A method for evaluating urinary system diseases based on urine testing

Abstract

The present invention provides a method for evaluating urinary system diseases based on urine detection, which belongs to the technical field of urology. The method for evaluating urinary system diseases based on urine detection comprises: adding corresponding reagents to three urine samples in sequence according to a preset time point sequence, and obtaining urine sample images after each addition of the reagents; performing image preprocessing on each acquired urine sample image, extracting image features including color changes, sediment morphology and distribution; establishing a time series feature data set, arranging and combining the image features corresponding to each time point in time sequence; using a pre-trained image recognition model to analyze the image features in the time series feature data set, and obtaining feature vectors; the method can solve the problems of high examination cost, great trauma, long examination time, limited diagnostic accuracy, etc. that the existing means of diagnosing urinary system diseases generally have.

Description

Technical Field

[0001] The present invention belongs to the technical field of urology, and in particular, relates to a method for evaluating urinary system diseases based on urine testing. Background Art

[0002] The urinary system is one of the most important organ systems in the human body. Its main functions include filtering blood, excreting metabolic waste, and maintaining fluid balance. However, since the urinary system is susceptible to various diseases, it is particularly important to diagnose and evaluate urinary system diseases in a timely and accurate manner. At present, the commonly used methods for diagnosing urinary system diseases in clinical practice mainly include routine urine examination, imaging examination (such as B-ultrasound, CT, MRI, etc.), biochemical examination, cytological examination, and histological examination. Among them, routine urine examination, as the most basic and routine examination method, can reflect the basic pathological changes of the urinary system, such as proteinuria, glucosuria, occult blood, etc. However, traditional routine urine examination can usually only perform simple qualitative or semi-quantitative tests on the physical and chemical indicators of urine samples, and it is difficult to make an in-depth analysis of the development dynamics and complex pathological mechanisms of the disease. Imaging examination and biochemical examination can provide more abundant information, but these examinations usually require professional instruments and technicians, are expensive, and are somewhat invasive. In addition, although cytological and histological examinations, as gold standards, can directly observe the pathological changes of diseased tissue cells, the operation is relatively complicated, takes a long time, and may cause certain trauma.

[0003] In general, the existing diagnostic methods for urinary system diseases generally have problems such as high examination costs, high trauma, long examination time, and limited diagnostic accuracy. This urgently requires the development of a new detection method that can quickly, simply, and non-invasively evaluate the pathological changes of the urinary system and provide a more reliable basis for clinical diagnosis. Summary of the invention

[0004] In view of this, the present invention provides a method for evaluating urinary system diseases based on urine testing, aiming to solve the problems of existing urinary system disease diagnosis methods, such as high examination cost, high trauma, long examination time, and limited diagnostic accuracy.

[0005] The present invention is achieved in that:

[0006] The present invention provides a method for evaluating urinary system diseases based on urine testing, which comprises the following steps:

[0007] S10, adding corresponding reagents to the three urine samples in sequence according to a preset time point sequence, and acquiring a urine sample image after each addition of the reagent;

[0008] S20, performing image preprocessing on each acquired urine sample image to extract image features including color change, sediment morphology and distribution;

[0009] S30, establishing a temporal feature data set, arranging and combining the image features corresponding to each time point in chronological order;

[0010] S40, using a pre-trained image recognition model to analyze the image features in the time series feature data set to obtain a feature vector;

[0011] S50, inputting the feature vector into a preset disease assessment model, where the disease assessment model is trained by a large amount of clinical case data;

[0012] S60, obtaining the disease type and its corresponding probability value output by the disease assessment model;

[0013] S70, determining the credibility of the evaluation result according to the probability value, and screening the evaluation results with high credibility according to a preset threshold;

[0014] S80, comparing the highly reliable evaluation results with clinical diagnostic criteria to generate a urinary system disease evaluation report;

[0015] S90, adding the urinary system disease assessment report and its corresponding time series feature data to the sample library for subsequent model optimization.

[0016] On the basis of the above technical solution, the method for evaluating urinary system diseases based on urine testing of the present invention can also be improved as follows:

[0017] The three urine samples are first-pass urine, mid-pass urine and final-pass urine:

[0018] The first-pass urine refers to the first 10 ml of urine at the beginning of urination, which mainly reflects the condition of the urethra.

[0019] The midstream urine refers to the middle part of urine after the first-stream urine is removed, which is 100 to 200 ml. This urine can best reflect the condition of the bladder and the urinary system above it.

[0020] The terminal urine refers to the last 10 ml of urine at the end of urination, which can reflect the condition of the prostate.

[0021] Furthermore, the corresponding reagents are added in sequence at the preset time points. Specifically, after collecting three urine samples, urine test paper, urine sediment stain, urine protein reagent, urine sugar detection reagent and urobilinogen detection reagent are added to each sample in sequence at five time points: 0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes respectively.

[0022] Furthermore, the method for extracting image features including color changes, sediment morphology and distribution is specifically to use a high-power microscope to perform 360-degree rotational photography of each urine sample to obtain characteristic parameters such as sample color, transparency, sediment size, shape, distribution density, etc.

[0023] Furthermore, the method of pre-training the image recognition model is specifically to collect more than 10,000 urine sample images of patients with clinically confirmed urinary system diseases, annotate each image, establish an image database, and use a deep learning algorithm to train the model so that the model can accurately identify the urine characteristics corresponding to different diseases.

[0024] Furthermore, the step of constructing the training data set of the disease assessment model specifically collects complete medical records of patients with clinically confirmed urinary system diseases, including symptom descriptions, examination results, diagnostic conclusions, etc., and associates them with corresponding urine test results to form a standardized training data set.

