Implicit anomaly monitoring method and system for monitoring image in intensive care medicine department

Abstract

The invention provides an implicit anomaly monitoring method and system for a monitoring image in an intensive care medicine department, and relates to the technical field of image processing, the method comprises the following steps: obtaining a non-dominant abnormal soft tissue type of a CT monitoring image in an intensive care department, the dominant abnormal soft tissue type being a tissue type which does not meet a set standard of a physician end; based on the medical history and the basic sign information of the target user, indexing the adjacent image set, and executing dominant anomaly image proportion analysis on the non-dominant anomaly soft tissue type to obtain shape anomaly density, texture anomaly density and color anomaly density; and based on the non-dominant abnormal soft tissue type of which any abnormal density of the shape abnormal density, the texture abnormal density and the color abnormal density is greater than or equal to a density threshold, constructing an implicit abnormal region, performing reference abnormal color rendering, and sending the implicit abnormal region to a physician end for display. The technical problem that in the prior art, implicit anomaly monitoring of monitoring images in the intensive care medicine department is not timely is solved.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method and system for monitoring implicit anomalies in monitoring images within a critical care medicine department. Background Art

[0002] In CT imaging diagnosis in critical care medicine, some lesions appear subtle in the early stages on CT images and have significant individual differences, such as mildly disordered texture in liver fibrosis and slightly uneven density of ground-glass nodules in lung adenocarcinoma. Relying solely on the patient's single CT image feature analysis, it is difficult to accurately identify certain non-explicit abnormalities, resulting in untimely monitoring of latent abnormalities. Summary of the Invention

[0003] The present invention aims to solve the technical problem in the prior art of untimely monitoring of implicit abnormalities in monitoring images in a critical care department and provides a method and system for monitoring implicit abnormalities in monitoring images in a critical care department.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a method for detecting implicit abnormalities in monitoring images within a critical care unit, comprising: Obtaining non-overt abnormal soft tissue types from critical care CT monitoring images, where overt abnormal soft tissue types refer to tissue types that do not meet the criteria set by the physician; Based on the target user's medical history and basic physical sign information, the adjacent image set is indexed, and the proportion of obvious abnormal images is analyzed for non-obvious abnormal soft tissue types to obtain shape abnormality density, texture abnormality density, and color abnormality density; Based on the non-overt abnormal soft tissue type whose abnormal density among the shape abnormal density, the texture abnormal density and the color abnormal density is greater than or equal to the density threshold, a hidden abnormal area is constructed, reference abnormal color rendering is performed, and it is sent to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

[0005] In a second aspect, the present invention provides a system for monitoring hidden abnormalities in monitoring images within a critical care unit, comprising: The non-overt abnormality screening module is used to obtain the non-overt abnormal soft tissue types in the critical internal CT monitoring images, where the overt abnormal soft tissue types refer to tissue types that do not meet the standards set by the physician; The abnormality density calculation module is used to index the adjacent image set based on the target user's medical history and basic physical sign information, perform explicit abnormality image ratio analysis on non-explicit abnormal soft tissue types, and obtain shape abnormality density, texture abnormality density, and color abnormality density; The rendering and sending module is used to construct a hidden abnormal area based on the non-overt abnormal soft tissue type whose abnormal density of any one of the shape abnormality density, the texture abnormality density and the color abnormality density is greater than or equal to the density threshold, perform reference abnormal color rendering, and send it to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

[0006] The beneficial effects of the present invention are: Compared with the existing technology, the present application first obtains the non-overt abnormal soft tissue types of critical internal CT monitoring images, wherein the overt abnormal soft tissue type refers to the tissue type that does not meet the standards set by the physician side, obtains the overt abnormal area, and provides the necessary data basis for the subsequent identification and analysis of non-overt abnormal soft tissue. Secondly, based on the target user's medical history and basic physical sign information, the adjacent image set is indexed, and the overt abnormal image ratio analysis is performed on the non-overt abnormal soft tissue type to obtain the shape abnormality density, texture abnormality density and color abnormality density, which provides a reliable quantitative basis for judging the potential risk of the current non-overt abnormal area and avoids missed diagnosis due to single threshold limitation. Finally, based on the non-overt abnormal soft tissue type with any abnormal density greater than or equal to the density threshold among the shape abnormality density, texture abnormality density and color abnormality density, the implicit abnormal area is constructed, the reference abnormal color rendering is performed, and it is sent to the physician side for display, extracting the possible implicit abnormal area with individual value, providing further reference for the doctor's subsequent diagnosis, and realizing early detection and intervention of the condition of critically ill patients.

[0007] Through the above technical solution, this application identifies the obvious abnormal area by multi-valued processing of various features of soft tissue, and then collects adjacent image sets based on the user's medical history and basic physical sign information, calculates the shape abnormality density, texture abnormality density and color abnormality density of the non-obvious abnormal area, and when any one of the shape abnormality density, texture abnormality density and color abnormality density of the non-obvious abnormal area is greater than or equal to the density threshold, implicit abnormality identification is performed, and the image with the obvious abnormal area and the implicit abnormal area is sent to the user end. In this way, the possible implicit abnormal area with individual value is extracted, which provides further reference for the doctor's subsequent diagnosis and improves the timeliness of implicit abnormality monitoring of monitoring images in the critical care department. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flowchart of a method for monitoring hidden abnormalities in monitoring images within a critical care unit provided by the present invention; Figure 2 This is a structural diagram of a system for monitoring hidden anomalies in monitoring images within a critical care medicine department provided by the present invention.

[0009] In the accompanying drawings, the components represented by the reference numerals are as follows: Non-explicit anomaly screening module 11, anomaly density calculation module 12, rendering and sending module 13. DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0011] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0012] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0013] Example 1, as Figure 1 As shown, an embodiment of the present invention provides a method for monitoring implicit abnormalities in monitoring images in a critical care medicine department, comprising: S10: Obtaining non-overt abnormal soft tissue types from critical care CT monitoring images, wherein overt abnormal soft tissue types refer to tissue types that do not meet the criteria set by the physician; Patients may have undoubted obvious abnormalities or non-obvious abnormalities. In order to avoid missing early lesions or latent complications due to a single dimensional feature not reaching the explicit threshold, it is necessary to simultaneously identify non-obvious abnormalities to achieve more comprehensive and earlier accurate monitoring of the condition of critically ill patients, thereby gaining a valuable time window for clinical intervention.

