Visual monitoring platform for postoperative muscle recovery process of gastric cancer patient and multi-terminal cooperation method

Abstract

The invention discloses a stomach cancer patient postoperative muscle recovery process visual monitoring platform and a multi-terminal cooperation method, and the method comprises the following steps: setting an initial scanning layer thickness for scanning according to the specific condition of a patient before CT examination; the CT scanning equipment is used for scanning and imaging the cross section of the postoperative muscular tissue of a gastric cancer patient, continuous dynamic image data are obtained by collecting scanning layer data in real time, and the real-time change condition in the muscular tissue is recorded. According to the intelligent prediction result of the convolutional neural network, the CT scanning layer thickness is automatically optimized in real time, and particularly, the scanning layer thickness is actively reduced when a high-risk gas sign is found, so that the image acquisition precision is remarkably improved, an abnormal gas region with relatively small volume and relatively low density in postoperative muscular tissue is better captured, and the postoperative muscular tissue quality is improved. Therefore, the space average effect and the missed diagnosis risk caused by traditional fixed layer thickness scanning are remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of postoperative muscle recovery, and particularly to a visual monitoring platform for the muscle recovery process of gastric cancer patients after surgery and a multi-terminal collaboration method. Background Art

[0002] The visual monitoring and multi-terminal collaboration of the muscle recovery process of gastric cancer patients after surgery refer to the collaborative work of multiple intelligent devices and systems to achieve dynamic monitoring, data analysis, and visual presentation of the muscle recovery situation of gastric cancer patients after surgery, so as to assist clinical decision-making and personalized rehabilitation management. Specifically, the system may include wearable devices (such as electromyography sensors, smart bracelets), imaging analysis techniques (such as CT, MRI), biomechanical modeling, artificial intelligence algorithms, and telemedicine platforms, etc. Through multi-modal data fusion, key indicators such as muscle mass, strength, and mobility of patients are collected in real time, and are synchronously displayed and interacted on multiple terminals (such as doctor terminals, patient terminals, and rehabilitation institution terminals). Multi-terminal collaboration means that doctors can remotely view the patient's recovery progress and adjust the rehabilitation plan, patients can independently understand their own recovery situation, and rehabilitation institutions can optimize the training plan based on the data, thereby improving the rehabilitation effect and reducing the risk of postoperative complications.

[0003] During the muscle recovery process of gastric cancer patients after surgery, the main role of CT (Computed Tomography) is to quantitatively evaluate muscle mass, muscle atrophy, and the degree of fat infiltration, providing objective data support for postoperative rehabilitation management. CT can accurately measure the cross-sectional area of skeletal muscle (SMA) through high-resolution cross-sectional imaging to evaluate muscle volume changes and detect the presence of sarcopenia. In addition, CT can also analyze muscle density to judge intramuscular fat infiltration (muscle fatty degeneration), and these indicators are closely related to the patient's postoperative survival rate, physical function recovery, and complication risk. By performing regular CT scans before and after surgery, doctors can dynamically monitor muscle mass changes, evaluate the rehabilitation efficacy, and formulate personalized nutrition and rehabilitation exercise plans for patients with muscle atrophy or fatty degeneration, thereby improving the postoperative prognosis and quality of life.

[0004] During the muscle recovery process of gastric cancer patients after surgery, abnormal low-density gas shadows may form inside the muscle tissue, especially in patients with postoperative infection or hemodynamic instability. Currently, when obtaining muscle cross-sectional images through CT scans, clinicians mostly use a relatively large and fixed scan layer thickness (such as 5 mm or more). Such a thick scan layer may be difficult to capture small-volume gas embolisms during dynamic imaging, resulting in the omission or underestimation of the volume of gas embolisms. Once a gas embolism fails to be accurately identified in a timely manner, it may enter the venous system and flow into the pulmonary circulation with the blood, thereby triggering pulmonary embolism, and in severe cases, even leading to fatal air embolism.

[0005] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a visual monitoring platform for the muscle recovery process after gastric cancer surgery and a multi-terminal collaboration method. According to the intelligent prediction results of the convolutional neural network, the CT scan layer thickness is automatically and real-time optimized, especially when high-risk gas signs are detected, the scan layer thickness is actively reduced, so that the imaging acquisition accuracy is significantly improved, and it is better to capture small-volume and low-density abnormal gas areas in the postoperative muscle tissue, thereby significantly reducing the spatial averaging effect and missed diagnosis risk brought by traditional fixed-layer thickness scanning, so as to solve the problems in the above background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solutions: A multi-terminal collaboration method for visual monitoring of the muscle recovery process after gastric cancer surgery, including the following steps:

[0008] Before the CT examination starts, set the starting scan layer thickness for the scan according to the specific situation of the patient;

[0009] Use a CT scanning device to scan and image the cross-section of the muscle tissue after gastric cancer surgery. By real-time collecting the scan layer data, continuous dynamic image data is obtained, and the real-time changes inside the muscle tissue are recorded;

[0010] Perform real-time processing on the images collected from each layer to obtain the density values of different pixels on the cross-section of the muscle tissue, capture the suspected parts with significantly reduced density, and mark the suspected parts as potential risk areas;

[0011] For the potential risk areas, extract the suspected gas image features. After in-depth analysis of the extracted features, the analyzed gas image data is used as a feature vector and input into the trained convolutional neural network, and the convolutional neural network is used to intelligently predict the gas presence risk in the muscle after gastric cancer surgery;

[0012] When it is recognized that there is a high-risk gas distribution in the muscle after gastric cancer surgery, an adaptive regulation mechanism is automatically triggered, and the actual scan layer thickness during CT scanning is adaptively regulated according to the prediction results of the convolutional neural network. When it is recognized that there is a gas risk in the muscle after gastric cancer surgery, the actual scan layer thickness during CT scanning is automatically reduced.