[0025] The step S10 specifically includes: placing three urine samples (first-pass urine, mid-pass urine and final-pass urine) in three separate containers respectively; then, according to the preset time point sequence (0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes), adding corresponding reagents to each container in turn, including urine test paper, urine sediment stain, urine protein reagent, urine sugar detection reagent and urobilinogen detection reagent; after each reagent is added, immediately rotating each urine sample 360 ​​degrees to obtain a complete sample image. In this way, the color change, sediment morphology and distribution characteristics of urine samples at different time points can be analyzed in time series, thereby reflecting the overall condition of the urinary system.

[0026] The step S20 specifically includes: first, preprocessing each acquired urine sample image to remove noise and irrelevant information in the image; then, using an image analysis algorithm to extract characteristic parameters such as color (such as RGB value, HSV value, etc.), transparency (such as gray value or optical density value), sediment characteristics (such as area, perimeter, major-minor axis ratio, and other geometric parameters) of the sample image. These characteristic parameters can reflect the physical and chemical properties and pathological changes of the urine sample, and provide an important basis for subsequent disease assessment.

[0027] The step S30 specifically includes: arranging the image features of each time point extracted in step S20 in chronological order to construct a time series feature data set. In this way, a matrix composed of multi-dimensional feature vectors can be obtained, in which each row corresponds to a time point and each column corresponds to an image feature. This time series feature data set can fully describe the feature evolution law of urine samples in the process of time change, and provide more valuable information for subsequent disease identification.

[0028] The step S40 specifically includes: using a pre-trained deep learning-based convolutional neural network (CNN) image recognition model, inputting the time series feature data set constructed in step S30 into the model for analysis. The model is obtained by collecting more than 10,000 urine sample images of patients with clinically confirmed urinary system diseases, annotating each image, establishing an image database, and then training using an iterative optimization method. After training, the model can accurately identify urine features corresponding to different diseases and convert them into feature vector outputs.

[0029] The step S50 specifically includes: taking the feature vector outputted in step S40 as input, and inputting it into a pre-trained disease assessment model for analysis. The disease assessment model is established by collecting complete medical records (including symptom descriptions, examination results, diagnosis conclusions, etc.) of a large number of clinically diagnosed patients with urinary system diseases, associating them with corresponding urine test results, forming a standardized training data set, and then training it using a machine learning algorithm. The model can mine the complex association between urine features and disease types, and output the probability value of each disease.

[0030] The step S60 specifically includes: obtaining preliminary results of disease assessment according to the disease type and its corresponding probability value outputted in step S50. It should be noted that the probability values ​​of different diseases may vary greatly, and their credibility needs to be further analyzed.

[0031] The step S70 specifically includes: screening the disease assessment results obtained in step S60 according to a preset credibility threshold. When the probability value of a certain disease exceeds 85%, the assessment result can be considered to have a high credibility; when the probability value of a certain disease is between 70% and 85%, the assessment result can be considered to have a medium credibility; when the probability values ​​of all diseases are lower than 70%, the assessment result can be considered to have a low credibility. Only the assessment results with high credibility will enter the next step of processing to ensure the accuracy and reliability of the diagnosis.

[0032] The step S80 specifically includes: comparing and analyzing the high-reliability disease assessment results screened out in step S70 with the clinical diagnosis criteria. If the assessment results are completely consistent with the clinical diagnosis, a urinary system disease assessment report is directly generated; if there is a difference, further verification and adjustment of the assessment model are required. The purpose of this step is to verify the effectiveness of the established disease assessment method and continuously optimize the model to improve its diagnostic accuracy.

[0033] The step S90 specifically includes: adding the urinary system disease assessment report generated in step S80 and the corresponding time series feature data set to the sample library to provide a basis for subsequent model optimization and iteration. By continuously accumulating clinical diagnosis data and urine sample data, the performance and reliability of the disease assessment model can be further improved.

[0034] Furthermore, the three urine samples are collected at one time using a segmented urine collection device.

[0035] Furthermore, the urine segmented collection device comprises: a main frame, a shunt assembly, a collection assembly and a sealing assembly; the main frame comprises a top plate, a bottom plate and four vertical support columns, and the four vertical support columns are respectively fixedly connected to the four corners of the top plate and the bottom plate; a liquid inlet hole is opened in the center of the top plate, and a funnel-shaped urine collecting basin is arranged above the liquid inlet hole; the shunt assembly is fixedly installed in the main frame and is located directly below the liquid inlet hole; the collection assembly is arranged below the shunt assembly; the sealing assembly is arranged on the bottom plate and connected to the collection assembly; the shunt assembly comprises:

[0036] A rotating support frame, a rotating shaft, a diverter plate and a plurality of guide grooves; the rotating support frame is fixed on four vertical support columns by bolts; the rotating shaft is vertically installed at the center of the rotating support frame by bearings; the diverter plate is fixedly installed on the rotating shaft; three diverter grooves are evenly arranged radially on the upper surface of the diverter plate, the inner end of each diverter groove is connected to the center of the diverter plate, and the outer end extends to the edge of the diverter plate; the collecting assembly includes: three collecting bottles, a bottle seat and a driving motor; the bottle seat is fixedly installed on the bottom plate in a circular ring shape, and three mounting holes are evenly arranged on the bottle seat along the circumferential direction; the three collecting bottles are respectively fixed on the bottle seat through the mounting holes, and the positions of the collecting bottles correspond to the liquid outlet positions of the diverter grooves one by one; the driving motor is fixedly installed on the rotating support frame, and the output shaft of the driving motor is coaxially connected to the rotating shaft; three groups of infrared sensors are arranged on the rotating support frame for detecting the rotation position of the diverter plate; each group of infrared sensors corresponds to the position of a diverter groove, and a signal is triggered when the diverter groove rotates to the corresponding collecting bottle position.