[0014] In response to the above problems, this application collects CT monitoring images, image feature baseline values, consistency deviation thresholds, target user medical history and basic physical sign information, performs semantic segmentation on CT monitoring images based on soft tissue types, determines several soft tissue type CT monitoring images, and calculates and obtains explicit abnormality coefficients of several soft tissue types, extracts soft tissue types whose explicit abnormality coefficients are greater than or equal to the explicit abnormality coefficient threshold, constructs explicit abnormal areas, and adds soft tissue types that do not have explicit abnormal areas into non-explicit abnormal soft tissue types.

[0015] Specifically, step S10 in the method includes: Receive critical care CT monitoring images from the detection instrument, and receive image feature baseline values, consistency deviation thresholds, target user medical history, and basic vital signs information from the physician end; performing semantic segmentation on the CT monitoring image based on the soft tissue type to determine a plurality of CT monitoring images of soft tissue types; Traverse several soft tissue type CT monitoring images, calculate the deviation vector of the CT monitoring image feature value and the image feature baseline value, and calculate the ratio of the deviation vector to the consistency deviation threshold as the dominant abnormality coefficient of several soft tissue types, extract the soft tissue types whose dominant abnormality coefficient is greater than or equal to the dominant abnormality coefficient threshold, construct the dominant abnormality area, perform deterministic abnormality color rendering, use the dominant abnormality coefficient as an accompanying label, and send it to the physician end for display, wherein the dominant abnormality coefficient threshold is ≥1, and the deterministic abnormality refers to an undoubted abnormality; The soft tissue types that do not have the obvious abnormal area are added to the non-obvious abnormal soft tissue types.

[0016] In this embodiment of the present application, a critical care internal CT monitoring image is first received from a detection instrument, and image feature baseline values, consistency deviation thresholds, and the target user's medical history and basic vital signs information (including various basic indicators such as height, weight, and gender) are received from the physician. The image feature baseline values ​​reflect the standard parameters of normal soft tissue in terms of shape, texture, and color, while the consistency deviation threshold defines the fluctuation range within the normal range. The target user's medical history and basic vital signs information provide important reference value for the subsequent comprehensive analysis of abnormal conditions.

[0017] Secondly, semantic segmentation is performed on the CT monitoring image based on the soft tissue type to determine several soft tissue type CT monitoring images. Specifically, semantic segmentation technology uses a deep learning algorithm to accurately classify different types of soft tissue in CT images, thereby determining several soft tissue type CT monitoring images, for example, extracting specific tissues such as the liver and lungs.

[0018] Again, traverse several soft tissue type CT monitoring images, calculate the deviation vector of the CT monitoring image feature value and the image feature baseline value, and calculate the ratio of the deviation vector to the consistency deviation threshold as the explicit abnormality coefficient of several soft tissue types, extract the soft tissue types whose explicit abnormality coefficient is greater than or equal to the explicit abnormality coefficient threshold, construct the explicit abnormality area, perform deterministic abnormality color rendering (for example, rendering in red), use the explicit abnormality coefficient as an accompanying label, and send it to the physician end for display, wherein the explicit abnormality coefficient threshold ≥1, deterministic abnormality refers to an undoubted abnormality, and the explicit abnormality coefficient threshold ≥1 is because when the characteristic value of the soft tissue deviates from the critical value of the normal range, it is 1. When the explicit abnormality coefficient threshold ≥1, it means that the degree of deviation of the characteristic value of the soft tissue has exceeded the clinically allowed fluctuation threshold, indicating that there is a deterministic abnormality (explicit abnormality) in the area, allowing physicians to quickly detect areas with obvious problems.

[0019] Finally, the soft tissue types that do not have the obvious abnormal areas are added to the non-obvious abnormal soft tissue types, waiting for further exploration of potential abnormalities.

[0020] Specifically, the “performing semantic segmentation on the CT monitoring image based on the soft tissue type to determine a plurality of CT monitoring images of soft tissue types” includes: Based on a first preset soft tissue type, collecting multiple sets of image data, wherein any set of the multiple sets of image data has image data including the preset soft tissue type and a label identifying the coordinates of the contour of the preset soft tissue type; Based on the image data of the preset soft tissue type and the label identifying the contour coordinates of the preset soft tissue type, training a first preset soft tissue type identifier and adding the identifier to the soft tissue type identifier set; The soft tissue type is calibrated according to the CT scanning area, the soft tissue type identifier set is selectively scheduled and connected in series, the critical care CT monitoring image is processed, and the plurality of soft tissue type CT monitoring images are output.

[0021] In the embodiment of the present application, a complex CT image is semantically segmented according to the soft tissue type and decomposed into several single soft tissue type CT monitoring images, as follows: First, data collection and labeling are performed. Specifically, based on the first preset soft tissue type, multiple sets of image data are collected, wherein any set of the multiple sets of image data has image data including the preset soft tissue type and labels identifying the contour coordinates of the preset soft tissue type. A labeled training data set is formed. Data collection and labeling link. The system needs to collect multiple sets of image data (such as 10,000+ CT monitoring images) around the first preset soft tissue type (such as liver, lungs and other specific tissues). Each set of image data contains image data of the preset soft tissue type (i.e., the original image information obtained by CT scanning) and labels identifying the contour coordinates of the preset soft tissue type. The labels of the contour coordinates can be obtained by manual labeling. In this way, the boundary range of the soft tissue in the image is accurately defined, and a labeled training data set is formed to provide the necessary data basis for subsequent model learning. Secondly, a first preset soft tissue type identifier is trained. Specifically, based on the image data of the preset soft tissue type and the label identifying the contour coordinates of the preset soft tissue type, the first preset soft tissue type identifier is trained and added to the soft tissue type identifier set. Exemplarily, the first preset soft tissue type identifier can adopt a deep learning (e.g., U-Net network) architecture, mainly including a data preprocessing layer, a feature extraction encoder, an attention mechanism enhancement module, a semantic segmentation decoder, and an output layer. The data preprocessing layer standardizes the original CT data to eliminate differences between different scanning devices; the feature extraction encoder can adopt ResNet-50 as the backbone network, gradually extract image features through 4 downsampling modules, and output feature maps of 64, 128, 256, and 512 channels respectively; the attention mechanism enhancement module generates weights for each feature channel through the channel attention module through global average pooling and a two-layer fully connected network, and the spatial attention module uses a convolutional layer to generate a spatial mask, focusing on the tissue boundary area in the image; the semantic segmentation decoder gradually restores the image resolution through 4 upsampling modules, and at the same time, combined with the jump connections of each stage of the encoder, retains feature information at different levels; the output layer uses 1×1 convolution to map the features to a number of channels equal to the number of preset soft tissue types, and then outputs the probability distribution of each pixel belonging to each type of tissue through the softmax activation function. Furthermore, the trained first preset soft tissue type identifier is added to the soft tissue type identifier set, wherein the trained first preset soft tissue type identifier can identify specific soft tissue and output a CT monitoring image of a specific soft tissue type. The soft tissue type identifier set composed of several first preset soft tissue type identifiers can identify several soft tissue types and output several soft tissue type CT monitoring images.

