Hemolysis risk assessment method and system based on red blood cell morphology

Abstract

The embodiment of the invention provides a hemolysis risk assessment method and system based on erythrocyte morphology, and belongs to the technical field of biomedical treatment. The method comprises the following steps: carrying out image segmentation on a blood sample smear image to obtain a corresponding target segmentation result; performing target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image; calculating the difference degree between the first classification result and the second classification result to obtain a difference value; determining a target classification result of the target segmentation result by using the first classification result and the second classification result according to the difference value; determining a first number corresponding to normal erythrocytes and a second number corresponding to abnormal erythrocytes in the target object according to the target classification result; and performing hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood sample smear image and the blood item test parameters to obtain a target assessment result corresponding to the target object.

Description

Technical Field

[0001] The present invention relates to the field of biomedical technology, and in particular to a hemolysis risk assessment method and system based on red blood cell morphology. Background Art

[0002] Red blood cells are the most numerous type of blood cells in the blood. These cells have specific structural characteristics in morphology: mature, resting red blood cells are biconcave discs with a thin center and thick periphery. The deformability of red blood cells is essential to their function. Once the shape of red blood cells changes abnormally, such as becoming spherical or other atypical forms, their deformability will be significantly weakened, making it difficult to pass smoothly through narrow parts of blood vessels. In this case, red blood cells are prone to mechanical damage, which leads to hemolysis - that is, premature rupture of red blood cells, release of hemoglobin, and thus affect the normal function of blood. Mechanical hemolysis is a pathological phenomenon that seriously affects blood function and needs to be prevented and treated through timely detection and intervention. In the existing technical system, although morphological characteristics such as the size, shape and deformability of red blood cells are essential for the assessment of hemolysis risk, traditional assessment methods often rely on a single morphological indicator. This limitation leads to inaccurate assessment results and makes it difficult to provide accurate diagnostic basis and treatment recommendations for clinicians.

[0003] Therefore, it is urgent to design a technical solution to solve at least one of the above technical problems. Summary of the invention

[0004] The main purpose of the embodiments of the present invention is to provide a hemolysis risk assessment method and system based on red blood cell morphology, aiming to solve the problem that the related technology relies on a single morphological indicator for hemolysis risk assessment, resulting in inaccurate assessment results, and thus making it difficult to provide accurate diagnostic basis and treatment recommendations for clinical practice.

[0005] In a first aspect, an embodiment of the present invention provides a method for assessing hemolysis risk based on red blood cell morphology, comprising:

[0006] Obtaining a blood sample smear image of red blood cells corresponding to the target object and blood test parameters corresponding to the target object;

[0007] Performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result;

[0008] Determine a first reference image and a second reference image corresponding to the red blood cells, and perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image;

[0009] Calculating the degree of difference between the first classification result and the second classification result to obtain a difference value between the first classification result and the second classification result;

[0010] Determine a target classification result corresponding to the target segmentation result using the first classification result and the second classification result according to the difference value;

[0011] Determine a first number corresponding to normal red blood cells and a second number corresponding to abnormal red blood cells in the target object according to the target classification result;

[0012] A hemolysis risk assessment is performed on the target object according to the first quantity, the second quantity, the blood smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object.

[0013] In a second aspect, an embodiment of the present invention provides a hemolysis risk assessment system based on red blood cell morphology, comprising:

[0014] A data acquisition module, used to obtain a blood sample smear image of red blood cells corresponding to the target object and blood test parameters corresponding to the target object;

[0015] A data segmentation module, used for performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result;

[0016] a data classification module, used to determine a first reference image and a second reference image corresponding to the red blood cells, and to perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image;

[0017] a difference calculation module, used to calculate the difference between the first classification result and the second classification result, and obtain the difference value between the first classification result and the second classification result;

[0018] A classification determination module, configured to determine a target classification result corresponding to the target segmentation result by using the first classification result and the second classification result according to the difference value;

[0019] a quantity determination module, configured to determine a first quantity corresponding to normal red blood cells and a second quantity corresponding to abnormal red blood cells in the target object according to the target classification result;

[0020] The risk assessment module is used to perform hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood smear image and the blood test parameters to obtain a target assessment result corresponding to the target object.

[0021] In a third aspect, an embodiment of the present invention further provides a terminal device, comprising a processor, a memory, a computer program stored in the memory and executable by the processor, and a data bus for realizing connection and communication between the processor and the memory, wherein when the computer program is executed by the processor, the steps of any one of the hemolytic risk assessment methods based on red blood cell morphology provided in the specification of the present invention are implemented.

[0022] In a fourth aspect, an embodiment of the present invention further provides a storage medium for computer-readable storage, characterized in that the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the hemolytic risk assessment methods based on red blood cell morphology provided in the specification of the present invention.

[0023] The embodiment of the present invention provides a method and system for hemolysis risk assessment based on red blood cell morphology, the method comprising: obtaining a blood sample smear image of red blood cells corresponding to a target object and a blood test parameter corresponding to the target object; performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result; determining a first reference image and a second reference image corresponding to the red blood cells, and performing target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image; calculating the degree of difference between the first classification result and the second classification result to obtain a difference value between the first classification result and the second classification result; determining a target classification result corresponding to the target segmentation result using the first classification result and the second classification result according to the difference value; determining a first number corresponding to normal red blood cells and a second number corresponding to abnormal red blood cells in the target object according to the target classification result; performing hemolysis risk assessment on the target object according to the first number, the second number, the blood sample smear image and the blood test parameter to obtain a target assessment result corresponding to the target object. The method can more accurately identify normal red blood cells and abnormal red blood cells by performing image segmentation and classification on the blood sample smear image, thereby reducing the subjective error of manual diagnosis. The first reference image and the second reference image are used for classification, so as to calibrate and verify the accuracy of the classification result and ensure the reliability of the diagnosis. Thus, by calculating the difference value between the first classification result and the second classification result, the consistency and reliability of the classification result can be quantified, and the scientificity and accuracy of the analysis result can be ensured. Then, according to the target classification result, the first number corresponding to the normal red blood cells and the second number corresponding to the abnormal red blood cells in the target object are determined. Finally, according to the first number and the second number, a comprehensive evaluation is performed in combination with the blood sample smear image and the blood test parameters, so as to comprehensively analyze the health status of the target object, improve the scientificity and comprehensiveness of the hemolysis risk assessment, and obtain accurate target assessment results, thereby providing a scientific basis for diagnosis for clinicians. It also solves the problem that the hemolysis risk assessment in the related technology relies on a single morphological indicator, resulting in inaccurate assessment results, and thus it is difficult to provide accurate diagnostic basis and treatment recommendations for clinicians. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1A schematic diagram of a process for hemolysis risk assessment based on red blood cell morphology provided in an embodiment of the present invention;

[0026] Figure 2 A schematic diagram of the module structure of a hemolysis risk assessment system based on red blood cell morphology provided in an embodiment of the present invention;

[0027] Figure 3 A schematic block diagram of the structure of a terminal device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0030] It should be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0031] The embodiment of the present invention provides a hemolysis risk assessment method and system based on red blood cell morphology. The hemolysis risk assessment method based on red blood cell morphology can be applied to a terminal device, which can be an electronic device such as a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0032] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0033] Please refer to Figure 1 , Figure 1 A schematic flow chart of a hemolysis risk assessment method based on red blood cell morphology provided in an embodiment of the present invention.