[0013] Preferably, the steps for performing real-time processing on the images collected from each layer to obtain the density values of different pixels on the cross-section of the muscle tissue, capture the suspected parts with significantly reduced density, and mark the suspected parts as potential risk areas are as follows:

[0014] Preprocess the original image data collected in real time at each level during the CT scan, including denoising, filtering, and image enhancement, to improve the image quality;

[0015] Extract the density numerical values at the pixel or voxel level for the preprocessed cross-sectional images layer by layer. By calculating the Hounsfield Unit value corresponding to each pixel, accurately obtain the density distribution inside the muscle tissue;

[0016] Automatically compare and analyze the density values at each position with the normal range density values of the muscle tissue, and use the set density difference threshold to identify the abnormally low-density areas;

[0017] Automatically mark the detected areas with significantly reduced density, and record and visually display them as potential abnormal gas presence areas, and mark this suspected part as a potential risk area.

[0018] Preferably, for the potential risk area, extract the suspected gas image features. Among them, the extracted features include the morphological complexity and edge sharpness of the potential risk area. After in-depth analysis of the morphological complexity and edge sharpness, generate the regional fractal dimension index and edge sharpness index respectively. Evaluate the complexity of the internal structure and the sharpness of the outer edge of this area through the regional fractal dimension index and edge sharpness index, and intuitively reflect the presence of abnormal gas in the muscles of gastric cancer patients after surgery.

[0019] Preferably, input the regional fractal dimension index and edge sharpness index after in-depth analysis as feature vectors into the trained convolutional neural network, and output the gas detection confidence index through the convolutional neural network. Based on the gas detection confidence index, intelligently predict the presence of gas in the muscles of gastric cancer patients after surgery.

[0020] Preferably, when the convolutional neural network that has been trained is used to intelligently predict the gas presence risk in the muscles of gastric cancer patients after surgery, compare and analyze the gas detection confidence index generated with the pre-set confidence reference threshold to identify the gas presence risk in the muscles of gastric cancer patients after surgery. The specific steps are as follows:

[0021] If the gas detection confidence index is greater than the confidence reference threshold, it is determined that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery; if the gas detection confidence index is less than or equal to the confidence reference threshold, it is determined that the image in the muscles of gastric cancer patients after surgery is normal and there is no high-risk gas distribution.

[0022] Preferably, the specific steps for generating the regional fractal dimension index after in-depth analysis of the morphological complexity of the potential risk area are as follows:

[0023] When detecting the morphological complexity of potential risk areas, multiple groups of morphological dilation scales are sequentially set, and the boundaries of potential risk areas are decomposed and scanned layer by layer. For each morphological scale i, with a value range from 1 to k, where k represents the total number of morphological scales, a morphological dilation factor is defined to quantify the size of the structural element selected when performing the dilation operation; at the same time, record the number of connected boundaries of the potential risk area at the morphological scale i , that is, after the i-th morphological operation, how many non-overlapping independent connected components the remaining boundary is divided into; by repeating this process at multiple scales, the complex extension characteristics of the internal structure of the potential risk area as manifested by the change in morphological scale can be captured layer by layer;

[0024] After completing the multi-scale morphological decomposition, calculate the regional fractal dimension index. The calculation formula of the regional fractal dimension index is: , where: is the regional fractal dimension index.

[0025] Preferably, the calculation expression of the morphological dilation factor is: , where: is the perimeter of the region at the i-th dilation scale, is the area of the region at the i-th dilation scale, , are morphological adjustment parameters used to adjust the relative influence weights of the perimeter and area of the region, is the boundary mutation response coefficient used to adjust the sensitivity of morphological changes, is the cumulative boundary change amplitude, indicating the degree of mutation of the region boundary at each dilation scale.

[0026] Preferably, the specific steps for generating the edge sharpness index after deeply analyzing the edge sharpness of the potential risk area are as follows:

[0027] Extract the boundary of the potential risk area through morphological operations or region growing algorithms. Subsequently, use a discrete gradient operator to calculate the local gradient magnitude of each boundary point. The calculation expression is: , where: represents the local gradient magnitude of the boundary point , represents the gray value of the image at the coordinate , and correspond to the adjacent points above, below, left, and right of the current pixel respectively;

[0028] After obtaining the local gradients of all boundary points, construct the edge sharpness index through the following formula to measure the "outer edge sharpness" level of the potential risk area as a whole. The generation expression of the edge sharpness index is: , where: is the edge sharpness index, B is the set of boundary pixels, represents the product operation for all boundary points.

[0029] Preferably, when a high-risk gas distribution is detected in the muscle after gastric cancer surgery, an adaptive regulation mechanism is automatically triggered, and the actual scan slice thickness during CT scanning is adaptively regulated according to the result predicted by the convolutional neural network. The specific steps are as follows:

[0030] When the convolutional neural network detects a high-risk gas distribution in the muscle after gastric cancer surgery, an adaptive regulation mechanism is automatically triggered to dynamically adjust the slice thickness of CT scanning and calculate the new scan slice thickness. The calculation expression of the new scan slice thickness is: , where: is the actual scan slice thickness of the adjusted CT, which is used to optimize the detection ability of small-volume bubbles, is the starting scan slice thickness, is the adaptive regulation coefficient, which determines the sensitivity of slice thickness adjustment, is the gas detection confidence index, which is calculated by the convolutional neural network, is the set minimum confidence index, is the set maximum confidence index;

[0031] After the calculation of the adjusted actual scan slice thickness is completed, the CT scan slice number will be further optimized to adapt to the new slice thickness setting and improve the accurate detection ability of small-volume bubbles. The optimization formula of the CT scan slice number is: , where: is the new CT scan slice number, that is, the number of cross-sectional images to be obtained under the adjusted scan slice thickness, is the starting CT scan slice number.