[0022] Among them, the training process of the first preset soft tissue type identifier can be achieved through the following technical paths: 1. Data preparation: The image data of the preset soft tissue type and the labels identifying the contour coordinates of the preset soft tissue type are divided into a training set, a validation set, and a test set according to a ratio of 7:1.5:1.5. 2. Model training: Based on the training set, a combined loss function is adopted, combining Dice loss and cross entropy loss. Data enhancement uses a variety of transformation methods, including random rotation, scaling, elastic deformation, and Gaussian noise, to greatly expand the diversity of training samples. The optimizer can use the Adam algorithm and use cosine annealing scheduling to dynamically adjust the learning rate. L2 regularization and Dropout (0.3) technology are used to prevent model overfitting. 3. Model convergence: Until the output accuracy on the test set reaches more than 95%, it is considered converged, and the first preset soft tissue type identifier is obtained after training.

[0023] Finally, based on the CT scanning region, the soft tissue types are calibrated, and the soft tissue type identifier set is selectively scheduled and connected in series to process the critical care CT monitoring image, and output the plurality of soft tissue type CT monitoring images. For example, based on the CT scanning region, the soft tissue types contained therein are first calibrated to identify the targets to be identified. For example, based on the CT scanning region (such as the abdomen), multiple soft tissue types such as the liver, spleen, and intestines are calibrated. Then, based on the different soft tissue types, selective scheduling is performed from the soft tissue type identifier set to select relevant identifiers, which are then connected in series to collaboratively process the critical care CT monitoring image, and output the plurality of soft tissue type CT monitoring images.

[0024] In this way, several soft tissue type CT monitoring images are output, which realizes the refined segmentation of the original CT image and decomposes the complex image into images of a single soft tissue type, providing a reliable basis for subsequent abnormality judgment.

[0025] Furthermore, the step of “traversing a plurality of soft tissue type CT monitoring images, calculating a deviation vector between a CT monitoring image feature value and the image feature reference value, and calculating a ratio of the deviation vector to the consistency deviation threshold as the dominant abnormality coefficients of the plurality of soft tissue types” includes: Extracting a first soft tissue type CT monitoring image from the plurality of soft tissue type CT monitoring images; Extracting a first soft tissue type shape descriptor reference value, a first soft tissue type texture descriptor reference value, and a first soft tissue type color descriptor reference value from the image feature reference value; Extracting a first soft tissue type shape descriptor deviation threshold, a first soft tissue type texture descriptor deviation threshold, and a first soft tissue type color descriptor deviation threshold from the consistency deviation threshold; Based on the first soft tissue type shape descriptor benchmark value, the first soft tissue type texture descriptor benchmark value, and the first soft tissue type color descriptor benchmark value, in combination with the first soft tissue type shape descriptor deviation threshold, the first soft tissue type texture descriptor deviation threshold, and the first soft tissue type color descriptor deviation threshold, the first soft tissue type CT monitoring image is processed to calculate the first soft tissue type shape dominant abnormality coefficient, the first soft tissue type texture dominant abnormality coefficient, and the first soft tissue type color dominant abnormality coefficient, and add them to the first soft tissue type dominant abnormality coefficient; The dominant abnormality coefficient of the first soft tissue type is added to the dominant abnormality coefficients of the plurality of soft tissue types.

[0026] In the embodiment of the present application, based on the image feature baseline value, the consistency deviation threshold and several soft tissue type CT monitoring images, the degree of dominant abnormality of the soft tissue type is quantitatively analyzed as follows: First, a first soft tissue type CT monitoring image is extracted from the plurality of soft tissue type CT monitoring images. For example, a first soft tissue type CT monitoring image (such as a liver CT monitoring image) is extracted from the plurality of soft tissue type CT monitoring images.

[0027] Next, a first soft tissue type shape descriptor benchmark value, a first soft tissue type texture descriptor benchmark value, and a first soft tissue type color descriptor benchmark value are extracted from the image feature benchmark values. For example, the first soft tissue type shape descriptor benchmark value, the first soft tissue type texture descriptor benchmark value, and the first soft tissue type color descriptor benchmark value are extracted from the liver image feature benchmark values. This is because the shape, texture, and color of soft tissue undergo significant changes when a lesion occurs. By analyzing these quantifiable parameters, it is possible to assess whether the soft tissue is abnormal. The first soft tissue type shape descriptor benchmark values ​​include: an area benchmark value (e.g., 200 cm²), a circularity benchmark value (e.g., 0.85), and an aspect ratio benchmark value (e.g., 1.5). The first soft tissue type texture descriptor benchmark values ​​include: a gray-level co-occurrence matrix contrast benchmark value (e.g., 9) and an entropy benchmark value (e.g., 3.5). The first soft tissue type color descriptor benchmark values ​​include: a mean HU value benchmark value (e.g., 50 HU) and a grayscale standard deviation benchmark value (e.g., 12 HU). These benchmark values ​​represent standard characteristic parameters of the soft tissue type under normal conditions.

[0028] Next, a first soft tissue type shape descriptor deviation threshold, a first soft tissue type texture descriptor deviation threshold, and a first soft tissue type color descriptor deviation threshold are extracted from the consistency deviation threshold. For example, the first soft tissue type shape descriptor deviation threshold, the first soft tissue type texture descriptor deviation threshold, and the first soft tissue type color descriptor deviation threshold are extracted from the consistency deviation threshold of the liver. The first soft tissue type shape descriptor deviation threshold includes an area deviation threshold (e.g., ±15), a circularity deviation threshold (e.g., ±0.08), and an aspect ratio deviation threshold (e.g., ±0.2). The first soft tissue type texture descriptor deviation threshold includes a contrast deviation threshold (e.g., ±20%) and an entropy deviation threshold (e.g., ±0.8). The first soft tissue type color descriptor deviation threshold includes a mean HU value deviation threshold (e.g., ±8 HU) and a grayscale standard deviation deviation threshold (e.g., ±5 HU). These thresholds define the fluctuation limits within the normal range.