[0037] Furthermore, the sealing assembly includes: three sealing joints and three hoses; the sealing joints are respectively installed at the bottle mouths of three collecting bottles; one end of the three hoses is respectively connected to the three sealing joints, and the other end is respectively sealed and connected to the liquid outlet end of the guide groove.

[0038] Furthermore, the inner surface of the funnel-shaped urine collecting basin is coated with a hydrophobic coating, and the diameter of the liquid outlet at the bottom of the urine collecting basin is 15 mm; the diameter of the diversion plate is 300 mm and the thickness is 20 mm; the cross-sections of the three guide grooves are all semicircular, with a groove depth of 5 mm and a groove width of 8 mm; the volume of the three collecting bottles is 50 ml, and the bottle body is made of transparent material and marked with scale lines; the outer diameter of the bottle base is 400 mm and the inner diameter is 200 mm.

[0039] Compared with the prior art, the method for evaluating urinary system diseases based on urine testing provided by the present invention has the following beneficial effects: The present invention proposes a method for evaluating urinary system diseases based on urine testing, which can accurately evaluate the health status of the patient's urinary system by analyzing the changes in the physical and chemical indicators of urine samples in time series and combining the recognition ability of deep learning models. Compared with the existing diagnostic methods for urinary system diseases, this method has the following advantages:

[0040] 1. Simple operation and non-invasive: Only urine samples from patients need to be collected, and first-pass urine, mid-pass urine and final-pass urine can be quickly obtained through the automated segmented collection device, without any invasive examination of the patient. This greatly reduces the difficulty of the examination and the damage to the patient.

[0041] 2. Low examination cost: Except for collecting urine samples, the entire testing process is completely completed automatically by instruments and equipment, without the need for professional doctors to participate, which greatly reduces the cost of diagnosis.

[0042] 3. High diagnostic accuracy: Through in-depth analysis of the changes in physical and chemical indicators of urine samples in the time dimension, combined with the powerful recognition ability of deep learning algorithms, it can more comprehensively and accurately reflect the pathological state of the urinary system, and the diagnostic accuracy is much higher than traditional routine urine examinations.

[0043] 4. Strong dynamic monitoring capability: This method can continuously monitor the changing characteristics of urine samples in a time series, which is conducive to timely discovering the progression of the disease and providing more valuable information for clinical diagnosis and treatment.

[0044] In conclusion, this urine-based urinary system disease assessment method makes full use of the current advanced image analysis and machine learning technologies, overcomes the limitations of existing diagnostic methods, and achieves rapid, accurate, and non-invasive disease diagnosis. It has broad prospects in clinical applications and is expected to become a new diagnostic tool for urinary system diseases. It can solve the problems of high examination costs, high trauma, long examination time, and limited diagnostic accuracy that are common in existing urinary system disease diagnostic methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0046] Figure 1 A flowchart of a method for evaluating urinary system diseases based on urine testing;

[0047] Figure 2 It is a schematic diagram of the overall structure of the urine segmented collection device;

[0048] Figure 3 It is a top view of the flow distribution component;

[0049] Figure 4 The representative images of Zhang's first-pass urine, middle-pass urine and final-pass urine at different time points in the embodiment;

[0050] Figure 5 The characteristic diagram of the changes of the physical and chemical indicators of the three urine samples in the embodiment during the time change process;

[0051] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0052] 1. Main frame; 11. Top plate; 12. Bottom plate; 13. Support column; 2. Urine collecting basin; 3. Diversion assembly; 31. Rotating support frame; 32. Rotating shaft; 33. Diversion plate; 34. Guide groove; 4. Collection assembly; 41. Collection bottle; 42. Bottle holder; 43. Drive motor; 5. Sealing assembly; 51. Sealing joint; 52. Hose. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0054] like Figure 1 FIG. 1 is a flow chart of a method for evaluating urinary system diseases based on urine testing provided by the present invention. The method comprises the following steps:

[0055] S10, adding corresponding reagents to the three urine samples in sequence according to a preset time point sequence, and acquiring a urine sample image after each addition of the reagent;

[0056] S20, performing image preprocessing on each acquired urine sample image to extract image features including color change, sediment morphology and distribution;

[0057] S30, establishing a temporal feature data set, arranging and combining the image features corresponding to each time point in chronological order;

[0058] S40, using a pre-trained image recognition model to analyze image features in the time series feature data set to obtain a feature vector;

[0059] S50, inputting the feature vector into a preset disease assessment model, where the disease assessment model is trained by a large amount of clinical case data;

[0060] S60, obtaining the disease type and its corresponding probability value output by the disease assessment model;

[0061] S70, determining the credibility of the evaluation result according to the probability value, and screening the evaluation results with high credibility according to a preset threshold;

[0062] S80. Compare the high-reliability evaluation results with the clinical diagnostic criteria to generate a urinary system disease evaluation report;

[0063] S90. Add the urinary system disease assessment report and its corresponding time series feature data to the sample library for subsequent model optimization.

[0064] The specific implementation method of step S10 is: first, after collecting urine samples from three different parts (first-pass urine, mid-pass urine and final-pass urine), the three urine samples are placed in three separate containers. Then, according to the preset time point sequence (0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes), the corresponding reagents are added to each container in turn. Specifically, at 0 minutes, a urine test paper is added for rapid detection of basic physical and chemical indicators of urine (such as pH value, specific gravity, etc.); at 5 minutes, a urine sediment dye is added for dyeing urine sediment and observing its morphology; at 10 minutes, a urine protein reagent is added for quantitative detection of protein content in urine; at 20 minutes, a urine sugar detection reagent is added for quantitative detection of glucose content in urine; at 30 minutes, a urobilinogen detection reagent is added for quantitative detection of bilirubin content in urine. After each reagent is added, each urine sample is immediately rotated 360 degrees to obtain a complete sample image. The purpose of this is to reflect the overall condition of the urinary system by analyzing the color changes, sediment morphology and distribution characteristics of urine samples at different time points.