[0032] The visualization monitoring platform for the muscle recovery process after gastric cancer surgery includes a scan parameter initialization module, an image acquisition and dynamic monitoring module, an image processing and risk area marking module, a gas feature analysis and intelligent prediction module, and a scan slice thickness adaptive regulation module;

[0033] The scan parameter initialization module sets the starting scan slice thickness for the scan according to the specific situation of the patient before the CT examination;

[0034] The image acquisition and dynamic monitoring module uses a CT scanning device to scan and image the cross-section of the muscle tissue after gastric cancer surgery, obtains continuous dynamic image data by real-time collecting scan slice data, and records the real-time changes inside the muscle tissue;

[0035] The image processing and risk area marking module performs real-time processing on the images collected at each layer, obtains the density values of different pixels of the muscle tissue on the cross-section, captures the suspected parts with significantly reduced density, and marks the suspected parts as potential risk areas;

[0036] The gas feature analysis and intelligent prediction module extracts the suspected gas image features for the potential risk areas. After deeply analyzing the extracted features, the analyzed gas image data is used as a feature vector and input into the trained convolutional neural network. The convolutional neural network is used to intelligently predict the gas presence risk in the muscles of gastric cancer patients after surgery;

[0037] The scan layer thickness adaptive regulation module automatically triggers the adaptive regulation mechanism when it identifies a high-risk gas distribution in the muscles of gastric cancer patients after surgery. According to the prediction result of the convolutional neural network, it adaptively regulates the actual scan layer thickness during CT scanning. When it identifies a gas risk in the muscles of gastric cancer patients after surgery, it automatically reduces the actual scan layer thickness during CT scanning.

[0038] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0039] According to the intelligent prediction result of the convolutional neural network, the present invention automatically and real-time optimizes the CT scan layer thickness. Especially when detecting high-risk gas signs, it actively reduces the scan layer thickness, significantly improving the image acquisition accuracy, better capturing the abnormal gas areas with smaller volume and lower density in the postoperative muscle tissue, thereby significantly reducing the spatial averaging effect and missed diagnosis risk brought by traditional fixed layer thickness scanning. In addition, through the gas detection confidence index output by the convolutional neural network, it can realize the intelligent quantitative evaluation and prediction of the gas presence risk, effectively reducing the errors and misjudgments caused by doctors' subjective judgments, further improving the sensitivity and accuracy of early identification and intervention of abnormal gases, timely detecting gas embolism and making intervention decisions, avoiding abnormal gases from entering the venous system and causing serious complications such as pulmonary embolism or fatal air embolism, and maximizing the medical safety and rehabilitation quality of gastric cancer patients after surgery. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is the method flow chart of the multi-terminal collaborative method for visual monitoring of the muscle recovery process of gastric cancer patients after surgery according to the present invention.

[0042] Figure 2This is a schematic diagram of the modules of the visual monitoring platform for the postoperative muscle recovery process of gastric cancer patients in the present invention. Specific implementation manners

[0043] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be more thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art.

[0044] The present invention provides a Figure 1 visual monitoring multi - terminal collaboration method for the postoperative muscle recovery process of gastric cancer patients as shown in the following steps:

[0045] Before the CT examination starts, set the starting scan slice thickness according to the specific conditions of the patient (including postoperative status, individual body shape, surgical site, etc.).

[0046] The starting scan slice thickness is a relatively moderate initial scan slice thickness (for example, 3 - 5 mm) rather than blindly choosing a too large scan slice thickness. The purpose of this is to balance the imaging quality and the early detection accuracy, minimize the "averaging" effect on small - volume bubbles, and reduce the possibility of misdiagnosis or missed diagnosis. If the slice thickness is too large, it is often difficult to accurately capture tiny bubbles; if it is too small, although the recognition degree is high, both the scanning time and the radiation dose will increase. By optimizing the slice thickness at the initial stage, room for subsequent adaptive regulation can be reserved, enabling the system to maintain the basic image quality while having more flexible adjustment space.

[0047] Use a CT scanning device to scan and image the cross - section of the muscle tissue of a gastric cancer patient after surgery. By real - time collecting the data of the scanned layer, obtain continuous dynamic image data and record the real - time changes inside the muscle tissue.

[0048] These image data can cover multi - level and multi - angle tissue structure information, thereby obtaining a comprehensive, clear data set for subsequent analysis. By obtaining dynamic data in real - time, sufficient data support can be provided for identifying abnormal density changes (such as abnormal bubbles), which helps to improve the accuracy of subsequent analysis and diagnosis.

[0049] Perform real - time processing on the images collected from each layer to obtain the density values of different pixels (or voxels) of the muscle tissue in the cross - section, capture the suspected parts with significantly reduced density, and mark the suspected parts as potential risk areas.

[0050] The specific implementation process can be refined into the following 4 steps:

[0051] First, preprocess the original image data collected in real time at each level during the CT scan, including denoising, filtering, and image enhancement, to improve the image quality and facilitate the subsequent accurate extraction of density values;

[0052] Secondly, layer by layer, perform density value extraction at the pixel or voxel level on the preprocessed cross-sectional images. By calculating the Hounsfield Unit (HU) value corresponding to each pixel (or voxel), accurately obtain the density distribution within the muscle tissue;

[0053] Next, automatically compare and analyze the density values at each position with the normal range density values of the muscle tissue. Using the set density difference threshold (such as significantly lower than the normal HU value of the muscle tissue, for example, lower than -500HU or even lower), quickly identify the abnormally low-density areas;

[0054] Finally, automatically mark these detected areas with significantly reduced density and record and visually display them as potential areas with abnormal gas presence, and mark this suspected part as a potential risk area, thus providing key evidence for subsequent in-depth diagnosis, analysis, and decision-making.

