging data, an anisotropic diffusion filter is used in the current denoising stage, and an initial number of iterations is set. After completing this stage, the system first obtains the computation time of 500 milliseconds and monitors the instantaneous CPU core utilization at 85%, memory access locality at medium, and data cache hit rate at 70%. Subsequently, the system matches the 500-millisecond computation time with a preset uptime benchmark table. This benchmark table shows that, under the current hardware configuration, a 500-millisecond computation time should typically achieve an 80% noise suppression effect. However, the actual noise suppression effect obtained is only 75%. Therefore, the calculated suppression effect deviation is -5% (the actual effect is lower than expected).

[0040] Simultaneously, the system matches the monitored CPU core instantaneous utilization, memory access locality, and data cache hit rate with the characteristic fluctuation pattern. The matching results show that the current pattern closely matches the "efficiency degradation pattern," which may mean that although CPU utilization is high, the actual processing efficiency has not reached its optimal level due to low memory access efficiency or low cache hit rate. Based on the suppression effect deviation of -5% and the target fluctuation pattern being the "efficiency degradation pattern," the system assesses the degree of change in computational efficiency as "significantly decreased." In view of this, the system will adjust the denoising processing strategy. Specifically, to improve overall efficiency, the system decides to reduce the number of iterations in the next denoising stage from the original 10 to 7, and considers switching to a bilateral filtering method with slightly lower computational complexity. At the same time, the lower limit of image quality is slightly relaxed to prioritize processing efficiency within an acceptable denoising quality range. In this way, the system can continue to process cardiovascular imaging data with better computational efficiency in subsequent denoising stages, thereby accelerating the process of myocardial death region image detection.

[0041] Through the above technical solution, this embodiment enables more refined and intelligent adjustments to the denoising strategy. This embodiment not only focuses on noise suppression effectiveness but also incorporates computational efficiency metrics such as computation time, instantaneous CPU core utilization, memory access locality, and data cache hit rate. Therefore, it effectively avoids the problems of excessive resource consumption or low processing efficiency that may occur when pursuing the ultimate denoising effect, thereby significantly improving processing efficiency and system resource utilization while ensuring the denoising quality of cardiovascular imaging data. This embodiment makes the denoising process more adaptable to different computing environments and real-time requirements, providing a more efficient and stable data foundation for subsequent myocardial death region image detection, thereby improving the timeliness and reliability of the overall diagnostic process.

[0042] In some embodiments, the step S105 of generating a noise reduction quality level may include, but is not limited to, the following steps: Local regional frequency analysis was performed on the noise-suppressed cardiovascular imaging data to identify the type and spatial distribution of residual noise. Edge analysis was performed on the noise-suppressed cardiovascular imaging data to assess the sharpness of local edges. Based on the noise suppression effect, local edge sharpness, type and spatial distribution of residual noise, a noise reduction quality level is generated.

[0043] In some embodiments, local frequency analysis can be performed on the noise-suppressed cardiovascular imaging data to identify the type and spatial distribution of residual noise. The aim is to identify the characteristics of any residual noise that may exist in the image by analyzing the distribution of different frequency components. Specifically, methods such as Fourier transform, wavelet analysis, or short-time Fourier transform can be used to perform spectral analysis on local areas of the image, thereby revealing the spatial distribution pattern of noise (e.g., uniform distribution, speckled distribution, or high-frequency noise) and its frequency characteristics (e.g., high-frequency noise or low-frequency noise). The purpose is to comprehensively understand the noise components still present in the denoised image, providing refined data support for subsequent quality assessment.

[0044] Then, edge analysis is performed on the noise-suppressed cardiovascular imaging data to assess local edge sharpness, aiming to evaluate the sharpness of local edges of important structures in the image (such as myocardial boundaries and vascular contours). During denoising, excessive smoothing may lead to blurred edges, affecting subsequent diagnostic accuracy. By employing edge detection algorithms such as the Sobel operator, Canny operator, or Laplacian operator, the sharpness and intensity of image edges can be quantified, thereby assessing the impact of denoising on image detail preservation. The goal is to ensure that while suppressing noise, key diagnostic information in the image is not lost or distorted due to edge blurring.

[0045] Then, based on the noise suppression effect, local edge sharpness, and the type and spatial distribution of residual noise, a denoising quality level is generated. This can comprehensively consider the evaluated noise suppression effect, local edge sharpness, and the type and spatial distribution of residual noise. For example, a multi-parameter weighted model can be constructed, a rule-based expert system can be used, or even a machine learning model can be trained, using these indicators as input to output a quantified denoising quality level. This quality level can be a discrete level (such as "Excellent," "Good," "Medium," "Poor") or a continuous score to comprehensively reflect the overall performance of the denoising process.

