Remote ultrasonic diagnosis support system based on cloud computing

Abstract

The invention relates to the technical field of remote medical treatment, in particular to a remote ultrasonic diagnosis support system based on cloud computing, which comprises an echo feature analysis module, a dynamic focus screening module, a case image comparison module, a focus trend classification module and a remote diagnosis support module. According to the method, by analyzing echo signals in a remote ultrasonic image data stream and calculating the echo change rate between image frames, the dynamic characteristic capturing capacity of a lesion area is improved, calculation of a gray gradient change value is combined, a suspected lesion area conforming to a change rate stable interval is screened, lesion shape edge curve contour parameters are utilized, and the dynamic characteristic capturing capacity of the lesion area is improved. The method comprises the following steps: performing matching analysis with an ultrasonic case image database, enhancing the accuracy of lesion category attribution, analyzing the morphological change trend of a lesion area, combining trend stability index calculation, realizing lesion evolution classification, ensuring the reasonability of lesion type division, and marking a progressive lesion area, so that remote diagnosis is more intuitive; and the auxiliary decision-making capability is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of telemedicine, and in particular to a remote ultrasonic diagnosis support system based on cloud computing. Background Art

[0002] The field of telemedicine technology involves the use of information and communication technology to achieve remote interaction and data sharing of medical services. The core content includes remote diagnosis, remote surgery, remote monitoring and health management, etc., aiming to overcome the problem of uneven geographical distribution of medical resources and improve the accessibility and efficiency of medical services. In telemedicine systems, technologies such as wireless communication, cloud computing, big data processing and artificial intelligence analysis are usually used to enable patients to receive real-time diagnosis and health monitoring from medical institutions in other places. At the same time, this field involves multiple aspects such as medical image transmission, data encryption and privacy protection, remote control and interactive diagnosis to ensure the security and real-time nature of medical data. It is widely used in scenarios such as ultrasound imaging diagnosis, electrocardiogram monitoring, and pathological analysis, and supports cross-regional medical image analysis and clinical decision assistance.

[0003] Among them, the cloud computing-based remote ultrasound diagnosis support system refers to an ultrasound image acquisition, transmission, processing and analysis system built on the basis of cloud computing technology. The system covers the digital acquisition of ultrasound signals, efficient encoding and transmission of ultrasound images, cloud-based medical image storage and management, remote ultrasound image analysis and diagnostic support. Specifically, the echo signal is converted into digital data using an ultrasound probe, and is transmitted to the cloud after reducing bandwidth occupancy through data compression methods. The cloud server uses medical image storage formats to store and classify ultrasound images, and performs image analysis through feature extraction and template matching. Doctors can access ultrasound images through the remote access interface, and use image segmentation and pattern recognition methods to diagnose and interpret the lesion area. In addition, the system uses medical image archiving protocols for data management, supports multi-terminal real-time access and remote consultation, and provides auxiliary decision-making basis for ultrasound diagnosis.

[0004] In the existing remote ultrasound diagnosis process, the recognition of dynamic changes of lesions relies on single-frame images or simple inter-frame comparisons, which makes it difficult to accurately obtain the tiny dynamic features of the lesion area, resulting in some lesion areas being difficult to distinguish from normal tissues. The lesion screening method uses a fixed feature threshold and lacks refined calculation of the lesion change trend, resulting in the failure to effectively identify early lesion areas. Case image matching mainly relies on global features and lacks accurate comparison of morphological edge contours, which affects the accuracy of lesion attribution. The time series analysis capability is insufficient and the evolution trend of the lesion cannot be fully quantified, making the assessment of the lesion development status uncertain. The telemedicine terminal equipment lacks intuitive progressive lesion annotations in the presentation of lesion classification results, making it difficult for doctors to quickly grasp the development of lesions during remote diagnosis, affecting the diagnostic efficiency of ultrasound images. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a remote ultrasound diagnosis support system based on cloud computing.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A remote ultrasound diagnosis support system based on cloud computing includes: The echo feature analysis module obtains echo signal data in the remote ultrasound image data stream, extracts the echo intensity value of the lesion area in the image frame, calculates the echo change rate between consecutive image frames, analyzes the echo intensity at different angles through cloud computing, calculates the gradient change in the time series, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and outputs the echo change rate feature data; The dynamic lesion screening module extracts the gray gradient change value of the lesion area in the continuous image frames based on the echo change rate characteristic data, calculates the change rate standard deviation, screens the suspected lesion area, and generates the lesion dynamic screening area; The case image comparison module calculates the lesion morphology edge curve contour parameters based on the lesion dynamic screening area, analyzes the morphological structure similarity of the screening area in combination with the ultrasound case image database, makes a lesion category determination, and generates a lesion image matching record; The lesion trend classification module analyzes the morphological change trend of the lesion area in the ultrasound image time series according to the lesion image matching record, obtains the slope of the change trend within the time period, calculates the trend stability index, divides the lesion type, and generates a lesion evolution classification result.

[0007] As a further scheme of the present invention, the echo change rate characteristic data includes echo intensity distribution data, echo change gradient data, and abnormal echo area; the lesion dynamic screening area includes suspected lesion boundaries, grayscale gradient stability areas, and change rate standard deviation ranges; the lesion image matching records include lesion morphological characteristic parameters, edge curve contour matching, and ultrasound case image similarity analysis results; the lesion evolution classification results include lesion morphological change trends, trend stability indexes, and lesion type classification information.

[0008] As a further solution of the present invention, the echo feature analysis module includes: The echo signal extraction submodule obtains the echo signal data in the remote ultrasound image data stream, detects the echo intensity value of the lesion area in the image frame, arranges the echo signals of the continuous image frames in time series, and generates the echo signal data of the lesion area; The echo change calculation submodule adopts the formula based on the echo signal data of the lesion area: ; Calculate the echo intensity change rate between adjacent image frames , generate echo rate of change data, where Representative The echo strength value of the frame, Representative The echo strength value of the frame, Representative The echo strength value of the frame, Represents the number of image frames; The dynamic gradient screening submodule calculates the gradient change value in the time series based on the echo change rate data, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and generates echo change rate characteristic data.