[0065] The specific implementation of step S20 is: first, pre-process each acquired urine sample image to remove noise and irrelevant information in the image. Then, use the image analysis algorithm to extract the characteristic parameters of the sample image, such as color, transparency, sediment size, shape, and distribution density. Among them, color features include RGB values, HSV values, etc.; transparency features can be represented by grayscale values ​​or optical density values; sediment features can be extracted through regional segmentation algorithms. Geometric parameters such as area, perimeter, and major-minor axis ratio. These characteristic parameters can reflect the physical and chemical properties and pathological changes of urine samples, and provide important basis for subsequent disease assessment.

[0066] The specific implementation of step S30 is: the image features of each time point extracted in step S20 are arranged in chronological order to construct a time series feature data set. In this way, a matrix composed of multi-dimensional feature vectors can be obtained, in which each row corresponds to a time point and each column corresponds to an image feature. This time series feature data set can fully describe the feature evolution law of urine samples in the process of time change, and provide more valuable information for subsequent disease identification.

[0067] The specific implementation method of step S40 is: using a pre-trained image recognition model, inputting the time series feature data set constructed in step S30 into the model for analysis. The image recognition model used here is a convolutional neural network (CNN) model based on deep learning. The model is obtained by collecting more than 10,000 urine sample images of patients with clinically confirmed urinary system diseases, annotating each image, establishing an image database, and then using an iterative optimization method to train the model. After the training is completed, the model can accurately identify the urine features corresponding to different diseases and convert them into feature vector outputs. These feature vectors contain rich urine image information, laying the foundation for subsequent disease evaluation.

[0068] The specific implementation of step S50 is: the feature vector outputted from step S40 is used as input to a pre-trained disease assessment model for analysis. This disease assessment model is established by collecting complete medical records (including symptom descriptions, test results, diagnosis conclusions, etc.) of a large number of clinically diagnosed patients with urinary system diseases, associating them with corresponding urine test results, forming a standardized training data set, and then training it using a machine learning algorithm. The model can mine the complex association between urine features and disease types, and output the probability value of each disease.

[0069] The specific implementation of step S60 is: according to the disease type and its corresponding probability value outputted in step S50, a preliminary result of disease assessment is obtained. It should be noted here that the probability values ​​of different diseases may vary greatly, and their credibility needs to be further analyzed.

[0070] The specific implementation of step S70 is: according to the preset credibility threshold, the disease assessment results obtained in step S60 are screened. Specifically, when the probability value of a certain disease exceeds 85%, the assessment result can be considered to have a high credibility; when the probability value of a certain disease is between 70% and 85%, the assessment result can be considered to have a medium credibility; when the probability values ​​of all diseases are lower than 70%, the assessment result can be considered to have a low credibility. Only the assessment results with high credibility will enter the next step of processing. The purpose of this is to ensure the accuracy and reliability of the disease assessment.

[0071] The specific implementation of step S80 is: compare and analyze the high-confidence disease assessment results screened out in step S70 with the clinical diagnosis criteria. If the assessment results are completely consistent with the clinical diagnosis, a urinary system disease assessment report is directly generated; if there is a difference, further verification and adjustment of the assessment model are required. The purpose of this step is to verify the effectiveness of the established disease assessment method and continuously optimize the model to improve its diagnostic accuracy.

[0072] The specific implementation of step S90 is: adding the urinary system disease assessment report generated in step S80 and the corresponding time series feature data set to the sample library to provide a basis for subsequent model optimization and iteration. By continuously accumulating clinical diagnosis data and urine sample data, the performance and reliability of the disease assessment model can be further improved.

[0073] In addition, the present invention also provides a urine segmentation collection device to achieve automatic segmentation collection of first-pass urine, middle-pass urine and last-pass urine. The device includes the following main components:

[0074] The main frame includes a top plate, a bottom plate and four vertical support columns, wherein a liquid inlet hole is opened in the center of the top plate, and a funnel-shaped urine collecting basin is arranged above.

[0075] Diverter assembly: located below the liquid inlet, it consists of a rotating support frame, a rotating shaft, a diverter plate and several guide grooves. The diverter plate is evenly provided with three guide grooves along the radial direction, the inner end of each guide groove is connected to the center of the diverter plate, and the outer end extends to the edge of the diverter plate.

[0076] Collection component: It is set below the diversion component and includes three collection bottles, a bottle holder and a drive motor. The three collection bottles are fixedly mounted on the bottle holder, and their positions correspond to the liquid outlet ends of the diversion trough. The drive motor drives the rotating shaft and the diversion plate to rotate through the output shaft.

[0077] Sealing assembly: includes three sealing joints and three hoses, which are used to seal the liquid outlet end of the diversion trough with three collection bottles.

[0078] Among them, in the above technical solution, the three urine samples are first-pass urine, mid-pass urine and last-pass urine:

[0079] Among them, the first-pass urine refers to the first 10 ml of urine at the beginning of urination, which mainly reflects the condition of the urethra;

[0080] Among them, midstream urine refers to the middle part of urine after the first-stream urine is eliminated, which is 100 to 200 ml. This urine can best reflect the condition of the bladder and the urinary system above it.

[0081] Among them, the terminal urine refers to the last 10 ml of urine at the end of urination, which can reflect the condition of the prostate.

[0082] Furthermore, in the above technical solution, corresponding reagents are added in sequence at preset time points. Specifically, after collecting three urine samples, urine test paper, urine sediment stain, urine protein reagent, urine sugar detection reagent and urobilinogen detection reagent are added to each sample in sequence at five time points: 0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes respectively.