[0046] This embodiment introduces local region frequency analysis and edge analysis to perform a more detailed and comprehensive quality assessment of the denoised cardiovascular imaging data. By identifying the type and spatial distribution of residual noise, this embodiment can understand the specific characteristics of the noise and avoid misjudgments caused by different noise types. By evaluating local edge sharpness, it can ensure that the boundary information of key anatomical structures and pathological features of the image is effectively preserved while removing noise. This multi-dimensional and refined evaluation method enables the generated denoising quality level to more accurately reflect the true quality of the image, thus providing more reliable input for subsequent image detection of myocardial death areas.

[0047] Through the above technical solution, this embodiment overcomes the limitations of evaluating denoising effectiveness with a single index, providing a more comprehensive and accurate denoising quality level. This quality level considers not only the degree of noise suppression but also the preservation of image details, especially edge information crucial for diagnosis. Therefore, it effectively avoids the loss of key information due to over-denoising or diagnostic interference due to insufficient denoising, significantly improving the reliability and accuracy of cardiovascular imaging data in subsequent myocardial death region image detection, thereby enhancing the overall diagnostic precision.

[0048] In some embodiments, step S106, based on the denoising quality level, performs myocardial death region image detection on the noise-suppressed cardiovascular imaging data, which may include, but is not limited to, the following steps: Step S201: Select multiple candidate boundary recognition strategies from the boundary recognition strategy library according to the denoising quality level; Step S202: According to the preset diagnostic priority rules, perform conflict resolution on multiple candidate boundary recognition strategies to obtain conflict resolution results; Step S203: Based on the conflict resolution result, multiple candidate boundary recognition strategies are fused to obtain the target boundary recognition strategy; Step S204: Based on the target boundary recognition strategy, perform myocardial death region image detection on the cardiovascular imaging data after noise suppression processing.

[0049] In some embodiments, multiple candidate boundary recognition strategies can be selected from a boundary recognition strategy library based on the denoising quality level. Several potentially effective strategies suitable for the given denoising quality level can be intelligently selected from the pre-built boundary recognition strategy library. The boundary recognition strategy library refers to a collection of image segmentation or edge detection algorithms designed for different image characteristics, noise levels, or pathological features, such as threshold-based segmentation, region growing, level set methods, and deep learning models. Its purpose is to ensure that subsequent detection processes are based on strategies that match the current image quality, thereby improving the initial adaptability of the detection.

[0050] Then, based on preset diagnostic priority rules, conflict resolution is performed on multiple candidate boundary identification strategies to obtain the conflict resolution result. After selecting multiple candidate strategies, actual diagnostic needs and clinical priorities can be considered, and these strategies can be evaluated and coordinated using preset diagnostic priority rules. Preset diagnostic priority rules may include preference settings for detection sensitivity, specificity, computational efficiency, robustness to specific artifacts, etc. The conflict resolution process aims to resolve potential differences in detection results or different focuses between different candidate strategies. For example, one strategy may perform well in sensitivity but have slightly lower specificity, while another strategy may be the opposite. Through resolution, a comprehensive decision result can be obtained to guide subsequent strategy fusion.

[0051] Based on the conflict resolution results, multiple candidate boundary recognition strategies are then fused to obtain the target boundary recognition strategy. Multiple candidate strategies that have undergone resolution can be organically combined to form a more comprehensive, robust, and diagnostically sound target boundary recognition strategy. Fusion methods can include, but are not limited to, voting mechanisms, weighted averaging, feature-level fusion, or decision-level fusion. The aim is to comprehensively utilize the advantages of each candidate strategy, compensate for the shortcomings of a single strategy, and thus generate an optimal detection scheme under the current denoising quality level and diagnostic priority.

[0052] Finally, based on the target boundary recognition strategy, myocardial death region image detection is performed on the noise-suppressed cardiovascular imaging data. The finalized target boundary recognition strategy can be applied to the noise-suppressed cardiovascular imaging data to accurately identify and delineate the boundaries of the myocardial death region. For example, this strategy can be used for pixel-level classification or region segmentation of the image, ultimately outputting an image with clearly marked myocardial death regions. The aim is to provide high-precision and high-reliability myocardial death region image detection results, providing strong support for clinical diagnosis.

[0053] This embodiment effectively addresses the issues of insufficient detection accuracy and robustness when processing cardiovascular imaging data with different denoising quality levels by introducing dynamic selection, conflict resolution, and fusion mechanisms for boundary recognition strategies. Specifically, the denoising quality level is used to guide the selection of candidate strategies, ensuring the matching between the detection strategy and image quality; the introduction of preset diagnostic priority rules allows the detection process to fully consider actual clinical needs, avoiding the one-sidedness of a single technical indicator; and conflict resolution and strategy fusion further improve the overall performance and reliability of the detection results, enabling the final target boundary recognition strategy to better adapt to complex and ever-changing diagnostic scenarios.