[0034] like Figure 1As shown, the hemolysis risk assessment method based on red blood cell morphology includes steps S101 to S107.

[0035] Step S101, obtaining a blood smear image of red blood cells corresponding to a target object and blood test parameters corresponding to the target object.

[0036] Exemplarily, a blood sample is collected from a target subject, and then smeared on a glass slide using a standard smear technique to form a blood smear, and then a blood smear image corresponding to the blood smear is captured using a high-resolution microscope and an image acquisition device.

[0037] Exemplarily, part of the blood sample is sent to a laboratory for routine blood testing, so that an automated blood analyzer is used to analyze and measure various parameters of the blood sample to obtain blood test parameters such as hemoglobin concentration, platelet count, etc.

[0038] Step S102: performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result.

[0039] For example, the edge of red blood cells is detected by using methods such as Canny edge detection and Sobel operator, and an edge image is generated, so that the blood smear image is segmented according to the edge image to obtain the corresponding target segmentation result. Alternatively, a trained convolutional neural network (such as U-Net, Mask R-CNN) is used to segment the blood smear image to obtain the corresponding target segmentation result.

[0040] In some embodiments, the image segmentation of the blood smear image to obtain the corresponding target segmentation result includes: filtering the blood smear image using non-local mean filtering to obtain a target filtered image; comparing the target filtered image and the blood smear image bit by bit to obtain an image correlation value between the target filtered image and the blood smear image, wherein the image correlation value is used to characterize the reliability of the blood smear image; determining a target window, and obtaining a window image corresponding to the blood smear image according to the target window; obtaining gradient change information corresponding to the window image, and determining a growth direction corresponding to the window image according to the gradient change information; determining pixel correlation values ​​between associated pixels in the window image according to the growth direction; determining a target correlation coefficient corresponding to the blood smear image according to the pixel correlation value and the image correlation value, wherein the target correlation coefficient is used to characterize the degree of correlation between neighboring pixels in the blood smear image; constructing a target fuzzy factor according to the target correlation coefficient, and performing image segmentation on the blood smear image using a fuzzy mean clustering algorithm according to the target fuzzy factor to obtain the corresponding target segmentation result.

[0041] Exemplarily, noise is reduced by calculating the weighted average of each pixel in the blood smear image and the surrounding pixels, where the weight is determined by the similarity between pixels, so as to set the parameters of the filter, such as the search window size, similarity threshold and filter strength, to ensure that the noise is effectively removed while retaining the image details, and then after filtering processing, the target filtered image is obtained.

[0042] Exemplarily, the target filtered image is compared bit by bit with the blood sample smear image to obtain the difference between the pixels under the bit by bit comparison, and the normalization is performed according to the grayscale level 255 to obtain the normalized result, and then the index value corresponding to the normalized result with natural numbers as the base is taken as the index, and then the reciprocal of the index value is taken to determine the image correlation value. The larger the image correlation value, the greater the similarity between the blood sample smear image and the filtered target filtered image, thereby indicating that the reliability of the blood sample smear image is high, and the smaller the image correlation value, the smaller the similarity between the blood sample smear image and the filtered target filtered image, thereby indicating that the reliability of the blood sample smear image is low.

[0043] Exemplarily, a target window is determined on the blood smear image. The target window is consistent with the average size of red blood cells, so the target window is sequentially slid in the blood smear image to extract the corresponding window image.

[0044] Exemplarily, the change rate of the pixel values ​​in the window image along different directions is calculated to obtain the gradient change information, and then the growth direction of the window image is determined based on the gradient change information. This direction usually corresponds to the direction with the largest gradient change, indicating the main extension direction of the structure or feature in the window image.

[0045] Exemplarily, in the window image, if the gradient change information between the associated pixels is gradually increasing or decreasing according to the growth direction, the pixel correlation value between the associated pixels is larger; if the gradient change information between the associated pixels is gradually increasing and then gradually decreasing, or gradually decreasing and then gradually increasing according to the growth direction, the pixel correlation value between the associated pixels corresponding to the gradual increase or gradual decrease in the window image is larger, but the pixel correlation value between the associated pixels corresponding to the gradual increase and the gradual decrease is smaller. Thus, the pixel correlation value between the associated pixels is determined according to the pixel difference between the associated pixels and the growth direction between the associated pixels to quantify the similarity or correlation between them.

[0046] Exemplarily, the image correlation value and the pixel correlation value are weighted and averaged to obtain the target correlation coefficient of the blood sample smear image. The target correlation coefficient is used to characterize the degree of association between neighboring pixels in the blood sample smear image.

[0047] Exemplarily, a target fuzzy factor is constructed according to the target correlation coefficient. The target fuzzy factor reflects the fuzziness of the association between pixels in the blood smear image and is used to guide the subsequent clustering process. Thus, the fuzzy mean clustering algorithm is applied, combined with the target fuzzy factor, to perform cluster analysis on the blood smear image. In this process, the pixels in the image are divided into different categories or clusters. Thus, according to the clustering result, the target segmentation result of the blood smear image is obtained.

[0048] In some embodiments, obtaining the gradient change information corresponding to the window image and determining the growth direction corresponding to the window image based on the gradient change information includes: obtaining a first pixel position corresponding to a minimum pixel and a second pixel position corresponding to a maximum pixel in the window image; calculating the gradient change information of the window image between the first pixel position and the second pixel position; determining a gradient change sequence corresponding to the window image based on the gradient change information, and determining the growth direction corresponding to the window image based on the gradient change sequence.

[0049] Exemplarily, in the window image, a point with a minimum pixel value is found and the corresponding pixel position when the pixel value is the minimum is determined as the first pixel position. And in the window image, a point with a maximum pixel value is found and the corresponding pixel position when the pixel value is the maximum is determined as the second pixel position.

[0050] Exemplarily, a suitable gradient algorithm (such as a Sobel operator, a Prewitt operator, or a Laplace operator) is used to calculate the gradient change information from the first pixel position to the second pixel position in the window image.

[0051] Exemplarily, a gradient change sequence is generated based on the calculated gradient change information. This gradient change sequence reflects the change of the gradient value from the first pixel position to the second pixel position. When the gradient change information is a positive number, the corresponding value in the gradient change sequence is 1; when the gradient change information is a negative number, the corresponding value in the gradient change sequence is 0.

[0052] Exemplarily, when there are both 1 and 0 in the gradient change sequence, the growth direction of the window image is determined to be gradually increasing and then gradually decreasing, or gradually decreasing and then gradually increasing. When there is only 1 in the gradient change sequence, the growth direction of the window image is determined to be gradually increasing; when there is only 0 in the gradient change sequence, the growth direction of the window image is determined to be gradually decreasing.

