Artificial intelligence-based hematuria analysis method, apparatus and device, and storage medium

Abstract

The invention discloses a hematuria analysis method, device and equipment based on artificial intelligence and a storage medium. The method comprises the following steps: acquiring a urine visible component graph of urine to be analyzed and an interference state of the urine to be analyzed; selecting a hematuria analysis network model corresponding to the interference state from a preset corresponding relation set of the interference state and the hematuria analysis network model; the urine visible component graph is input into the hematuria analysis network model, the number of red blood cells corresponding to the to-be-analyzed urine is determined through the urine visible component graph, the hematuria category corresponding to the to-be-analyzed urine is determined based on the number of red blood cells, and the hematuria category comprises pseudo hematuria or true hematuria. According to the interference state of the to-be-analyzed urine, by constructing a plurality of corresponding hematuria analysis network models, misjudgment of the red blood cells in the to-be-analyzed urine can be reduced, accurate counting of the red blood cells is achieved, and therefore the accuracy of true hematuria detection is improved.

Description

Technical Field

[0001] This application relates to the field of medical technology, and in particular to a blood and urine analysis method, device, equipment and storage medium based on artificial intelligence. Background Technology

[0002] With advancements in technology and increased health awareness, more and more people are prioritizing physical examinations. During routine urinalysis, occult blood or microscopic hematuria are frequently detected. Because the causes of hematuria are highly complex, doctors often need to rely on the patient's examination information to further diagnose and determine the specific cause. This places a high demand on the doctor's medical skills and experience, potentially affecting the objectivity and accuracy of identifying the exact cause of hematuria. Furthermore, many interfering factors in clinical practice can lead to misdiagnosis of true hematuria as false hematuria. For example, medications and dietary factors can cause misdiagnosis. Therefore, the accuracy of true hematuria detection needs improvement. Summary of the Invention

[0003] The technical problem to be solved by this application is to provide a blood and urine analysis method, device, equipment and storage medium based on artificial intelligence, which addresses the shortcomings of the existing technology.

[0004] To address the aforementioned technical problems, the first aspect of this application provides an artificial intelligence-based hematuria analysis method, wherein the artificial intelligence-based hematuria analysis method specifically includes:

[0005] Obtain the urine formation profile and interference state of the urine to be analyzed;

[0006] In the preset set of correspondences between interference states and blood and urine analysis network models, the blood and urine analysis network model corresponding to the interference state is selected.

[0007] The formed element map of urine is input into the hematuria analysis network model. The number of red blood cells in the urine to be analyzed is determined by the formed element map of urine. Based on the number of red blood cells, the hematuria category of the urine to be analyzed is determined. The hematuria category includes pseudo-hematuria or true hematuria.

[0008] In one embodiment, according to the AI-based hematuria analysis method, the correspondence set includes several hematuria analysis network models, each of which corresponds to a disturbance state, and the disturbance states corresponding to each hematuria analysis network model are different from each other. Each hematuria analysis network model is a trained deep learning network model, and the training analysis urine in the training dataset corresponding to each hematuria analysis network model has the disturbance state corresponding to that hematuria analysis network model.

[0009] In one embodiment, according to the AI-based hematuria analysis method, each of the hematuria analysis network models has the same model structure. The hematuria analysis network model includes a convolutional module, several cascaded residual modules, an average pooling layer, and a fully connected layer. The convolutional module is connected to the first residual module, the last residual module is connected to the average pooling layer, and the average pooling layer is connected to the fully connected layer. The convolutional module includes a convolutional layer and a pooling layer, the convolutional layer is connected to the pooling layer, and the pooling layer is connected to the first residual module.

[0010] In one embodiment, according to the AI-based hematuria analysis method, the process of acquiring the interference state specifically includes:

[0011] The number of bacteria in the urine sample to be analyzed was detected.

[0012] When the number of bacteria exceeds the preset number, the urine containing bacteria will be considered as an interference state of the urine to be analyzed.

[0013] When the number of bacteria is less than or equal to a preset number, sterile urine is considered as an interference state of the urine to be analyzed.

[0014] In one embodiment, according to the AI-based hematuria analysis method, the method further includes:

[0015] When the type of hematuria is true hematuria, the proportion of abnormal red blood cells in the urine fraction graph is obtained;

[0016] The hematuria classification corresponding to the urine to be analyzed is predicted based on the proportion of abnormal red blood cells, wherein the hematuria classification includes glomerular hematuria or non-glomerular hematuria.