[0009] As a further solution of the present invention, the dynamic lesion screening module comprises: The grayscale gradient calculation submodule detects the grayscale value of the lesion area through continuous image frames based on the echo change rate characteristic data, calculates the grayscale change amount between adjacent frames, obtains the grayscale gradient change value of the lesion area in the image frame sequence, and calculates and outputs the grayscale gradient change rate; The change rate standard deviation calculation submodule calculates the mean of the gray gradient change rate according to the gray gradient change rate, counts the square sum of the deviations of each image frame relative to the mean, and calculates the change rate standard deviation; The stable interval screening submodule sets the upper and lower thresholds of the stable interval based on the standard deviation of the change rate, screens the lesion areas of all image frames, and extracts the suspected lesion areas that meet the range of the stable interval, using the formula: ; Calculation of screening stability index , screen the suspected lesion areas that meet the stable rate range and generate the lesion dynamic screening area, where, Representative The gray gradient change rate of the frame, Represents the mean gray gradient change rate, Represents the standard deviation of the gray gradient change rate, Represents the weight coefficient of the current frame in the sequence, Represents the number of image frames.

[0010] As a further solution of the present invention, the case image comparison module includes: The morphological edge calculation submodule detects the edge contour of the lesion area based on the dynamic lesion screening area, calculates the curvature change, contour closure and area ratio of the edge curve, and obtains the lesion morphological edge parameters; The structural similarity analysis submodule compares the lesion morphological edge parameters with the morphological parameters in the ultrasound case image database using the formula: ; Calculate the output lesion morphology matching ,in, Representative The morphological parameters of the lesions in the group database, Represents the morphological parameters of the current lesion area, Represents the curvature value of the lesion area, represents the curvature value of the database matching lesion, Represents the number of morphological parameter comparison items in the database, represents the number of lesion curvature comparison items; The lesion category attribution submodule sets a category determination threshold based on the lesion morphology matching degree, screens the lesion category with the highest matching degree, obtains the corresponding category attribution information, and establishes a lesion image matching record.

[0011] As a further solution of the present invention, the lesion trend classification module includes: The lesion morphology trend calculation submodule detects the morphological changes of the lesion area in the time series based on the lesion image matching record, calculates the rate of change of the lesion contour in each time period, fits the morphological change trend curve of the lesion area, and obtains the change trend slope data; The trend stability analysis submodule calculates the standard deviation of the trend slope of each segment in the time series according to the change trend slope, and uses the formula: ; Calculating the Trend Stability Index ,in, Representative The slope of the trend of change in a time period, Represents the mean value of the slope of the changing trend, represents the rate of change of lesion morphology, represents the change rate of lesion morphology matching, represents the number of time periods in the time series, represents the number of lesion matching change items; The lesion type classification submodule sets the change rate and stability classification criteria based on the trend stability index, screens the lesion types that meet the specific trend pattern, and establishes the lesion evolution classification results.

[0012] As a further solution of the present invention, the system further includes a remote diagnosis support module; The remote diagnosis support module associates the remote ultrasound terminal device based on the lesion evolution classification result, pushes the classification result to the terminal device, marks the progressive lesion area, and generates remote diagnosis reference data; The remote diagnosis reference data includes the marking results of lesion progressive changes, the classification push results of the remote ultrasound terminal, and the remote diagnosis auxiliary analysis information.

[0013] As a further solution of the present invention, the remote diagnosis support module includes: The classification result push submodule detects the matching remote ultrasound terminal device based on the lesion evolution classification result, extracts the unique identification information of the terminal device, pushes the lesion classification result to the target terminal device, and obtains the remote push data; The lesion region marking submodule detects the progressive lesion region in the classification result according to the remote push data, obtains the lesion region coordinate information, calculates the region boundary features, performs image annotation, and generates a progressive lesion mark; The remote diagnosis data integration submodule extracts the historical classification information of the lesion area based on the progressive lesion mark, matches the image data of similar lesions, counts the frequency of changes in the lesion area, integrates the image matching information, and establishes remote diagnosis reference data.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by parsing the echo signal in the remote ultrasound image data stream, the echo intensity value of the lesion area is extracted, and the echo change rate between image frames is calculated, so as to improve the dynamic feature capture capability of the lesion area. Based on the multi-angle analysis of the echo intensity and the gradient change in the calculation time series, the area with abnormal change rate can be screened out, so that the identification of the lesion area is more targeted. Combined with the calculation of the gray gradient change value, the suspected lesion area that meets the stable change rate interval is screened out to avoid misscreening due to individual differences or image noise. The lesion morphology edge curve contour parameters are used to match and analyze with the ultrasound case image database to enhance the accuracy of lesion classification. Based on the ultrasound image time series analysis of the morphological change trend of the lesion area, combined with the calculation of the trend stability index, the lesion evolution classification is realized to ensure the rationality of the lesion type division. The classification results are obtained in combination with the remote ultrasound terminal equipment, and the progressive lesion area is marked to make the remote diagnosis more intuitive and improve the auxiliary decision-making ability of ultrasound diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the echo characteristic analysis module of the present invention; Figure 3 This is a flow chart of the dynamic lesion screening module of the present invention; Figure 4 This is a flow chart of the case image comparison module of the present invention; Figure 5 This is a flow chart of the lesion trend classification module of the present invention; Figure 6 This is a flow chart of the remote diagnosis support module of the present invention. DETAILED DESCRIPTION

[0016] 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.

[0017] 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.