[0029] Furthermore, based on the first soft tissue type shape descriptor baseline value, the first soft tissue type texture descriptor baseline value, and the first soft tissue type color descriptor baseline value, combined with the first soft tissue type shape descriptor deviation threshold, the first soft tissue type texture descriptor deviation threshold, and the first soft tissue type color descriptor deviation threshold, the first soft tissue type CT monitoring image is processed to calculate the first soft tissue type shape dominant anomaly coefficient, the first soft tissue type texture dominant anomaly coefficient, and the first soft tissue type color dominant anomaly coefficient, and add them to the first soft tissue type dominant anomaly coefficient. The calculation method of the three types of anomaly coefficients is the same, as shown in the following formula: in, is the abnormal coefficient, is the characteristic value of the i-th CT monitoring image, is the i-th image feature reference value, is the i-th consistency deviation threshold, and n is the total number of eigenvalues.

[0030] Exemplarily, the first soft tissue type shape descriptor benchmark value of the liver is an area benchmark value (such as 200 cm²), a circularity benchmark value (such as 0.85), and an aspect ratio benchmark value (such as 1.5); the first soft tissue type shape descriptor deviation threshold is an area deviation threshold (such as ±15), a circularity deviation threshold (such as ±0.08), and an aspect ratio deviation threshold (such as ±0.2); the extracted CT monitoring image feature value is an area of ​​209 cm², a circularity of 0.95, and an aspect ratio of 1.9. Substituting into the above formula to calculate the first soft tissue type shape dominant abnormality coefficient is 1.28, which is added to the first soft tissue type dominant abnormality coefficient.

[0031] Finally, the dominant abnormality coefficient of the first soft tissue type is added to the dominant abnormality coefficients of the plurality of soft tissue types. This provides an accurate quantitative basis for determining whether the soft tissue type has a dominant abnormality, and provides reliable support for subsequent marking and analysis of abnormal areas.

[0032] In summary, compared to the existing technology, this application collects CT monitoring images, image feature baseline values, consistency deviation thresholds, target user medical history and basic physical sign information, and performs semantic segmentation on CT monitoring images based on soft tissue type. It determines several soft tissue type CT monitoring images, calculates and obtains the dominant abnormality coefficients of several soft tissue types, extracts soft tissue types with dominant abnormality coefficients greater than or equal to the dominant abnormality coefficient threshold, constructs dominant abnormal regions, and adds soft tissue types without dominant abnormal regions to non-dominant abnormal soft tissue types. In this way, dominant abnormal regions are obtained, providing the necessary data foundation for the subsequent identification and analysis of non-dominant abnormal soft tissue.

[0033] S20: Based on the target user's medical history and basic physical sign information, index the adjacent image set, perform explicit abnormal image ratio analysis on the non-explicit abnormal soft tissue type, and obtain the shape abnormality density, texture abnormality density, and color abnormality density; Non-obvious abnormal soft tissue lesions (such as subtle tissue changes caused by early-stage tumors, inflammation, or metabolic abnormalities) are difficult to identify directly with the naked eye or conventional algorithms because they lack distinct morphological features (e.g., unclear boundaries and insignificant density changes). These lesions are easily missed or misdiagnosed by traditional single-case imaging analysis. Furthermore, if non-obvious abnormalities in some serious diseases (such as early-stage malignancies and advanced acute inflammation) are not promptly identified, the treatment window may be delayed.

[0034] To address the above issues, this application indexes the adjacent image set based on the target user's medical history and basic physical sign information, performs an analysis of the proportion of overt abnormal images for non-overt abnormal soft tissue types, and obtains the shape abnormality density, texture abnormality density, and color abnormality density.

[0035] Specifically, step S20 in the method includes: Retrieving a CT record image set of the same severe sample user set whose medical history is consistent with the target user's medical history and whose sample vital sign information is consistent with the basic vital sign information, and storing the set as the adjacent image set; Traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type to obtain the shape abnormality density; Traversing the adjacent image set, performing texture-dominant abnormal image ratio analysis on the non-dominant abnormal soft tissue type to obtain the texture abnormality density; The adjacent image set is traversed, and the color-dominant abnormal image ratio analysis is performed on the non-dominant abnormal soft tissue type to obtain the color abnormality density.

[0036] In an embodiment of the present application, a CT image set of a sample user with the same severe condition whose medical history is consistent with the target user's medical history (e.g., previous illnesses, surgical history), and whose sample vital signs information is consistent with the basic vital signs information (including various basic indicators such as age, gender, height, and weight), is first retrieved from a blockchain distributed node and stored as the adjacent image set. For example, based on the target user's medical history (e.g., a history of chronic hepatitis B for ≥10 years, no antiviral treatment or substandard treatment) and basic vital signs information (e.g., age 50, male, 175 cm, 70 kg, etc.), multiple CT image sets of a sample user with the same severe condition are retrieved from multiple blockchain distributed nodes and stored as the adjacent image set.

[0037] Next, the adjacent image sets are traversed, and the proportion of shape-dominant abnormality images is analyzed for non-dominant abnormal soft tissue types to obtain the shape abnormality density. Specifically, the shape abnormality density is obtained by traversing multiple adjacent image sets, analyzing the proportion of shape-dominant abnormality images for non-dominant abnormal soft tissue types, obtaining multiple initial shape abnormality densities, and then calculating the mode density of these multiple initial shape abnormality densities to obtain the shape abnormality density, where the initial shape abnormality density = the number of shape-dominant abnormal images / the total number of adjacent images.

[0038] Next, the adjacent image sets are traversed, and the proportion of texture-dominant abnormal images is analyzed for non-dominant abnormal soft tissue types to obtain the texture abnormality density. Specifically, the shape abnormality density is calculated by traversing multiple adjacent image sets, analyzing the proportion of texture-dominant abnormal images for non-dominant abnormal soft tissue types, obtaining multiple initial texture abnormality densities, and then calculating the mode density of these multiple initial texture abnormality densities to obtain the texture abnormality density, where initial texture abnormality density = number of texture-dominant abnormal images / total number of adjacent images.