[0055] For the potential risk area, extract the suspected gas image features. After deeply analyzing the extracted features, input the analyzed gas image data as feature vectors into the trained convolutional neural network, and use the convolutional neural network to intelligently predict the risk of gas presence in the muscles after gastric cancer surgery;

[0056] For the potential risk area, extract the suspected gas image features. Among them, the extracted features include the morphological complexity and edge sharpness of the potential risk area. After deeply analyzing the morphological complexity and edge sharpness, generate the regional fractal dimension index and edge sharpness index respectively. Evaluate the complexity of the internal structure and the sharpness of the outer edge of this area through the regional fractal dimension index and edge sharpness index, intuitively reflect the presence of abnormal gas in the muscles after gastric cancer surgery, and input the deeply analyzed regional fractal dimension index and edge sharpness index as feature vectors into the trained convolutional neural network. Output the gas detection confidence index through the convolutional neural network, and intelligently predict the gas presence situation in the muscles after gastric cancer surgery based on the gas detection confidence index.

[0057] When the morphological complexity of the potential risk area within the muscle tissue of a gastric cancer patient after surgery is found to be relatively high through CT image analysis, it usually indicates that the patient has a real gas risk hazard. Specifically, normal exudate, mild inflammatory edema, or a small amount of residual gas after surgery in the muscle tissue usually presents as a structure with relatively regular morphology and smooth edges; while pathological abnormal gas (such as bubbles generated by infection, gas accompanying muscle necrosis, or abnormal gas embolism in blood vessels) often shows a complex, irregular, scattered, and multi-directionally extended structure when diffusing in the tissue space, presenting as a blurred, irregular, fragmented, branched, or dendritic complex morphology on the image. Therefore, when the morphological complexity of the potential risk area calculated through CT image processing and analysis is relatively high, it usually indicates that there is a higher real pathological gas risk hazard in the patient's muscle.

[0058] The specific steps for generating the regional fractal dimension index through in-depth analysis of the morphological complexity of the potential risk area are as follows:

[0059] When detecting the morphological complexity of the potential risk area, first set multiple groups of morphological dilation scales in sequence (for example, perform closing and opening operations with structural elements of different sizes), decompose and scan the boundary of the potential risk area layer by layer. For each morphological scale i, with the value range from 1 to k, where k represents the total number of morphological scales, define the morphological dilation factor used to quantify the size of the structural element selected when performing the dilation operation; at the same time, record the number of connected boundaries of the potential risk area at the morphological scale i , that is, after the i-th morphological operation, how many non-overlapping independent connected components the remaining boundary is divided into; by repeating this process at multiple scales, the complex extension characteristics of the internal structure of the potential risk area with the change of morphological scale can be captured layer by layer;

[0060] The function of this step is to expose the different levels of tortuosity and the number of branches of each potential risk area at different dilation scales, thus laying a multi-scale observation foundation for the subsequent calculation of the fractal dimension.

[0061] The calculation expression of the morphological dilation factor is: , where: is the perimeter of the area at the i-th dilation scale (i.e., the length of the outer boundary of the dilated area), is the area of the area at the i-th dilation scale (i.e., the number of pixels or voxels of the dilated area), , are morphological adjustment parameters used to adjust the relative influence weights of the perimeter and area of the area (usually taking , ), is the boundary mutation response coefficient, which is used to adjust the sensitivity of morphological changes (it can be set to ), is the cumulative boundary change amplitude, which represents the degree of mutation of the regional boundary at each expansion scale.

[0062] After completing the multi-scale morphological decomposition, calculate the regional fractal dimension index. The calculation formula of the regional fractal dimension index is: , where: is the regional fractal dimension index;

[0063] By accumulating the sum of "the number of connected boundaries × expansion factor" at each scale and then comparing it with the logarithm of the scale number k, the degree of maintaining irregularity and fragmentation of this potential risk area at different scales can be quantified. If the value of the regional fractal dimension index is higher, it indicates that the potential gas area shows high complexity at multiple scales (such as variable edges and branching extensions), thus suggesting significant morphological complexity, and further indicating that this area is more likely to be a real abnormal gas aggregation area and requires further attention.

[0064] As can be seen from the regional fractal dimension index, the larger the value of the regional fractal dimension index generated after in-depth analysis of the morphological complexity of the potential risk area usually means the higher the morphological complexity of the potential risk area, thus suggesting a greater risk of abnormal gas in the muscles of gastric cancer patients after surgery; conversely, if the value of the regional fractal dimension index is small, it indicates that the morphology of this area tends to be regular, which may be normal postoperative tissue changes and the gas risk is low. This is because in imaging analysis, normal muscle tissue and postoperative inflammatory reaction areas usually show relatively regular morphologies, while abnormal gas areas show characteristics of complex, irregular boundaries and strong fractal structures, such as branched, honeycomb-like or fragmented distribution morphologies. The regional fractal dimension index can effectively capture these abnormal features by quantifying morphological complexity, edge mutation conditions and multi-scale morphological changes. Therefore, when the value of the regional fractal dimension index is high, it indicates that this area may contain free gas or gas bubble aggregation.

[0065] For potential risk areas, if the edges in the CT images show a high degree of sharpness, it usually indicates the potential risk of abnormal gas in the muscles of gastric cancer patients after surgery. Specifically, abnormal gas appears as an area with a density significantly lower than that of the surrounding tissues in the CT images. The density difference at the interface between the gas and the muscle tissue is significant, so it usually has relatively sharp and clear edge features. In contrast, the boundaries of postoperative effusion or inflammatory edema areas are mostly in a blurred and gradual state, with a lower degree of edge sharpness and no obvious sense of clear demarcation. Therefore, when the edge sharpness of a potential risk area is found to be high, it indicates that this area is more likely to be a real gas embolism area, meaning that there may be a real gas risk in the muscles of gastric cancer patients after surgery. This requires clinicians to pay immediate attention and consider further examinations or intervention measures to prevent the spread of gas embolism into the blood circulation system and cause serious complications.