[0018] See also Figure 1 , a remote ultrasound diagnosis support system based on cloud computing includes: The echo feature analysis module obtains echo signal data in the remote ultrasound image data stream, extracts the echo intensity value of the lesion area in the image frame, calculates the echo change rate between consecutive image frames, analyzes the echo intensity at different angles through cloud computing, calculates the gradient change in the time series, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and outputs the echo change rate feature data; The dynamic lesion screening module extracts the grayscale gradient change value of the lesion area in the continuous image frames based on the echo change rate feature data, calculates the standard deviation of the change rate, screens the suspected lesion area that meets the stable change rate interval, and generates the dynamic lesion screening area; The case image comparison module calculates the lesion morphology edge curve contour parameters based on the dynamic lesion screening area, analyzes the morphological structure similarity of the screening area in combination with the ultrasound case image database, determines the lesion category of the screening area, and generates lesion image matching records; The lesion trend classification module analyzes the morphological change trend of the lesion area in the ultrasound image time series according to the lesion image matching records, obtains the slope of the change trend within the time period, calculates the trend stability index, divides the lesion type according to the change rate and trend stability index, and generates the lesion evolution classification result; The remote diagnosis support module associates the remote ultrasound terminal device based on the lesion evolution classification results, pushes the classification results to the terminal device, marks the progressive lesion area, and generates remote diagnosis reference data.

[0019] The echo change rate characteristic data include echo intensity distribution data, echo change gradient data, and abnormal echo area. The lesion dynamic screening area includes the suspected lesion boundary, gray gradient stable area, and change rate standard deviation range. The lesion image matching record includes the lesion morphological characteristic parameters, edge curve contour matching degree, and ultrasound case image similarity analysis results. The lesion evolution classification results include lesion morphological change trend, trend stability index, and lesion type classification information. The remote diagnosis reference data includes the lesion progressive change marking results, the remote ultrasound terminal classification push results, and the remote diagnosis auxiliary analysis information.

[0020] See also Figure 2 , the echo feature analysis module includes: The echo signal extraction submodule obtains the echo signal data in the remote ultrasound image data stream, detects the echo intensity value of the lesion area in the image frame, arranges the echo signals of the continuous image frames in time series, and generates the echo signal data of the lesion area; In the process of ultrasonic diagnosis, it is first necessary to obtain the echo signal data in the remote ultrasonic image data stream. The ultrasonic probe emits ultrasonic waves at a certain frequency (such as 3.5MHz). When the ultrasonic waves propagate in the human tissue and encounter tissue interfaces of different densities, they are reflected to form echo signals. These echo signals are received by the probe and converted into electrical signals. Subsequently, the analog signals are converted into digital signals through an analog-to-digital converter to form an ultrasonic image data stream. Assume that in a scanning process, the system acquires 1000 frames of images, each frame of the image contains 1024×768 pixels, and each pixel corresponds to an echo intensity value, indicating the tissue characteristics of the position. After acquiring the image data, it is necessary to detect the echo intensity value of the lesion area in the image frame. First, the image segmentation method is used. The lesion area in the image frame is distinguished from the normal tissue area. Assuming that in a certain frame of image, the lesion area occupies 200×150 pixels. Next, the echo intensity value of each pixel in the area is extracted to form a matrix containing 200×150 elements, which represents the echo intensity distribution of the lesion area. In order to make the echo signal data more stable, signal noise reduction processing is required, such as using median filtering to smooth the echo signal to reduce the interference of random noise. Assuming that the echo intensity values ​​of a pixel in 5 frames of images are 0.78, 0.81, 0.80, 0.79, and 0.82 respectively, the echo intensity value after median filtering is 0.80, thereby improving the stability of the data. Finally, the echo signal data of the lesion area is generated.

[0021] The echo change calculation submodule is based on the echo signal data of the lesion area and uses the formula: ; Calculate the echo intensity change rate between adjacent image frames , generate echo rate of change data, where Representative The echo strength value of the frame, Representative The echo strength value of the frame, Representative The echo strength value of the frame, Represents the number of image frames; Based on the echo signal data of the lesion area, the echo intensity change rate between adjacent image frames is calculated, and a time series echo gradient change model is constructed. Specifically, for the lesion area in each frame of the image, the echo intensity difference of the corresponding pixel points in adjacent frames is calculated. Assuming that Frame and The echo intensity of a certain pixel point in the frame is and , then the echo intensity change rate of the pixel point can be expressed as: ; The echo intensity change rate of all pixels in the lesion area is averaged to obtain the average echo change rate of the frame image. Assume that in a certain frame, the average echo change rate in the lesion area is 0.05.

[0022] In order to enhance the accuracy of calculation, the second-order gradient calculation is introduced to measure the dynamic change trend of echo intensity, and the echo change rate is calculated using the formula.

[0023] In order to verify the practical application of the calculation method, it is assumed that the echo intensity of a lesion area in continuous image frames changes as follows: Table 1.1 Data on echo intensity changes in the lesion area of ​​the image frame (unit: dB)

[0024] According to the data in Table 1.1, calculate the echo change rate: ; ; The calculation results show that the echo change rate of the lesion area is 0.0475dB, which can be used to further analyze the dynamic evolution of the lesion and ultimately generate echo change rate data.

[0025] The dynamic gradient screening submodule calculates the gradient change value in the time series based on the echo change rate data, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and generates echo change rate characteristic data; Based on the echo change rate data, the gradient change in the time series is calculated to screen out the echo signal area whose change rate exceeds the dynamic echo gradient reference value. Specifically, for each analysis angle, the gradient of its echo intensity change over time is calculated. Assuming that at an angle of 0°, the echo intensity values ​​at the 1st second, 2nd second, and 3rd second are 0.80, 0.82, and 0.85, respectively, the gradient change is calculated as follows: ; In order to screen out areas with abnormal echo intensity changes, it is necessary to set a dynamic echo gradient baseline value. The setting of this baseline value is based on large-scale statistical analysis of ultrasound image data. By extracting the echo gradient change range of normal tissue areas and lesion areas, calculating their mean and standard deviation, the baseline value is finally determined. Specifically, 1,000 normal ultrasound image data were selected, echo signals were extracted in different tissue areas (such as muscle, fat, liver tissue, etc.), the gradient change rate in the time series was calculated, and the mean was obtained. and standard deviation In normal tissue, the mean of the echo gradient change rate is usually between 0.012 and 0.018, and the standard deviation is about 0.002. Combined with statistical analysis, the benchmark value is set to ,Right now: ; The changing trend of this value is related to tissue type and imaging parameters. If the frequency of the ultrasound probe is increased, for example, from 3.5 MHz to 7.5 MHz, the local tissue gradient change will be more obvious due to the improved resolution, and the baseline value may need to be adjusted. On the contrary, if the image resolution is reduced or the tissue interface echo signal is weak, the baseline value may be reduced to 0.018 accordingly.