[0039] Finally, the adjacent image sets are traversed, and the percentage of color-dominant abnormal images is analyzed for non-dominant abnormal soft tissue types to obtain the color abnormality density. Specifically, the color abnormality density is obtained by traversing multiple adjacent image sets, performing a color-dominant abnormal image percentage analysis for non-dominant abnormal soft tissue types, obtaining multiple initial color abnormality densities, and then calculating the mode density of these multiple initial color abnormality densities to obtain the color abnormality density, where the initial color abnormality density = the number of color-dominant abnormal images / the total number of adjacent images.

[0040] For example, based on the target user's medical history (such as a history of chronic hepatitis B for ≥10 years, no antiviral treatment or irregular treatment) and basic physical signs information (such as age 50 years, male, 175 cm, 70 kg, etc.), 100 qualified CT record image sets of users with the same severe condition are retrieved in the blockchain distributed nodes and stored as an adjacent image set. By traversing multiple adjacent image sets, multiple initial shape abnormality densities, initial texture abnormality densities, and initial color abnormality densities are calculated respectively, and then the mode density of multiple initial shape abnormality densities, initial texture abnormality densities, and initial color abnormality densities is calculated to obtain a shape abnormality density of 0.3, a texture abnormality density of 0.45, and a color abnormality density of 0.25. In this way, although the target user has a single-dimensional abnormality and the shape, color, and texture explicit abnormality coefficients do not clearly meet the standards, the shape, texture, and color abnormality densities calculated through the adjacent image sets are 0.3, 0.45, and 0.25, respectively, which are significantly higher than the baseline level of the normal population, consistent with the pathological evolution of liver fibrosis (the early stage is mainly characterized by rough texture and mild edge abnormalities), suggesting the presence of early liver fibrosis (such as S2-S3 stages), which requires further confirmation by liver elasticity imaging (such as FibroScan) or pathological biopsy. Furthermore, the step of “traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type, and obtaining the shape abnormality density” includes: Traversing the plurality of adjacent image sets, performing shape-overt abnormality image ratio analysis on non-overt abnormal soft tissue types to obtain a plurality of initial shape abnormality densities; The mode density of the multiple initial shape anomaly densities is calculated to obtain the shape anomaly density.

[0041] In this embodiment of the present application, the plurality of adjacent image sets are first traversed, and a shape-dominant abnormality image ratio analysis is performed on non-dominant abnormal soft tissue types to obtain a plurality of initial shape abnormality densities. The adjacent image sets include a plurality of adjacent image sets collected from a plurality of blockchain distributed nodes, as medical data is typically stored in a dispersed manner across a plurality of blockchain distributed nodes. Specifically, for each blockchain distributed node, an independent analysis is performed on the shape-dominant abnormality image ratio of non-dominant abnormal soft tissue types to obtain a plurality of initial shape abnormality densities. For example, 10 adjacent image sets are collected from 10 blockchain distributed nodes. For example, there are 100 adjacent images in adjacent image set 1, of which 30 have shape-dominant abnormal images. Then, the initial shape abnormality density of adjacent image set 1 = 30 / 100 = 0.3; there are 60 adjacent images in adjacent image set 2, of which 45 have shape-dominant abnormal images. Then, the initial shape abnormality density of adjacent image set 2 = 15 / 60 = 0.25. In this way, the 10 adjacent image sets are traversed respectively to calculate and obtain multiple initial shape abnormality densities.

[0042] Secondly, the mode density of the multiple initial shape abnormality densities is calculated to obtain the shape abnormality density. Specifically, the mode density is calculated based on the multiple initial shape abnormality densities to obtain the shape abnormality density. This is because the multiple initial shape abnormality densities obtained by calculation may fluctuate due to differences in the data of each node (such as different sample sizes, uneven data quality, etc.), and the mode is the value that appears the most times in a set of data. By finding the density interval with the highest frequency among the multiple initial shape abnormality densities and calculating the midpoint value of the interval as the final shape abnormality density, it is possible to effectively resist the interference of outliers and avoid erroneous judgments caused by deviations in individual node data. For example, if the initial shape abnormality densities calculated for most nodes are concentrated in the range of 0.2 - 0.25, the midpoint of the interval, 0.225, is taken as the shape abnormality density to ensure that the analysis results can better reflect the actual abnormal situation of the group and provide a more reliable reference for clinical diagnosis.

[0043] In summary, compared to existing technologies, this application indexes adjacent image sets based on the target user's medical history and basic physical sign information, performs an analysis of the proportion of overt abnormal images for non-overt abnormal soft tissue types, and obtains shape abnormality density, texture abnormality density, and color abnormality density. These abnormality density values ​​can reflect the probability of shape, texture, or color abnormalities in a specific soft tissue type in a patient group with similar clinical conditions. This provides a reliable quantitative basis for judging the potential risk of the current non-overt abnormal area and avoids missed diagnoses due to single threshold limitations.

[0044] S30: Based on the non-overt abnormal soft tissue type whose abnormal density among the shape abnormal density, the texture abnormal density and the color abnormal density is greater than or equal to the density threshold, construct a hidden abnormal area, perform reference abnormal color rendering, and send it to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

[0045] In CT imaging diagnosis in critical care medicine, since some lesions have subtle manifestations in the early stages, it is difficult to accurately identify non-obvious abnormalities by relying solely on image feature analysis of a single case, which can easily lead to missed diagnosis or delay of the disease. Therefore, a comprehensive assessment can be performed based on the density of shape abnormalities, texture abnormalities, and color abnormalities.

[0046] In response to the above problems, this application constructs a hidden abnormal area based on the non-overt abnormal soft tissue type whose abnormal density of shape abnormality, texture abnormality and color abnormality is greater than or equal to the density threshold, performs reference abnormal color rendering, and sends it to the physician side for display, wherein the reference abnormality refers to a suspected abnormality. Specifically, when any one of the shape abnormality density, texture abnormality density and color abnormality density is greater than or equal to the density threshold dynamically determined according to the type of severe disease and the degree of data discreteness, it will be determined that the corresponding non-overt abnormal soft tissue type has a suspected abnormal risk, and then a hidden abnormal area is constructed, and the hidden abnormal area is rendered using a reference abnormal color (such as orange or yellow), and sent to the physician side. This visual presentation method can help physicians quickly locate suspected abnormal parts, and combine clinical experience and other examination results to further determine whether the patient has a lesion, thereby achieving early detection and intervention of the condition of critically ill patients, improving the efficiency and accuracy of diagnosis and treatment, and reducing the risk of missed diagnosis.

