Medical image automatic analysis system and method

Abstract

The invention relates to the technical field of medical image processing, in particular to a medical image automatic analysis system and method, and the system comprises a time sequence dynamic monitoring module, a lesion area weighted fusion module, a boundary energy optimization module, a multi-channel filtering processing module, and a time sequence feature construction module. According to the method, time sequence tracking analysis of medical images is realized through time sequence dynamic monitoring, change salient region screening and precise lesion evolution trend positioning, lesion data at different moments are integrated by means of weighted fusion processing, the accuracy of time sequence analysis is improved, data fluctuation errors are reduced, lesion boundaries are optimized in combination with gradient analysis and gray scale comparison, and the accuracy of medical image analysis is improved. Noise interference is eliminated, lesion contours are clearer, different frequency features of images are screened through multichannel filtering, lesion differentiation information is highlighted, irrelevant interference is reduced, feature expression is enhanced, a complete lesion time sequence analysis framework is constructed, and a comprehensive image analysis basis is provided for precise medical treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a medical image automatic analysis system and method. Background Art

[0002] The field of medical image processing technology includes the use of computer technology to automatically process, analyze and interpret medical images. The core content of this field involves the application of technical means such as recognition, enhancement, segmentation, registration, classification, and diagnosis of medical images. Medical image processing can help doctors detect diseases, provide diagnostic support, and plan treatment. With the rapid development of computer vision and artificial intelligence technologies, medical image processing has been widely used in the field of image diagnosis, especially in automated and intelligent analysis. Significant progress has been made. This field includes image processing algorithms, machine learning models, deep learning technologies, etc., aiming to improve the accuracy and efficiency of medical images and their scope of application.

[0003] Among them, the automatic medical image analysis system refers to a system that uses computer-aided analysis to automatically identify and process medical images. The system mainly focuses on the analysis and interpretation of data in medical images. Specifically, it uses image processing technology to realize functions such as preprocessing, feature extraction and classification of medical images. The system includes image denoising, contrast enhancement, target area segmentation and other processes, and then marks and analyzes the diseased areas in the image. The patent subject completes image recognition and analysis in an automated manner, thereby supporting the auxiliary diagnosis function of medical images, reducing manual intervention and improving analysis accuracy.

[0004] Although the existing technology can realize automatic identification and analysis, it has limitations in the dynamic monitoring of the lesion area. It mainly relies on image analysis at a single time point, lacks continuous tracking of the evolution trend of the lesion, and is difficult to accurately evaluate the development trend of the lesion. In the process of lesion boundary identification, it is easily affected by image noise and uneven grayscale, resulting in blurred boundaries, affecting the precise positioning of the lesion area and reducing the reliability of automatic analysis. In the feature extraction process, the existing image processing method relies on a fixed parameter filtering method, which is difficult to take into account the specificity of different image data, resulting in some key features that cannot be effectively highlighted, affecting the accuracy of image analysis. The identification of the lesion area mainly relies on static images, and fails to make full use of time series image data for multi-dimensional analysis, resulting in difficulty in comprehensively evaluating the dynamic evolution characteristics of the lesion during the diagnosis process, affecting the accurate judgment of the disease process. In the process of automated interpretation of medical images, the existing technology has not yet established an effective time series feature analysis framework, and cannot fully reflect the development trend of the lesion, which limits the clinical guidance value of image analysis. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a medical image automatic analysis system and method.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A medical image automatic analysis system comprises: The time series dynamic monitoring module extracts the volume, shape, and grayscale value features of the lesion area based on the medical imaging data at different time points, calculates the change rate of the lesion area between adjacent time points, screens the image sequences with prominent change rates, determines the time series change features in the medical imaging data, and obtains the dynamic change trend of the lesion; The lesion area weighted fusion module extracts the lesion change rate at each time point based on the lesion dynamic change trend, calculates the weight value of the lesion change rate in the image time series, screens the time series image data, and obtains weighted fusion lesion area data; The boundary energy optimization module identifies the grayscale contrast value between the lesion area and the surrounding tissue based on the weighted fusion lesion area data, analyzes the energy distribution of the lesion area boundary, screens the regional boundary features affected by noise in the medical imaging data, and obtains the optimized lesion area boundary; The multi-channel filtering processing module extracts the frequency distribution information in the medical image data based on the optimized lesion area boundary, screens the lesion area features in the differentiated frequency range, performs filtering processing on the medical image, and obtains the multi-channel filtered lesion features.

[0007] As a further solution of the present invention, the dynamic change trend of the lesion includes a volume change trend, a shape change trend, and a grayscale value change trend; the weighted fused lesion area data includes weighted volume data, weighted shape data, and weighted grayscale data; the optimized lesion area boundary includes a denoised boundary, an enhanced contrast boundary, and a smooth boundary; the multi-channel filtered lesion features include high-frequency features, low-frequency features, and medium-frequency features.

[0008] As a further solution of the present invention, the time series dynamic monitoring module includes: The lesion region feature recognition submodule extracts the segmentation information of the lesion region based on the medical imaging data at different time points, identifies the volume, shape, and grayscale value features of the lesion region, and counts the pixel distribution parameters of the lesion region at each time point to obtain multi-dimensional feature data of the lesion region; The change rate calculation submodule calls the multi-dimensional feature data of the lesion area, extracts the feature change values ​​of the lesion area at adjacent time points, and uses the formula: ; The change rate of the lesion area is calculated; in, Represents the rate of change of the lesion area, Representing time point The volume of the lesion area, Representing time point The volume of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point Gray value characteristics of the lesion area, Representing time point Gray value characteristics of the lesion area; The dynamic change pattern screening submodule calls the change rate of the lesion area, screens the time points with prominent change rates, extracts the time series of the lesion area with prominent changes, compares the trends of the lesion areas at differentiated time points, and obtains the dynamic change trend of the lesion.