[0017] In one embodiment, according to the AI-based hematuria analysis method, the method further includes:

[0018] When the hematuria is classified as non-glomerular hematuria, the subject information corresponding to the urine to be analyzed is obtained, wherein the subject information includes at least medical history data and symptom data;

[0019] The crystallization data, bacterial data, and white blood cell data in the urine fraction map are read. The crystallization data, bacterial data, white blood cell data, and subject information are input into a trained first etiological analysis model. The first etiological analysis model outputs the hematuria subclass corresponding to the urine to be analyzed. The hematuria subclass includes stone-type hematuria, inflammatory hematuria, or hematuria of unknown cause.

[0020] In one embodiment, according to the AI-based hematuria analysis method, the method further includes:

[0021] When the hematuria subclass is hematuria of unknown cause, epithelial cells in the urine to be analyzed are obtained, and the cell characteristics of each obtained epithelial cell are identified.

[0022] The cell type of each epithelial cell is determined based on the identified cell characteristics;

[0023] Based on the identified cell types and subject information, the urine sample to be analyzed was determined to be hematuria suspected of being caused by bladder cancer or other types of hematuria.

[0024] A second aspect of this application provides an artificial intelligence-based hematuria analysis device, wherein the artificial intelligence-based hematuria analysis device specifically includes:

[0025] The acquisition module acquires the urine formation pattern of the urine to be analyzed and the interference state of the urine to be analyzed;

[0026] The determination module selects the blood and urine analysis network model corresponding to the interference state from a preset set of correspondences between interference states and blood and urine analysis network models.

[0027] The classification module inputs the urine formed element map into the hematuria analysis network model, determines the number of red blood cells in the urine to be analyzed based on the urine formed element map, and determines the hematuria category corresponding to the urine to be analyzed based on the number of red blood cells, wherein the hematuria category includes pseudo-hematuria or true hematuria.

[0028] A third aspect of this application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the artificial intelligence-based hematuria analysis method described above.

[0029] A fourth aspect of this application provides a terminal device, which includes: a processor and a memory;

[0030] The memory stores a computer-readable program that can be executed by the processor;

[0031] When the processor executes the computer-readable program, it implements the steps in any of the above-described artificial intelligence-based hematuria analysis methods.

[0032] Beneficial effects: Compared with existing technologies,

[0033] (1) Based on the interference state of the urine to be analyzed, this application constructs several blood and urine analysis network models corresponding to the interference state, which can reduce the misjudgment of red blood cells in the urine to be analyzed, achieve accurate counting of red blood cells, thereby improving the accuracy of true hematuria detection and the accuracy of hematuria category detection.

[0034] (2) For urine samples with true hematuria, the hematuria classification of the urine sample is further detected based on the proportion of abnormal red blood cells in the urine sample. When the hematuria is classified as non-glomerular hematuria, the subclassification of hematuria can be further obtained by constructing a first etiological analysis model and combining it with the subject information, which can improve the accuracy of subclassification identification in the urine sample. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A flowchart illustrating a preferred embodiment of the artificial intelligence-based hematuria analysis method provided in this application.

[0037] Figure 2 The principle of a preferred embodiment of the artificial intelligence-based hematuria analysis method provided in this application.

[0038] Figure 3 This is a schematic diagram of the model principle of step S20 in a preferred embodiment of the artificial intelligence-based hematuria analysis method provided in this application.

[0039] Figure 4 A schematic diagram of the AI-based hematuria analysis device provided in this application.

[0040] Figure 5 A schematic diagram of the terminal device provided in this application. Detailed Implementation

[0041] This application provides a method, apparatus, device, and storage medium for blood and urine analysis based on artificial intelligence. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0042] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0043] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0044] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0045] The inventors discovered through research that with advancements in technology and increased health awareness, more and more people are paying attention to physical examinations, and hematuria is frequently found in routine urinalysis results. The causes of hematuria are complex, including but not limited to acute nephritis, chronic kidney disease, urinary tract stones, urinary tract infections, and urinary system tumors. Currently, doctors often need to conduct further diagnosis based on the examinee's examination information to determine the specific cause of hematuria. It is worth noting that in clinical practice, cases where false hematuria is misdiagnosed as true hematuria are frequently found due to interfering factors. These interfering factors include the examinee's physiological and disease state, medications, and dietary factors.

[0046] Bacteria in urine are a significant interfering factor in the misdiagnosis of hematuria. For example, bacteria in urine can cause false-positive results for occult blood in urine using dry chemistry methods, thus misdiagnosing false hematuria as true hematuria.