[0083] Furthermore, in the above technical solution, the method for extracting image features including color changes, sediment morphology and distribution is to use a high-power microscope to rotate each urine sample 360 ​​degrees to obtain characteristic parameters such as sample color, transparency, sediment size, shape, distribution density, etc.

[0084] Furthermore, in the above technical solution, the method of pre-training the image recognition model is specifically to collect more than 10,000 urine sample images of patients with clinically diagnosed urinary system diseases, annotate each image, establish an image database, and use a deep learning algorithm to train the model so that the model can accurately identify the urine characteristics corresponding to different diseases.

[0085] Furthermore, in the above technical solution, the step of constructing a training data set for the disease assessment model is to collect complete medical records of patients with clinically confirmed urinary system diseases, including symptom descriptions, examination results, diagnostic conclusions, etc., and associate them with corresponding urine test results to form a standardized training data set.

[0086] Furthermore, in the above technical solution, the three urine samples are collected at one time using a segmented urine collection device.

[0087] Further, in the above technical solution, the urine segmented collection device includes: a main frame, a shunt assembly, a collection assembly and a sealing assembly; the main frame includes a top plate, a bottom plate and four vertical support columns, and the four vertical support columns are respectively fixedly connected to the four corners of the top plate and the bottom plate; a liquid inlet hole is opened in the center of the top plate, and a funnel-shaped urine collecting basin is arranged above the liquid inlet hole; the shunt assembly is fixedly installed in the main frame, located directly below the liquid inlet hole; the collection assembly is arranged below the shunt assembly; the sealing assembly is arranged on the bottom plate and connected to the collection assembly; the shunt assembly includes:

[0088] A rotating support frame, a rotating shaft, a diverter plate and a plurality of guide grooves; the rotating support frame is fixed on four vertical support columns by bolts; the rotating shaft is vertically installed at the center of the rotating support frame through a bearing; the diverter plate is fixedly installed on the rotating shaft; three guide grooves are evenly arranged radially on the upper surface of the diverter plate, the inner end of each guide groove is connected to the center of the diverter plate, and the outer end extends to the edge of the diverter plate; the collecting component includes: three collecting bottles, a bottle seat and a driving motor; the bottle seat is fixedly installed on the bottom plate in a circular ring shape, and three mounting holes are evenly arranged on the bottle seat along the circumferential direction; the three collecting bottles are fixed on the bottle seat through the mounting holes respectively, and the positions of the collecting bottles correspond to the liquid outlet positions of the diverter grooves one by one; the driving motor is fixedly installed on the rotating support frame, and the output shaft of the driving motor is coaxially connected to the rotating shaft; three groups of infrared sensors are arranged on the rotating support frame for detecting the rotation position of the diverter plate; each group of infrared sensors corresponds to the position of a diverter groove, and a signal is triggered when the diverter groove rotates to the corresponding collecting bottle position.

[0089] Urine flows from the urine collecting basin into the liquid inlet hole, and enters three collection bottles through the guide groove of the diversion component. Driven by the drive motor, the diversion disk rotates gradually, so that the first-stage urine, middle-stage urine and last-stage urine are collected in three different collection bottles in turn. The infrared sensor detects the rotation position of the diversion disk to ensure that each segment of urine is accurately diverted. The device can realize fast and accurate segmented urine collection and provide valuable samples for subsequent detection and analysis. The start and stop time of the drive motor is determined based on multiple historical urine collection experiments, and the historical urine collection experiments are specifically the start and stop time of the drive motor required to collect urine of different segments.

[0090] The key to the entire urine testing and evaluation method is that by analyzing the changes in the physical and chemical indicators of urine samples in time series, combined with the recognition ability of the deep learning model, the patient's urinary system health status can be more accurately assessed. This method overcomes the limitations of single-time point detection and can more comprehensively reflect the development dynamics of the disease. At the same time, the use of an automated urine segmentation collection device can obtain more accurate and reliable samples, laying the foundation for subsequent testing and analysis.

[0091] Furthermore, in the above technical solution, the sealing assembly includes: three sealing joints and three hoses; the sealing joints are respectively installed at the bottle mouths of the three collecting bottles; one end of the three hoses is respectively connected to the three sealing joints, and the other end is respectively sealed and connected to the liquid outlet end of the guide groove.

[0092] Furthermore, in the above technical solution, the inner surface of the funnel-shaped urine collecting basin is coated with a hydrophobic coating, and the diameter of the liquid outlet at the bottom of the urine collecting basin is 15 mm; the diameter of the diversion plate is 300 mm and the thickness is 20 mm; the cross-sections of the three guide grooves are all semicircular, with a groove depth of 5 mm and a groove width of 8 mm; the capacity of the three collecting bottles is 50 ml, and the bottle body is made of transparent material and marked with scale lines; the outer diameter of the bottle base is 400 mm and the inner diameter is 200 mm.

[0093] Specifically, the principle of the present invention is: the core idea of ​​the technical solution design of the present invention is that by conducting in-depth analysis of the changes in the physical and chemical characteristics of urine samples in a time series, combined with the image recognition capability based on deep learning, the patient's urinary system disease condition can be evaluated more comprehensively and accurately.

[0094] Specifically, the method first uses a urine segmentation collection device to obtain urine samples from three different parts at one time: first-pass urine, mid-pass urine, and final-pass urine. This can better reflect the pathological changes in different parts such as the urethra, bladder, and prostate. Then, according to the preset time point sequence, different types of chemical reagents are added to each urine sample in turn, and the sample is immediately rotated 360 degrees after each addition of the reagent to obtain a complete sample image. These images contain rich physical and chemical information such as urine color changes, sediment morphology and distribution, and can reflect the pathological changes of the urinary system in the time dimension.