[0009] As a further solution of the present invention, the lesion area weighted fusion module includes: The lesion rate calculation submodule extracts the lesion area volume and edge contour change data in the image time series based on the dynamic change trend of the lesion, identifies the lesion change rate at each time point, and adjusts the rate in combination with the lesion evolution trend, using the formula: ; The rate of change of the lesion was calculated; in, Representative The rate of change of the lesion at each time point, Representative The volume of the lesion area at each time point, Representative The volume of the lesion area at each time point, Representative The real time corresponding to the time point, Representative The real time corresponding to each time point, Representative Time point The lesion area with edge contour points, No. Time point The lesion area with edge contour points, Represents the total number of lesion edge contour points, represents the lesion morphology adjustment factor; The image data screening submodule calls the lesion change rate, screens the image time series data, removes the image data with a change rate lower than a threshold, and obtains the screened image time series data; The lesion region fusion submodule identifies the weight of the lesion change rate based on the screened image time series data, performs weighted fusion on the lesion region according to the weight, and obtains weighted fused lesion region data.

[0010] As a further solution of the present invention, the boundary energy optimization module includes: The boundary gradient extraction submodule calculates the grayscale gradient in the image data based on the weighted fusion lesion area data, identifies the boundary gradient change rate, analyzes the local gradient change of the pixel grayscale value in the area, screens the boundary gradient mutation point, and obtains the boundary gradient distribution data; The lesion area comparison analysis submodule calls the boundary gradient distribution data to compare the grayscale of the lesion area with the surrounding tissues using the formula: ; Calculate the grayscale contrast index of the lesion area boundary, filter the boundary area, and obtain the contrast characteristics of the lesion area; in, The grayscale contrast index representing the boundary of the lesion area, Represents the first The gray value of a pixel, Representing the lesion area The mean grayscale value of the inner pixels, Represents the surrounding tissue area The mean grayscale value of the inner pixels, Represents the number of pixels in the boundary gradient area; The noise impact screening submodule calls the lesion area contrast feature, analyzes the noise impact area, identifies the energy distribution of the lesion area boundary, screens the abnormal area in the energy distribution, removes the noise impact boundary, and obtains the optimized lesion area boundary.

[0011] As a further solution of the present invention, the multi-channel filtering processing module includes: The spatial frequency analysis submodule extracts the pixel grayscale value based on the optimized lesion area boundary, analyzes the grayscale change trend of the differentiated position, extracts the spatial frequency information, and determines the frequency distribution ratio using the formula: ; Calculate the frequency distribution weight value of the pixel point to obtain the frequency distribution information; in, Represents the frequency distribution weight value of the pixel, Represents pixel The gray value of Represents the global grayscale mean of the image, Represents the total number of pixels in the image. Represents pixel The local weight coefficient of Represents the index of the pixel; The lesion region identification submodule calls the frequency distribution information, compares the frequency characteristics of the image region, screens the frequency abnormality region, identifies the frequency deviation, and obtains the frequency characteristics of the lesion region; The filtering feature recognition submodule calls the frequency features of the lesion area, adjusts the weights of the image frequency components, uses filtering processing, strengthens and weakens the frequency information of the key area, and obtains multi-channel filtering lesion features.

[0012] As a further solution of the present invention, the system also includes a time series feature construction module: The time series feature construction module identifies the lesion area information at different time points in the medical image based on the multi-channel filtered lesion features, extracts the key lesion features in the image sequence, and obtains the lesion time series image analysis data; The lesion time series image analysis data includes key lesion information, change patterns, and time series feature distribution.

[0013] As a further solution of the present invention, the timing feature construction module includes: The filtered lesion signal calculation submodule analyzes the signal changes in the lesion area based on the multi-channel filtered lesion characteristics, calculates the image gradient value, screens the lesion response channel, and obtains the lesion signal gradient value; The lesion time point determination submodule identifies the gradient change rate of the time series image based on the lesion signal gradient value, analyzes the time increment of the lesion area, and obtains the lesion time series change amount; The lesion feature extraction submodule identifies the morphological characteristics of the lesion area based on the lesion time series change, analyzes the boundary changes at the time point, and uses the formula: ; Calculate the characteristic change value of the lesion area and obtain the time series image analysis data of the lesion; in, Represents the characteristic change value of the lesion area, Represents the gradient change of adjacent time points in the lesion area, Represents the pixel offset of the lesion area at adjacent time points, represents the rate of change of the lesion feature vector at adjacent time points, Represents the change in the lesion boundary between adjacent time points.