[0047] When using dry chemistry to determine hematuria, this method utilizes the peroxidase-like activity of heme in hemoglobin. This activity causes the decomposition of hydrogen peroxide and anisin to release reactive oxygen species, which oxidize receptors (such as benzidine), causing them to change color. Therefore, in addition to detecting intact red blood cells and free hemoglobin, urine analyzers can also detect substances with peroxidase-like activity, such as bacteria that produce peroxidase, all of which will produce a positive reaction. In urinary tract infections, most Gram-negative bacteria and some Gram-positive bacteria may release hydroperoxidase-active substances or synthesize catalase, peroxidase, and superoxide dismutase during their reproduction for metabolic needs. These substances and enzymes can cause false positives when using dry chemistry to measure urinary red blood cells, thus misdiagnosing false hematuria as true hematuria.

[0048] When using a urine formed element analyzer to measure hematuria, the analyzer detects hematuria by identifying and counting red blood cells in the urine. However, because some types of bacteria and red blood cells share similarities in size, morphology, and optical properties, it can be difficult to distinguish them, affecting the accurate identification of urine formed elements and causing bacteria to be mistaken for red blood cells.

[0049] On the other hand, further intelligent analysis of true hematuria is also an urgent clinical need. The etiology of hematuria is complex, and clinically, it can be divided into glomerular origin and non-glomerular origin (stones, inflammation, tumors, etc.). Routine urinalysis only provides information on indicators such as pH, protein, glucose, ketone bodies, bilirubin, urobilinogen, red blood cells or hemoglobin, nitrite, white blood cells, and specific gravity, as well as counts of formed elements such as red blood cells, white blood cells, bacteria, parasites, yeast, squamous epithelial cells, non-squamous epithelial cells, crystals, hyaline casts, unclassified casts, mucus filaments, and sperm. However, current urinalysis data on the subcategories of formed elements in urine are insufficient. According to the "Expert Consensus on the Nomenclature and Result Reporting of Formed Elements in Urine Examination" published in the *Chinese Journal of Laboratory Medicine* in 2021, red blood cells can be further subdivided into 11 subcategories, white blood cells into 8 subcategories, epithelial cells into 15 subcategories, casts into 21 subcategories, and crystals into 24 subcategories. Frontline primary care physicians are busy, and due to the limited information on formed elements in urinalysis, the analysis of true hematuria often relies heavily on the physician's subjective judgment. This places a high demand on the physician's medical skills and experience, thus affecting the objectivity and accuracy of identifying the specific cause of hematuria.

[0050] In conclusion, the misdiagnosis of true hematuria due to interfering factors is an important issue. Furthermore, how to further identify subclasses in true hematuria and improve the accuracy of subclass identification is also an important issue.

[0051] To address the aforementioned issues, this embodiment of the application acquires a urine formed element map and the interference state of the urine to be analyzed. From a preset set of correspondences between interference states and hematuria analysis network models, a hematuria analysis network model corresponding to the interference state is selected. The urine formed element map is input into the hematuria analysis network model, and the number of red blood cells in the urine to be analyzed is determined based on the urine formed element map. The hematuria category is then determined based on the number of red blood cells, where the hematuria category includes false hematuria or true hematuria. This application, by constructing several corresponding hematuria analysis network models based on the interference state of the urine to be analyzed, can reduce misjudgments of red blood cells in the urine to be analyzed, achieve accurate red blood cell counting, and thus improve the accuracy of hematuria category detection.

[0052] The application content will be further explained below with reference to the accompanying drawings and the description of the embodiments.

[0053] This embodiment provides a blood and urine analysis method based on artificial intelligence, such as... Figure 1 As shown, the method includes:

[0054] S10. Obtain the urine fraction diagram and interference status of the urine to be analyzed.

[0055] Specifically, the urine to be analyzed is the subject's hematuria, such as urine with positive occult blood or microscopic hematuria detected in a routine urinalysis. The urine formed element image is a formed element image of hematuria, such as a formed element image of sterile urine with positive occult blood, a formed element image of bacteria urine with positive occult blood, etc. The urine formed element image can be obtained by an automated urine analyzer, a digital camera, or by microscopic imaging. Of course, in practical applications, the urine formed element image can also be obtained by other methods, such as by using a urine formed element analyzer.

[0056] The interfering state of the urine to be analyzed refers to the urine state corresponding to the interfering factors that affect the judgment of hematuria. The interference factors causing the misjudgment of false hematuria as true hematuria are those resulting from the interfering factors. In practical applications, false hematuria and true hematuria are determined by detecting red blood cells in the urine. That is, if the urine to be analyzed contains a certain number of red blood cells, it indicates that the subject does indeed have urinary bleeding, and thus the urine to be analyzed is true hematuria; if the urine to be analyzed does not contain red blood cells or contains only a small number of red blood cells, although the urine color is red, the subject does not have urinary bleeding, and thus the urine to be analyzed is false hematuria. Therefore, the interfering factors can be factors that affect the red blood cell count in the urine to be analyzed, such as bacteria in the urine, the urine collection and preservation method, the urine concentration and pH, and the subject's physiological and disease state. The urine state can be the urine state corresponding to the interfering factors.