[0066] The specific steps for generating the edge sharpness index after a deep analysis of the edge sharpness of potential risk areas are as follows:

[0067] Extract the boundary of the potential risk area through morphological operations or region growing algorithms. Subsequently, use the discrete gradient operator to calculate the local gradient magnitude of each boundary point. The calculation expression is: , where: represents the boundary point 's local gradient magnitude, represents the gray value of the image at the coordinate location, and correspond to the adjacent points above, below, left, and right of the current pixel respectively;

[0068] The purpose of this step is to accurately identify the "outer edge" of the suspected gas area and quantify the gradient intensity of each boundary point, laying a solid foundation for calculating the overall edge sharpness of the area.

[0069] After obtaining the local gradients of all boundary points, construct the edge sharpness index through the following formula to measure the "outer edge sharpness" level of the potential risk area as a whole. The generation expression of the edge sharpness index is: , where: is the edge sharpness index, B is the set of boundary pixels (or voxels), П represents the product operation for all boundary points, The product is then mapped to a logarithmic space that is easier to compare and analyze. Through this product-logarithmic form, each high gradient value on the boundary will significantly amplify the impact in the final edge sharpness index. Therefore, when the edge of the potential risk area shows a higher sharpness, the edge sharpness index will increase accordingly, indicating that the area is more likely to be an abnormal area where gas actually exists. The purpose of this step is to integrate the gradient information of discrete boundary points into an overall edge sharpness index, fully highlighting the obvious density difference between gas and tissue, and providing a clear basis for subsequent diagnosis or intervention.

[0070] From the edge sharpness index, it can be seen that the larger the performance value of the edge sharpness index generated after in-depth analysis of the edge sharpness of the potential risk area, the higher the edge sharpness of the potential risk area, indicating that the tissue density change in the area is more drastic, which is usually closely related to the presence of abnormal gas. Since the density of gas in CT images is extremely low (close to -1000 HU), it forms a significant density contrast with the surrounding muscle tissue (30~70 HU). Therefore, if gas does exist, the grayscale gradient of the regional boundary will be higher, and the edge sharpness index will also increase accordingly, indicating that the risk of gas in the muscle tissue of gastric cancer patients after surgery is higher; on the contrary, if the edge sharpness index is low, it indicates that the gradient of the regional boundary changes slowly, which may only be tissue edema, hematoma or inflammatory exudate, rather than a real gas area. Therefore, the size of the edge sharpness index can effectively evaluate the possibility of abnormal gas in postoperative muscle and provide an important quantitative reference for clinical intervention.

[0071] A trained convolutional neural network refers to a stable model obtained after sufficient iteration and adjustment on a large-scale and highly targeted medical image dataset, which can exhibit good prediction and generalization capabilities on new input data. Specifically, researchers first need to construct or collect a CT image dataset containing various postoperative gas distribution patterns, including normal muscle images and various possible abnormal situations (such as large-area gas, fine bubbles, different infection degrees, etc.), and accurately label them, indicating which areas in the images actually contain gas and which areas are only pseudo-low-density such as postoperative edema or hematoma. Subsequently, the labeled data is used to repeatedly train the convolutional neural network (CNN): the sample images and their corresponding labels (whether containing gas, gas location, area or severity) are input into the CNN, and the convolutional layer and fully connected layer in the network will perform hierarchical feature extraction and weighted calculation on the images according to the current parameters, and output a preliminary prediction (for example, predicting the probability that the area belongs to "containing gas" or "not containing gas"). After obtaining the prediction result, the system will calculate the deviation or loss value between the prediction and the labeled true value, and then continuously correct the network weights through backpropagation and optimization algorithms (such as stochastic gradient descent or Adam, etc.) to enable the CNN to gradually learn to extract feature maps that can better distinguish gas features. As the number of training times increases, various indicators such as accuracy, recall rate, and F1 score of the model on the validation set or test set are continuously improved; when the prediction effect of the network reaches the expectation or no longer rises significantly after multiple validations, it can be regarded as entering the "convergent" state. At this time, the CNN that has undergone multiple rounds of iteration and optimization is regarded as "trained", and the internal convolutional kernels, biases, and weight distributions of the model have learned to make relatively accurate judgments for different types of gas patterns or similar low-density areas, and also have good recognition capabilities when facing new samples.

[0072] During the above training process, through continuous adaptive learning of the convolutional kernel, the CNN gradually masters a large number of implicit gas detection rules. Starting from simple edge detection or directional texture discrimination at the beginning, it gradually condenses the key elements for judging "whether there is gas" and "the morphological characteristics of the gas area". These elements not only include simple pixel gray level or density value discrimination, but also involve higher-level spatial distribution, regional connectivity, edge sharpness, and the correlation characteristics between organizational structures. For this solution, by inputting high-order feature vectors such as "regional fractal dimension index" and "edge sharpness index" into the already trained CNN, the network will further fuse, weight, and perform pattern matching on these input features based on the convolutional feature maps learned during the training stage, and output a "gas detection confidence index". This gas detection confidence index essentially reflects the "confidence" level of the network in whether the input sample contains gas and its severity, and can comprehensively evaluate the gas situation in the patient's postoperative muscles in combination with other-level imaging information. Since the CNN repeatedly learns various gas distribution patterns during the training process and has a deep understanding of the data change trends, abnormal features, and differences from normal muscle tissues, the final gas detection confidence index can usually match the clinical reality well, thus helping medical staff or automated systems detect and intervene in potential gas embolism risks at the earliest time point. This not only improves the prevention and treatment efficiency of postoperative complications, but also provides more accurate medical support for the patient's postoperative rehabilitation to a certain extent.