[0014] The automatic medical image analysis method is performed based on the automatic medical image analysis system, and includes the following steps: S1: Based on the medical imaging data at different time points, extract the lesion area information in the image, calculate the volume, morphology and grayscale value of the lesion area, arrange the data in chronological order, compare the lesion area characteristics at adjacent time points, identify the lesion change rate, and screen the image sequence with the change rate; S2: based on the image sequence with significant change rate, extract the change rate of the lesion at each time point, calculate the rate weight of the time point, select the image data at the optimal time point, combine the lesion areas at multiple time points, and obtain weighted fusion lesion area data; S3: Based on the weighted fusion lesion area data, extract the boundary gradient information in the image, calculate the grayscale contrast value between the lesion area and the surrounding tissue, screen the boundary points of the grayscale contrast, analyze the distribution of the boundary points in the image, analyze the energy distribution of the region boundary, eliminate the low-energy boundary points affected by noise, and obtain the optimized lesion area boundary; S4: extracting frequency distribution information in the image based on the optimized lesion area boundary, screening lesion area features in a differentiated frequency range, filtering the image data, and obtaining multi-channel filtered lesion features; S5: Based on the multi-channel filtered lesion features, compare the lesion area of ​​the differentiated time series images, identify the key morphological features of the lesion area, extract the key point information of the lesion area at the differentiated time points, and obtain the lesion time series image analysis data.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through the dynamic monitoring of time series, the characteristic information such as volume, morphology, grayscale, etc. of the lesion area is extracted, the rate of change between adjacent images is calculated according to the time axis sequence, the area with significant change amplitude is screened, the dynamic evolution trend of the lesion is accurately located, and the time-series tracking analysis of medical images is realized. With the help of weighted fusion processing, the lesion data at different times are comprehensively processed according to the weight calculation of the lesion change rate, the accuracy of image time series analysis is improved, the error caused by data fluctuation is reduced, and the boundary information of the lesion area is optimized by gradient analysis and grayscale contrast means, the noise interference is eliminated, the edge recognition ability of the lesion area is improved, and the lesion morphology is clearer. The different frequency features in the image are screened by multi-channel filtering technology, the differentiated information of the lesion area is highlighted, the irrelevant interference in the image is effectively reduced, the expression ability of the target lesion characteristics is enhanced, the key features are extracted from the image time series data, a complete lesion time series image analysis framework is constructed, the automatic interpretation ability of medical images is improved, the evolution trend of the lesion area is made more intuitive, and a more comprehensive image analysis basis is provided for precision medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the time series dynamic monitoring module in the present invention; Figure 3 This is a flow chart of the weighted fusion module of the lesion area in the present invention; Figure 4 This is a flow chart of the boundary energy optimization module in the present invention; Figure 5 This is a flow chart of the multi-channel filtering processing module in the present invention; Figure 6 This is a flow chart of the timing feature construction module in the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0019] See also Figure 1 The present invention provides a technical solution: a medical image automatic analysis system comprising: The time series dynamic monitoring module extracts the volume, shape, and grayscale value features of the lesion area based on the medical imaging data at different time points, arranges the data in chronological order, calculates the change rate of the lesion area between adjacent time points, screens the image sequences with prominent change rates, compares the dynamic change patterns of the lesion area, determines the time series change features in the medical imaging data, and obtains the dynamic change trend of the lesion; The lesion area weighted fusion module extracts the lesion change rate at each time point based on the dynamic change trend of the lesion, calculates the weight value of the lesion change rate in the image time series, screens the time series image data, and obtains the weighted fusion lesion area data; The boundary energy optimization module extracts boundary gradient information from medical imaging data based on weighted fusion of lesion area data, identifies the grayscale contrast value between the lesion area and surrounding tissues, analyzes the energy distribution of the lesion area boundary, screens the boundary features of the area affected by noise in the medical imaging data, and obtains the optimized lesion area boundary; The multi-channel filtering processing module extracts the frequency distribution information in the medical image data based on the optimized lesion area boundary, screens the lesion area features in the differentiated frequency range, performs filtering processing on the medical image, and obtains the multi-channel filtering lesion features; The time series feature construction module is based on multi-channel filtering of lesion features, identifies lesion area information at differentiated time points in medical images, extracts key lesion features in image sequences, and obtains lesion time series image analysis data.

[0020] The dynamic change trends of lesions include volume change trends, shape change trends, and gray value change trends. The weighted fusion lesion area data include weighted volume data, weighted shape data, and weighted gray value data. The optimized lesion area boundaries include denoised boundaries, enhanced contrast boundaries, and smooth boundaries. The multi-channel filtered lesion features include high-frequency features, low-frequency features, and medium-frequency features. The lesion time series image analysis data includes key lesion information, change patterns, and time series feature distribution.

[0021] See also Figure 2 , the time series dynamic monitoring module includes: The lesion region feature recognition submodule extracts the segmentation information of the lesion region based on the medical imaging data at different time points, identifies the volume, shape, and grayscale value features of the lesion region, and counts the pixel distribution parameters of the lesion region at each time point to obtain multi-dimensional feature data of the lesion region; Medical imaging data at different time points are collected by high-precision scanners. For example, in cranial MRI detection, the brain tumor lesion area is scanned at continuous time points, and serial images are collected. The lesion area is segmented from the image at each time point and its boundary is identified through image processing software and edge detection algorithms. This process involves the use of pixel-level comparison and grayscale threshold segmentation technology to ensure accurate segmentation. The volume and shape of the segmented lesion area are quantitatively measured. The volume is obtained by converting the number of pixels to the known scan volume ratio, and the shape characteristics are described by calculating geometric parameters such as the curvature radius of the boundary line. The gray value feature is calculated by counting the grayscale average value of all pixels in the lesion area and the standard deviation of its distribution. The data is compared with the patient's historical imaging data to analyze the change trend of the lesion area over time, and finally obtain multi-dimensional feature data of the lesion area.

[0022] The change rate calculation submodule calls the multi-dimensional feature data of the lesion area and extracts the feature change values ​​of the lesion area at adjacent time points using the formula: ; The change rate of the lesion area is calculated; in, Represents the rate of change of the lesion area, Representing time point The volume of the lesion area, Representing time point The volume of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point Gray value characteristics of the lesion area, Representing time point Gray value characteristics of the lesion area; The volume, shape, and grayscale values ​​at consecutive time points are calculated and compared to obtain the rate of change of the lesion area. For example, suppose a patient has ) in an MRI scan, the volume of the lesion area is , the shape feature (such as the comprehensive value of boundary curvature) is The average gray value is , at the second time point ( ) in an MRI scan, the volume of the lesion area grew to , the shape characteristics change to , the average gray value changes to ; Calculate the rate of change of the lesion area and substitute the specific value into the formula for calculation: ; The calculated change rate of the lesion area is 1.50, indicating that the volume of the lesion area has increased significantly during this time interval. Considering the changes in shape characteristics and grayscale values, the change rate exceeds the threshold of 1.0. Therefore, this time point can be marked as a time point when the lesion changes significantly, and its clinical significance needs to be further analyzed. The innovation of the formula for calculating the change rate of the lesion area lies in that it avoids the limitations of a single volume calculation by integrating volume changes, shape feature changes and grayscale value changes. It can more accurately reflect the dynamic change trend of the lesion area and improve the sensitivity to the development of the lesion. The calculation result can be directly used to determine whether the lesion area has entered the abnormal growth stage, helping doctors decide whether further intervention or adjustment of the treatment plan is needed.