[0057] In one implementation of this embodiment, the interfering factor can be bacteria in the urine, and the interfering state can be either bacteria-containing urine or sterile urine. Specifically, bacteria-containing urine refers to urine in which bacteria present can affect the red blood cell count, while sterile urine refers to urine in which bacteria present cannot affect the red blood cell count. It is understood that when the interfering state is bacteria-containing urine, the bacteria in the urine may affect the red blood cell count, potentially misdiagnosing false hematuria as true hematuria; conversely, when the interfering state is sterile urine, the bacteria in the urine will not affect the red blood cell count, and false hematuria will not be misdiagnosed as true hematuria due to the presence of bacteria in the urine.

[0058] In one implementation of this embodiment, such as Figure 2 As shown, by inputting the formed component image of the urine to be analyzed into a trained interference state analysis network model, the interference state of the formed component image of the urine is obtained.

[0059] Specifically, the preset interference state is an interference state pre-set according to the interference factor. For example, when the interference factor is bacteria in urine, the preset interference state can be sterile urine or sterile urine.

[0060] The interference state analysis network model is a trained deep learning model used to identify interference states in urine formed element images. The training dataset for this network model includes urine samples with interference factors corresponding to preset interference states. For example, when the interference factor is bacteria in the urine, the training dataset can contain formed element images of sterile urine with positive occult blood and formed element images of bacteria-containing urine with positive occult blood. This embodiment uses an interference state analysis model to obtain the interference states in the urine formed element images, reducing misdiagnosis of hematuria due to interference states.

[0061] In one implementation of this embodiment, according to the artificial intelligence-based hematuria analysis method, the process of acquiring the interference state specifically includes:

[0062] The number of bacteria in the urine sample to be analyzed was detected.

[0063] When the number of bacteria exceeds the preset number, the urine containing bacteria will be considered as an interference state of the urine to be analyzed.

[0064] When the number of bacteria is less than or equal to a preset number, sterile urine is considered as an interference state of the urine to be analyzed.

[0065] Specifically, the preset quantity is a pre-set bacterial count based on actual conditions, used to determine whether the bacterial count in the urine to be analyzed will affect the red blood cell count, that is, whether the bacterial count in the urine to be analyzed will cause false hematuria to be misdiagnosed as true hematuria. When the bacterial count is greater than the preset quantity, it means that the bacterial count will cause false hematuria to be misdiagnosed as true hematuria, and the urine containing bacteria is considered an interference state of the urine to be analyzed; when the bacterial count is less than or equal to the preset quantity, it means that the bacterial count will not cause false hematuria to be misdiagnosed as true hematuria, and the sterile urine is considered an interference state of the urine to be analyzed.

[0066] S20. From the preset set of correspondences between interference states and hematuria analysis network models, select the hematuria analysis network model corresponding to the interference state.

[0067] Specifically, the correspondence set includes several hematuria analysis network models. Each hematuria analysis network model corresponds to a different interference state. Each hematuria analysis network model is a trained deep learning network model, and the training analysis urine samples in the training dataset corresponding to each hematuria analysis network model all have the interference state corresponding to that model. For example, when the interference state is sterile urine, there is only one hematuria analysis network model corresponding to it in the preset correspondence set; this model can be the first hematuria analysis network model. When the interference state is sterile urine, there is also only one hematuria analysis network model corresponding to it in the preset correspondence set; this model can be the second hematuria analysis network model. In this embodiment, the training dataset for the first hematuria analysis network model consists entirely of sterile urine samples. For example, the training dataset for the first hematuria analysis network model could include formed element images of sterile urine samples with positive occult blood and formed element images of sterile urine samples with false positive occult blood. Similarly, the training dataset for the second hematuria analysis network model consists entirely of sterile urine samples. For example, the training dataset for the second hematuria analysis network model could include formed element images of sterile urine samples with positive occult blood and formed element images of sterile urine samples with false positive occult blood. This embodiment improves the accuracy of each hematuria analysis network model in detecting hematuria categories by ensuring that the training urine samples in the training dataset for each hematuria analysis network model have the corresponding interference states.

[0068] In one implementation of this embodiment, the blood and urine analysis network models in the plurality of blood and urine analysis network models have the same model structure. Each blood and urine analysis network model includes a convolutional module, several cascaded residual modules, an average pooling layer, and a fully connected layer. The convolutional module is connected to the first residual module, the last residual module is connected to the average pooling layer, and the average pooling layer is connected to the fully connected layer. Each convolutional module includes a convolutional layer and a pooling layer, with the convolutional layer connected to the pooling layer, and the pooling layer connected to the first residual module. Furthermore, the model structure of the disturbance state analysis network model is the same as that of the blood and urine analysis network model.