[0073] The convolutional neural network is not specifically limited here, and any deep learning model that can achieve comprehensive analysis of the regional fractal dimension index and the edge sharpness index to generate a gas detection confidence index is acceptable. To implement the technical solution of the present invention, the present invention provides a specific implementation method; the expression for generating the gas detection confidence index is: , where, , are respectively the preset proportionality coefficients of the regional fractal dimension index and the edge sharpness index , and , are both greater than 0.

[0074] As can be seen from the gas detection confidence index, the larger the performance value of the regional fractal dimension index generated after in-depth analysis of the morphological complexity of the potential risk area, and the larger the performance value of the edge sharpness index generated after in-depth analysis of the edge sharpness of the potential risk area. That is, when the trained convolutional neural network conducts intelligent prediction on the gas presence risk in the muscles of gastric cancer patients after surgery, the larger the performance value of the gas detection confidence index generated, the greater the risk of gas presence in the muscles of gastric cancer patients after surgery. Conversely, it indicates that the risk of gas presence in the muscles of gastric cancer patients after surgery is smaller.

[0075] Compare and analyze the gas detection confidence index generated when the trained convolutional neural network conducts intelligent prediction on the gas presence risk in the muscles of gastric cancer patients after surgery with the pre-set confidence reference threshold to identify the gas presence risk in the muscles of gastric cancer patients after surgery. The specific steps are as follows:

[0076] If the gas detection confidence index is greater than the confidence reference threshold, it is determined that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery; if the gas detection confidence index is less than or equal to the confidence reference threshold, it is determined that the image of the muscles of gastric cancer patients after surgery is normal and there is no high-risk gas distribution.

[0077] When it is identified that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery, the adaptive regulation mechanism is automatically triggered to adaptively regulate the actual scan layer thickness during CT scanning according to the prediction result of the convolutional neural network. When it is identified that there is a gas risk in the muscles of gastric cancer patients after surgery, the actual scan layer thickness during CT scanning is automatically reduced to accurately identify small-volume gas embolisms;

[0078] When it is identified that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery, the adaptive regulation mechanism is automatically triggered to adaptively regulate the actual scan layer thickness during CT scanning according to the prediction result of the convolutional neural network. The specific steps are as follows:

[0079] When the convolutional neural network detects a high-risk gas distribution in the muscles of gastric cancer patients after surgery, the adaptive regulation mechanism is automatically triggered to dynamically adjust the layer thickness of CT scanning to ensure the use of finer imaging parameters in the high-risk area to enhance the visualization ability of small-volume bubbles. To calculate the new scan layer thickness, the following formula is defined: , where: is the actual scan layer thickness of the adjusted CT, which is used to optimize the detection ability of small-volume bubbles, is the starting scan layer thickness, that is, the initially set default layer thickness value, usually between 3mm and 5mm, is the adaptive regulation coefficient, which determines the sensitivity of the layer thickness adjustment, and the range is usually set at , the larger the value, the more obvious the adjustment amplitude, is the gas detection confidence index, calculated by a convolutional neural network. The larger the value, the higher the likelihood of high-risk gases. is the set minimum confidence index (if it is 0, it means there is no gas risk). is the set maximum confidence index, usually representing a gas area with extremely high risk;

[0080] This step uses the normalized gas detection confidence index to dynamically adjust the scan slice thickness. When the gas confidence index is high, the scan slice thickness is reduced, thereby improving the spatial resolution and enhancing the detection ability for small-volume bubbles. At the same time, the adjustment amplitude is controlled by an adaptive regulation coefficient to ensure that the CT scan slice thickness does not change suddenly, so as to balance the imaging quality and radiation dose.

[0081] After calculating the adjusted actual scan slice thickness, the number of CT scan slices will be further optimized to adapt to the new slice thickness setting and improve the precise detection ability for small-volume bubbles. The optimization formula for the number of CT scan slices is: , where: is the new number of CT scan slices, that is, the number of cross-sectional images to be obtained under the adjusted scan slice thickness, is the starting number of CT scan slices, usually depending on the patient's body size and the scan area, and the initial value may be between 50 and 100 slices;

[0082] This step ensures that when the scan slice thickness is reduced, the number of CT scans is correspondingly increased to maintain the complete coverage of the original scan range, while improving the spatial resolution and ensuring the continuous tracking ability for small bubbles. This adaptive adjustment mechanism can accurately identify small-volume gas embolisms without increasing the patient's additional radiation exposure and improve the sensitivity of postoperative gas detection.

[0083] The above-mentioned multi-terminal collaborative method for visual monitoring of the muscle recovery process after gastric cancer surgery can effectively improve the accurate recognition ability of small-volume gas embolism through the intelligent adaptive regulation mechanism of CT scan slice thickness assisted by deep learning. Specifically, this method first automatically and real-time optimizes the CT scan slice thickness according to the intelligent prediction results of the convolutional neural network. Especially when high-risk gas signs are detected, it actively reduces the scan slice thickness, significantly improving the image acquisition accuracy, better capturing the abnormal gas areas with small volume and low density in the postoperative muscle tissue, thus significantly reducing the spatial averaging effect (Partial Volume Effect) and missed diagnosis risk brought by traditional fixed slice thickness scanning. In addition, through the gas detection confidence index output by the convolutional neural network, the intelligent quantitative evaluation and prediction of the gas presence risk can be realized, effectively reducing the errors and misjudgments caused by the subjective judgment of doctors, further improving the sensitivity and accuracy of early recognition and intervention of abnormal gas, timely detecting gas embolism and making intervention decisions, avoiding the entry of abnormal gas into the venous system and causing serious complications such as pulmonary embolism or fatal air embolism, and maximizing the medical safety and rehabilitation quality of gastric cancer patients after surgery.