[0053] Step S103, determining a first reference image and a second reference image corresponding to the red blood cells, and performing target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image.

[0054] Exemplarily, the first reference image is an image corresponding to normal red blood cells, and the second reference image is an image corresponding to abnormal red blood cells, or the first reference image is an image corresponding to abnormal red blood cells, and the second reference image is an image corresponding to normal red blood cells.

[0055] Exemplarily, an image library containing red blood cells in various states, such as normal red blood cells and abnormal red blood cells, is established. A first reference image representing normal red blood cells and a second reference image representing abnormal red blood cells are selected from the image library.

[0056] Exemplarily, a feature extraction algorithm, such as gray-level co-occurrence matrix (GLCM), local binary pattern (LBP) or a deep learning feature extraction model (such as a convolutional neural network) is used from the target segmentation result to extract the first key features of the red blood cells, such as shape, size, color, texture, etc.

[0057] Exemplarily, a feature extraction algorithm such as a gray level co-occurrence matrix (GLCM), a local binary pattern (LBP) or a deep learning feature extraction model (such as a convolutional neural network) is used to extract a second key feature of the first reference image, such as shape, size, color, texture, etc. And a feature extraction algorithm such as a gray level co-occurrence matrix (GLCM), a local binary pattern (LBP) or a deep learning feature extraction model (such as a convolutional neural network) is used to extract a third key feature of the second reference image, such as shape, size, color, texture, etc.

[0058] Exemplarily, a first similarity value between the first key feature and the second key feature and a second similarity value between the first key feature and the third key feature are calculated, so as to obtain a first classification result corresponding to the target segmentation result under the first reference image according to the first similarity value, and to obtain a second classification result corresponding to the target segmentation result under the second reference image according to the second similarity value.

[0059] For example, when the first similarity value is s1, the first classification result corresponding to the target segmentation result under the first reference image is that the probability of the target segmentation result being a normal red blood cell is s1, and the probability of the target segmentation result being an abnormal red blood cell is 1 minus s1. When the second similarity value is s2, the second classification result corresponding to the target segmentation result under the second reference image is that the probability of the target segmentation result being an abnormal red blood cell is s2, and the probability of the target segmentation result being a normal red blood cell is 1 minus s2.

[0060] In some embodiments, the target classification of the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image includes: using a first feature extraction layer of a first classification model to extract features from the first reference image to obtain a first feature map corresponding to the first reference image; using a first convolution layer of the first classification model to convolve the first feature map and the target segmentation result to obtain a first similarity map; using a first processing layer of the first classification model to normalize the first similarity map to obtain a second similarity map; using a second feature extraction layer of the first classification model to extract features from the target segmentation result to obtain first feature information; using a first enhancement layer of the first classification model to add the second similarity map to the first feature information through a second convolution layer to obtain first enhancement information; using the first classification layer of the first classification model to The first enhancement information is used to perform target classification on the target segmentation result to obtain the first classification result corresponding to the first reference image; the third feature extraction layer of the second classification model is used to perform feature extraction on the second reference image to obtain the second feature map corresponding to the second reference image; the third convolution layer of the second classification model is used to convolve the second feature map and the target segmentation result to obtain a third similarity map; the second processing layer of the second classification model is used to normalize the third similarity map to obtain a fourth similarity map; the fourth feature extraction layer of the second classification model is used to perform feature extraction on the target segmentation result to obtain second feature information; the second enhancement layer of the second classification model is used to add the fourth similarity map to the second feature information through a fourth convolution layer to obtain second enhancement information; the second classification layer of the second classification model is used to perform target classification on the target segmentation result according to the second enhancement information to obtain the second classification result corresponding to the second reference image.

[0061] Exemplarily, the first feature extraction layer of the first classification model is used to extract features from the first reference image to obtain a first feature map corresponding to the first reference image. The first feature map is convolved with the target segmentation result to generate a first similarity map. The convolution operation is intended to detect similar features between the target segmentation result and the first reference image. The first similarity map is normalized using the first processing layer to generate a second similarity map. Normalization helps to eliminate scale differences between data and improve the stability and comparability of features.

[0062] Exemplarily, the second feature extraction layer of the first classification model is used to extract features from the target segmentation result to obtain the first feature information. The normalized second similarity graph is added to the first feature information through the second convolution layer to generate the first enhancement information. This step enhances the associated features between the target segmentation result and the first reference image. The target segmentation result is classified according to the first enhancement information to obtain the first classification result under the first reference image. The classification layer uses the enhanced features to make decisions and outputs the classification label of the target segmentation result under the first reference image. When the target segmentation result is similar to the normal red blood cell type corresponding to the first reference image, the first enhancement information will increase the probability of the target segmentation result being identified as a normal red blood cell. Conversely, when the target segmentation result is not similar to the first reference image, the first enhancement information will reduce the probability of the target segmentation result being identified as a normal red blood cell.

[0063] Exemplarily, the third feature extraction layer of the second classification model is used to extract features from the second reference image to obtain a second feature map corresponding to the second reference image. The second feature map is convolved with the target segmentation result to generate a third similarity map. The convolution operation is intended to detect similar features between the target segmentation result and the second reference image. The third similarity map is normalized by the second processing layer to generate a fourth similarity map. Normalization helps to eliminate scale differences between data and improve the stability and comparability of features. The fourth feature extraction layer is used to extract features from the target segmentation result by using the fourth feature extraction layer of the second classification model to obtain second feature information. Thus, the normalized fourth similarity map is added to the second feature information through the fourth convolution layer to generate second enhanced information. This step enhances the associated features between the target segmentation result and the second reference image. Finally, the target segmentation result is classified by the second enhanced information according to the second classification layer to obtain the second classification result under the second reference image. The second classification layer makes decisions using the enhanced features and outputs the classification label of the target segmentation result under the second reference image. When the target segmentation result is similar to the type of abnormal red blood cells corresponding to the second reference image, the second enhancement information increases the probability that the target segmentation result is identified as an abnormal red blood cell. Conversely, when the target segmentation result is not similar to the second reference image, the second enhancement information reduces the probability that the target segmentation result is identified as an abnormal red blood cell.

[0064] Step S104: Calculate the degree of difference between the first classification result and the second classification result to obtain a difference value between the first classification result and the second classification result.

[0065] Exemplarily, the difference between the first classification result and the second classification result is calculated using Euclidean distance, Manhattan distance, cosine similarity, etc., so as to obtain the difference value between the first classification result and the second classification result.