[0069] In one implementation, such as Figure 3As shown, the convolutional module includes one convolutional layer and one pooling layer. The convolutional layer can use a 7×7 convolutional kernel and have 64 channels, while the pooling layer can use a 3×3 pooling kernel. The residual module can sequentially include three first residual modules, four second residual modules, six third residual modules, and three fourth residual modules. Each residual module can be composed of two stacked 3×3 convolutional layers, and the number of channels can be set according to actual needs. For example, the first residual module can be composed of two stacked 3×3 convolutional layers, each with 64 channels. The second residual module can have 128 channels, the third residual module can have 256 channels, and the fourth residual module can have 512 channels. The second, third, and fourth residual modules have the same structure as the first residual module and can all be composed of two stacked 3×3 convolutional layers. The first residual module is the foremost residual module and is connected to the pooling layer. The last residual module is the last residual module and is connected to the average pooling layer. The average pooling layer is connected to the fully connected layer. The average pooling layer can use a 7×7 pooling kernel.

[0070] S30. Input the urine formed element map into the hematuria analysis network model, determine the number of red blood cells corresponding to the urine to be analyzed through the urine formed element map, and determine the hematuria category corresponding to the urine to be analyzed based on the number of red blood cells.

[0071] Specifically, the hematuria analysis network model is used to detect the hematuria category in the formed fraction map of the urine, wherein the hematuria category includes false hematuria or true hematuria. The hematuria analysis network model is selected from the correspondence set based on the interference category of the urine to be analyzed. For example, when the interference category of the urine to be analyzed is sterile urine, a first hematuria analysis network model is selected; when the interference category of the urine to be analyzed is sterile urine, a second hematuria analysis network model is selected. This embodiment selects the hematuria analysis network model corresponding to the interference state based on the interference category of the urine to be analyzed through the correspondence set, which can reduce the possibility of false hematuria detection due to interference factors, thereby improving the accuracy of true hematuria detection.

[0072] In one implementation of this embodiment, determining the number of red blood cells in the urine sample to be analyzed based on the urine formed element map, and determining the hematuria category of the urine sample to be analyzed based on the number of red blood cells, specifically involves:

[0073] Obtain the number of red blood cells in the urine sample to be analyzed;

[0074] When the number of red blood cells is greater than or equal to the preset number of red blood cells, the type of hematuria with a formed fraction in the urine is true hematuria;

[0075] When the number of red blood cells is less than a preset number, the hematuria category in the urine formation profile is pseudohematuria.

[0076] Specifically, the number of red blood cells in the urine to be analyzed can be identified and obtained through the corresponding hematuria analysis network model. The corresponding hematuria analysis network model can be either the first hematuria analysis network model or the second hematuria analysis network model. That is, the first or second hematuria analysis network model can identify red blood cells in the formed elements map of the urine and count them to obtain the number of red blood cells in the urine to be analyzed. The hematuria category corresponding to the urine to be analyzed is determined by obtaining the number of red blood cells in the urine.

[0077] The preset red blood cell count is a number of red blood cells pre-set according to actual conditions, and is a threshold value used to determine the type of hematuria. For example, the preset red blood cell count can be 3 centrifuged red blood cells per high-power field or 8000 centrifuged red blood cells per milliliter. In other words, when there are greater than or equal to 3 centrifuged red blood cells per high-power field or greater than or equal to 8000 centrifuged red blood cells per milliliter, the hematuria with formed patterns in the urine is true hematuria; when there are less than 3 centrifuged red blood cells per high-power field or less than 8000 centrifuged red blood cells per milliliter, the hematuria with formed patterns in the urine is pseudo-hematuria.

[0078] In one implementation of this embodiment, according to the artificial intelligence-based hematuria analysis method, the method further includes:

[0079] When the type of hematuria is true hematuria, the proportion of abnormal red blood cells in the urine fraction graph is obtained;

[0080] The hematuria classification corresponding to the urine to be analyzed is predicted based on the proportion of abnormal red blood cells, wherein the hematuria classification includes glomerular hematuria or non-glomerular hematuria.

[0081] Specifically, the hematuria classification is based on the source of true hematuria, which can be divided into glomerular hematuria and non-glomerular hematuria. Glomerular hematuria refers to true hematuria originating from the kidneys, such as hematuria caused by glomerular diseases, kidney stones, nephritis, or kidney tumors. Non-glomerular hematuria refers to true hematuria originating from sites other than the kidneys, such as hematuria caused by ureteral, bladder, or urethral infections, urinary tract infections, stones, or tumors.