[0095] Next, the image analysis algorithm is used to extract parameters such as color, transparency, and sediment characteristics of each sample image to construct a time series feature dataset. This time series feature dataset can fully describe the characteristic evolution law of urine samples during time changes. Then, the time series feature dataset is input into the pre-trained convolutional neural network (CNN) image recognition model for analysis. The model can accurately identify the urine characteristics corresponding to different diseases and convert them into feature vectors.

[0096] Finally, the above feature vectors are input into the pre-trained disease assessment model for analysis. This assessment model collects complete medical records of a large number of patients with clinically diagnosed urinary system diseases, establishes a complex association between urine characteristics and disease types, and outputs the probability value of each disease. Based on these probability values, combined with the preset credibility threshold, high-credibility diagnostic results can be screened out and compared with clinical standards for verification.

[0097] Compared with the existing urinary system disease diagnosis methods, the key innovations of the present invention are:

[0098] 1) It makes full use of the physical and chemical change characteristics of urine samples in a time series, rather than static indicators at a single time point, which helps to more comprehensively reflect the development dynamics of the disease.

[0099] 2) It uses advanced deep learning image recognition technology, which can accurately extract key features related to the disease from urine sample images, greatly improving the accuracy of diagnosis.

[0100] 3) A standardized disease assessment model was established, and the complex relationship between urine characteristics and diseases was mined through machine learning algorithms to achieve quantitative disease risk assessment.

[0101] 4) An automated urine segmentation collection device is used, which can quickly and accurately obtain urine samples from different parts of the body, providing reliable basic data for subsequent analysis.

[0102] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below. A 30-year-old male patient named Zhang was admitted to the Urology Department of a certain hospital. He complained of symptoms such as frequent urination, painful urination, and hematuria for 2 weeks. The patient was previously healthy and had no obvious history of chronic diseases. After preliminary examination, the doctor initially suspected that Zhang might have diseases such as cystitis or urinary tract infection. In order to further clarify the diagnosis, the doctor decided to use the urinary system disease assessment method based on urine testing provided by the present invention to examine him.

[0103] First, according to step S10 of the present invention, the medical staff used the urine segmentation collection device to collect urine samples from three different parts of Zhang's urine at one time: first-pass urine, mid-pass urine and final-pass urine. Specifically, first-pass urine is the first 10 mL of urine at the beginning of urination, which mainly reflects the condition of the urethra; mid-pass urine is the 100-200 mL of urine after the first-pass urine is eliminated, which can best reflect the condition of the bladder and the urinary system above; final-pass urine is the last 10 mL of urine at the end of urination, which can reflect the condition of the prostate.

[0104] Then, according to the preset time point sequence (0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes), the following reagents were added to each urine sample in sequence:

[0105] At 0 minutes, add urine test paper to test basic physical and chemical indicators such as pH value and specific gravity of urine;

[0106] At 5 minutes, add urine sediment dye to observe the morphology of urine sediment;

[0107] At 10 minutes, urine protein reagent was added to quantitatively detect the protein content in urine;

[0108] At 20 minutes, add urine sugar detection reagent to quantitatively detect the glucose content in urine;

[0109] After 30 minutes, urobilinogen detection reagent was added to quantitatively detect the bilirubin content in urine.

[0110] After adding each reagent, medical staff immediately rotated each urine sample 360 ​​degrees to obtain a complete sample image. Figure 4 Shown are representative images of Zhang's first-pass urine, mid-pass urine, and last-pass urine at different time points.

[0111] Then, according to step S20 of the present invention, the urine sample images obtained above are processed and feature extracted. First, the image preprocessing algorithm is used to remove noise and irrelevant information, and then the color (RGB value, HSV value), transparency (grayscale value), sediment characteristics (area, perimeter, major-minor axis ratio, etc.) and other parameters of each image are extracted by color analysis, region segmentation and other methods. Table 1 lists some characteristic parameters of Zhang's three urine samples at five time points.

[0112] According to step S30, the above time series feature data are arranged in chronological order to construct a time series feature data set. In this way, a matrix composed of multi-dimensional feature vectors is formed, each row represents a time point, and each column represents an image feature. This time series feature data set can comprehensively describe the feature evolution law of urine samples in the process of time change.

[0113] Next, according to step S40, the time series feature data set is input into the pre-trained CNN image recognition model for analysis. The model is trained based on urine sample images of more than 10,000 confirmed cases and can accurately identify urine features corresponding to different diseases. After model analysis, the feature vectors of Zhang's three urine samples are shown in Table 2.

[0114] Then, according to step S50, the above feature vector is input into the pre-trained disease assessment model for analysis. The assessment model is established by collecting a large amount of clinical case data, and can mine the complex association between urine characteristics and disease types. The results of the model output are shown in Table 3, showing that the probability of Zhang suffering from cystitis is 90%, the probability of urinary tract infection is 75%, and the probability of other diseases is less than 50%.

[0115] According to step S70, since the probability value of cystitis exceeds 85%, which is a high-confidence result, it is determined that Zhang is most likely to have cystitis. The probability value of urinary tract infection is between 70% and 85%, which is a medium-confidence result. Taking all factors into consideration, the doctor determines that Zhang is very likely to have cystitis, and there is also a certain probability of urinary tract infection.

[0116] Finally, according to step S80, the doctor compares and verifies the above diagnosis results with the clinical diagnosis criteria. Through further urine routine examination, cystoscopy and other means, it is confirmed that Zhang does have the symptoms and signs of cystitis, and the diagnosis results are completely consistent with the evaluation results of the method of the present invention. The doctor formulates a corresponding treatment plan based on this, and includes the diagnosis report and time series feature data of this case into the sample library according to step S90 to provide support for subsequent model optimization.