[0054] To illustrate this technical solution more clearly, a specific example is used below. Suppose that after a certain denoising stage, the cardiovascular imaging data is assessed as having a "medium" denoising quality level. Based on this level, the system selects multiple candidate boundary recognition strategies from a boundary recognition strategy library, such as "gradient-based edge detection strategies," "texture-feature-based region growing strategies," and "deep learning-based semantic segmentation models." Subsequently, according to preset diagnostic priority rules (e.g., prioritizing high sensitivity in early diagnosis and high specificity in precise lesion assessment), these candidate strategies are conflict-resolved. For example, if the current diagnostic scenario focuses more on detecting all possible lesions (high sensitivity), the decision might favor strategies that perform better in terms of sensitivity. Based on this decision, these strategies are fused, for example, through weighted voting or ensemble learning, to form a comprehensive target boundary recognition strategy. Finally, this target boundary recognition strategy is applied to the denoised cardiovascular imaging data to accurately identify and delineate the myocardial death region, thus providing a detection result that considers both image quality and clinical diagnostic priorities.

[0055] Through the above technical solution, this embodiment can dynamically adjust and optimize the detection strategy for myocardial death regions based on the specific denoising quality level of cardiovascular imaging data, significantly improving the accuracy and robustness of detection. This embodiment not only effectively addresses the challenges posed by different noise levels and image quality, but also incorporates diagnostic priority rules, making the detection results more clinically valuable and helping doctors make more accurate diagnostic judgments, thereby improving the overall efficiency and reliability of cardiovascular disease diagnosis.

[0056] In some embodiments, step S204, performing myocardial death region image detection on the noise-suppressed cardiovascular imaging data according to the target boundary recognition strategy, may include, but is not limited to, the following steps: Adjust the gradient sensitivity and smoothing constraint weights according to the target boundary identification strategy; Texture features were extracted from the noise-suppressed cardiovascular imaging data to obtain the local texture of myocardial tissue. Calculate texture similarity based on local texture and microtexture template of myocardial tissue; Based on gradient sensitivity, smoothing constraint weights, and texture similarity, myocardial death region detection is performed on noise-suppressed cardiovascular imaging data.

[0057] In some embodiments, the gradient sensitivity and smoothness constraint weights can be adjusted first according to the target boundary recognition strategy. Gradient sensitivity refers to a parameter used in image processing to control the degree to which the edge detection algorithm responds to changes in image brightness. By adjusting the gradient sensitivity, the detection algorithm can be made more sensitive to weak or blurred boundaries in areas of myocardial death, or its sensitivity can be reduced when the boundaries are clear to avoid over-segmentation. Smoothness constraint weights refer to parameters used to balance boundary smoothness and image data fit during image segmentation or boundary extraction. Higher smoothness constraint weights help generate smoother, more continuous boundaries, while lower weights allow the boundaries to better fit image details. These parameter adjustments are based on the aforementioned target boundary recognition strategy to ensure that the detection process can adapt to specific diagnostic needs and image features.

[0058] Then, texture features are extracted from the noise-suppressed cardiovascular imaging data to obtain the local texture of myocardial tissue. The structural information of myocardial tissue can be quantified by analyzing the local spatial arrangement and intensity variation patterns of image pixels. For example, methods such as gray-level co-occurrence matrix, local binary pattern, or Gabor filter can be used to extract local texture features of myocardial tissue. The local texture of myocardial tissue is a collection of these extracted features, reflecting the microstructure and texture of the myocardial region.

[0059] Texture similarity is then calculated based on the local texture of myocardial tissue and the microtexture template of myocardial tissue. The microtexture template of myocardial tissue is a pre-established texture feature pattern representing healthy myocardial tissue or typical areas of myocardial death. These templates can be trained and learned based on a large amount of clinical imaging data. The probability of the current area belonging to healthy tissue or a dead area can be quantified by comparing the similarity between the local texture of myocardial tissue and the microtexture template of myocardial tissue. For example, Euclidean distance, cosine similarity, or correlation coefficient can be used to calculate texture similarity.

[0060] Finally, based on gradient sensitivity, smoothness constraint weights, and texture similarity, myocardial death region detection is performed on the noise-suppressed cardiovascular imaging data. For example, these parameters can be used as weighting factors in the energy function to drive iterative segmentation using active contour models or level set methods. Gradient sensitivity affects the attractiveness of the contour to edges, smoothness constraint weights control the shape regularity of the contour, and texture similarity provides additional regional information, guiding the contour to converge towards regions with specific texture features.

[0061] This embodiment effectively addresses the limitations of relying solely on a single boundary recognition strategy when processing complex myocardial images by introducing adjustments to gradient sensitivity and smoothness constraint weights, as well as texture similarity calculations for local and micro-texture templates of myocardial tissue. Specifically, once the target boundary recognition strategy is determined, its inherent prior knowledge is used to guide the dynamic adjustment of gradient sensitivity and smoothness constraint weights. For example, if the target strategy tends to detect blurred boundaries, the gradient sensitivity will be increased accordingly; if the strategy requires smoother segmentation results, the smoothness constraint weights will be increased. Simultaneously, by extracting texture features from cardiovascular imaging data and calculating texture similarity, boundary recognition can be aided by deeper tissue structural information. Myocardial death regions are often accompanied by changes in micro-texture, which may not be obvious in grayscale gradients but can be effectively captured through texture similarity. Therefore, combining gradient information, smoothness constraints, and texture information enables the detection algorithm to understand image content more comprehensively and accurately, thereby achieving precise identification of myocardial death regions even under complex pathological conditions.