[0082] The deformed red blood cells are those formed when red blood cells in glomerular hematuria are compressed and deformed as they pass through narrow and tortuous renal tubules. Red blood cells in non-glomerular hematuria do not pass through renal tubules, therefore they are not compressed and deformed, thus forming the deformed red blood cells.

[0083] When the hematuria category is true hematuria, the proportion of abnormal red blood cells in the formed elements map of the urine is obtained, and the hematuria category corresponding to the urine to be analyzed is predicted based on the proportion of abnormal red blood cells. Specifically, by constructing a third hematuria analysis network model and a fourth hematuria analysis network model, the proportion of abnormal red blood cells in the formed elements map of the urine is obtained, and the hematuria category corresponding to the urine to be analyzed is obtained.

[0084] Specifically, both the third and fourth hematuria analysis network models are among the several hematuria analysis network models, and both are used to classify true hematuria. That is, both the third and fourth hematuria analysis network models are used to detect whether true hematuria is glomerular or non-glomerular in origin. The only difference is that the interference states corresponding to the third and fourth hematuria analysis network models are different; therefore, the training datasets for the third and fourth hematuria analysis network models are different. For example, if the interference state corresponding to the third hematuria analysis network model is bacterial urine, then the training dataset for the third hematuria analysis network model can be images of glomerular hematuria and non-glomerular hematuria formed elements from images of bacterial urine with positive occult blood. If the interference state corresponding to the fourth hematuria analysis network model is sterile urine, then the training dataset for the fourth hematuria analysis network model can be images of glomerular hematuria and non-glomerular hematuria formed elements from images of sterile urine with positive occult blood. In other words, when the interference state is bacterial urine and the hematuria category is true hematuria, the hematuria category is obtained through the third hematuria analysis network model; when the interference state is sterile urine and the hematuria category is true hematuria, the hematuria category is obtained through the fourth hematuria analysis network model.

[0085] The specific steps for predicting the hematuria classification of the urine sample based on the proportion of abnormal red blood cells are as follows:

[0086] When the proportion of abnormal red blood cells is greater than or equal to a preset proportion, the hematuria is classified as glomerular hematuria.

[0087] When the proportion of abnormal red blood cells is less than a preset proportion, the hematuria is classified as non-glomerular hematuria.

[0088] Specifically, the preset ratio is a pre-set proportion of abnormal red blood cells based on actual conditions, used to determine the hematuria classification of the true hematuria. In other words, the preset ratio is used to determine whether the true hematuria is glomerular hematuria or non-glomerular hematuria.

[0089] This embodiment further classifies the hematuria in urine samples that show true hematuria as either glomerular or non-glomerular. By selecting the appropriate hematuria analysis network model based on the interference state of the urine sample, the accuracy of hematuria classification in the analysis can be improved.

[0090] In one implementation of this embodiment, according to the artificial intelligence-based hematuria analysis method, the method further includes:

[0091] When the hematuria is classified as non-glomerular hematuria, the subject information corresponding to the urine to be analyzed is obtained, wherein the subject information includes at least medical history data and symptom data;

[0092] The crystallization data, bacterial data, and white blood cell data in the urine fraction map are read. The crystallization data, bacterial data, white blood cell data, and subject information are input into a trained first etiological analysis model. The first etiological analysis model outputs the hematuria subclass corresponding to the urine to be analyzed. The hematuria subclass includes stone-type hematuria, inflammatory hematuria, or hematuria of unknown cause.

[0093] The first etiology analysis model is a trained deep learning network model used to detect the hematuria subclass corresponding to the formed element images of urine. The hematuria subclass can include calculous hematuria, inflammatory hematuria, or hematuria of unknown cause. The training dataset for the first etiology analysis model can be formed element images of urine from non-glomerular hematuria sources. For example, the deep learning network model can be a cost-sensitive residual convolutional neural network. Cost-sensitive residual convolutional neural networks can solve the classification problem of imbalanced datasets, ensuring that the classification results are not biased towards the majority class, thereby guaranteeing the accuracy of the classification results.

[0094] This embodiment constructs a first etiological analysis model and combines it with the subject's information to perform hematuria subclassification detection for non-glomerular hematuria. This enables fine subclassification identification of the urine to be analyzed, improving the accuracy of subclassification identification of the urine to be analyzed.

[0095] In one implementation of this embodiment, according to the artificial intelligence-based hematuria analysis method, the method further includes:

[0096] When the hematuria subclass is hematuria of unknown cause, epithelial cells in the urine to be analyzed are obtained, and the cell characteristics of each obtained epithelial cell are identified.