[0084] The present invention provides a Figure 2 visual monitoring platform for the muscle recovery process after gastric cancer surgery as shown, including a scan parameter initialization module, an image acquisition and dynamic monitoring module, an image processing and risk area marking module, a gas feature analysis and intelligent prediction module, and a scan slice thickness adaptive regulation module;

[0085] The scan parameter initialization module sets the initial scan slice thickness for the scan according to the specific situation of the patient before the CT examination starts;

[0086] The image acquisition and dynamic monitoring module uses a CT scan device to scan and image the cross-section of the muscle tissue of a gastric cancer patient after surgery. By real-time collecting the scan slice data, continuous dynamic image data is obtained, and the real-time changes inside the muscle tissue are recorded;

[0087] The image processing and risk area marking module performs real-time processing on the images collected at each slice, obtains the density values of different pixels of the muscle tissue on the cross-section, captures the suspected parts with significantly reduced density, and marks the suspected parts as potential risk areas;

[0088] The gas feature analysis and intelligent prediction module extracts the suspected gas image features for the potential risk areas. After deeply analyzing the extracted features, the analyzed gas image data is used as a feature vector and input into the trained convolutional neural network, and the convolutional neural network is used to intelligently predict the gas presence risk in the muscle of a gastric cancer patient after surgery;

[0089] The scan layer thickness adaptive regulation module automatically triggers the adaptive regulation mechanism when it recognizes the existence of high-risk gas distribution in the muscles of gastric cancer patients after surgery, and adaptively regulates the actual scan layer thickness during CT scanning according to the results predicted by the convolutional neural network. When it recognizes the gas risk in the muscles of gastric cancer patients after surgery, it automatically reduces the actual scan layer thickness during CT scanning.

[0090] The multi-terminal collaborative method for visual monitoring of the muscle recovery process of gastric cancer patients after surgery provided by the embodiments of the present invention is implemented through the above-mentioned visual monitoring platform for the muscle recovery process of gastric cancer patients after surgery. The specific methods and processes of the visual monitoring platform for the muscle recovery process of gastric cancer patients after surgery are detailed in the embodiments of the above-mentioned multi-terminal collaborative method for visual monitoring of the muscle recovery process of gastric cancer patients after surgery, and will not be elaborated here.

[0091] The above formulas are all dimensionless and take their numerical calculations. The formula is a formula obtained by collecting a large amount of data for software simulation to get the closest to the real situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0092] As mentioned above, only the specific implementation manners of the present application are described, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claimed rights.

[0093] Only some exemplary embodiments of the present invention have been described by way of illustration above. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. Visual monitoring and multi-terminal collaboration method for the muscle recovery process of gastric cancer patients after surgery, characterized in that, Including the following steps: Before the CT examination starts, set the starting scan slice thickness according to the specific conditions of the patient; Use the CT scanning device to scan and image the cross-section of the muscle tissue after gastric cancer surgery. By collecting the scan slice data in real time, obtain continuous dynamic image data and record the real-time changes inside the muscle tissue; Perform real-time processing on the images collected at each slice to obtain the density values of different pixels on the cross-section of the muscle tissue, capture the suspected areas with significantly reduced density, and mark the suspected areas as potential risk areas; For the potential risk areas, extract the suspected gas image features. After deeply analyzing the extracted features, use the analyzed gas image data as feature vectors and input them into the trained convolutional neural network. Through the convolutional neural network, intelligently predict the gas presence risk in the muscle after gastric cancer surgery; When it is recognized that there is a high-risk gas distribution in the muscle after gastric cancer surgery, automatically trigger the adaptive regulation mechanism, and adaptively regulate the actual scan slice thickness during CT scanning according to the results predicted by the convolutional neural network. When it is recognized that there is a gas risk in the muscle after gastric cancer surgery, automatically reduce the actual scan slice thickness during CT scanning.

2. The multi - terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 1, wherein, Perform real-time processing on the images collected at each slice to obtain the density values of different pixels on the cross-section of the muscle tissue, capture the suspected areas with significantly reduced density, and mark the suspected areas as potential risk areas. The specific steps are as follows: Preprocess the original image data collected in real time at each slice during the CT scanning process, including denoising, filtering, and image enhancement to improve the image quality; Extract the density numerical values at the pixel or voxel level layer by layer for the preprocessed cross-sectional images. By calculating the Hounsfield Unit value corresponding to each pixel, accurately obtain the density distribution inside the muscle tissue; Automatically compare and analyze the density values at each position with the density values in the normal range of the muscle tissue, and use the set density difference threshold to identify the abnormally low-density areas; Automatically mark the detected areas with significantly reduced density, and record and visually display them as potential abnormal gas presence areas, and mark the suspected areas as potential risk areas.

3. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 1, characterized in that, For the potential risk areas, extract the suspected gas image features. Among them, the extracted features include the morphological complexity and edge sharpness of the potential risk areas. After deeply analyzing the morphological complexity and edge sharpness, generate the regional fractal dimension index and edge sharpness index respectively. Evaluate the complexity of the internal structure and the sharpness of the outer edge of the area through the regional fractal dimension index and edge sharpness index, and intuitively reflect the presence of abnormal gas in the muscle after gastric cancer surgery.

4. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 3, characterized in that, Use the deeply analyzed regional fractal dimension index and edge sharpness index as feature vectors and input them into the trained convolutional neural network. Through the convolutional neural network, output the gas detection confidence index, and intelligently predict the gas presence situation in the muscle after gastric cancer surgery based on the gas detection confidence index.

5. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 4, wherein, When the gas detection confidence index generated by the trained convolutional neural network for the intelligent prediction of the risk of gas presence in the muscles of gastric cancer patients after surgery is compared and analyzed with the pre-set confidence reference threshold, the risk of gas presence in the muscles of gastric cancer patients after surgery is identified. The specific steps are as follows: If the gas detection confidence index is greater than the confidence reference threshold, it is determined that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery; if the gas detection confidence index is less than or equal to the confidence reference threshold, it is determined that the imaging of the muscles of gastric cancer patients after surgery is normal and there is no high-risk gas distribution.

6. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 3, wherein The specific steps for generating the regional fractal dimension index after in-depth analysis of the morphological complexity of the potential risk area are as follows: When detecting the morphological complexity of potential risk areas, multiple groups of morphological dilation scales are set in sequence, and the boundaries of potential risk areas are decomposed and scanned layer by layer. For each morphological scale i, with a value range from 1 to k, where k represents the total number of morphological scales, a morphological dilation factor is defined to quantify the size of the structural element selected when performing the dilation operation; at the same time, record the number of connected boundaries of the potential risk area at this morphological scale i , that is, after the i-th morphological operation, how many non-overlapping independent connected components the remaining boundaries are divided into; by repeating this process at multiple scales, the complex extension characteristics of the internal structure of the potential risk area with the change of morphological scale can be captured layer by layer; After completing the multi-scale morphological decomposition, calculate the regional fractal dimension index. The calculation formula for the regional fractal dimension index is as follows: , where: is the regional fractal dimension index.

7. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 6, wherein, The calculation expression of the morphological dilation factor is as follows: , where: is the perimeter of the region at the i-th dilation scale, is the area of the region at the i-th dilation scale, , are morphological adjustment parameters used to adjust the relative influence weights of the perimeter and area of the region, is the boundary mutation response coefficient used to adjust the sensitivity of morphological changes, is the cumulative boundary change amplitude, indicating the degree of mutation of the region boundary at each dilation scale.

8. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 3, wherein, The specific steps for generating the edge sharpness index after in-depth analysis of the edge sharpness of the potential risk area are as follows: The boundaries of potential risk areas are extracted through morphological operations or region growing algorithms. Subsequently, the local gradient magnitude of each boundary point is calculated using a discrete gradient operator, and the calculation expression is: , where: represents the boundary point of the local gradient magnitude, represents the gray value of the image at the coordinate , and correspond to the adjacent points above, below, to the left, and to the right of the current pixel respectively; After obtaining the local gradients of all boundary points, an edge sharpness index is constructed by the following formula to measure the "outer edge sharpness" level of the potential risk area as a whole. The generation expression of the edge sharpness index is: , where: is the edge sharpness index, B is the set of boundary pixels, represents the product operation on all boundary points.

9. The multi-terminal collaborative method for visual monitoring of the muscle recovery process after surgery for gastric cancer patients according to claim 5, characterized in that When it is identified that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery, an adaptive regulation mechanism is automatically triggered, and the actual scan slice thickness during CT scanning is adaptively regulated according to the results predicted by the convolutional neural network. The specific steps are as follows: When the convolutional neural network detects a high-risk gas distribution in the muscles of a gastric cancer patient after surgery, it automatically triggers an adaptive regulation mechanism to dynamically adjust the slice thickness of the CT scan and calculate the new slice thickness. The calculation expression for the new slice thickness is: , where: is the actual slice thickness of the adjusted CT, which is used to optimize the detection ability of small-volume bubbles, is the starting slice thickness, is the adaptive regulation coefficient, which determines the sensitivity of the slice thickness adjustment, is the gas detection confidence index, which is calculated by the convolutional neural network, is the set minimum confidence index, is the set maximum confidence index; After the adjusted actual scan slice thickness is calculated, the number of CT scan slices will be further optimized to adapt to the new slice thickness setting and improve the accurate detection ability of small-volume bubbles. The optimization formula for the number of CT scan slices is as follows: , where: is the new number of CT scan slices, that is, the number of cross-sectional images to be obtained under the adjusted scan slice thickness, is the starting number of CT scan slices.

10. Visual monitoring platform for the muscle recovery process after gastric cancer surgery, which is used to implement the visual monitoring multi-terminal collaboration method for the muscle recovery process after gastric cancer surgery described in any one of the above claims 1-9, characterized in that, It includes a scan parameter initialization module, an image acquisition and dynamic monitoring module, an image processing and risk area marking module, a gas feature analysis and intelligent prediction module, and a scan slice thickness adaptive regulation module; The scan parameter initialization module sets the starting scan slice thickness for the scan according to the specific situation of the patient before the CT examination starts; The image acquisition and dynamic monitoring module uses a CT scanning device to scan and image the cross-section of the muscle tissue of gastric cancer patients after surgery. By collecting the scan slice data in real time, continuous dynamic image data is obtained, and the real-time changes inside the muscle tissue are recorded; The image processing and risk area marking module performs real-time processing on the images collected at each slice, obtains the density values of different pixels of the muscle tissue in the cross-section, captures the suspected parts with significantly reduced density, and marks the suspected parts as potential risk areas; The gas feature analysis and intelligent prediction module extracts the suspected gas image features for the potential risk area. After in-depth analysis of the extracted features, the analyzed gas image data is used as a feature vector and input into the trained convolutional neural network, and the convolutional neural network is used to intelligently predict the risk of gas presence in the muscles of gastric cancer patients after surgery; The scan slice thickness adaptive regulation module automatically triggers the adaptive regulation mechanism when it is identified that there is a high-risk gas distribution in the muscles of gastric cancer patients after surgery, and adaptively regulates the actual scan slice thickness during CT scanning. When it is identified that there is a gas risk in the muscles of gastric cancer patients after surgery, the actual scan slice thickness during CT scanning is automatically reduced.

Images