[0047] Furthermore, the density threshold determination process in the method includes: According to the type of severe illness, a set of images to be analyzed with consistent medical history and physical signs is retrieved; Perform similarity clustering on the image set to be analyzed, obtain the number of clusters, calculate the ratio of the number of clusters to the total number of image sets to be analyzed, and set it as the discrete ratio; The density threshold is negatively correlated with the discrete ratio.

[0048] In this embodiment, first, based on the severity type, a set of images to be analyzed with consistent medical history and physical signs is retrieved. For example, based on the severity type of the patient (e.g., coma), a set of images to be analyzed with completely consistent medical history (e.g., history of chronic hepatitis B ≥ 10 years) and physical signs (e.g., age 50, male, 175 cm, 70 kg, etc.) is retrieved. These images come from a population with a similar clinical condition to the current patient and are highly comparable.

[0049] Next, the image set to be analyzed is clustered based on similarity, and the number of clusters is calculated. This number is then divided into clusters based on the degree of feature similarity. The ratio of the number of clusters to the total number of images in the set is then calculated as the dispersion ratio. Specifically, a clustering algorithm is used to perform similarity clustering on the image set to be analyzed, dividing the images into clusters based on feature similarity. The dispersion ratio is then calculated by dividing the number of clusters by the total number of images in the set to obtain the dispersion ratio. The dispersion ratio can intuitively reflect the degree of dispersion of the data: a higher dispersion ratio indicates greater differences in image features among similar critically ill patients, leading to more diverse abnormalities. A lower dispersion ratio indicates more concentrated image features and a relatively simple abnormal pattern.

[0050] Finally, the density threshold is negatively correlated with the discrete ratio, which is determined based on the characteristics of the discrete ratio. When the discrete ratio is high, the density threshold is lowered to more sensitively capture various potential abnormalities and avoid missing early or atypical lesions. When the discrete ratio is low, the density threshold is appropriately raised to reduce false positives due to individual differences or image noise, ensuring the reliability of the test results. In this way, by dynamically adjusting the threshold, it can better adapt to the characteristics of different severe types and improve the accuracy and clinical practicality of non-overt abnormality detection.

[0051] In summary, compared to existing technologies, this application constructs implicit abnormal regions based on non-overt abnormal soft tissue types with any one of the abnormal density of shape abnormality, texture abnormality, and color abnormality greater than or equal to the density threshold, performs reference abnormal color rendering, and sends it to the physician for display. In this way, it extracts potential implicit abnormal regions with individualized value, providing further reference for doctors' subsequent diagnosis and achieving early detection and intervention for critically ill patients.

[0052] In summary, the embodiments of the present application have at least the following technical effects: Compared with the existing technology, this application first collects CT monitoring images, image feature baseline values, consistency deviation thresholds, target user medical history and basic physical sign information, and then performs semantic segmentation on CT monitoring images based on soft tissue types. It determines several soft tissue type CT monitoring images, calculates and obtains the dominant abnormality coefficients of several soft tissue types, extracts soft tissue types with dominant abnormality coefficients greater than or equal to the dominant abnormality coefficient threshold, constructs dominant abnormal areas, and adds soft tissue types without dominant abnormal areas to non-dominant abnormal soft tissue types. In this way, dominant abnormal areas are obtained, providing the necessary data basis for the subsequent identification and analysis of non-dominant abnormal soft tissue.

[0053] Secondly, based on the target user's medical history and basic physical signs, this application indexes adjacent image sets and performs an analysis of the proportion of overt abnormal images for non-overt abnormal soft tissue types to obtain shape abnormality density, texture abnormality density, and color abnormality density. These abnormality density values ​​can reflect the probability of shape, texture, or color abnormalities in a specific soft tissue type in a patient group with similar clinical conditions. This provides a reliable quantitative basis for judging the potential risk of the current non-overt abnormal area and avoids missed diagnosis due to single threshold limitations.

[0054] Finally, this application constructs implicit abnormal regions based on non-overt abnormal soft tissue types with any one of the abnormal density of shape abnormality, texture abnormality, and color abnormality greater than or equal to the density threshold, performs reference abnormal color rendering, and sends it to the physician for display. In this way, the potential existence of implicit abnormal regions with individualized value is extracted, providing further reference for the doctor's subsequent diagnosis, and achieving early detection and intervention of the condition of critically ill patients.

[0055] Through the above technical solution, this application identifies the obvious abnormal area by multi-valued processing of various features of soft tissue, and then collects adjacent image sets based on the user's medical history and basic physical sign information, calculates the shape abnormality density, texture abnormality density and color abnormality density of the non-obvious abnormal area, and when any one of the shape abnormality density, texture abnormality density and color abnormality density of the non-obvious abnormal area is greater than or equal to the density threshold, implicit abnormality identification is performed, and the image with the obvious abnormal area and the implicit abnormal area is sent to the user end. In this way, the possible implicit abnormal area with individual value is extracted, which provides further reference for the doctor's subsequent diagnosis and improves the timeliness of implicit abnormality monitoring of monitoring images in the critical care department.

[0056] Example 2, as Figure 2 As shown, based on the same inventive concept as the method for monitoring implicit abnormalities in monitoring images in a critical care department provided in Example 1, an embodiment of the present invention further provides a system for monitoring implicit abnormalities in monitoring images in a critical care department, comprising: The non-overt abnormality screening module 11 is used to obtain the non-overt abnormal soft tissue type in the critical internal CT monitoring image, wherein the overt abnormal soft tissue type refers to the tissue type that does not meet the standard set by the physician; The abnormality density calculation module 12 is used to index the adjacent image set based on the target user's medical history and basic physical sign information, perform explicit abnormality image ratio analysis on the non-explicit abnormal soft tissue type, and obtain shape abnormality density, texture abnormality density, and color abnormality density; The rendering and sending module 13 is used to construct a hidden abnormal area based on the non-overt abnormal soft tissue type whose abnormal density of any one of the shape abnormality density, the texture abnormality density and the color abnormality density is greater than or equal to the density threshold, perform reference abnormal color rendering, and send it to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

[0057] The non-overt abnormality screening module 11 is specifically used for: Receive critical care CT monitoring images from the detection instrument, and receive image feature baseline values, consistency deviation thresholds, target user medical history, and basic vital signs information from the physician end; performing semantic segmentation on the CT monitoring image based on the soft tissue type to determine a plurality of CT monitoring images of soft tissue types; Traverse several soft tissue type CT monitoring images, calculate the deviation vector of the CT monitoring image feature value and the image feature baseline value, and calculate the ratio of the deviation vector to the consistency deviation threshold as the dominant abnormality coefficient of several soft tissue types, extract the soft tissue types whose dominant abnormality coefficient is greater than or equal to the dominant abnormality coefficient threshold, construct the dominant abnormality area, perform deterministic abnormality color rendering, use the dominant abnormality coefficient as an accompanying label, and send it to the physician end for display, wherein the dominant abnormality coefficient threshold is ≥1, and the deterministic abnormality refers to an undoubted abnormality; The soft tissue types that do not have the obvious abnormal area are added to the non-obvious abnormal soft tissue types.