[0062] To illustrate this technical solution more clearly, a specific example is used below. Suppose we need to detect the myocardial death region in a cardiovascular imaging dataset. In this image, the boundary of the myocardial death region is diffuse and has low grayscale contrast with the surrounding healthy tissue, making it difficult to accurately identify using traditional edge detection methods alone. First, based on the denoising quality level, a target boundary identification strategy is obtained through a boundary identification strategy library, diagnostic priority rules, and conflict resolution. This strategy may indicate the need to identify diffuse boundaries. Based on this target boundary identification strategy, the gradient sensitivity is adjusted to medium-high to ensure that subtle gradient changes can be captured, while the smoothing constraint weight is set appropriately to prevent the contour from overfitting noise.

[0063] Subsequently, local texture features are extracted from the image data, for example, using local binary pattern descriptors to quantify the microstructure of myocardial tissue. The extracted local textures of the myocardial tissue are then compared with pre-established microtexture templates representing typical myocardial death regions. If the texture similarity of a region highly matches the death region template, even if its gradient is not significant, that region is assigned a high probability of being a death region. Finally, combining the adjusted gradient sensitivity, smoothness constraint weights, and the calculated texture similarity, an active contour model based on an energy function is used for iterative segmentation. In this model, texture similarity is used as a weight for the region term, guiding the contour to converge towards regions with texture features similar to the death region template, while gradient sensitivity and smoothness constraint weights control the edge attractiveness and smoothness of the contour, respectively. In this way, even with blurred boundaries and subtle texture variations, myocardial death regions can be accurately detected and segmented.

[0064] Through the above technical solutions, this embodiment can significantly improve the accuracy and robustness of myocardial death region image detection. By dynamically adjusting gradient sensitivity and smoothing constraint weights according to the target boundary recognition strategy, the detection process can better adapt to different image qualities and pathological features, avoiding under-segmentation or over-segmentation problems that may be caused by fixed parameters. In addition, the introduction of similarity calculation between local texture of myocardial tissue and micro-texture template provides important supplementary information for boundary recognition. Especially when the boundary is blurred or the gradient information is not obvious, texture features can effectively distinguish healthy tissue from dead tissue, thereby reducing the misdiagnosis rate and missed diagnosis rate. Therefore, this embodiment can achieve more refined and reliable myocardial death region identification when faced with complex and variable cardiovascular imaging data, providing a more accurate basis for clinical diagnosis.

[0065] In some embodiments, step S202, according to a preset diagnostic priority rule, performs conflict resolution on multiple candidate boundary identification strategies to obtain a conflict resolution result, which may include, but is not limited to, the following steps: Step S301: Obtain the patient's target pathological stage information and disease risk factor information; Step S302: Adjust the preset diagnostic priority rules based on the target pathological stage information and disease risk factor information; Step S303: According to the adjusted preset diagnostic priority rules, assign weights to multiple candidate boundary recognition strategies to obtain the weight assignment results; Step S304: Based on the weight allocation results and the adjusted preset diagnostic priority rules, conflict resolution is performed on multiple candidate boundary recognition strategies to obtain conflict resolution results.

[0066] In some embodiments, the pre-defined diagnostic priority rules may not adequately take into account individual patient differences, such as different pathological stages and disease risk factors, which may lead to inaccurate conflict resolution results and affect the accuracy and clinical applicability of the final myocardial death area image detection.

[0067] To this end, information on the patient's target pathological stage and disease risk factors can be obtained first. Stage-specific data related to the patient's current disease state can be collected, such as the stage of myocardial infarction and the grade of heart failure. This information can be obtained through various means, including medical history records, clinical examination reports, and imaging assessments. Disease risk factors include patient-specific or acquired factors that may affect the identification of myocardial death areas, such as hypertension, diabetes, hyperlipidemia, smoking history, and family history. These factors can be extracted from the patient's electronic medical records, consultation records, or laboratory test results.

[0068] Then, based on the target pathological stage information and disease risk factor information, the pre-defined diagnostic priority rules are adjusted. Existing, general priority rules can be dynamically modified or optimized according to the patient's specific situation. For example, for patients with acute myocardial infarction, a boundary identification strategy that can quickly and accurately identify the infarct core area may be prioritized; while for patients with chronic ischemic heart disease, a strategy that identifies fibrotic or scar tissue may be more emphasized. This adjustment can be achieved through predefined rule sets, machine learning models, or expert systems.

[0069] Then, based on the adjusted preset diagnostic priority rules, weights are assigned to multiple candidate boundary recognition strategies, yielding the weight allocation results. Each candidate boundary recognition strategy can be assigned a numerical value reflecting its importance or applicability in the specific context of the current patient. The weight allocation result is a set of these values. For example, under the adjusted diagnostic priority rules, a certain strategy might be assigned a higher weight because it is more suitable for the current patient's pathological characteristics.