[0066] For example, a first probability value that the target segmentation result belongs to normal red blood cells in the first reference image is obtained according to the first classification result, and a second probability value that the target segmentation result belongs to normal red blood cells in the second reference image is obtained according to the second classification result; and a third probability value that the target segmentation result belongs to abnormal red blood cells in the first reference image is obtained according to the first classification result, and a fourth probability value that the target segmentation result belongs to abnormal red blood cells in the second reference image is obtained according to the second classification result, so as to calculate the first distance corresponding to the first probability value and the second probability value according to the Euclidean distance and the second distance corresponding to the third probability value and the fourth probability value according to the Euclidean distance, and then average the first distance and the second distance to obtain the target distance between the first classification result and the second classification result, and finally, determine the difference value between the first classification result and the second classification result according to the target distance, the larger the difference value, the greater the classification difference between the first classification result and the second classification result; the smaller the difference value, the smaller the classification difference between the first classification result and the second classification result, the smaller the difference value, the greater the reliability of the fusion of the first classification result and the second classification result, and the larger the difference value, the smaller the reliability of the fusion of the first classification result and the second classification result.

[0067] In some embodiments, the calculating the degree of difference between the first classification result and the second classification result to obtain the difference value between the first classification result and the second classification result includes: obtaining from the first classification result a first probability corresponding to when the target segmentation result is determined as the first type under the first reference image and a second probability corresponding to when the target segmentation result is the second type; obtaining from the second classification result a third probability corresponding to when the target segmentation result is determined as the first type under the second reference image and a fourth probability corresponding to when the target segmentation result is the second type; determining a first similarity value of the target segmentation result between the first classification result and the second classification result according to the first probability and the third probability; determining a second similarity value of the target segmentation result between the first classification result and the second classification result according to the second probability and the fourth probability; determining the difference value between the first classification result and the second classification result according to the first similarity value and the second similarity value; wherein the first similarity value and the second similarity value are obtained according to the following formula:

[0068]

[0069] Wherein, sim1 represents the first similarity value, sim2 represents the second similarity value, and p 1i represents the first probability corresponding to the i-th target segmentation result when it is determined to be the first type under the first reference image, h 1irepresents the third probability corresponding to the i-th target segmentation result when it is determined to be the first type under the second reference image, represents the average value corresponding to the first probability, represents the average value corresponding to the third probability, n represents the number of segmentations corresponding to the target segmentation result, abs represents the absolute value, p 2i represents the second probability corresponding to the i-th target segmentation result when it is determined to be the second type under the first reference image; h 2i represents the fourth probability corresponding to the i-th target segmentation result when it is determined to be the second type under the second reference image, represents the average value corresponding to the second probability, represents the average value corresponding to the fourth probability.

[0070] Exemplarily, a first probability when the target segmentation result is determined as a first type such as normal red blood cells in the first reference image such as normal red blood cells and a second probability when the target segmentation result is a second type such as abnormal red blood cells are extracted from the first classification result. And a third probability when the target segmentation result is determined as a first type such as normal red blood cells in the second reference image such as abnormal red blood cells and a fourth probability when the target segmentation result is a second type such as abnormal red blood cells are extracted from the second classification result.

[0071] Exemplarily, according to the first probability and the third probability, the first similarity value between the target segmentation result and the first classification result and the second classification result is calculated using the following formula.

[0072]

[0073] Among them, sim1 represents the first similarity value, p 1i represents the first probability corresponding to the first type of the i-th target segmentation result under the first reference image, h 1i represents the third probability corresponding to the i-th target segmentation result when it is determined as the first type under the second reference image, represents the average value corresponding to the first probability, It represents the average value corresponding to the third probability, n represents the number of segmentations corresponding to the target segmentation result, and abs represents the absolute value.

[0074] For example, by calculating the first similarity value, the consistency between the classification results of the target segmentation results under different reference images can be quantified. A larger first similarity value indicates that the classification results under the two reference images are more consistent, which helps to ensure the stability of the classification model. By calculating the first similarity value using the average and absolute values ​​of the first probability and the third probability, the influence of a single sample or extreme value on the result can be effectively reduced. This averaging and standardization process enhances the robustness of the model, making the classification results more stable and reliable in different data sets and scenarios.

[0075] Exemplarily, according to the second probability and the fourth probability, the second similarity value between the target segmentation result and the first classification result and the second classification result is calculated using the following formula.

[0076]

[0077] Among them, sim2 represents the second similarity value, n represents the number of segmentations corresponding to the target segmentation result, abs represents the absolute value, and p 2i represents the second probability corresponding to the i-th target segmentation result when it is determined as the second type under the first reference image; h 2i represents the fourth probability corresponding to the i-th target segmentation result when it is determined to be the second type under the second reference image, represents the average value corresponding to the second probability, Represents the average value corresponding to the fourth probability.

[0078] For example, by calculating the second similarity value, the consistency between the classification results of the target segmentation results under different reference images can be quantified. A larger second similarity value indicates that the classification results under the two reference images are more consistent, which helps to ensure the stability of the classification model. By calculating the second similarity value using the average and absolute values ​​of the second probability and the fourth probability, the influence of a single sample or extreme value on the result can be effectively reduced. This averaging and standardization process enhances the robustness of the model, making the classification results more stable and reliable in different data sets and scenarios.

[0079] Exemplarily, the first similarity value and the second similarity value are summed and averaged to obtain the average similarity value, and then the difference value between the first classification result and the second classification result is obtained by subtracting the average similarity value from 1. The smaller the difference value, the greater the reliability of the fusion of the first classification result and the second classification result, and the larger the difference value, the lower the reliability of the fusion of the first classification result and the second classification result.

[0080] Step S105 : determining a target classification result corresponding to the target segmentation result by using the first classification result and the second classification result according to the difference value.

[0081] Exemplarily, the calculated difference value is evaluated. The difference value indicates the degree of difference between the first classification result and the second classification result. A larger difference value indicates that there is a significant difference between the two classification results, and a smaller difference value indicates that the first classification result and the second classification result are closer. A threshold is determined. The threshold is used to determine whether the size of the difference value is sufficient to distinguish the first classification result from the second classification result.

[0082] For example, if the difference value is less than or equal to the set threshold, it means that the difference between the first classification result and the second classification result is not significant. At this time, you can choose to combine the classification results of the two, take the average or other weighted average methods to determine the final target classification result. If the difference value is greater than the set threshold, it means that there is a significant difference between the first classification result and the second classification result. It is necessary to re-acquire the blood smear image and re-segment and classify it until the difference value is less than or equal to the set threshold.

[0083] Exemplarily, if the difference value is less than or equal to the set threshold, the probability values ​​of the same type in the first classification result and the second classification result are summed and averaged, and the red blood cell type corresponding to the maximum probability is determined as the target classification result corresponding to the target segmentation result. When the calculated difference value is less than or equal to the set threshold, it indicates that the difference between the first classification result and the second classification result is not significant. In this case, it can be considered that the two classification results are similar to a certain extent, so it is necessary to comprehensively consider the two results to determine the final target classification result. For example, first, the probability value of the same type is extracted from the first classification result and the second classification result. Assume that there are two types in the first classification result and the second classification result: the first type and the second type. Extract the first probability of the first type in the first classification result and the third probability in the second classification result, and the second probability of the second type in the first classification result and the fourth probability in the second classification result. Thus, the first probability and the second probability under the first type are summed and averaged to obtain the first target probability corresponding to the first type, and the third probability and the fourth probability under the second type are summed and averaged to obtain the second target probability corresponding to the second type. Finally, the type with the largest probability value in the first target probability and the second target probability is used as the target classification result corresponding to the target segmentation result. That is, if the first target probability of the first type is greater than the second target probability of the second type, the first type is determined as the target classification result corresponding to the target segmentation result; otherwise, the second type is determined as the target classification result corresponding to the target segmentation result.