[0097] The cell type of each epithelial cell is determined based on the identified cell characteristics;

[0098] Based on the identified cell types and subject information, the urine sample to be analyzed was determined to be hematuria suspected of being caused by bladder cancer or other types of hematuria.

[0099] Specifically, the cell types of the epithelial cells include renal tubular epithelial cells, columnar epithelial cells, high-grade urothelial carcinoma cells, suspected high-grade urothelial carcinoma cells, atypical urothelial cells, and low-grade urothelial tumor cells. The cellular characteristics of each epithelial cell are its internal texture features; in other words, the cell type of the epithelial cells can be identified based on the differences in their internal texture features.

[0100] The method for determining the cell type of each epithelial cell based on the identified cell features involves using a trained second etiology analysis model. This second etiology analysis model is a trained deep learning network model, such as a trained cost-sensitive residual convolutional neural network. The training dataset for this second etiology analysis model can be formed element images of urine from cases where hematuria is subclassified as non-glomerular hematuria of unknown cause.

[0101] The second etiological analysis model is used to detect the epithelial cell subclassification corresponding to the formed element image of the urine, wherein the epithelial cell subclassification includes hematuria suspected of bladder cancer or other hematuria. Based on the determined cell types and subject information, the epithelial cell subclassification of the urine to be analyzed is determined, that is, the urine to be analyzed is determined to be hematuria suspected of bladder cancer or other hematuria. For example, if the epithelial cells in the obtained urine to be analyzed contain high-grade urothelial carcinoma cells or suspected high-grade urothelial carcinoma cells, combined with the subject information, it can be determined whether the urine to be analyzed is hematuria suspected of bladder cancer or other hematuria. This embodiment targets non-glomerular hematuria. Through the second etiological analysis module and combined with the subject information, it can achieve precise identification of the epithelial cell subclassification of the urine to be analyzed, improving the accuracy of the identification of the epithelial cell subclassification of the urine to be analyzed.

[0102] In summary, this embodiment provides an artificial intelligence-based hematuria analysis method. The method includes: acquiring a urine formed element map of the urine to be analyzed and the interference states of the urine; selecting a hematuria analysis network model corresponding to a preset set of correspondences between interference states and hematuria analysis network models; inputting the urine formed element map into the hematuria analysis network model; determining the number of red blood cells in the urine to be analyzed based on the urine formed element map; and determining the hematuria category based on the number of red blood cells, wherein the hematuria category includes false hematuria or true hematuria. This application, by constructing several hematuria analysis network models corresponding to the interference states of the urine to be analyzed, can reduce misjudgments of red blood cells in the urine to be analyzed, achieve accurate red blood cell counting, and thus improve the accuracy of true hematuria detection and the accuracy of hematuria category detection. This application targets urine samples with true hematuria. Based on the proportion of abnormal red blood cells in the urine sample, it further detects the hematuria classification. When the hematuria is classified as non-glomerular hematuria, a first etiological analysis model is constructed and combined with the subject's information to further obtain the hematuria subclassification, which can improve the accuracy of subclassification identification in the urine sample.

[0103] Based on the above-mentioned AI-based hematuria analysis method, this embodiment provides an AI-based hematuria analysis device, such as... Figure 4 As shown, the artificial intelligence-based hematuria analysis device specifically includes:

[0104] The acquisition module 100 acquires the urine formation pattern of the urine to be analyzed and the interference state of the urine to be analyzed;

[0105] The determination module 200 selects the blood and urine analysis network model corresponding to the interference state from the preset set of correspondences between interference states and blood and urine analysis network models.

[0106] The classification module 300 inputs the urine formed element map into the hematuria analysis network model, determines the number of red blood cells corresponding to the urine to be analyzed through the urine formed element map, and determines the hematuria category corresponding to the urine to be analyzed based on the number of red blood cells, wherein the hematuria category includes false hematuria or true hematuria.

[0107] Based on the above-described AI-based hematuria analysis method, this embodiment provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the AI-based hematuria analysis method described in the above embodiment.

[0108] Based on the aforementioned AI-based hematuria analysis method, this application also provides a terminal device, such as... Figure 5 As shown, it includes at least one processor 20; a display screen 21; and a memory 22, and may also include a communications interface 23 and a bus 24. The processor 20, display screen 21, memory 22, and communications interface 23 can communicate with each other via the bus 24. The display screen 21 is configured to display a preset user guide interface in the initial setup mode. The communications interface 23 can transmit information. The processor 20 can invoke logical instructions in the memory 22 to execute the methods described in the above embodiments.

[0109] Furthermore, the logical instructions in the aforementioned memory 22 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0110] The memory 22, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 22, thereby implementing the methods in the above embodiments.