[0070] Finally, based on the weight allocation results and the adjusted preset diagnostic priority rules, conflict resolution is performed on multiple candidate boundary identification strategies to obtain the conflict resolution result. Conflict resolution refers to determining the final and optimal boundary identification strategy or combination thereof based on the adjusted diagnostic priority rules and weight allocation results when there are differences in identification results or priority conflicts among multiple candidate strategies. This can involve various methods such as voting mechanisms, weighted averaging, and fuzzy logic reasoning.

[0071] To illustrate this technical solution more clearly, a specific example is used below. Suppose a patient's cardiovascular imaging data requires myocardial necrosis region detection. First, the system acquires the patient's target pathological stage information, for example, the patient is diagnosed with "subacute myocardial infarction with mild left ventricular dysfunction," and also acquires their disease risk factor information, such as "ten-year history of hypertension and poorly controlled diabetes." Next, the system adjusts the preset diagnostic priority rules based on this patient-specific information. For example, the original rules might treat all patients equally, but after adjustment, the system will increase the priority of boundary recognition strategies that can accurately identify infarct edge regions and assess myocardial viability, while decreasing the priority of strategies that only focus on large areas of necrosis, because the subacute phase requires more refined tissue assessment. Subsequently, the system assigns weights to multiple candidate strategies selected from the boundary recognition strategy library (e.g., region-growing, level-set-based, deep learning-based segmentation, etc.) according to the adjusted diagnostic priority rules. For example, a deep learning model that excels at identifying subacute infarct edges might be assigned a higher weight, while a traditional threshold-based region growing algorithm might be assigned a lower weight. Ultimately, based on these weighted allocations and adjusted priority rules, the system resolves conflicts to arrive at a target boundary identification strategy best suited to the patient's current pathological state and risk factors. For example, the resolution might indicate a combination of a weighted fusion deep learning model and a level set algorithm to balance recognition accuracy and edge smoothness. In this way, the image detection results of the myocardial death region will more closely reflect the patient's actual clinical situation.

[0072] Through the above technical solution, this embodiment can personalize the diagnostic priority rules according to the patient's specific pathological stage and disease risk factors, thereby making the weight allocation and conflict resolution process of the candidate boundary recognition strategy more accurate and intelligent. This embodiment can significantly improve the accuracy and clinical applicability of myocardial death region image detection, provide doctors with more reliable diagnostic evidence, help to develop more individualized treatment plans, and thus improve the patient's treatment outcome.

[0073] In some embodiments, obtaining the patient's target pathological stage information in step S301 may include, but is not limited to, the following steps: Obtain initial pathological stage information for multi-source heterogeneity; The initial pathological stage information of multi-source heterogeneity is standardized in format. Semantic mapping is performed on the initial pathological stage information after format standardization. Based on the confidence level of the data source, the reliability of the initial pathological stage information after semantic mapping is screened. Based on the information timeliness rules, the initial pathological stage information after reliability screening is further screened for validity to obtain the target pathological stage information.

[0074] In some embodiments, initial pathological stage information from multiple sources can be obtained first. Patient pathological stage-related data are collected from various sources, including different medical systems, electronic medical records, imaging reports, and laboratory test results. This data may exist in different formats, structures, and terminology, hence the term "multi-source heterogeneity."

[0075] Then, the initial pathological stage information from multiple heterogeneous sources is standardized to eliminate format differences between different data sources. For example, all date formats can be unified to "YYYY-MM-DD", text descriptions can be converted to predefined codes or classifications, or numerical units can be standardized. The aim is to provide a unified and standardized data foundation for subsequent data processing and analysis.

[0076] Next, semantic mapping is performed on the initial pathological stage information after format standardization. This allows terms or codes from different sources that express the same or similar concepts to be mapped to a unified semantic representation. For example, different hospitals may use different internal codes for the same pathological stage; semantic mapping can unify these codes into internationally recognized standard codes, ensuring a consistent understanding of the pathological stages.

[0077] Based on the confidence level of the data source, the initial pathological stage information after semantic mapping is subjected to reliability screening. The confidence level of the data source can be assessed based on factors such as the authority of the data source, the accuracy of data collection, and the reliability of historical data. For example, diagnostic reports from authoritative medical institutions may have higher confidence levels, while patient self-reported information may have lower confidence levels. Through reliability screening, the influence of low-confidence data can be eliminated or reduced, thereby improving the overall reliability of the information.

[0078] Finally, based on the information timeliness rules, the initial pathological stage information after reliability screening is further filtered for validity to obtain the target pathological stage information. The information timeliness rules aim to ensure that the acquired pathological stage information is currently or recently valid. For example, a time window can be set to retain only pathological information updated or confirmed within a certain past period, avoiding the use of outdated or inaccurate data. Through validity screening, it can be ensured that the final target pathological stage information is the most up-to-date and relevant.