[0026] Assume that the calculation result of a certain area is 0.025. Therefore, this area is judged as the echo signal area with a change rate exceeding , and finally, the echo change rate characteristic data is generated.

[0027] See also Figure 3 , the dynamic lesion screening module includes: The grayscale gradient calculation submodule detects the grayscale value of the lesion area through continuous image frames based on the echo change rate feature data, calculates the grayscale change between adjacent frames, obtains the grayscale gradient change value of the lesion area in the image frame sequence, and calculates and outputs the grayscale gradient change rate; When analyzing continuous image frames, it is necessary to first obtain the grayscale value of the lesion area. For a sequence containing multiple image frames, it is first necessary to extract the grayscale value of each frame and set the image frame sequence. ,in Indicates Frame image data, each frame of image is composed of a pixel matrix, each pixel Has a gray value In practical applications, such as analysis of ultrasound images, the grayscale value range of each pixel is usually , where 0 represents the darkest area and 255 represents the brightest area. For example, in a lung CT image, the lesion area may have high-density features, and its grayscale value is often higher than the grayscale value of the surrounding healthy tissue. Therefore, by detecting the grayscale value of the pixel points in the lesion area frame by frame, a grayscale data matrix can be established. .

[0028] Calculate the grayscale change between adjacent frames using the formula: ; in, Represents pixel The grayscale difference between two adjacent frames. For example, in a set of image sequences, suppose a pixel is The grayscale value of the frame is 120, and The grayscale value of the frame is 135, so the grayscale change of the pixel is If the grayscale changes of all pixels are calculated continuously, the grayscale change data set of the entire image frame sequence can be obtained. .

[0029] In order to further extract the grayscale gradient change value of the lesion area, it is necessary to calculate the average grayscale gradient of the entire lesion area: ; in, Represents the total number of pixels in the lesion area. For example, if the lesion area contains 100 pixels, and the calculated grayscale changes are , then its mean is: ; This mean value can be used for subsequent analysis to ultimately obtain the grayscale gradient change rate.

[0030] Table 2.1 Image frame grayscale change rate data

[0031] As shown in Table 2.1, the grayscale gradient change rate data of continuous image frames are listed, which are used to calculate the standard deviation of the change rate and screen the stable interval.

[0032] The change rate standard deviation calculation submodule calculates the mean of the gray gradient change rate according to the gray gradient change rate, and calculates the square sum of the deviations of each image frame relative to the mean to calculate the change rate standard deviation; Call the gray gradient change rate calculated above and set a time series , corresponding to the acquisition time of the image frame, each time point Has a gray gradient change rate In order to analyze the discrete degree of the change rate, it is necessary to calculate the mean of the gray gradient change rate: ; in, is the number of image frames. For example, if the gray gradient change rates calculated in 10 frames are , then its mean: ; Next, the sum of squared deviations of all frames relative to the mean is calculated: ; In the above example: ; Calculate the standard deviation: ; Substituting the values: ; Finally, the standard deviation of the rate of change is obtained.

[0033] The stable interval screening submodule sets the upper and lower thresholds of the stable interval based on the standard deviation of the change rate, screens the lesion areas of all image frames, and extracts the suspected lesion areas that meet the range of the stable interval using the formula: ; Calculation of screening stability index , screen the suspected lesion areas that meet the stable rate range and generate the lesion dynamic screening area, where, Representative The gray gradient change rate of the frame, Represents the mean gray gradient change rate, Represents the standard deviation of the gray gradient change rate, Represents the weight coefficient of the current frame in the sequence, Represents the number of image frames; According to the standard deviation of the change rate calculated above, set the upper and lower thresholds of the stable interval, and set the threshold range as follows: ; ; in, It is an empirical setting factor, which is set based on the discrete degree of the gray gradient change rate of the lesion area in the image frame sequence. It is usually set using the standard normal distribution in statistics, that is, the data is above and below the mean. The interval covers about 95.4% of the data points. If the data has a large degree of variation (i.e., standard deviation higher), you can adjust it appropriately value to expand the screening range. If the data distribution is relatively concentrated ( smaller), it should be appropriately reduced value to avoid over-filtering.

[0034] In this example, the total number of image frames is 10 frames, and the calculated average , standard deviation ,set up When , the screening interval is calculated as follows: ; ; The theoretical basis for this setting is that the grayscale gradient change rate of most lesion areas in the image frame sequence should be within this range. or higher , it indicates that the frame may be interfered by external factors or belongs to an unstable lesion area. Filter all Image frame data.

[0035] Grayscale gradient change rate that meets the range , removing image frames that exceed the threshold range .