[0117] It can be seen from the above implementation process that the method of the present invention can comprehensively and accurately evaluate the patient's urinary system disease status. Its key advantages are reflected in the following aspects:

[0118] The segmented collection of urine can better reflect the pathological changes in different parts of the urinary system.

[0119] like Figure 5 As shown in the figure, Zhang's first-pass urine, mid-pass urine and final-pass urine showed obvious differences in the course of time. The first-pass urine is turbid in color and contains more sediment, which may reflect urethritis; the mid-pass urine is more transparent and contains less sediment, which basically reflects the condition of the bladder and above urinary system; the final-pass urine is darker in color and contains more concentrated sediment, which may indicate prostatitis. These differences provide valuable information for subsequent disease analysis.

[0120] Time series feature analysis can more comprehensively reflect the development dynamics of the disease.

[0121] As can be seen from Table 1, the color, transparency, and sediment characteristics of Zhang's three urine samples changed significantly over time. These changes provide an important basis for evaluating the progression of the disease. For example, the sediment area of ​​the first-stream urine continued to increase from 0 to 30 minutes, reflecting that urethral inflammation may be gradually worsening; while the transparency of the midstream urine continued to decrease from 5 to 20 minutes, indicating that bladder inflammation may be worsening.

[0122] Table 1 Characteristic parameters of Zhang's three urine samples at five time points

[0123]

[0124]

[0125] Deep learning model recognition has greatly improved the accuracy of diagnosis.

[0126] As shown in Table 2, through the feature extraction of the CNN model, the feature vectors of Zhang's three urine samples showed obvious differences. These feature vectors contain rich disease information, laying the foundation for subsequent disease evaluation. As can be seen from Table 3, based on the analysis of these feature vectors, the model can accurately output the probability values ​​of cystitis and urinary tract infection, meeting the requirements of clinical diagnosis.

[0127] Automated sampling equipment improves the convenience and standardization of inspections.

[0128] The urine segmentation collection device can quickly and accurately obtain the first-pass urine, middle-pass urine and final-pass urine, avoiding the errors that may occur during manual collection. At the same time, the device has good sealing and automation, ensuring the purity and stability of the sample, and providing reliable basic data for subsequent detection and analysis.

[0129] In summary, by combining modern image analysis, machine learning and other technical means with urine testing, the method of the present invention can quickly, accurately and non-invasively evaluate the patient's urinary system disease status, and has broad prospects in clinical application. For patients with cystitis or urinary tract infection like Zhang, this method can detect and diagnose the disease in a timely manner, providing a basis for clinicians to formulate personalized diagnosis and treatment plans.

[0130] The following data analysis charts more intuitively demonstrate the application effect of the method of the present invention in this embodiment:

[0131] Figure 5 Representative images of Zhang's three urine samples at different time points are shown. As can be seen from the figure, the color, turbidity, and sediment distribution of each urine sample have changed significantly over time, reflecting the dynamic characteristics of lesions in various parts of the urinary system.

[0132] Figure 5 The characteristics of the changes in the physical and chemical indicators of three urine samples over time are shown. Among them, the color of the first-pass urine gradually deepens, reflecting the worsening of urethral inflammation; the turbidity of the mid-pass urine continues to increase, indicating that bladder inflammation is aggravating; the sediment area of ​​the last-pass urine continues to increase, indicating the possibility of prostate inflammation. These time series characteristics provide important basis for subsequent disease evaluation.

[0133] Table 1 lists some characteristic parameters of Zhang's three urine samples at five time points. It can be seen that the color (RGB value), transparency (grayscale value), and sediment characteristics (area, perimeter, and major-minor axis ratio) of each urine sample have changed significantly over time. These rich temporal features lay the foundation for building a disease assessment model.

[0134] Table 2 lists the feature vectors of Zhang's three urine samples after CNN model analysis. It can be seen that the feature vectors of each urine sample show certain differences, reflecting the pathological changes of the urinary system in different parts. For example, in the feature vector of first-pass urine, the feature value related to urethritis is higher; the feature vector of mid-pass urine reflects more the lesions of the bladder and above urinary system; the feature vector of last-pass urine highlights the features related to the prostate. These differential features provide valuable basis for subsequent disease evaluation.

[0135] Table 2 Feature vectors of Zhang's three urine samples

[0136] feature First-pass urine Midstream urine Terminal urine Symptoms associated with urethritis 0.85 0.62 0.75 Symptoms associated with cystitis 0.70 0.88 0.65 Prostatitis-related symptoms 0.60 0.55 0.90 Symptoms associated with pyelonephritis 0.45 0.75 0.50 Characteristics of urinary stones 0.35 0.40 0.55

[0137] Finally, Table 3 gives the disease assessment results based on the above feature vectors. It can be seen that the probability of Zhang suffering from cystitis is as high as 90%, which is a high-reliability result; at the same time, he also has a certain probability (75%) of urinary tract infection. This is completely consistent with the doctor's subsequent clinical diagnosis.

[0138] Table 3 Zhang's disease assessment results

[0139]

[0140]

[0141] In summary, through the urinary system disease assessment method based on urine testing provided by the present invention, the doctor can quickly and accurately diagnose Zhang's possible cystitis and urinary tract infection and other problems, providing strong support for subsequent treatment measures.