[0079] This embodiment ensures the accuracy, consistency, and timeliness of the acquired patient target pathological stage information. First, by acquiring multi-source, heterogeneous initial pathological stage information, comprehensive patient pathological data can be collected, avoiding information omissions. Second, format standardization and semantic mapping resolve the data heterogeneity issue between different data sources, enabling subsequent data processing to be conducted within a unified framework and guaranteeing data consistency. Furthermore, data source confidence screening and information timeliness rule screening rigorously control information from both reliability and timeliness dimensions, effectively excluding low-quality, outdated, or inaccurate data, thereby ensuring that the final target pathological stage information is of high quality and high reliability. The synergistic effect of these steps allows subsequent adjustments to the preset diagnostic priority rules to be based on accurate and reliable patient pathological stage information, thereby improving the accuracy of the entire myocardial death region image detection method.

[0080] Through the above technical solution, this embodiment overcomes the problems that may be encountered when acquiring patient pathological stage information, such as complex data sources, inconsistent formats, semantic inconsistencies, and difficulties in ensuring information reliability and timeliness. This embodiment, through the integration of multi-source heterogeneous information, format standardization, semantic mapping, data source confidence filtering, and information timeliness filtering, can obtain more comprehensive, accurate, consistent, and timely patient target pathological stage information. This high-quality pathological stage information provides a solid foundation for subsequent adjustments to preset diagnostic priority rules, significantly improving the rationality and accuracy of diagnostic priority rule adjustments. This, in turn, makes the overall process of myocardial death region image detection more precise and reliable, helping doctors make more accurate clinical diagnoses.

[0081] In some embodiments, in step S302, adjusting the preset diagnostic priority rules based on the target pathological stage information and disease risk factor information may include, but is not limited to, the following steps: Obtain information on historical pathological stages, historical risk factors, and historical diagnostic priority rules; Analyze historical pathological stage information, historical risk factor information, and historical diagnostic priority rules to construct a rule mapping table; The target pathological stage information and disease risk factor information are matched with the rule mapping table to obtain the matching results; Based on the matching results, the preset diagnostic priority rules are adjusted.

[0082] In some embodiments, since rule adjustments are made solely based on current pathological stage information and disease risk factor information, historical experience and the evolution of complex cases may not be adequately considered, resulting in the adjusted priority rules lacking sufficient robustness and adaptability, which in turn affects the accuracy of boundary identification strategy conflict adjudication.

[0083] To this end, one can first obtain historical pathological stage information, historical risk factor information, and historical diagnostic prioritization rules. This data, including past patients' pathological stage data, disease risk factor data, and the diagnostic prioritization rules used at the time, can be collected from medical databases, electronic medical record systems, or clinical research literature. This historical data can be stored in structured databases or unstructured documents, with the aim of providing a rich empirical foundation for subsequent rule adjustments.

[0084] Then, information on historical pathological stages, historical risk factors, and historical diagnostic priority rules are analyzed to construct a rule mapping table. Data mining, machine learning, or expert systems can be used to perform in-depth analysis of the collected historical data, identifying the association patterns between different pathological stages, risk factor combinations, and corresponding diagnostic priority rules. For example, algorithms such as decision trees, association rule learning, or neural networks can be used to map specific historical pathological stages and risk factor combinations to corresponding diagnostic priority rule adjustment suggestions, thus forming a queryable and updatable rule mapping table. This rule mapping table aims to transform historical experience into actionable rules adjustment criteria.

[0085] The target pathological stage information and disease risk factor information are then matched with the rule mapping table to obtain the matching results. The current patient's target pathological stage information and disease risk factor information can be used to perform queries or pattern matching within the constructed rule mapping table. For example, based on the current patient's characteristics, the system will search for the most similar or matching historical case patterns in the rule mapping table and extract corresponding rule adjustment suggestions, thus obtaining the matching results. The purpose is to provide personalized and experience-based guidance for adjusting the current patient's diagnostic priority rules.

[0086] Finally, the preset diagnostic priority rules are adjusted based on the matching results. The initial preset diagnostic priority rules can be refined, modified, or have their weights adjusted based on the matching results obtained from the rule mapping table. For example, if the matching results indicate that a certain boundary identification strategy should be given higher priority under specific pathological stages and risk factors, the system will correspondingly increase the weight of that strategy or introduce new rules to optimize the decision-making process. The aim is to ensure that the diagnostic priority rules can dynamically adapt to the specific circumstances of the patient, improving the accuracy and reliability of the diagnosis.

[0087] To illustrate this technical solution more clearly, a specific example is used below. Suppose a patient is diagnosed with early post-myocardial infarction and has risk factors such as hypertension and diabetes. First, the system acquires a large amount of historical pathological stage information (e.g., early, middle, and late post-myocardial infarction), historical risk factor information (e.g., hypertension, diabetes, and hyperlipidemia), and historical diagnostic priority rules used in these historical contexts. Next, this historical data is input into a machine learning model, such as a rule-based expert system or a decision tree model. This model analyzes this historical data, learns, and constructs a rule mapping table. For example, this mapping table might contain rules such as: "If the patient is in the early post-myocardial infarction stage and has hypertension, a boundary recognition strategy based on perfusion defects should be prioritized, and sensitivity to morphological features should be reduced."