[0036] The formula is used to further calculate the screening stability index, where the weight coefficient Set the time interval during the acquisition process based on the frame sequence, for example: ; Bring in the above data to calculate: ; The calculation results in: ; In order to determine whether the image frame sequence belongs to the stable rate range, it is necessary to set the screening stability threshold , which is used to distinguish between stable and unstable lesion areas. The setting of is based on the statistical characteristics of the overall grayscale changes in the image frame sequence. Usually, a value is selected that keeps the proportion of stable lesion areas above about 90%. The selection of this value needs to be analyzed in combination with multiple sets of image frame data. In this example, by analyzing 10 sets of image frame data of different lesion areas, the average screening stability index of different areas is calculated, and then different values ​​are set. The lesion area screening results under the values ​​are shown in Table 2.2: Table 2.2 Different Filter results under settings

[0037] From Table 2.2, we can see that when , 80% of the image frames passed the screening, and only 2 frames were removed, which shows that this threshold can remove noise frames while ensuring a sufficient number of frame data, which meets the actual screening requirements. Therefore, in this example, set: ; Under this threshold, if the calculated , it means that the gray gradient of the image frame sequence changes stably and can be judged as a suspected lesion area. , it means that the image frame sequence may contain many abnormal frames or noise, and the screening conditions need to be further adjusted.

[0038] Finally, remove (Grayscale change is too small, may belong to the background area) and above The frames with grayscale changes too fast may belong to unstable lesion areas or image noise, and the frame data that meet the stability requirements are summarized to generate a dynamic lesion screening area.

[0039] See also Figure 4 , the case image comparison module includes: The morphological edge calculation submodule detects the edge contour of the lesion area based on the dynamic lesion screening area, calculates the curvature change, contour closure and area ratio of the edge curve, and obtains the lesion morphological edge parameters; Based on the dynamic screening area of ​​the lesion, it is first necessary to extract the edge contour of the lesion area from the continuous image frames, obtain the change trend of the grayscale distribution, use the gradient calculation method to identify the edge pixel point set, and smooth the edge curve to reduce noise interference. Taking a certain ultrasound image data as an example, an image with a resolution of 512×512 pixels is selected, of which the lesion area accounts for about 16%. The edge point coordinate set is extracted by the gradient calculation method, and the contour closure is calculated based on the coordinate data. The contour closure can be defined as: ; in, is the length of the closed contour, is the total length of the lines between all edge points. If the calculation result , then the lesion contour is considered to be close to closed. Further, the curvature change of the edge curve is calculated, several measuring points are set, and the normal vector change rate of adjacent pixel points is measured to obtain the curvature distribution characteristics. Taking a lesion area with a diameter of 20mm as an example, among the 50 sampling points, if 30 points have a curvature change of less than 0.1, the lesion morphology tends to be regular, and if more than 50% of the points have a curvature change of more than 0.3, it means that the lesion edge is irregular. This setting is based on the smoothness of the normal tissue boundary. Usually, the boundary of benign lesions changes smoothly and the curvature changes little, while malignant lesions often have lobed or burr structures, resulting in local curvature mutations. For the standard of 30 points with a curvature change of less than 0.1, it is based on the statistical distribution of the curvature of the normal tissue boundary. More than 90% of the benign lesions in the database meet this standard, while the standard of more than 50% of the points with a curvature change of more than 0.3 is based on the statistical characteristics of malignant lesions, and more than 85% of malignant lesions have this feature. Finally, the area ratio is calculated, that is, the ratio of the lesion area to its convex hull area: ; in, is the area of ​​the lesion, is the area of ​​its convex hull. If the area ratio is lower than 0.75, the lesion morphology may have irregular expansion characteristics. This setting is based on the analysis of morphological parameters of database statistics. The area ratio of benign lesions is usually between 0.80-0.95, while the area ratio of malignant lesions is lower than 0.75 due to the presence of invasive growth characteristics. In the statistics of 200 cases, 78% of lesions with an area ratio <0.75 were diagnosed as malignant lesions, while only 12% of lesions with a ratio >0.8 were diagnosed as malignant. Finally, the lesion morphology edge parameters are obtained.

[0040] The structural similarity analysis submodule compares the lesion morphological edge parameters with the morphological parameters in the ultrasound case image database using the formula: ; Calculate the output lesion morphology matching ,in, Representative The morphological parameters of the lesions in the group database, Represents the morphological parameters of the current lesion area, Represents the curvature value of the lesion area, represents the curvature value of the database matching lesion, Represents the number of morphological parameter comparison items in the database, represents the number of lesion curvature comparison items; The lesion morphology edge parameters are called, and the lesion data with high morphological similarity are extracted from the ultrasound case image database. The similarity of the contour morphology is compared, and the morphological structure similarity of the matching lesions is calculated. The database is set to contain 3000 ultrasound lesion images, and 10 key morphological parameters are extracted from each image. The similarity between the morphological parameters is calculated using the Euclidean distance.

[0041] Taking the lesion morphology parameters as an example, the closure degree of a lesion is set to C=0.82, matching the closest data point in the database sample, assuming that its closure mean =0.85, then the closure deviation of the lesion is: ; Similarly, if the average curvature of the lesion area is 0.12 and the average curvature of the database matching lesions is 0.11, then calculate: ; Morphological matching It can be calculated as: ; Set the morphology matching threshold Indicates that the lesions are highly similar. Indicates that the lesions are partially similar. Indicates that there are significant differences in morphology. This setting is based on the analysis of 1,000 image comparisons. The statistical results show that the morphological matching degree is Among the lesions, 95% had the same pathological category. Among the lesions, about 70% had the same pathological category. In order to ensure the rationality, 30 cases with high matching degree ( ) The image results of the lesion were compared with the pathological diagnosis data, and the matching category accuracy was found to be 96%. In a specific case, if the morphological parameters of the database matching lesions are shown in the following table: Table 3.1 Morphological matching calculation results

[0042] As shown in Table 3.1, lesions with a matching degree lower than 0.1 can be considered as highly similar samples, and lesions with a matching degree higher than 0.2 may belong to different types. Finally, the lesion morphology matching degree is obtained.