[0111] The memory 22 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 22 may include high-speed random access memory (RAM) and non-volatile memory. Examples include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, as well as transient storage media.

[0112] Furthermore, the specific process of loading and executing multiple instruction processors in the aforementioned storage medium and terminal device has been described in detail in the above method, and will not be repeated here.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A hematuria analysis method based on artificial intelligence, characterized in that, The aforementioned AI-based hematuria analysis method specifically includes: Obtain the urine formation profile and interference state of the urine to be analyzed; In the preset set of correspondences between interference states and blood and urine analysis network models, the blood and urine analysis network model corresponding to the interference state is selected. The formed element map of urine is input into the hematuria analysis network model. The number of red blood cells in the urine to be analyzed is determined by the formed element map of urine. Based on the number of red blood cells, the hematuria category of the urine to be analyzed is determined. The hematuria category includes pseudo-hematuria or true hematuria.

2. The artificial intelligence-based hematuria analysis method according to claim 1, characterized in that, The correspondence set includes several blood and urine analysis network models. Each blood and urine analysis network model corresponds to a disturbance state, and the disturbance states corresponding to each blood and urine analysis network model are different from each other. Each blood and urine analysis network model is a trained deep learning network model, and the training analysis urine in the training dataset corresponding to each blood and urine analysis network model has the disturbance state corresponding to that blood and urine analysis network model.

3. The artificial intelligence-based hematuria analysis method according to claim 2, characterized in that, The blood and urine analysis network models in the plurality of blood and urine analysis network models have the same model structure. The blood and urine analysis network model includes a convolutional module, a plurality of cascaded residual modules, an average pooling layer and a fully connected layer. The convolutional module is connected to the first residual module, the last residual module is connected to the average pooling layer, and the average pooling layer is connected to the fully connected layer. The convolutional module includes a convolutional layer and a pooling layer. The convolutional layer is connected to the pooling layer, and the pooling layer is connected to the first residual module.

4. The artificial intelligence-based hematuria analysis method according to claim 1, characterized in that, The process of obtaining the interference state specifically includes: The number of bacteria in the urine sample to be analyzed was detected. When the number of bacteria exceeds the preset number, the urine containing bacteria will be considered as an interference state of the urine to be analyzed. When the number of bacteria is less than or equal to a preset number, sterile urine is considered as an interference state of the urine to be analyzed.

5. The artificial intelligence-based hematuria analysis method according to claim 1, characterized in that, The method further includes: When the type of hematuria is true hematuria, the proportion of abnormal red blood cells in the urine fraction graph is obtained; The hematuria classification corresponding to the urine to be analyzed is predicted based on the proportion of abnormal red blood cells, wherein the hematuria classification includes glomerular hematuria or non-glomerular hematuria.

6. The artificial intelligence-based hematuria analysis method according to claim 5, characterized in that, The method further includes: When the hematuria is classified as non-glomerular hematuria, the subject information corresponding to the urine to be analyzed is obtained, wherein the subject information includes at least medical history data and symptom data; The crystallization data, bacterial data, and white blood cell data in the urine fraction map are read. The crystallization data, bacterial data, white blood cell data, and subject information are input into a trained first etiological analysis model. The first etiological analysis model outputs the hematuria subclass corresponding to the urine to be analyzed. The hematuria subclass includes stone-type hematuria, inflammatory hematuria, or hematuria of unknown cause.

7. The artificial intelligence-based hematuria analysis method according to claim 6, characterized in that, The method further includes: When the hematuria subclass is hematuria of unknown cause, epithelial cells in the urine to be analyzed are obtained, and the cell characteristics of each obtained epithelial cell are identified. The cell type of each epithelial cell is determined based on the identified cell characteristics; Based on the identified cell types and subject information, the urine sample to be analyzed was determined to be hematuria suspected of being caused by bladder cancer or other types of hematuria.

8. A blood and urine analysis device based on artificial intelligence, characterized in that, The aforementioned AI-based hematuria analysis device specifically includes: The acquisition module acquires the urine formation pattern of the urine to be analyzed and the interference state of the urine to be analyzed; The determination module selects the blood and urine analysis network model corresponding to the interference state from the preset correspondence between interference states and blood and urine analysis network models. The classification module inputs the urine formed element map into the hematuria analysis network model, determines the number of red blood cells in the urine to be analyzed based on the urine formed element map, and determines the hematuria category corresponding to the urine to be analyzed based on the number of red blood cells, wherein the hematuria category includes pseudo-hematuria or true hematuria.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the artificial intelligence-based hematuria analysis method as described in any one of claims 1-7.

10. A terminal device, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the artificial intelligence-based hematuria analysis method as described in any one of claims 1-7.

Images