[0088] Subsequently, when the patient's target pathological stage information (early post-myocardial infarction) and disease risk factor information (hypertension, diabetes) are input, the system uses this information to match it against a pre-constructed rule mapping table. The matching results may indicate that, for such patients, the priority weight of the "perfusion defect-based boundary recognition strategy" should be increased by 20%, while the priority weight of the "local texture analysis-based boundary recognition strategy" should be increased by 10%, and the "smoothing constraint weight" should be fine-tuned. Finally, based on these matching results, the system dynamically adjusts the preset diagnostic priority rules. For example, during the conflict resolution phase, the system assigns weights and resolves conflicts among multiple candidate boundary recognition strategies according to the adjusted priority rules, thereby obtaining a target boundary recognition strategy that better suits the current patient's specific situation to guide the final image detection of the myocardial death region.

[0089] Through the above technical solution, this embodiment enables more refined and intelligent adjustment of preset diagnostic priority rules. By fully utilizing historical data, the system can learn and adapt to the optimal diagnostic strategy under different pathological stages and risk factors, avoiding the bias or inaccuracy that may result from adjustments based solely on current information. Therefore, the adjusted diagnostic priority rules are more clinically instructive and robust, enabling more accurate selection and fusion of the most suitable boundary recognition strategy for the current patient situation in subsequent boundary recognition strategy conflict adjudication. This significantly improves the accuracy and reliability of myocardial death region image detection, providing stronger technical support for clinical diagnosis.

[0090] The beneficial effects of implementing the embodiments of the present invention include: First, cardiovascular imaging data is acquired. Then, in the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular imaging data. Next, the noise suppression effect is evaluated based on the denoised cardiovascular imaging data, and the denoising strategy is adjusted according to the noise suppression effect to execute the next denoising stage. After the cardiovascular imaging data completes the noise suppression process, a denoising quality level is generated. Finally, based on the denoising quality level, myocardial death region image detection is performed on the noise-suppressed cardiovascular imaging data. Thus, by dividing the data into multiple denoising stages, fine-grained denoising processing can be performed, enhancing the denoising effect and achieving image detection, thereby improving accuracy and reliability.

[0091] like Figure 2 As shown, this embodiment of the invention also provides an image detection system for myocardial death regions based on cardiovascular imaging, comprising: Data acquisition module 401 is used to acquire cardiovascular imaging data; The denoising processing module 402 is used to perform denoising processing on cardiovascular image data using an anisotropic diffusion filter in the current denoising stage. The effect evaluation module 403 is used to evaluate the noise suppression effect based on the denoised cardiovascular imaging data; The strategy adjustment module 404 is used to adjust the denoising processing strategy according to the noise suppression effect. The denoising processing strategy includes iteration number adjustment, filtering method switching and image quality lower limit adjustment, which is used to execute the next denoising stage. The denoising quality level generation module 405 is used to generate a denoising quality level after the cardiovascular imaging data has undergone noise suppression processing. The noise suppression processing includes multiple denoising stages. Image detection module 406 is used to detect the myocardial death region in cardiovascular imaging data after noise suppression based on the noise reduction quality level.

[0092] The content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0093] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

Claims

1. A method for detecting myocardial death regions based on cardiovascular imaging, characterized in that, Includes the following steps: Acquire cardiovascular imaging data; In the current denoising stage, an anisotropic diffusion filter is used to denoise the cardiovascular image data; The noise suppression effect was evaluated based on the denoised cardiovascular imaging data. Based on the noise suppression effect, the denoising strategy is adjusted, which includes adjusting the number of iterations, switching the filtering method, and adjusting the lower limit of image quality, in order to execute the next denoising stage. After the cardiovascular imaging data has undergone noise suppression processing, a denoising quality level is generated. The noise suppression processing includes multiple denoising stages. Based on the denoising quality level, myocardial death region image detection is performed on the cardiovascular imaging data after noise suppression processing.

2. The method according to claim 1, characterized in that, The evaluation of noise suppression effectiveness based on the denoised cardiovascular imaging data includes: Obtain the local variance and peak signal-to-noise ratio of the previous denoising stage; Based on the preset sliding window size and the denoised cardiovascular image data, calculate the local variance of the current denoising stage; The degree of smoothness optimization is calculated based on the local variance of the previous denoising stage and the local variance of the current denoising stage. Calculate the peak signal-to-noise ratio for the current denoising stage based on the denoised cardiovascular imaging data; The degree of signal-to-noise ratio optimization is calculated based on the peak signal-to-noise ratio of the previous denoising stage and the peak signal-to-noise ratio of the current denoising stage. The noise suppression effect is evaluated based on the degree of smoothness optimization and the degree of signal-to-noise ratio optimization.