[0043] The lesion category attribution submodule sets a category determination threshold based on the lesion morphology matching degree, selects the lesion category with the highest matching degree, obtains the corresponding category attribution information, and establishes a lesion image matching record; Based on the lesion morphology matching, compare the classified lesion images in the database and set the matching threshold As the category attribution standard. Assume that there are 50 different categories of lesions marked in the database, and each category contains 50-200 case images. According to the lesion morphology matching, select the top 5 most similar database lesions, count the category proportions, and use the maximum probability attribution principle to determine the lesion category. For example, if 3 of the 5 matching lesions belong to category A and 2 belong to category B, the current lesion is classified as A. If there is a tie, a weighted voting method is used for decision making. Suppose that the match between a lesion and the database is as follows:

[0044] In this case, category A accounts for 3 / 5 and category B accounts for 2 / 5, so the lesion is finally determined to be category A. Finally, a lesion image matching record is established.

[0045] See also Figure 5 , the lesion trend classification module includes: The lesion morphology trend calculation submodule detects the morphological changes of the lesion area in the time series based on the lesion image matching records, calculates the rate of change of the lesion contour in each time period, fits the morphological change trend curve of the lesion area, and obtains the change trend slope data; Based on the lesion image matching records, historical image data are called and the contour features of the lesion area are extracted. The image data are sorted in time series, and image frames with uniform time intervals are selected. The contour parameters of the lesion area, including area, perimeter, and convexity, are measured. The piecewise linear regression method is used to calculate the change rate of the lesion morphology in each time period to obtain the slope of the change trend.

[0046] To refine the calculation process, a lesion image matching record of a case is selected. The area of ​​the lesion at the initial moment is 45.3 mm² and the perimeter is 28.1 mm. At the second time node, the area increases to 50.8 mm² and the perimeter increases to 30.5 mm. Assuming that the time interval is 30 days, the morphological change rate is calculated as follows: ; ; The calculated morphological change rates of the lesion were 0.1833 mm² / day and 0.08 mm / day, respectively. The morphological change rates of all time periods were fitted to generate the slope of the change trend.

[0047] The trend stability analysis submodule calculates the standard deviation of the trend slope of each segment in the time series according to the changing trend slope, and uses the formula: ; Calculating the Trend Stability Index ,in, Representative The slope of the trend of change in a time period, Represents the mean value of the slope of the changing trend, represents the rate of change of lesion morphology, represents the change rate of lesion morphology matching, represents the number of time periods in the time series, represents the number of lesion matching change items; Based on the slope of the changing trend, the standard deviation of the trend slope of each segment in the time series is calculated, and the mean deviation of the slope change is obtained to measure the stability of the trend change. The formula is used for calculation.

[0048] Suppose that a total of H=6 time points are collected in the time series of a lesion, and its slope change is calculated: , , , , , .

[0049] Calculate the mean slope: ; ; ; ; ; ; Set the lesion morphology change rate data: , ; , ; , ; , ; , ; , ; calculate: ; ; ; ; ; ; ; ; Calculate the final trend stability index ; Table 4.1: Lesion trend slope calculation table

[0050] Table 4.2: Calculation table of lesion morphological change rate

[0051] As shown in Table 4.2, the morphological change rate at each time point was calculated, and the trend stability index was calculated based on the slope of the change trend.

[0052] The lesion type classification submodule sets the change rate and stability classification criteria based on the trend stability index, screens the lesion types that meet the specific trend pattern, and establishes the lesion evolution classification results; The classification standard of lesion types was set based on the long-term trend data analysis of 500 lesion samples in the ultrasound imaging database, and combined with the correlation between the image change rate and clinical pathological diagnosis: Stable: Trend Stability Index In the follow-up data of ultrasound images, the average change rate of the morphological edge curve of this type of lesion is less than 0.05mm² / day compared with the lesion morphological parameters in the image matching record, and the standard deviation is Less than 0.015, the corresponding pathological test data showed that this type of lesion was mainly benign hyperplasia, cystic lesions or inflammatory reactions.

[0053] Slowly progressive type: The average rate of morphological change of such lesions ranged from 0.05 mm² / day to 0.12 mm² / day, with a standard deviation of It is between 0.015 and 0.03, and the matching lesions in its ultrasound image database are mostly low-grade malignant or locally proliferative lesions, such as focal nodules.

[0054] Rapidly progressive type: The morphological change rate of such lesions in continuous imaging data is higher than 0.12 mm² / day, and the standard deviation Greater than 0.03. Compared with the imaging database, more than 70% of the lesion samples that meet this standard were confirmed by histopathology to be highly malignant, such as invasive carcinoma or rapidly spreading tumors.

[0055] Assume that the trend stability index of a lesion at 6 time points is as follows: ; Calculate its mean: ; Calculate the standard deviation: ; ; ; ; Compared with the classification criteria, the mean trend stability index of the lesion Less than 0.02, standard deviation It is less than 0.015, so the lesion is classified as stable, indicating that its morphological changes are small in the imaging time series, which is consistent with the imaging characteristics of benign hyperplasia or inflammatory response, and is stored in the database to generate the lesion evolution classification results.

[0056] Table 4.3: Classification criteria for lesion types

[0057] As shown in Table 4.3, by calculating the trend stability index and its standard deviation, the evolution category of the lesion is determined, and finally the lesion evolution classification result is established.

[0058] See also Figure 6 , remote diagnosis support modules include: The classification result push submodule detects the matching remote ultrasound terminal device based on the lesion evolution classification result, extracts the unique identification information of the terminal device, pushes the lesion classification result to the target terminal device, and obtains the remote push data; To detect matching remote ultrasound terminal devices based on the lesion evolution classification results, it is first necessary to extract specific lesion type data from the lesion evolution classification results, such as static lesions, progressive lesions or regressing lesions, etc. Then, call the device management database in the telemedicine system to obtain a list of registered remote ultrasound terminal devices and screen out devices that meet the current lesion matching requirements. These devices mainly include portable handheld ultrasound devices, hospital fixed ultrasound scanners, remote robotic ultrasound devices, etc. According to specific clinical application requirements, the matching principles of these devices are also different. For example, portable handheld ultrasound devices are suitable for primary hospitals or community clinics to facilitate doctors to follow up on suspicious lesions, while hospital fixed ultrasound scanners are mainly used in large medical institutions and are suitable for high-resolution detailed examinations. Remote robotic ultrasound devices are suitable for telemedicine scenarios, and ultrasound scanning is performed by remote doctors controlling the equipment.