[0023] The dynamic change pattern screening submodule calls the change rate of the lesion area, screens the time points with prominent change rates, extracts the time series of the lesion area with prominent changes, compares the trends of the lesion area at different time points, and obtains the dynamic change trend of the lesion; Threshold screening technology is used to identify time points with significant change rates. For example, for periodic CT examinations of cancer patients, by comparing the change rates of consecutive examinations, time points with abnormal rates are screened out. The screening of time points is based on a preset clinical significance threshold. If the rate exceeds 0.2, it is considered clinically significant and requires immediate treatment or further examination. During the screening process, the data at each time point is statistically analyzed and compared with the rate difference at the previous time point to identify the dynamic change pattern of the lesion area. Comparative analysis of data at different time points can reveal the response pattern of the lesion area as the treatment progresses, as well as real-time feedback on the treatment effect, and ultimately obtain the dynamic change trend of the lesion.

[0024] See also Figure 3 , the lesion area weighted fusion module includes: The lesion rate calculation submodule extracts the lesion area volume and edge contour change data in the image time series based on the dynamic change trend of the lesion, identifies the lesion change rate at each time point, and adjusts the rate based on the lesion evolution trend. The formula is: ; The rate of change of the lesion was calculated; in, Representative The rate of change of the lesion at each time point, Representative The volume of the lesion area at each time point, Representative The volume of the lesion area at each time point, Representative The real time corresponding to each time point, Representative The real time corresponding to each time point, Representative Time point The lesion area with edge contour points, No. Time point The lesion area with edge contour points, Represents the total number of lesion edge contour points, represents the lesion morphology adjustment factor; The calculation logic of the "lesion change rate" is completed in the time series dynamic monitoring module. First, the system extracts the volume, shape, grayscale value and other features of the lesion area based on the medical imaging data at different times, and calculates the change rate of the lesion area by comparing the lesion area features between adjacent time points. Specifically, the lesion change rate is evaluated by comprehensively evaluating the volume change, shape feature change and grayscale value change of the lesion area. In the calculation formula, the lesion change rate is calculated by the difference in the lesion volume, shape parameters and grayscale features at each time point, thereby obtaining the change of the lesion in different time periods. This multi-dimensional calculation method can avoid the errors that may be caused by a single feature, thereby more accurately reflecting the dynamic evolution trend of the lesion; First, the volume of the lesion area at each time point is obtained. The data is obtained by MRI or CT scan and quantified into volume values ​​by related software. Assume that the MRI scan of a patient shows that the volume of the lesion area is 200 at the first time point and 210 at the second time point, and the time interval is 24 hours (1 day). Then calculate the rate of change of the lesion, using the difference in the change of the volume of the lesion area between adjacent time points, divided by the time difference. Represents the time interval between two adjacent time points. , ; Substituting into numerical calculation, first calculate the first part, that is, the rate of change of lesion volume: ; Secondly, consider the change of the edge contour of the lesion. Assume that there are 10 edge points and each point moves an average of is 0.2, 0.18, 0.22, 0.19, 0.21, 0.20, 0.23, 0.21, 0.19, 0.20, then the second term is calculated as follows: ; Assumed lesion morphology adjustment factor If it is set to 0.5 (adjusted according to the lesion type and tissue morphology), the final calculated lesion change rate is: ; This value indicates that at the second time point, the volume growth rate of the lesion area plus the change in the edge contour made the overall lesion change rate reach 10.102, which better reflects the dynamic evolution trend of the lesion than calculating the volume change rate of 10 alone. This calculation method ensures that the morphological changes of the lesion are included in the analysis, making the data more comprehensive and accurate.

[0025] The image data screening submodule calls the lesion change rate, screens the image time series data, removes the image data with a change rate lower than a threshold, and obtains the screened image time series data; A threshold for the lesion change rate is set, for example, the threshold is set to 0.5, which means that all time points with lesion change rates lower than this value will be regarded as slow changes and will not be included in the key analysis sequence. The setting of this threshold is based on the analysis of historical case data. By statistically analyzing the distribution of lesion change rates, the rate threshold that can reflect significant lesion progression is determined. Then, the lesion change rates at each time point are compared, and the time points with change rates higher than the threshold are screened out. The screened out time point data will be used for further analysis. For example, in a one-month periodic liver scan, the three time points with the highest change rates are selected, and the corresponding image data will be used for the next step of weighted fusion analysis to ensure that only the most representative data are used for the final lesion analysis, thereby improving the accuracy and efficiency of the analysis and obtaining the screened image time series data.

[0026] The lesion region fusion submodule identifies the weight of the lesion change rate based on the screened image time series data, performs weighted fusion on the lesion region according to the weight, and obtains weighted fusion lesion region data; The weight value of the lesion change rate at each time point in the image time series is calculated. The weight calculation is based on the proportion of the lesion change rate to the total rate change. Assuming that the change rate at a certain time point accounts for 10% of the total rate, its weight is also 10%. This weighting method ensures that the focus of the analysis is on the time point with the most significant change. The lesion area is weighted and fused according to the weight. For example, if the weight of a certain time point is 30%, the influence of the lesion image at that time point in the fused image is 30%. In this way, obtaining weighted fused lesion area data can more realistically reflect the overall development process of the lesion and provide important visual basis for the diagnosis of the lesion and the formulation of treatment plans.