[0058] Furthermore, the “performing semantic segmentation on the CT monitoring image based on the soft tissue type to determine a plurality of soft tissue type CT monitoring images” includes: Based on a first preset soft tissue type, collecting multiple sets of image data, wherein any set of the multiple sets of image data has image data including the preset soft tissue type and a label identifying the coordinates of the contour of the preset soft tissue type; Based on the image data of the preset soft tissue type and the label identifying the contour coordinates of the preset soft tissue type, training a first preset soft tissue type identifier and adding the identifier to the soft tissue type identifier set; The soft tissue type is calibrated according to the CT scanning area, the soft tissue type identifier set is selectively scheduled and connected in series, the critical care CT monitoring image is processed, and the plurality of soft tissue type CT monitoring images are output.

[0059] Furthermore, the step of “traversing a plurality of soft tissue type CT monitoring images, calculating a deviation vector between a CT monitoring image feature value and the image feature reference value, and calculating a ratio of the deviation vector to the consistency deviation threshold as the dominant abnormality coefficients of the plurality of soft tissue types” includes: Extracting a first soft tissue type CT monitoring image from the plurality of soft tissue type CT monitoring images; Extracting a first soft tissue type shape descriptor reference value, a first soft tissue type texture descriptor reference value, and a first soft tissue type color descriptor reference value from the image feature reference value; Extracting a first soft tissue type shape descriptor deviation threshold, a first soft tissue type texture descriptor deviation threshold, and a first soft tissue type color descriptor deviation threshold from the consistency deviation threshold; Based on the first soft tissue type shape descriptor benchmark value, the first soft tissue type texture descriptor benchmark value, and the first soft tissue type color descriptor benchmark value, in combination with the first soft tissue type shape descriptor deviation threshold, the first soft tissue type texture descriptor deviation threshold, and the first soft tissue type color descriptor deviation threshold, the first soft tissue type CT monitoring image is processed to calculate the first soft tissue type shape dominant abnormality coefficient, the first soft tissue type texture dominant abnormality coefficient, and the first soft tissue type color dominant abnormality coefficient, and add them to the first soft tissue type dominant abnormality coefficient; The dominant abnormality coefficient of the first soft tissue type is added to the dominant abnormality coefficients of the plurality of soft tissue types.

[0060] The abnormal density calculation module 12 is specifically used to: Retrieving a CT record image set of the same severe sample user set whose medical history is consistent with the target user's medical history and whose sample vital sign information is consistent with the basic vital sign information, and storing the set as the adjacent image set; Traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type to obtain the shape abnormality density; Traversing the adjacent image set, performing texture-dominant abnormal image ratio analysis on the non-dominant abnormal soft tissue type to obtain the texture abnormality density; The adjacent image set is traversed, and the color-dominant abnormal image ratio analysis is performed on the non-dominant abnormal soft tissue type to obtain the color abnormality density.

[0061] Furthermore, the step of “traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type, and obtaining the shape abnormality density” includes: Traversing the plurality of adjacent image sets, performing shape-overt abnormality image ratio analysis on non-overt abnormal soft tissue types to obtain a plurality of initial shape abnormality densities; The mode density of the multiple initial shape anomaly densities is calculated to obtain the shape anomaly density.

[0062] The rendering and sending module 13 is specifically configured to: Based on the non-overt abnormal soft tissue type whose abnormal density among the shape abnormal density, the texture abnormal density and the color abnormal density is greater than or equal to the density threshold, a hidden abnormal area is constructed, reference abnormal color rendering is performed, and it is sent to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

[0063] The process of determining the “density threshold” includes: According to the type of severe illness, a set of images to be analyzed with consistent medical history and physical signs is retrieved; Perform similarity clustering on the image set to be analyzed, obtain the number of clusters, calculate the ratio of the number of clusters to the total number of image sets to be analyzed, and set it as the discrete ratio; The density threshold is negatively correlated with the discrete ratio.

[0064] In summary, the embodiments of the present application have at least the following technical effects: Compared with the existing technology, this application first obtains the non-overt abnormal soft tissue types in critical internal CT monitoring images through a non-overt abnormality screening module, where the overt abnormal soft tissue type refers to the tissue type that does not meet the standards set by the physician end, obtains the overt abnormal area, and provides the necessary data basis for the subsequent identification and analysis of non-overt abnormal soft tissue. Secondly, through the abnormal density calculation module, based on the target user's medical history and basic physical sign information, indexes the adjacent image set, performs overt abnormal image ratio analysis on the non-overt abnormal soft tissue type, obtains the shape abnormality density, texture abnormality density, and color abnormality density, providing a reliable quantitative basis for judging the potential risk of the current non-overt abnormal area and avoiding missed diagnosis due to single threshold limitation. Finally, through the rendering and sending module, based on the non-overt abnormal soft tissue type with any abnormal density greater than or equal to the density threshold among the shape abnormality density, texture abnormality density, and color abnormality density, a hidden abnormal area is constructed, reference abnormal color rendering is performed, and it is sent to the physician end for display, extracting the possible existence of hidden abnormal areas with individual value, providing further reference for the doctor's subsequent diagnosis, and achieving early detection and intervention of the condition of critical patients. In this way, possible hidden abnormal areas with individualized value are extracted, providing further reference for doctors' subsequent diagnosis and improving the timeliness of hidden abnormality monitoring in monitoring images in the critical care department.