[0059] During the screening process, the device is screened based on its geographic location, hospital level, device model, online status and other parameters. For example, if the lesion comes from the long-term monitoring of a high-risk patient, the patient's exclusive monitoring device will be given priority for push. If the lesion type is a suspected lesion in the initial screening, it will be pushed to all ultrasound devices with diagnostic authority in the area. After the screening is completed, the system structures the lesion data into standardized push information, which includes the lesion number, matching image index, evolution trend category, etc., and formats the data. The converted data is encapsulated according to telemedicine standards such as HL7 or DICOM, and the remote communication protocol is called for data transmission. During the transmission process, the AES encryption algorithm is used to protect the security of lesion data and ensure data integrity and confidentiality. Finally, after the target terminal device receives the pushed data, it verifies the data integrity and generates a push record on the device side to ensure that the data is successfully transmitted and received, and the remote push data is obtained.

[0060] Table 5.1 Classification of remote ultrasound terminal equipment

[0061] As shown in Table 5.1, different remote ultrasound terminal devices are suitable for different scenarios, ensuring that the lesion classification results can be pushed and processed on appropriate devices.

[0062] The lesion region marking submodule detects the progressive lesion region in the classification results based on the remote push data, obtains the lesion region coordinate information, calculates the region boundary features, performs image annotation, and generates progressive lesion markings; According to the remote push data, the progressive lesion area in the classification result is detected. First, the remote terminal device parses the push data to extract the lesion number, image index and its corresponding morphological change parameters, such as lesion area, edge curvature, morphological matching degree, etc. Then, the terminal device calls the historical ultrasound image corresponding to the lesion from the local storage or remote database, and compares the lesion contour according to the time series, and calculates the lesion contour change rate in each frame image. For progressive lesions, it is necessary to mark the change area on the image. The specific marking method is to use pixel-level region segmentation technology to extract the lesion area and generate an isovalue curve, and then mark it in the image. The semi-transparent red area is superimposed for highlighting. At the same time, the dynamic changes of the lesion outline are displayed in the image in the form of a dynamic heat map, and the color gradient corresponding to different degrees of change is set. For example, the area with a change rate within the range of 0.01-0.05 is marked as yellow, the area with a change rate of 0.05-0.1 is marked as orange, and the area greater than 0.1 is marked as red. In addition, the annotated data also includes parameters such as the maximum diameter of the lesion and the rate of decrease of morphological similarity, so that doctors can quickly judge the changing trend of the lesion. The annotated data is stored in the terminal device database and can be transmitted back to the central database through a remote connection to finally generate a progressive lesion marker.

[0063] Table 5.2 Calculation results of lesion area change rate

[0064] As shown in Table 5.2, the contour change rates of different lesions correspond to different color markers, which are used for intuitive display on the image to assist doctors in quickly identifying the changing trend of lesions.

[0065] The remote diagnosis data integration submodule extracts historical classification information of the lesion area based on progressive lesion marking, matches the image data of similar lesions, counts the frequency of lesion area changes, integrates image matching information, and establishes remote diagnosis reference data; Based on the progressive lesion mark, the historical classification information of the lesion area is extracted. First, the storage system is called to query the historical classification records of the lesion at different time points, including whether it has been classified as progressive, static or regressive lesions. At the same time, the image data of similar lesions are matched to calculate the change frequency of the lesion area. The change frequency is calculated by counting the fluctuation amplitude of the lesion morphological matching degree over time and calculating its standard deviation. If the standard deviation exceeds the set threshold, the lesion is considered to have abnormal fluctuation characteristics. For example, the standard deviation of the morphological matching degree of a lesion in the past 30 days is calculated as follows: ; in, For the The matching degree of the lesion morphology of the day, is the mean matching value, is the number of days in the time series. Assume that the historical matching data of a lesion is as follows: ; Calculate its mean: ; Calculate the variance: ; ; ; ; If the standard deviation is greater than 0.05, the lesion is considered to have changed dramatically, otherwise it is considered to have changed steadily. The threshold is set based on the analysis of the evolutionary characteristics of the lesion, mainly referring to the fluctuation range of the lesion's image morphology, and combining the morphological matching data of a large number of clinical cases to observe the range of changes in matching over time. Clinical data show that the standard deviation of morphological matching for most stable lesions is usually between 0.02 and 0.05, while the standard deviation of progressive lesions is concentrated between 0.06 and 0.12. Therefore, 0.05 is selected as the threshold for drastic changes. This value will be adjusted with the image acquisition time interval, the lesion volume change rate, and tissue elasticity parameters. For example, if the lesion volume growth rate exceeds 5% / month, the standard deviation threshold can be adjusted to 0.045. If the image acquisition time interval is short (such as daily acquisition), the threshold can be appropriately relaxed to 0.055 to adapt to higher frequency data fluctuations.

[0066] As shown in Table 5.3, the calculation results of the morphological matching degree fluctuation of different lesions are given and used for subsequent remote diagnosis data integration.

[0067] Table 5.3 Lesion morphology matching degree fluctuation calculation table

[0068] As shown in Table 5.3, the standard deviation of the matching degree of different lesions determines whether they are classified as stable or fluctuating lesions. Stable lesions have small changes and can be used as follow-up lesions, while fluctuating lesions require further analysis to determine whether there is a trend of malignant change.