Claims

1. A method for evaluating urinary system diseases based on urine testing, characterized in that: The following steps are involved: S10, adding corresponding reagents to the three urine samples in sequence according to a preset time point sequence, and acquiring a urine sample image after each addition of the reagent; S20, performing image preprocessing on each acquired urine sample image to extract image features including color change, sediment morphology and distribution; S30, establishing a temporal feature data set, arranging and combining the image features corresponding to each time point in chronological order; S40, using a pre-trained image recognition model to analyze the image features in the time series feature data set to obtain a feature vector; S50, inputting the feature vector into a preset disease assessment model, where the disease assessment model is obtained by training a large amount of clinical case data; S60, obtaining the disease type and its corresponding probability value output by the disease assessment model; S70, determining the credibility of the evaluation result according to the probability value, and screening the evaluation results with high credibility according to a preset threshold; S80, comparing the highly reliable evaluation results with clinical diagnostic criteria to generate a urinary system disease evaluation report; S90, adding the urinary system disease assessment report and its corresponding time series feature data to the sample library for subsequent model optimization.

2. A method for evaluating urinary system diseases based on urine testing according to claim 1, characterized in that: The three types of urine samples are first-pass urine, mid-pass urine, and last-pass urine: The first-pass urine refers to the first 10 ml of urine at the beginning of urination, which mainly reflects the condition of the urethra. The midstream urine refers to the middle part of urine after the first-stream urine is removed, which is 100 to 200 ml. This urine can best reflect the condition of the bladder and the urinary system above it. The terminal urine refers to the last 10 ml of urine at the end of urination, which can reflect the condition of the prostate.

3. A method for evaluating urinary system diseases based on urine testing according to claim 2, characterized in that: The preset time point sequence adds corresponding reagents in sequence, specifically, after collecting three urine samples, urine test paper, urine sediment stain, urine protein reagent, urine sugar detection reagent and urobilinogen detection reagent are added to each sample in sequence at five time points of 0 minutes, 5 minutes, 10 minutes, 20 minutes and 30 minutes respectively.

4. A method for evaluating urinary system diseases based on urine testing according to claim 3, characterized in that: The method for extracting image features including color changes, sediment morphology and distribution is specifically to use a high-power microscope to rotate and photograph each urine sample 360 ​​degrees to obtain characteristic parameters such as sample color, transparency, sediment size, shape, distribution density, etc.

5. A method for evaluating urinary system diseases based on urine testing according to claim 4, characterized in that: The method for pre-training the image recognition model is specifically to collect more than 10,000 urine sample images of patients with clinically confirmed urinary system diseases, annotate each image, establish an image database, and use a deep learning algorithm to train the model so that the model can accurately identify the urine characteristics corresponding to different diseases.

6. A method for evaluating urinary system diseases based on urine testing according to claim 5, characterized in that: The step of constructing the training data set of the disease assessment model is specifically to collect complete medical records of patients with clinically confirmed urinary system diseases, including symptom descriptions, examination results, diagnostic conclusions, etc., and associate them with corresponding urine test results to form a standardized training data set.

7. A method for evaluating urinary system diseases based on urine testing according to claim 6, characterized in that: The three urine samples are collected at one time using a segmented urine collection device.

8. A method for evaluating urinary system diseases based on urine testing according to claim 7, characterized in that: The urine segmented collection device comprises: a main frame, a shunt assembly, a collection assembly and a sealing assembly; the main frame comprises a top plate, a bottom plate and four vertical support columns, and the four vertical support columns are respectively fixedly connected to the four corners of the top plate and the bottom plate; a liquid inlet hole is opened in the center of the top plate, and a funnel-shaped urine collecting basin is arranged above the liquid inlet hole; the shunt assembly is fixedly installed in the main frame and is located directly below the liquid inlet hole; the collection assembly is arranged below the shunt assembly; the sealing assembly is arranged on the bottom plate and connected to the collection assembly; the shunt assembly comprises: A rotating support frame, a rotating shaft, a diverter plate and a plurality of guide grooves; the rotating support frame is fixed on four vertical support columns by bolts; the rotating shaft is vertically installed at the center of the rotating support frame by bearings; the diverter plate is fixedly installed on the rotating shaft; three diverter grooves are evenly arranged radially on the upper surface of the diverter plate, the inner end of each diverter groove is connected to the center of the diverter plate, and the outer end extends to the edge of the diverter plate; the collecting assembly includes: three collecting bottles, a bottle seat and a driving motor; the bottle seat is fixedly installed on the bottom plate in a circular ring shape, and three mounting holes are evenly arranged on the bottle seat along the circumferential direction; the three collecting bottles are respectively fixed on the bottle seat through the mounting holes, and the positions of the collecting bottles correspond to the liquid outlet positions of the diverter grooves one by one; the driving motor is fixedly installed on the rotating support frame, and the output shaft of the driving motor is coaxially connected to the rotating shaft; three groups of infrared sensors are arranged on the rotating support frame for detecting the rotation position of the diverter plate; each group of infrared sensors corresponds to the position of a diverter groove, and a signal is triggered when the diverter groove rotates to the corresponding collecting bottle position.

9. A method for evaluating urinary system diseases based on urine testing according to claim 8, characterized in that: The sealing assembly includes: three sealing joints and three hoses; the sealing joints are respectively installed at the bottle mouths of three collecting bottles; one end of the three hoses is respectively connected to the three sealing joints, and the other end is respectively sealed and connected to the liquid outlet end of the guide groove.

10. A method for evaluating urinary system diseases based on urine testing according to claim 9, characterized in that: The inner surface of the funnel-shaped urine collecting basin is coated with a hydrophobic coating, and the diameter of the liquid outlet at the bottom of the urine collecting basin is 15 mm; the diameter of the diversion plate is 300 mm and the thickness is 20 mm; the cross-sections of the three guide grooves are all semicircular, with a groove depth of 5 mm and a groove width of 8 mm; the volume of the three collecting bottles is 50 ml, and the bottle body is made of transparent material and marked with scale lines; the outer diameter of the bottle base is 400 mm and the inner diameter is 200 mm.

Images