[0027] See also Figure 4 , the boundary energy optimization module includes: The boundary gradient extraction submodule calculates the grayscale gradient in the image data based on the weighted fusion lesion area data, identifies the boundary gradient change rate, analyzes the local gradient change of the pixel grayscale value in the area, screens the boundary gradient mutation points, and obtains the boundary gradient distribution data; Select the lesion area from the medical image and calculate the gray value change of each pixel in the area relative to its surrounding pixels. Use image processing technology to identify the rough boundary of the lesion area in the image, and then perform pixel-by-pixel analysis on the gray value of the boundary points. For example, in a 128*128 image, select a 10*10 area in the boundary area, calculate the gray value difference between each pixel and its adjacent pixels, in this way, you can get the gradient value of each point on the boundary, and then select the points with large gradient mutation as significant boundary feature points, the data will be used for subsequent lesion area identification and analysis. For example, if you find that the gradient value of a certain area suddenly jumps from 10 to 50 when processing brain CT images, this indicates that there is a lesion or important feature at the boundary, and obtain the boundary gradient distribution data, which can be directly used for the next step of lesion area comparison analysis.

[0028] The lesion area comparison analysis submodule calls the boundary gradient distribution data to compare the grayscale of the lesion area with the surrounding tissue using the formula: ; Calculate the grayscale contrast index of the lesion area boundary, filter the boundary area, and obtain the contrast characteristics of the lesion area; in, The grayscale contrast index representing the boundary of the lesion area, Represents the first The gray value of a pixel, Representing the lesion area The mean grayscale value of the inner pixels, Represents the surrounding tissue area The mean grayscale value of the inner pixels, Represents the number of pixels in the boundary gradient area; The calculation logic of the grayscale contrast index of the lesion area boundary is based on the evaluation of the grayscale difference between the lesion area and the surrounding tissue. The specific process is to first calculate the average grayscale value of the lesion area and the surrounding normal tissue area, and calculate the contrast through the difference in grayscale values. In the formula, the grayscale contrast index is calculated by comparing the grayscale values ​​of the lesion area and the surrounding tissue area to measure the degree of grayscale difference between the lesion area and the surrounding tissue. This calculation method involves the comparison of the grayscale value of each pixel in the lesion area with the grayscale mean of the surrounding tissue pixels, and the result is used to determine whether the boundary of the lesion area is obvious. In this calculation, the units of the grayscale values ​​in the formula mentioned in the manual have not been completely unified, resulting in mismatched units, which may affect the accuracy of the calculation results; First, determine the grayscale contrast between the lesion area and the surrounding tissue. This process requires calculating the average grayscale value of the lesion area and its adjacent normal tissue, and measuring the contrast intensity through a standardized method. Assuming a 512×512 pixel medical image, the pixel set size of the lesion area is m=400, and the pixel set size of the surrounding normal tissue is n=1000. Obtain the grayscale values ​​of all pixels in the lesion area and calculate their mean: ; Assume that the average gray value of the lesion area is calculated as , perform the same calculation on the pixels in the surrounding normal tissue area: ; Assume that the average gray value of the surrounding tissue is calculated to be , calculate the grayscale contrast index between the lesion area and the surrounding tissue ; For actual calculation, we select some pixel values ​​in the lesion area. For example, the grayscale values ​​of the pixels in the lesion area are 170, 190, 175, 185, and 200, respectively. Then we calculate the numerator: ; Calculate the denominator: ; Take the square root: ; Finally, the grayscale contrast index is calculated: ; This value indicates the grayscale difference between the lesion area and the surrounding tissue. The larger the value, the more obvious the difference between the lesion area and the surrounding tissue, and the easier it is to identify. If the contrast threshold is set to 0.3, then If the threshold is exceeded, the region is judged to have significant pathological features. If it is lower than the threshold, it belongs to the noise area or the area with unclear boundaries. The calculation is finally used to generate the contrast features of the lesion area, which can be used for further lesion area segmentation and feature extraction to assist doctors in making accurate diagnosis.

[0029] The noise impact screening submodule calls the lesion area comparison features, analyzes the noise impact area, identifies the energy distribution of the lesion area boundary, screens the abnormal area in the energy distribution, removes the noise impact boundary, and obtains the optimized lesion area boundary; The energy distribution of the boundary of the lesion area is calculated and analyzed to identify irregular noise points introduced by equipment noise or image processing algorithms. By setting a noise threshold, such as setting the noise contrast threshold to 0.5, if the energy value of a certain area is lower than this threshold, it is regarded as a noise-affected area and is eliminated. The energy distribution is calculated by weighted averaging the grayscale values ​​of the identified boundary points to obtain a global energy distribution map. This map reflects the visual prominence of each area. For brain scan images, this can help identify the error area caused by noise, thereby accurately depicting the true boundary of the lesion area and obtaining the optimized lesion area boundary. This is a key result data in the entire diagnostic process and is of decisive significance for further pathological analysis and subsequent treatment plans.