[0084] Step S106: Determine a first number corresponding to normal red blood cells and a second number corresponding to abnormal red blood cells in the target object according to the target classification result.

[0085] Exemplarily, the number of segmented regions corresponding to when the target classification result in the target segmentation result is normal red blood cells is determined as the first number corresponding to the normal red blood cells in the target object, and the number of segmented regions corresponding to when the target classification result in the target segmentation result is abnormal red blood cells is determined as the second number corresponding to the abnormal red blood cells in the target object.

[0086] Step S107: performing a hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object.

[0087] Exemplarily, a first ratio of normal red blood cells corresponding to the target object and a second ratio of abnormal red blood cells are determined based on the first quantity and the second quantity, and then the first ratio, the second ratio, the blood smear image and the blood test parameters are fused, and then the neural network model is used to perform hemolytic risk assessment on the parameter fusion result to obtain a target assessment result corresponding to the target object.

[0088] In some embodiments, the hemolysis risk assessment is performed on the target object based on the first quantity, the second quantity, the blood smear image and the blood test parameters to obtain a target assessment result corresponding to the target object, including: obtaining an abnormal area corresponding to the abnormal red blood cells from the blood smear image; determining the abnormal proportion corresponding to the abnormal red blood cells based on the first quantity and the second quantity; and using a risk assessment model to perform hemolysis risk assessment on the target object based on the abnormal area, the abnormal proportion and the blood test parameters to obtain the target assessment result corresponding to the target object.

[0089] Exemplarily, an abnormal region corresponding to abnormal red blood cells in the blood sample smear image is identified according to the target segmentation result, and an abnormal ratio of abnormal red blood cells to total red blood cells is calculated using the first number and the second number.

[0090] Exemplarily, a risk assessment model is constructed and trained using historical data and a verified hemolysis risk assessment algorithm. The model can perform risk assessment based on abnormal areas, abnormal proportions, and blood test parameters such as hemoglobin concentration, white blood cell count, platelet count, etc. Key features are extracted from abnormal areas and blood test parameters, such as morphological characteristics, distribution characteristics, hemoglobin concentration, and red blood cell osmotic fragility index of abnormal red blood cells. The extracted features are then input into the trained risk assessment model to perform hemolysis risk assessment, thereby generating a target assessment result corresponding to the target object. The target assessment result can be a specific hemolysis risk score, risk level, or probability value.

[0091] In some embodiments, the method further includes: obtaining an assessment accuracy corresponding to the target assessment result, and collecting the first quantity, the second quantity, the blood smear image, and the blood test parameters as incremental data based on the assessment accuracy; performing incremental training on the risk assessment model based on the incremental data to obtain an updated risk assessment model.

[0092] For example, the target evaluation result is compared with the reference standard to calculate the evaluation accuracy. Commonly used indicators include but are not limited to accuracy, precision, recall, F1 score, etc.

[0093] Exemplarily, when the evaluation accuracy is lower than a preset value, the first quantity, the second quantity, the blood smear image and the blood test parameters corresponding to the evaluation status corresponding to the target evaluation result when the evaluation is wrong are collected and determined as incremental data.

[0094] Exemplarily, when the amount of incremental data is greater than a preset amount, the risk assessment model is incrementally trained according to the incremental data to obtain an updated risk assessment model. Thus, after effectively evaluating the accuracy of the target assessment result, the incremental data is collected and used for incremental training of the risk assessment model to obtain an updated risk assessment model, thereby improving the accuracy and reliability of the model.

[0095] See also Figure 2 , Figure 2A hemolysis risk assessment system 200 based on red blood cell morphology is provided in an embodiment of the present application. The hemolysis risk assessment system 200 based on red blood cell morphology includes a data acquisition module 201, a data segmentation module 202, a data classification module 203, a difference calculation module 204, a classification determination module 205, a quantity determination module 206, and a risk assessment module 207, wherein the data acquisition module 201 is used to obtain a blood sample smear image corresponding to the red blood cells of a target object and a blood test parameter corresponding to the target object; the data segmentation module 202 is used to perform image segmentation on the blood sample smear image to obtain a corresponding target segmentation result; the data classification module 203 is used to determine a first reference image and a second reference image corresponding to the red blood cells, and perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain the first reference image. a first classification result corresponding to the image and a second classification result corresponding to the second reference image; a difference calculation module 204, used to calculate the difference between the first classification result and the second classification result, and obtain the difference value between the first classification result and the second classification result; a classification determination module 205, used to determine the target classification result corresponding to the target segmentation result using the first classification result and the second classification result according to the difference value; a quantity determination module 206, used to determine the first quantity corresponding to the normal red blood cells and the second quantity corresponding to the abnormal red blood cells in the target object according to the target classification result; a risk assessment module 207, used to perform hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood sample smear image and the blood test parameters, and obtain the target assessment result corresponding to the target object.

[0096] In some implementations, the data segmentation module 202 performs, during the process of performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result:

[0097] Using non-local mean filtering to filter the blood sample smear image to obtain a target filtered image;

[0098] Comparing the target filtered image and the blood smear image bit by bit to obtain an image correlation value between the target filtered image and the blood smear image, wherein the image correlation value is used to characterize the reliability of the blood smear image;

[0099] Determine a target window, and obtain a window image corresponding to the blood sample smear image according to the target window;

[0100] Obtaining gradient change information corresponding to the window image, and determining a growth direction corresponding to the window image according to the gradient change information;

[0101] Pixel correlation values ​​between associated pixels in the window image according to the growth direction;

[0102] Determining a target correlation coefficient corresponding to the blood smear image according to the pixel correlation value and the image correlation value, wherein the target correlation coefficient is used to characterize the degree of association between neighboring pixels in the blood smear image;

[0103] A target fuzzy factor is constructed according to the target correlation coefficient, and a fuzzy mean clustering algorithm is used to perform image segmentation on the blood sample smear image according to the target fuzzy factor to obtain the corresponding target segmentation result.

[0104] In some implementations, the data segmentation module 202, in the process of obtaining the gradient change information corresponding to the window image and determining the growth direction corresponding to the window image according to the gradient change information, performs:

[0105] Obtaining a first pixel position corresponding to a minimum pixel and a second pixel position corresponding to a maximum pixel in the window image;

[0106] Calculating the gradient change information of the window image between the first pixel position and the second pixel position;

[0107] A gradient change sequence corresponding to the window image is determined according to the gradient change information, and the growth direction corresponding to the window image is determined according to the gradient change sequence.