3. The method according to claim 1, characterized in that, The step of adjusting the noise reduction strategy based on the noise suppression effect includes: Obtain the computation time and computation resource usage pattern of the current denoising stage. The computation resource usage pattern includes the instantaneous utilization rate of the central processing unit core, memory access locality, and data cache hit rate. The computation time of the current denoising stage is matched with the normal operating time reference table to obtain the expected suppression effect; Calculate the suppression effect deviation based on the noise suppression effect and the expected suppression effect; The computing resource usage pattern is matched with the characteristic fluctuation pattern to obtain the target fluctuation pattern, which includes a normal high load pattern and an efficiency degradation pattern. The degree of change in computational efficiency is assessed based on the described suppression effect deviation and the described target fluctuation pattern; The denoising strategy is adjusted based on the degree of change in computational efficiency.

4. The method according to claim 1, characterized in that, The generated denoising quality level includes: Local regional frequency analysis was performed on the noise-suppressed cardiovascular imaging data to identify the type and spatial distribution of residual noise. Edge analysis was performed on the noise-suppressed cardiovascular imaging data to assess the sharpness of local edges. The noise reduction quality level is generated based on the noise suppression effect, the local edge sharpness, and the type and spatial distribution of the residual noise.

5. The method according to claim 1, characterized in that, The step of detecting myocardial death regions in the noise-suppressed cardiovascular imaging data based on the denoising quality level includes: Based on the denoising quality level, select multiple candidate boundary recognition strategies from the boundary recognition strategy library; According to the preset diagnostic priority rules, conflict resolution is performed on the multiple candidate boundary recognition strategies to obtain the conflict resolution result; Based on the conflict resolution result, the multiple candidate boundary identification strategies are fused to obtain the target boundary identification strategy; Based on the target boundary recognition strategy, myocardial death region image detection is performed on the cardiovascular imaging data after noise suppression processing.

6. The method according to claim 5, characterized in that, The step of detecting myocardial death regions in the noise-suppressed cardiovascular imaging data according to the target boundary recognition strategy includes: Adjust the gradient sensitivity and smoothing constraint weights according to the target boundary identification strategy; Texture features were extracted from the noise-suppressed cardiovascular imaging data to obtain the local texture of myocardial tissue. Calculate texture similarity based on the local texture of myocardial tissue and the microtexture template of myocardial tissue; Based on the gradient sensitivity, the smoothing constraint weight, and the texture similarity, myocardial death region detection is performed on the noise-suppressed cardiovascular imaging data.

7. The method according to claim 5, characterized in that, The step of performing conflict resolution on the multiple candidate boundary identification strategies according to preset diagnostic priority rules to obtain conflict resolution results includes: Obtain information on the patient's target pathological stage and disease risk factors; The preset diagnostic priority rules are adjusted based on the target pathological stage information and disease risk factor information. According to the adjusted preset diagnostic priority rules, the multiple candidate boundary recognition strategies are weighted and weighted to obtain the weight allocation results; Based on the weight allocation result and the adjusted preset diagnostic priority rule, conflict resolution is performed on the multiple candidate boundary identification strategies to obtain the conflict resolution result.

8. The method according to claim 7, characterized in that, The acquisition of the patient's target pathological stage information includes: Obtain initial pathological stage information for multi-source heterogeneity; The initial pathological stage information of the multi-source heterogeneity is processed for format standardization. Semantic mapping is performed on the initial pathological stage information after format standardization. Based on the confidence level of the data source, the reliability of the initial pathological stage information after semantic mapping is screened. Based on the information timeliness rules, the initial pathological stage information after reliability screening is subjected to validity screening to obtain the target pathological stage information.

9. The method according to claim 7, characterized in that, The step of adjusting the preset diagnostic priority rule based on the target pathological stage information and disease risk factor information includes: Obtain information on historical pathological stages, historical risk factors, and historical diagnostic priority rules; The historical pathological stage information, the historical risk factor information, and the historical diagnostic priority rules are analyzed to construct a rule mapping table; The target pathological stage information and the disease risk factor information are matched with the rule mapping table to obtain the matching result; Based on the matching results, the preset diagnostic priority rules are adjusted.

10. A system for detecting myocardial death regions based on cardiovascular imaging, characterized in that, include: The data acquisition module is used to acquire cardiovascular imaging data; The denoising module is used to denoise the cardiovascular image data using an anisotropic diffusion filter in the current denoising stage. The effect evaluation module is used to evaluate the noise suppression effect based on the denoised cardiovascular imaging data. The strategy adjustment module is used to adjust the denoising processing strategy according to the noise suppression effect. The denoising processing strategy includes iteration number adjustment, filtering method switching and image quality lower limit adjustment, and is used to execute the next denoising stage. The noise reduction quality level generation module is used to generate a noise reduction quality level after the cardiovascular imaging data has undergone noise suppression processing. The noise suppression processing includes multiple noise reduction stages. The image detection module is used to perform myocardial death region image detection on the noise-suppressed cardiovascular imaging data according to the denoising quality level.

Images