[0030] See also Figure 5 , the multi-channel filtering processing module includes: The spatial frequency analysis submodule extracts the pixel grayscale value based on the optimized lesion area boundary, analyzes the grayscale change trend of the differentiated position, extracts the spatial frequency information, and determines the frequency distribution ratio using the formula: ; Calculate the frequency distribution weight value of the pixel point to obtain the frequency distribution information; in, Represents the frequency distribution weight value of the pixel, Represents pixel The gray value of Represents the global grayscale mean of the image, Represents the total number of pixels in the image. Represents pixel The local weight coefficient of Represents the index of the pixel; Extracting pixel grayscale values ​​from image data involves digitizing the original image to extract grayscale information. In practical applications, such as cardiac ultrasound image analysis, the grayscale value of each pixel represents the intensity of the echo signal. Doctors can use the data to determine the density changes of myocardial tissue. For a cardiac ultrasound image, the total number of pixels is set to , the global grayscale mean is , select the pixels in a certain area for analysis. For example, set the grayscale values ​​of the five pixels in the area to be , , , , , and set the local weight coefficients of the pixels to be , , , , ; Step 1: Calculate the grayscale deviation value of each pixel; ; ; ; ; ; Step 2: Calculate the denominator (the square root of the sum of squares of gray value deviations); ; ; Step 3: Calculate the normalized frequency value and multiply it by the corresponding weight ; ; ; ; ; ; Finally calculate the frequency distribution weight value ; This value represents the frequency distribution weight of the area. Doctors can compare the value with the area to determine whether there is an abnormal grayscale change in the area. If the value deviates from the normal myocardial tissue range (such as the normal tissue range is within arrive If the myocardial morphology is too small, it indicates that there is a lesion in this area, and further analysis of the morphological characteristics of the myocardium is needed to more accurately locate the lesion area.

[0031] The lesion region identification submodule calls the frequency distribution information, compares the frequency characteristics of the image region, screens the frequency abnormality region, identifies the frequency deviation, and obtains the frequency characteristics of the lesion region; Detailed comparison of different areas in the image mainly involves the use of advanced image processing technology to analyze the frequency characteristics of each area. Taking cardiac ultrasound as an example, the diseased areas with abnormal frequencies can be screened out by comparing the frequency characteristics of different cardiac areas. For the calculation of frequency deviation values, doctors can set specific thresholds to determine which areas have frequency deviations beyond the normal range. The thresholds are set based on a large amount of clinical data and expert experience. The specific setting process involves statistical analysis of the frequency distribution of past cases, and on this basis, reasonable judgment criteria are formulated. In this way, the diseased areas in cardiac ultrasound can be effectively identified, and the specific location and range of the diseased areas can be further determined, thereby providing decision support for subsequent treatment and obtaining the frequency characteristics of the diseased areas.

[0032] The filter feature recognition submodule calls the frequency characteristics of the lesion area, adjusts the weight of the image frequency component, uses filtering processing, strengthens and weakens the frequency information of the key area, and obtains the multi-channel filtered lesion characteristics; Adjust and apply filtering processing technology. The processing process mainly involves selecting appropriate filters to strengthen or weaken the frequency information of specific areas. For the actual example of cardiac ultrasound images, if the frequency characteristics of a certain area indicate the presence of potential lesions, the diagnostic effect of the image can be enhanced by adjusting the weights of the frequency components of the area. The selection and parameter setting of the filter are based on the specific pathological characteristics and the expected image effect. The specific setting process includes analyzing the frequency characteristics of the lesion area, selecting the filter type that can maximize the characteristics, and then adjusting the filter strength according to the severity of the lesion. Such processing not only makes the lesion area more obvious in the image, but also facilitates doctors to make more accurate assessments. Finally, the multi-channel filtered lesion characteristics are obtained, which will directly affect the accuracy of the diagnosis results and the formulation of treatment plans. This provides doctors with a more accurate and personalized diagnostic tool.

[0033] See also Figure 6 , the temporal feature building blocks include: The filtered lesion signal calculation submodule analyzes the signal changes in the lesion area based on the multi-channel filtered lesion characteristics, calculates the image gradient value, screens the lesion response channel, and obtains the lesion signal gradient value; The multi-channel filtering results are called to calculate the lesion signal. First, the image data is obtained and filtered in different channels to calculate the lesion signal changes in each channel. For each channel, its image gradient value is analyzed, which involves differential operation on the grayscale value of each pixel to find the area with the largest grayscale change. Then the data is compared to select the channel with the strongest lesion reaction. For example, for a hypothetical lung CT scan, the lesion areas under different channels can be compared to find the morphological changes of lung nodules. This involves determining the boundary between the lesion area and the non-lesion area through the difference in grayscale values. Finally, the signal strength of the channel is calculated and analyzed for its correlation with the gradient change. This process includes using statistical software to perform regression analysis on the signal strength and the gradient change to determine whether there is a significant correlation, thereby obtaining the lesion signal gradient value.

[0034] The lesion time point determination submodule identifies the gradient change rate of the time series image based on the lesion signal gradient value, analyzes the time increment of the lesion area, and obtains the lesion time series change amount; First, the lesion gradient change rate of each frame image in the time series is calculated, which includes comparing the grayscale values ​​of the lesion area between consecutive frames and calculating its change rate. Then, the time increment of the lesion gradient change between image frames is analyzed, that is, the time interval of the grayscale gradient change of the lesion area from one frame to the next frame is calculated, and the change thresholds between different frames are compared. The time points with significant changes are screened out according to the set grayscale change thresholds. The time points are determined by calculation. For example, in liver MRI scans, the development stage of cirrhosis lesions can be calibrated. Finally, the time points with significant changes in lesion gradients are identified, and the change amplitude between time points is calculated. This can be achieved by comparing the grayscale statistical distribution of the lesion area at each time point to obtain the change in the lesion time series.