[0108] In some implementations, the data classification module 203, in the process of classifying the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image, performs:

[0109] Using a first feature extraction layer of a first classification model to perform feature extraction on the first reference image to obtain a first feature map corresponding to the first reference image;

[0110] Convolving the first feature map and the target segmentation result using the first convolutional layer of the first classification model to obtain a first similarity map;

[0111] Using the first processing layer of the first classification model to perform normalization processing on the first similarity graph to obtain a second similarity graph;

[0112] Using the second feature extraction layer of the first classification model to perform feature extraction on the target segmentation result to obtain first feature information;

[0113] Using the first enhancement layer of the first classification model, adding the second similarity graph to the first feature information through a second convolutional layer to obtain first enhancement information;

[0114] Using the first classification layer of the first classification model to perform target classification on the target segmentation result according to the first enhancement information to obtain the first classification result corresponding to the first reference image;

[0115] Using a third feature extraction layer of a second classification model to perform feature extraction on the second reference image to obtain a second feature map corresponding to the second reference image;

[0116] Using the third convolutional layer of the second classification model, convolving the second feature map and the target segmentation result to obtain a third similarity map;

[0117] Using the second processing layer of the second classification model to normalize the third similarity graph to obtain a fourth similarity graph;

[0118] Using the fourth feature extraction layer of the second classification model to perform feature extraction on the target segmentation result to obtain second feature information;

[0119] Using the second enhancement layer of the second classification model, adding the fourth similarity graph to the second feature information through a fourth convolutional layer to obtain second enhancement information;

[0120] The target segmentation result is subjected to target classification using the second classification layer of the second classification model according to the second enhancement information to obtain the second classification result corresponding to the second reference image.

[0121] In some implementations, the difference calculation module 204, in the process of calculating the difference between the first classification result and the second classification result to obtain the difference value between the first classification result and the second classification result, performs:

[0122] Obtaining from the first classification result a first probability corresponding to when the target segmentation result is determined as a first type under the first reference image and a second probability corresponding to when the target segmentation result is determined as a second type;

[0123] Obtaining from the second classification result a third probability corresponding to when the target segmentation result is determined to be the first type under the second reference image and a fourth probability corresponding to when the target segmentation result is the second type;

[0124] Determine a first similarity value of the target segmentation result between the first classification result and the second classification result according to the first probability and the third probability;

[0125] Determine a second similarity value of the target segmentation result between the first classification result and the second classification result according to the second probability and the fourth probability;

[0126] Determine the difference value between the first classification result and the second classification result according to the first similarity value and the second similarity value;

[0127] The first similarity value and the second similarity value are obtained according to the following formula:

[0128]

[0129] Wherein, sim1 represents the first similarity value, sim2 represents the second similarity value, and p 1i represents the first probability corresponding to the i-th target segmentation result when it is determined to be the first type under the first reference image, h 1i represents the third probability corresponding to the i-th target segmentation result when it is determined to be the first type under the second reference image, represents the average value corresponding to the first probability, represents the average value corresponding to the third probability, n represents the number of segmentations corresponding to the target segmentation result, abs represents the absolute value, p 2i represents the second probability corresponding to the i-th target segmentation result when it is determined to be the second type under the first reference image; h 2i represents the fourth probability corresponding to the i-th target segmentation result when it is determined to be the second type under the second reference image, represents the average value corresponding to the second probability, represents the average value corresponding to the fourth probability.

[0130] In some embodiments, the risk assessment module 207, in the process of performing hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood smear image, and the blood test parameters to obtain the target assessment result corresponding to the target object, executes:

[0131] Obtaining the abnormal area corresponding to the abnormal red blood cells from the blood sample smear image;

[0132] determining an abnormal proportion corresponding to the abnormal red blood cells according to the first number and the second number;

[0133] The risk assessment model is used to perform hemolysis risk assessment on the target object according to the abnormal area, the abnormal proportion and the blood test parameters to obtain the target assessment result corresponding to the target object.

[0134] In some embodiments, the risk assessment module 207 also performs:

[0135] Obtaining the evaluation accuracy corresponding to the target evaluation result, and collecting the first quantity, the second quantity, the blood sample smear image, and the blood test parameters according to the evaluation accuracy to determine as incremental data;

[0136] Incrementally train the risk assessment model according to the incremental data to obtain an updated risk assessment model.

[0137] In some embodiments, the red blood cell morphology-based hemolysis risk assessment system 200 can be applied to a terminal device.

[0138] It should be noted that, those skilled in the art can clearly understand that, for the sake of convenience and brevity of description, the specific working process of the hemolysis risk assessment system 200 based on red blood cell morphology described above can refer to the corresponding process in the aforementioned embodiment of the hemolysis risk assessment method based on red blood cell morphology, and will not be repeated here.

[0139] See also Figure 3 , Figure 3 A schematic block diagram of the structure of a terminal device provided in an embodiment of the present invention.

[0140] like Figure 3 As shown, the terminal device 300 includes a processor 301 and a memory 302, and the processor 301 and the memory 302 are connected via a bus 303, such as an I2C (Inter-integrated Circuit) bus.

[0141] Specifically, the processor 301 is used to provide computing and control capabilities to support the operation of the entire terminal device. The processor 301 can be a central processing unit (CPU), and the processor 301 can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0142] Specifically, the memory 302 may be a Flash chip, a read-only memory (ROM) disk, an optical disk, a USB flash drive, or a mobile hard disk.

[0143] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the embodiment of the present invention, and does not constitute a limitation on the terminal device to which the embodiment of the present invention is applied. The specific server may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0144] The processor is used to run a computer program stored in the memory, and implement any one of the hemolysis risk assessment methods based on red blood cell morphology provided by the embodiments of the present invention when executing the computer program.

[0145] In one embodiment, the processor is used to run a computer program stored in the memory, and implements the following steps when executing the computer program:

[0146] Obtaining a blood sample smear image of red blood cells corresponding to the target object and blood test parameters corresponding to the target object;

[0147] Performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result;

[0148] Determine a first reference image and a second reference image corresponding to the red blood cells, and perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image;

[0149] Calculating the degree of difference between the first classification result and the second classification result to obtain a difference value between the first classification result and the second classification result;

[0150] Determine a target classification result corresponding to the target segmentation result using the first classification result and the second classification result according to the difference value;

[0151] Determine a first number corresponding to normal red blood cells and a second number corresponding to abnormal red blood cells in the target object according to the target classification result;

[0152] A hemolysis risk assessment is performed on the target object according to the first quantity, the second quantity, the blood smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object.