[0069] 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 remote ultrasound diagnosis support system based on cloud computing, characterized in that: The system comprises: The echo feature analysis module obtains echo signal data in the remote ultrasound image data stream, extracts the echo intensity value of the lesion area in the image frame, calculates the echo change rate between consecutive image frames, analyzes the echo intensity at different angles through cloud computing, calculates the gradient change in the time series, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and outputs the echo change rate feature data; The dynamic lesion screening module extracts the gray gradient change value of the lesion area in the continuous image frames based on the echo change rate characteristic data, calculates the change rate standard deviation, screens the suspected lesion area, and generates the lesion dynamic screening area; The case image comparison module calculates the lesion morphology edge curve contour parameters based on the lesion dynamic screening area, analyzes the morphological structure similarity of the screening area in combination with the ultrasound case image database, makes a lesion category determination, and generates a lesion image matching record; The lesion trend classification module analyzes the morphological change trend of the lesion area in the ultrasound image time series according to the lesion image matching record, obtains the slope of the change trend within the time period, calculates the trend stability index, divides the lesion type, and generates a lesion evolution classification result.

2. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The echo change rate characteristic data includes echo intensity distribution data, echo change gradient data, and abnormal echo area. The lesion dynamic screening area includes the suspected lesion boundary, the grayscale gradient stable area, and the change rate standard deviation range. The lesion image matching record includes the lesion morphological characteristic parameters, edge curve contour matching degree, and ultrasound case image similarity analysis results. The lesion evolution classification results include the lesion morphological change trend, trend stability index, and lesion type classification information.

3. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The echo feature analysis module comprises: The echo signal extraction submodule obtains the echo signal data in the remote ultrasound image data stream, detects the echo intensity value of the lesion area in the image frame, arranges the echo signals of the continuous image frames in time series, and generates the echo signal data of the lesion area; The echo change calculation submodule adopts the formula based on the echo signal data of the lesion area: ； Calculate the echo intensity change rate between adjacent image frames , generate echo rate of change data, where Representative The echo strength value of the frame, Representative The echo strength value of the frame, Representative The echo strength value of the frame, Represents the number of image frames; The dynamic gradient screening submodule calculates the gradient change value in the time series based on the echo change rate data, screens the echo signal area whose change rate exceeds the dynamic echo gradient reference value, and generates echo change rate characteristic data.

4. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The dynamic lesion screening module comprises: The grayscale gradient calculation submodule detects the grayscale value of the lesion area through continuous image frames based on the echo change rate characteristic data, calculates the grayscale change amount between adjacent frames, obtains the grayscale gradient change value of the lesion area in the image frame sequence, and calculates and outputs the grayscale gradient change rate; The change rate standard deviation calculation submodule calculates the mean of the gray gradient change rate according to the gray gradient change rate, counts the square sum of the deviations of each image frame relative to the mean, and calculates the change rate standard deviation; The stable interval screening submodule sets the upper and lower thresholds of the stable interval based on the standard deviation of the change rate, screens the lesion areas of all image frames, and extracts the suspected lesion areas that meet the range of the stable interval, using the formula: ； Calculation of screening stability index , screen the suspected lesion areas that meet the stable rate range and generate the lesion dynamic screening area, where, Representative The gray gradient change rate of the frame, Represents the mean gray gradient change rate, Represents the standard deviation of the gray gradient change rate, Represents the weight coefficient of the current frame in the sequence, Represents the number of image frames.

5. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The case image comparison module includes: The morphological edge calculation submodule detects the edge contour of the lesion area based on the dynamic lesion screening area, calculates the curvature change, contour closure and area ratio of the edge curve, and obtains the lesion morphological edge parameters; The structural similarity analysis submodule compares the lesion morphological edge parameters with the morphological parameters in the ultrasound case image database using the formula: ； Calculate the output lesion morphology matching ,in, Representative The morphological parameters of the lesions in the group database, Represents the morphological parameters of the current lesion area, Represents the curvature value of the lesion area, represents the curvature value of the database matching lesion, Represents the number of morphological parameter comparison items in the database, represents the number of lesion curvature comparison items; The lesion category attribution submodule sets a category determination threshold based on the lesion morphology matching degree, screens the lesion category with the highest matching degree, obtains the corresponding category attribution information, and establishes a lesion image matching record.

6. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The lesion trend classification module includes: The lesion morphology trend calculation submodule detects the morphological changes of the lesion area in the time series based on the lesion image matching record, calculates the rate of change of the lesion contour in each time period, fits the morphological change trend curve of the lesion area, and obtains the change trend slope data; The trend stability analysis submodule calculates the standard deviation of the trend slope of each segment in the time series according to the change trend slope, and uses the formula: ； Calculating the Trend Stability Index ,in, Representative The slope of the trend in a period of time, Represents the mean value of the slope of the changing trend, represents the rate of change of lesion morphology, represents the change rate of lesion morphology matching, represents the number of time periods in the time series, represents the number of lesion matching change items; The lesion type classification submodule sets the change rate and stability classification criteria based on the trend stability index, screens the lesion types that meet the specific trend pattern, and establishes the lesion evolution classification results.

7. The cloud computing-based remote ultrasound diagnosis support system according to claim 1, characterized in that: The system also includes a remote diagnostic support module; The remote diagnosis support module associates the remote ultrasound terminal device based on the lesion evolution classification result, pushes the classification result to the terminal device, marks the progressive lesion area, and generates remote diagnosis reference data; The remote diagnosis reference data includes the marking results of lesion progressive changes, the classification push results of the remote ultrasound terminal, and the remote diagnosis auxiliary analysis information.

8. The cloud computing-based remote ultrasound diagnosis support system according to claim 7, characterized in that: The remote diagnosis support module includes: The classification result push submodule detects the matching remote ultrasound terminal device based on the lesion evolution classification result, extracts the unique identification information of the terminal device, pushes the lesion classification result to the target terminal device, and obtains the remote push data; The lesion region marking submodule detects the progressive lesion region in the classification result according to the remote push data, obtains the lesion region coordinate information, calculates the region boundary features, performs image annotation, and generates a progressive lesion mark; The remote diagnosis data integration submodule extracts the historical classification information of the lesion area based on the progressive lesion mark, matches the image data of similar lesions, counts the frequency of changes in the lesion area, integrates the image matching information, and establishes remote diagnosis reference data.

Images