[0035] The lesion feature extraction submodule identifies the morphological characteristics of the lesion area based on the change in the lesion time series, analyzes the boundary changes at time points, and uses the formula: ; Calculate the characteristic change value of the lesion area and obtain the time series image analysis data of the lesion; in, Represents the characteristic change value of the lesion area, Represents the gradient change of adjacent time points in the lesion area, Represents the pixel offset of the lesion area at adjacent time points, represents the rate of change of the lesion feature vector at adjacent time points, represents the amount of change in the lesion boundary between adjacent time points; Starting from the change in the lesion time series, the morphological characteristics of the lesion area over time are calculated, which involves performing morphological analysis on the lesion area at each time point, such as using an edge detection algorithm to determine the boundary change of the lesion area and calculating the pixel distribution offset. For a specific example, assuming that in consecutive MRI scan frames, the pixel position of a tumor moves from (100, 200) in the first frame to (110, 210) in the second frame, which means a lateral offset of 10 pixels and a longitudinal offset of 10 pixels. By establishing a lesion feature vector matrix, which is composed of the lesion feature vectors at each time point, the change trend of the lesion morphology is represented; Assume the specific values ​​are: , , , , , , , ; Substitute into the formula to calculate: ; The intensity of morphological changes in the lesion area between consecutive time points can be calculated. This value (2.62) represents the degree of characteristic change in the lesion area from one time point to another. It can be used to evaluate the treatment effect or lesion progression and obtain lesion time series imaging analysis data.

[0036] The automatic medical image analysis method is performed based on the automatic medical image analysis system, and includes the following steps: S1: Based on the medical imaging data at different time points, extract the lesion area information in the image, calculate the volume, morphology and grayscale value of the lesion area, arrange the data in chronological order, compare the lesion area characteristics at adjacent time points, identify the lesion change rate, and screen the image sequence with the change rate; S2: Based on the image sequence with significant change rate, the change rate of the lesion at each time point is extracted, the rate weight of the time point is calculated, the image data at the optimal time point is selected, and the lesion areas at multiple time points are combined to obtain the weighted fusion lesion area data; S3: Based on the weighted fusion of the lesion area data, the boundary gradient information in the image is extracted, the grayscale contrast value between the lesion area and the surrounding tissue is calculated, the boundary points of the grayscale contrast are screened, the distribution of the boundary points in the image is analyzed, the energy distribution of the regional boundary is analyzed, the low-energy boundary points affected by noise are eliminated, and the optimized lesion area boundary is obtained; S4: Based on the optimized lesion area boundary, extract the frequency distribution information in the image, screen the lesion area features in the differentiated frequency range, filter the image data, and obtain the multi-channel filtered lesion features; S5: Based on the multi-channel filtering lesion features, the lesion areas of the differentiated time series images are compared, the key morphological features of the lesion areas are identified, the key point information of the lesion areas at the differentiated time points is extracted, and the lesion time series image analysis data is obtained.

[0037] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A medical image automatic analysis system, characterized in that: The system comprises: The time series dynamic monitoring module extracts the volume, shape, and grayscale value features of the lesion area based on the medical imaging data at different time points, calculates the change rate of the lesion area between adjacent time points, screens the image sequences with prominent change rates, determines the time series change features in the medical imaging data, and obtains the dynamic change trend of the lesion; The lesion area weighted fusion module extracts the lesion change rate at each time point based on the lesion dynamic change trend, calculates the weight value of the lesion change rate in the image time series, screens the time series image data, and obtains weighted fusion lesion area data; The boundary energy optimization module identifies the grayscale contrast value between the lesion area and the surrounding tissue based on the weighted fusion lesion area data, analyzes the energy distribution of the lesion area boundary, screens the regional boundary features affected by noise in the medical imaging data, and obtains the optimized lesion area boundary; The multi-channel filtering processing module extracts the frequency distribution information in the medical image data based on the optimized lesion area boundary, screens the lesion area features in the differentiated frequency range, performs filtering processing on the medical image, and obtains the multi-channel filtered lesion features.

2. The automatic medical image analysis system according to claim 1, characterized in that: The dynamic change trend of the lesion includes a volume change trend, a shape change trend, and a grayscale value change trend. The weighted fusion lesion area data includes weighted volume data, weighted shape data, and weighted grayscale data. The optimized lesion area boundary includes a denoised boundary, an enhanced contrast boundary, and a smooth boundary. The multi-channel filtered lesion features include high-frequency features, low-frequency features, and medium-frequency features.

3. The automatic medical image analysis system according to claim 1, characterized in that: The time series dynamic monitoring module includes: The lesion region feature recognition submodule extracts the segmentation information of the lesion region based on the medical imaging data at different time points, identifies the volume, shape, and grayscale value features of the lesion region, and counts the pixel distribution parameters of the lesion region at each time point to obtain multi-dimensional feature data of the lesion region; The change rate calculation submodule calls the multi-dimensional feature data of the lesion area, extracts the feature change values ​​of the lesion area at adjacent time points, and uses the formula: ； The change rate of the lesion area is calculated; in, represents the rate of change of the lesion area, Representing time point The volume of the lesion area, Representing time point The volume of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point The shape characteristic parameters of the lesion area, Representing time point Gray value characteristics of the lesion area, Representing time point Gray value characteristics of the lesion area; The dynamic change pattern screening submodule calls the change rate of the lesion area, screens the time points with prominent change rates, extracts the time series of the lesion area with prominent changes, compares the trends of the lesion areas at differentiated time points, and obtains the dynamic change trend of the lesion.