[0153] In some implementations, the processor 301, during the process of performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result, executes:

[0154] Using non-local mean filtering to filter the blood sample smear image to obtain a target filtered image;

[0155] Comparing the target filtered image and the blood smear image bit by bit to obtain an image correlation value between the target filtered image and the blood smear image, wherein the image correlation value is used to characterize the reliability of the blood smear image;

[0156] Determine a target window, and obtain a window image corresponding to the blood sample smear image according to the target window;

[0157] Obtaining gradient change information corresponding to the window image, and determining a growth direction corresponding to the window image according to the gradient change information;

[0158] Pixel correlation values ​​between associated pixels in the window image according to the growth direction;

[0159] Determining a target correlation coefficient corresponding to the blood smear image according to the pixel correlation value and the image correlation value, wherein the target correlation coefficient is used to characterize the degree of association between neighboring pixels in the blood smear image;

[0160] A target fuzzy factor is constructed according to the target correlation coefficient, and a fuzzy mean clustering algorithm is used to perform image segmentation on the blood sample smear image according to the target fuzzy factor to obtain the corresponding target segmentation result.

[0161] In some implementations, the processor 301, in the process of obtaining the gradient change information corresponding to the window image and determining the growth direction corresponding to the window image according to the gradient change information, executes:

[0162] Obtaining a first pixel position corresponding to a minimum pixel and a second pixel position corresponding to a maximum pixel in the window image;

[0163] Calculating the gradient change information of the window image between the first pixel position and the second pixel position;

[0164] A gradient change sequence corresponding to the window image is determined according to the gradient change information, and the growth direction corresponding to the window image is determined according to the gradient change sequence.

[0165] In some implementations, the processor 301, in the process of performing target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image, performs:

[0166] Using a first feature extraction layer of a first classification model to perform feature extraction on the first reference image to obtain a first feature map corresponding to the first reference image;

[0167] Convolving the first feature map and the target segmentation result using the first convolutional layer of the first classification model to obtain a first similarity map;

[0168] Using the first processing layer of the first classification model to perform normalization processing on the first similarity graph to obtain a second similarity graph;

[0169] Using the second feature extraction layer of the first classification model to perform feature extraction on the target segmentation result to obtain first feature information;

[0170] Using the first enhancement layer of the first classification model, adding the second similarity graph to the first feature information through a second convolutional layer to obtain first enhancement information;

[0171] Using the first classification layer of the first classification model to perform target classification on the target segmentation result according to the first enhancement information to obtain the first classification result corresponding to the first reference image;

[0172] Using a third feature extraction layer of a second classification model to perform feature extraction on the second reference image to obtain a second feature map corresponding to the second reference image;

[0173] Using the third convolutional layer of the second classification model, convolving the second feature map and the target segmentation result to obtain a third similarity map;

[0174] Using the second processing layer of the second classification model to normalize the third similarity graph to obtain a fourth similarity graph;

[0175] Using the fourth feature extraction layer of the second classification model to perform feature extraction on the target segmentation result to obtain second feature information;

[0176] Using the second enhancement layer of the second classification model, adding the fourth similarity graph to the second feature information through a fourth convolutional layer to obtain second enhancement information;

[0177] The target segmentation result is subjected to target classification using the second classification layer of the second classification model according to the second enhancement information to obtain the second classification result corresponding to the second reference image.

[0178] In some implementations, the processor 301, in the process of calculating the degree of difference between the first classification result and the second classification result and obtaining the difference value between the first classification result and the second classification result, executes:

[0179] Obtaining from the first classification result a first probability corresponding to when the target segmentation result is determined as a first type under the first reference image and a second probability corresponding to when the target segmentation result is determined as a second type;

[0180] Obtaining from the second classification result a third probability corresponding to when the target segmentation result is determined to be the first type under the second reference image and a fourth probability corresponding to when the target segmentation result is the second type;

[0181] Determine a first similarity value of the target segmentation result between the first classification result and the second classification result according to the first probability and the third probability;

[0182] Determine a second similarity value of the target segmentation result between the first classification result and the second classification result according to the second probability and the fourth probability;

[0183] Determine the difference value between the first classification result and the second classification result according to the first similarity value and the second similarity value;

[0184] The first similarity value and the second similarity value are obtained according to the following formula:

[0185]

[0186] Wherein, sim1 represents the first similarity value, sim2 represents the second similarity value, and p 1i represents the first probability corresponding to the i-th target segmentation result when it is determined to be the first type under the first reference image, h 1i represents the third probability corresponding to the i-th target segmentation result when it is determined to be the first type under the second reference image, represents the average value corresponding to the first probability, represents the average value corresponding to the third probability, n represents the number of segmentations corresponding to the target segmentation result, abs represents the absolute value, p 2i represents the second probability corresponding to the i-th target segmentation result when it is determined to be the second type under the first reference image; h 2i represents the fourth probability corresponding to the i-th target segmentation result when it is determined to be the second type under the second reference image, represents the average value corresponding to the second probability, represents the average value corresponding to the fourth probability.

[0187] In some embodiments, the processor 301, in the process of performing hemolysis risk assessment on the target object according to the first number, the second number, the blood smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object, executes:

[0188] Obtaining the abnormal area corresponding to the abnormal red blood cells from the blood sample smear image;

[0189] determining an abnormal proportion corresponding to the abnormal red blood cells according to the first number and the second number;

[0190] The risk assessment model is used to perform hemolysis risk assessment on the target object according to the abnormal area, the abnormal proportion and the blood test parameters to obtain the target assessment result corresponding to the target object.

[0191] In some implementations, the processor 301 further executes:

[0192] Obtaining the evaluation accuracy corresponding to the target evaluation result, and collecting the first quantity, the second quantity, the blood sample smear image, and the blood test parameters according to the evaluation accuracy to determine as incremental data;

[0193] Incrementally train the risk assessment model according to the incremental data to obtain an updated risk assessment model.

[0194] It should be noted that technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the terminal device described above can refer to the corresponding process in the aforementioned embodiment of the hemolysis risk assessment method based on red blood cell morphology, and will not be repeated here.

[0195] An embodiment of the present invention also provides a storage medium for computer-readable storage, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the hemolytic risk assessment methods based on red blood cell morphology provided in the description of the embodiment of the present invention.

[0196] The storage medium may be an internal storage unit of the terminal device in the aforementioned embodiment, such as a hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the terminal device.

[0197] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0198] It should be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0199] The serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A hemolysis risk assessment method based on red blood cell morphology, characterized in that: The method comprises: Obtaining a blood sample smear image of red blood cells corresponding to the target object and blood test parameters corresponding to the target object; Performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result; Determine a first reference image and a second reference image corresponding to the red blood cells, and perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image; Calculating the degree of difference between the first classification result and the second classification result to obtain a difference value between the first classification result and the second classification result; Determine a target classification result corresponding to the target segmentation result using the first classification result and the second classification result according to the difference value; Determine a first number corresponding to normal red blood cells and a second number corresponding to abnormal red blood cells in the target object according to the target classification result; A hemolysis risk assessment is performed on the target object according to the first quantity, the second quantity, the blood smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object.