[0065] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0066] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0068] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0070] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0071] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting hidden abnormalities in monitoring images in a critical care unit, characterized in that: include: Obtaining non-overt abnormal soft tissue types from critical care CT monitoring images, where overt abnormal soft tissue types refer to tissue types that do not meet the criteria set by the physician; Based on the target user's medical history and basic physical sign information, the adjacent image set is indexed, and the proportion of obvious abnormal images is analyzed for non-obvious abnormal soft tissue types to obtain shape abnormality density, texture abnormality density, and color abnormality density; Based on the non-overt abnormal soft tissue type whose abnormal density among the shape abnormal density, the texture abnormal density and the color abnormal density is greater than or equal to the density threshold, a hidden abnormal area is constructed, reference abnormal color rendering is performed, and it is sent to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

2. The method according to claim 1, wherein Obtain non-obvious abnormal soft tissue types on CT monitoring images of critical care units, including: Receive critical care CT monitoring images from the detection instrument, and receive image feature baseline values, consistency deviation thresholds, target user medical history, and basic vital signs information from the physician end; performing semantic segmentation on the CT monitoring image based on the soft tissue type to determine a plurality of CT monitoring images of soft tissue types; Traverse several soft tissue type CT monitoring images, calculate the deviation vector of the CT monitoring image feature value and the image feature baseline value, and calculate the ratio of the deviation vector to the consistency deviation threshold as the dominant abnormality coefficient of several soft tissue types, extract the soft tissue types whose dominant abnormality coefficient is greater than or equal to the dominant abnormality coefficient threshold, construct the dominant abnormality area, perform deterministic abnormality color rendering, use the dominant abnormality coefficient as an accompanying label, and send it to the physician end for display, wherein the dominant abnormality coefficient threshold is ≥1, and the deterministic abnormality refers to an undoubted abnormality; The soft tissue types that do not have the obvious abnormal area are added to the non-obvious abnormal soft tissue types.

3. The method according to claim 2, wherein Based on the soft tissue type, the CT monitoring image is semantically segmented to determine several soft tissue type CT monitoring images, including: Based on a first preset soft tissue type, collecting multiple sets of image data, wherein any set of the multiple sets of image data has image data including the preset soft tissue type and a label identifying the coordinates of the contour of the preset soft tissue type; Based on the image data of the preset soft tissue type and the label identifying the contour coordinates of the preset soft tissue type, training a first preset soft tissue type identifier and adding the identifier to the soft tissue type identifier set; The soft tissue type is calibrated according to the CT scanning area, the soft tissue type identifier set is selectively scheduled and connected in series, the critical care CT monitoring image is processed, and the plurality of soft tissue type CT monitoring images are output.

4. The method according to claim 2, wherein Traversing several soft tissue type CT monitoring images, calculating the deviation vector of the CT monitoring image feature value and the image feature reference value, and calculating the ratio of the deviation vector to the consistency deviation threshold as the explicit abnormality coefficients of several soft tissue types, including: Extracting a first soft tissue type CT monitoring image from the plurality of soft tissue type CT monitoring images; Extracting a first soft tissue type shape descriptor reference value, a first soft tissue type texture descriptor reference value, and a first soft tissue type color descriptor reference value from the image feature reference value; Extracting a first soft tissue type shape descriptor deviation threshold, a first soft tissue type texture descriptor deviation threshold, and a first soft tissue type color descriptor deviation threshold from the consistency deviation threshold; Based on the first soft tissue type shape descriptor benchmark value, the first soft tissue type texture descriptor benchmark value, and the first soft tissue type color descriptor benchmark value, in combination with the first soft tissue type shape descriptor deviation threshold, the first soft tissue type texture descriptor deviation threshold, and the first soft tissue type color descriptor deviation threshold, the first soft tissue type CT monitoring image is processed to calculate the first soft tissue type shape dominant abnormality coefficient, the first soft tissue type texture dominant abnormality coefficient, and the first soft tissue type color dominant abnormality coefficient, and add them to the first soft tissue type dominant abnormality coefficient; The dominant abnormality coefficient of the first soft tissue type is added to the dominant abnormality coefficients of the plurality of soft tissue types.

5. The method according to claim 1, wherein Based on the target user's medical history and basic physical signs, the adjacent image set is indexed and the proportion of obvious abnormal images is analyzed for non-obvious abnormal soft tissue types to obtain shape abnormality density, texture abnormality density, and color abnormality density, including: Retrieving a CT record image set of the same severe sample user set whose medical history is consistent with the target user's medical history and whose sample vital sign information is consistent with the basic vital sign information, and storing the set as the adjacent image set; Traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type to obtain the shape abnormality density; Traversing the adjacent image set, performing texture-dominant abnormal image ratio analysis on the non-dominant abnormal soft tissue type to obtain the texture abnormality density; The adjacent image set is traversed, and the color-dominant abnormal image ratio analysis is performed on the non-dominant abnormal soft tissue type to obtain the color abnormality density.

6. The method according to claim 1, wherein Traversing the adjacent image set, performing shape-overt abnormality image ratio analysis on the non-overt abnormal soft tissue type to obtain the shape abnormality density, the adjacent image set including multiple adjacent image sets collected from multiple blockchain distributed nodes, including: Traversing the plurality of adjacent image sets, performing shape-overt abnormality image ratio analysis on non-overt abnormal soft tissue types to obtain a plurality of initial shape abnormality densities; The mode density of the multiple initial shape anomaly densities is calculated to obtain the shape anomaly density.

7. The method according to claim 1, wherein The process of determining the density threshold includes: According to the type of severe illness, a set of images to be analyzed with consistent medical history and physical signs is retrieved; Perform similarity clustering on the image set to be analyzed, obtain the number of clusters, calculate the ratio of the number of clusters to the total number of image sets to be analyzed, and set it as the discrete ratio; The density threshold is negatively correlated with the discrete ratio.

8. A system for monitoring hidden abnormalities in monitoring images in a critical care department, characterized in that: Used to perform the method according to any one of claims 1 to 7, comprising: The non-overt abnormality screening module is used to obtain the non-overt abnormal soft tissue types in the critical internal CT monitoring images, where the overt abnormal soft tissue types refer to tissue types that do not meet the standards set by the physician; The abnormality density calculation module is used to index the adjacent image set based on the target user's medical history and basic physical sign information, perform explicit abnormality image ratio analysis on non-explicit abnormal soft tissue types, and obtain shape abnormality density, texture abnormality density, and color abnormality density; The rendering and sending module is used to construct a hidden abnormal area based on the non-overt abnormal soft tissue type whose abnormal density of any one of the shape abnormality density, the texture abnormality density and the color abnormality density is greater than or equal to the density threshold, perform reference abnormal color rendering, and send it to the physician end for display, wherein the reference abnormality refers to a suspected abnormality.

Images