4. The automatic medical image analysis system according to claim 3, characterized in that: The lesion area weighted fusion module includes: The lesion rate calculation submodule extracts the lesion area volume and edge contour change data in the image time series based on the dynamic change trend of the lesion, identifies the lesion change rate at each time point, and adjusts the rate in combination with the lesion evolution trend, using the formula: ； The rate of change of the lesion was calculated; in, Representative The rate of change of the lesion at each time point, Representative The volume of the lesion area at each time point, Representative The volume of the lesion area at each time point, Representative The real time corresponding to the time point, Representative The real time corresponding to the time point, Representative Time point The lesion area with edge contour points, No. Time point The lesion area with edge contour points, Represents the total number of lesion edge contour points, represents the lesion morphology adjustment factor; The image data screening submodule calls the lesion change rate, screens the image time series data, removes the image data with a change rate lower than a threshold, and obtains the screened image time series data; The lesion region fusion submodule identifies the weight of the lesion change rate based on the screened image time series data, performs weighted fusion on the lesion region according to the weight, and obtains weighted fused lesion region data.

5. The automatic medical image analysis system according to claim 4, characterized in that: The boundary energy optimization module includes: The boundary gradient extraction submodule calculates the grayscale gradient in the image data based on the weighted fusion lesion area data, identifies the boundary gradient change rate, analyzes the local gradient change of the pixel grayscale value in the area, screens the boundary gradient mutation point, and obtains the boundary gradient distribution data; The lesion area comparison analysis submodule calls the boundary gradient distribution data to compare the grayscale of the lesion area with the surrounding tissues using the formula: ； Calculate the grayscale contrast index of the lesion area boundary, filter the boundary area, and obtain the contrast characteristics of the lesion area; in, The grayscale contrast index representing the boundary of the lesion area, Represents the first The gray value of a pixel, Representing the lesion area The mean grayscale value of the inner pixels, Represents the surrounding tissue area The mean grayscale value of the inner pixels, Represents the number of pixels in the boundary gradient area; The noise impact screening submodule calls the lesion area contrast feature, analyzes the noise impact area, identifies the energy distribution of the lesion area boundary, screens the abnormal area in the energy distribution, removes the noise impact boundary, and obtains the optimized lesion area boundary.

6. The automatic medical image analysis system according to claim 5, characterized in that: The multi-channel filtering processing module comprises: The spatial frequency analysis submodule extracts the pixel grayscale value based on the optimized lesion area boundary, analyzes the grayscale change trend of the differentiated position, extracts the spatial frequency information, and determines the frequency distribution ratio using the formula: ； Calculate the frequency distribution weight value of the pixel point to obtain the frequency distribution information; in, Represents the frequency distribution weight value of the pixel, Represents pixel The gray value of Represents the global grayscale mean of the image, Represents the total number of pixels in the image. Represents pixel The local weight coefficient of Represents the index of the pixel; The lesion region identification submodule calls the frequency distribution information, compares the frequency characteristics of the image region, screens the frequency abnormality region, identifies the frequency deviation, and obtains the frequency characteristics of the lesion region; The filtering feature recognition submodule calls the frequency features of the lesion area, adjusts the weights of the image frequency components, uses filtering processing, strengthens and weakens the frequency information of the key area, and obtains multi-channel filtering lesion features.

7. The automatic medical image analysis system according to claim 1, characterized in that: The system also includes a time series feature building block: The time series feature construction module identifies the lesion area information at different time points in the medical image based on the multi-channel filtered lesion features, extracts the key lesion features in the image sequence, and obtains the lesion time series image analysis data; The lesion time series image analysis data includes key lesion information, change patterns, and time series feature distribution.

8. The automatic medical image analysis system according to claim 7, characterized in that: The timing feature building module comprises: The filtered lesion signal calculation submodule analyzes the signal changes in the lesion area based on the multi-channel filtered lesion characteristics, calculates the image gradient value, screens the lesion response channel, and obtains the lesion signal gradient value; The lesion time point determination submodule identifies the gradient change rate of the time series image based on the lesion signal gradient value, analyzes the time increment of the lesion area, and obtains the lesion time series change amount; The lesion feature extraction submodule identifies the morphological characteristics of the lesion area based on the lesion time series change, analyzes the boundary changes at the time point, and uses the formula: ； Calculate the characteristic change value of the lesion area and obtain the time series image analysis data of the lesion; in, Represents the characteristic change value of the lesion area, Represents the gradient change of adjacent time points in the lesion area, Represents the pixel offset of the lesion area at adjacent time points, represents the rate of change of the lesion feature vector at adjacent time points, Represents the change in the lesion boundary between adjacent time points.

9. A method for automatic analysis of medical images, characterized in that: The automatic medical image analysis system according to any one of claims 1 to 8 comprises the following steps: S1: Based on the medical imaging data at different time points, extract the lesion area information in the image, calculate the volume, morphology and grayscale value of the lesion area, arrange the data in chronological order, compare the lesion area characteristics at adjacent time points, identify the lesion change rate, and screen the image sequence with the change rate; S2: Based on the image sequence with significant change rate, extract the change rate of the lesion at each time point, calculate the rate weight of the time point, select the image data at the optimal time point, combine the lesion areas at multiple time points, and obtain weighted fusion lesion area data; S3: Based on the weighted fusion lesion area data, extract the boundary gradient information in the image, calculate the grayscale contrast value between the lesion area and the surrounding tissue, screen the boundary points of the grayscale contrast, analyze the distribution of the boundary points in the image, analyze the energy distribution of the region boundary, eliminate the low-energy boundary points affected by noise, and obtain the optimized lesion area boundary; S4: extracting frequency distribution information in the image based on the optimized lesion area boundary, screening lesion area features in a differentiated frequency range, filtering the image data, and obtaining multi-channel filtered lesion features; S5: Based on the multi-channel filtered lesion features, compare the lesion area of ​​the differentiated time series images, identify the key morphological features of the lesion area, extract the key point information of the lesion area at the differentiated time points, and obtain the lesion time series image analysis data.

Images