2. The method according to claim 1, characterized in that The performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result includes: Using non-local mean filtering to filter the blood sample smear image to obtain a target filtered image; Comparing the target filtered image and the blood smear image bit by bit to obtain an image correlation value between the target filtered image and the blood smear image, wherein the image correlation value is used to characterize the reliability of the blood smear image; Determine a target window, and obtain a window image corresponding to the blood sample smear image according to the target window; Obtaining gradient change information corresponding to the window image, and determining a growth direction corresponding to the window image according to the gradient change information; Pixel correlation values ​​between associated pixels in the window image according to the growth direction; Determining a target correlation coefficient corresponding to the blood smear image according to the pixel correlation value and the image correlation value, wherein the target correlation coefficient is used to characterize the degree of association between neighboring pixels in the blood smear image; A target fuzzy factor is constructed according to the target correlation coefficient, and a fuzzy mean clustering algorithm is used to perform image segmentation on the blood sample smear image according to the target fuzzy factor to obtain the corresponding target segmentation result.

3. The method according to claim 2, characterized in that The obtaining of gradient change information corresponding to the window image, and determining a growth direction corresponding to the window image according to the gradient change information, includes: Obtaining a first pixel position corresponding to a minimum pixel and a second pixel position corresponding to a maximum pixel in the window image; Calculating the gradient change information of the window image between the first pixel position and the second pixel position; A gradient change sequence corresponding to the window image is determined according to the gradient change information, and the growth direction corresponding to the window image is determined according to the gradient change sequence.

4. The method according to claim 1, characterized in that: The performing target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image includes: Using a first feature extraction layer of a first classification model to perform feature extraction on the first reference image to obtain a first feature map corresponding to the first reference image; Convolving the first feature map and the target segmentation result using the first convolutional layer of the first classification model to obtain a first similarity map; Using the first processing layer of the first classification model to perform normalization processing on the first similarity graph to obtain a second similarity graph; Using the second feature extraction layer of the first classification model to perform feature extraction on the target segmentation result to obtain first feature information; Using the first enhancement layer of the first classification model, adding the second similarity graph to the first feature information through a second convolutional layer to obtain first enhancement information; Using the first classification layer of the first classification model to perform target classification on the target segmentation result according to the first enhancement information to obtain the first classification result corresponding to the first reference image; Using a third feature extraction layer of a second classification model to perform feature extraction on the second reference image to obtain a second feature map corresponding to the second reference image; Using the third convolutional layer of the second classification model, convolving the second feature map and the target segmentation result to obtain a third similarity map; Using the second processing layer of the second classification model to normalize the third similarity graph to obtain a fourth similarity graph; Using the fourth feature extraction layer of the second classification model to perform feature extraction on the target segmentation result to obtain second feature information; Using the second enhancement layer of the second classification model, adding the fourth similarity graph to the second feature information through a fourth convolutional layer to obtain second enhancement information; The target segmentation result is subjected to target classification using the second classification layer of the second classification model according to the second enhancement information to obtain the second classification result corresponding to the second reference image.

5. The method according to claim 1, characterized in that The calculating the difference between the first classification result and the second classification result to obtain the difference value between the first classification result and the second classification result includes: Obtaining from the first classification result a first probability corresponding to when the target segmentation result is determined as a first type under the first reference image and a second probability corresponding to when the target segmentation result is determined as a second type; Obtaining from the second classification result a third probability corresponding to when the target segmentation result is determined to be the first type under the second reference image and a fourth probability corresponding to when the target segmentation result is the second type; Determine a first similarity value of the target segmentation result between the first classification result and the second classification result according to the first probability and the third probability; Determine a second similarity value of the target segmentation result between the first classification result and the second classification result according to the second probability and the fourth probability; Determine the difference value between the first classification result and the second classification result according to the first similarity value and the second similarity value; The first similarity value and the second similarity value are obtained according to the following formula: Wherein, sim1 represents the first similarity value, sim2 represents the second similarity value, and p 1i represents the first probability corresponding to the i-th target segmentation result when it is determined to be the first type under the first reference image, h 1i represents the third probability corresponding to the i-th target segmentation result when it is determined to be the first type under the second reference image, represents the average value corresponding to the first probability, represents the average value corresponding to the third probability, n represents the number of segmentations corresponding to the target segmentation result, abs represents the absolute value, p 2i represents the second probability corresponding to the i-th target segmentation result when it is determined to be the second type under the first reference image; h 2i represents the fourth probability corresponding to the i-th target segmentation result when it is determined to be the second type under the second reference image, represents the average value corresponding to the second probability, represents the average value corresponding to the fourth probability.

6. The method according to claim 1, characterized in that The step of performing hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood sample smear image, and the blood test parameters to obtain a target assessment result corresponding to the target object includes: Obtaining the abnormal area corresponding to the abnormal red blood cells from the blood sample smear image; determining an abnormal proportion corresponding to the abnormal red blood cells according to the first number and the second number; The risk assessment model is used to perform hemolysis risk assessment on the target object according to the abnormal area, the abnormal proportion and the blood test parameters to obtain the target assessment result corresponding to the target object.

7. The method according to claim 6, characterized in that The method further comprises: Obtaining the evaluation accuracy corresponding to the target evaluation result, and collecting the first quantity, the second quantity, the blood sample smear image, and the blood test parameters according to the evaluation accuracy to determine as incremental data; Incrementally train the risk assessment model according to the incremental data to obtain an updated risk assessment model.

8. A hemolysis risk assessment system based on red blood cell morphology, characterized in that: include: A data acquisition module, used to obtain a blood sample smear image of red blood cells corresponding to the target object and blood test parameters corresponding to the target object; A data segmentation module, used for performing image segmentation on the blood sample smear image to obtain a corresponding target segmentation result; a data classification module, used to determine a first reference image and a second reference image corresponding to the red blood cells, and to perform target classification on the target segmentation result according to the first reference image and the second reference image to obtain a first classification result corresponding to the first reference image and a second classification result corresponding to the second reference image; a difference calculation module, used to calculate the difference between the first classification result and the second classification result, and obtain the difference value between the first classification result and the second classification result; A classification determination module, configured to determine a target classification result corresponding to the target segmentation result by using the first classification result and the second classification result according to the difference value; a quantity determination module, configured to determine a first quantity corresponding to normal red blood cells and a second quantity corresponding to abnormal red blood cells in the target object according to the target classification result; The risk assessment module is used to perform hemolysis risk assessment on the target object according to the first quantity, the second quantity, the blood smear image and the blood test parameters to obtain a target assessment result corresponding to the target object.

9. A terminal device, characterized in that: The terminal device includes a processor and a memory; The memory is used to store computer programs; The processor is configured to execute the computer program and implement the hemolysis risk assessment method based on red blood cell morphology according to any one of claims 1 to 7 when executing the computer program.

10. A computer storage medium for computer storage, characterized in that: The computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the hemolysis risk assessment method based on red blood cell morphology as described in any one of claims 1 to 7.

Images