Serum quality detection method by comparing image parameterization with database
By using image parameterization and database comparison, the high error rate of serum quality testing and the barcode occlusion problem in automated testing have been solved, achieving fully automated, low-cost, and efficient serum quality testing with high recognition accuracy and success rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE FIRST AFFILIATED HOSPITAL OF ZHENGZHOU UNIV
- Filing Date
- 2025-12-11
- Publication Date
- 2026-05-12
AI Technical Summary
Existing serum quality testing methods suffer from high error rates, high costs, low efficiency, strong subjectivity, lack of unified standards, and the obstruction of serum tube barcodes affects automated testing.
By employing image parameterization and database comparison, R, G, and B data of pixels are extracted from serum images to establish a comprehensive and sub-database, forming a color cube. The K-NN algorithm is then used to compare serum windows, and the best image is used for quality analysis.
It achieves fully automated, low-cost, and efficient serum quality testing. The sensitivity and specificity for identifying hemolysis, lipemia, and jaundice are 91.5%, 100.0%, and 61.2%, respectively, and the specificity is 99.5% and 60.7%, respectively. The AUC for identifying abnormal serum is 0.810, and the success rate is 94% to 98%.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention belongs to the fields of clinical medicine, computer science and information technology, and specifically relates to a method for serum quality detection through image parameterization and database comparison. Background Technology
[0002] Currently, the error rate in pre-serum analysis is very high, accounting for 60%-70% of errors in laboratory diagnostics. This can lead to inappropriate testing and unnecessary cost increases, and in some cases, even inappropriate treatment or treatment modifications. Hemolysis, jaundice, and lipemia samples account for the main proportion of errors in pre-serum analysis. Existing serum index review (SIS) methods, both domestically and internationally, rely on manual judgment or reagent testing, resulting in low efficiency, high cost, and strong subjectivity. There is an urgent need for standardization in serum testing; the lack of unified serum quality assessment standards leads to poor reproducibility of test results. Furthermore, barcodes on serum tubes can obscure serum, significantly interfering with the selection of serum images and subsequent analysis in fully automated laboratory (TLA) systems. Therefore, a fully automated, efficient, and low-cost method for serum quality review is urgently needed.
[0003] Jaundice results in elevated bilirubin levels in serum, lipemia results in abnormally high concentrations of triglycerides and other lipids in serum, and hemolysis results in the release of large amounts of hemoglobin from the rupture of red blood cells. Therefore, different degrees of jaundice, lipemia, and hemolysis can have multi-level and multi-layered effects on serum color, and the RGB attributes (R for Red, G for Green, and B for Blue) of pixels on serum images will be affected.
[0004] Currently, serum quality testing mainly relies on biochemical analyzers and spectrophotometry, in addition to direct visual observation. Biochemical analysis is expensive due to its reliance on equipment and reagents, while spectrophotometry has low batch processing efficiency and requires manual calibration. While these methods acknowledge the importance of serum color in serum quality testing, they haven't further parameterized serum color or integrated it with AI to achieve fully automated, efficient, and low-cost serum quality testing.
[0005] Patent CN 108562584 A utilizes RGB data of serum pixels, but this patent only uses the average value of R, G, and B data of serum pixels as a measure of the degree of jaundice, hemolysis, and lipemia, comparing it with their set values to determine whether the serum is abnormal, which has certain limitations. Therefore, developing a method that combines serum image parameterization with AI and database comparison will provide a new perspective for serum quality detection and reduce the error rate of pre-analysis, and will be of great significance. Summary of the Invention
[0006] In view of the problems and shortcomings of the existing technology, the present invention aims to provide a serum quality detection method by image parameterization and database comparison.
[0007] To achieve the objectives of this invention, the technical solution adopted is as follows: The first aspect of this invention provides a method for serum quality detection through image parameterization and database comparison, comprising the following steps: S1: Randomly select effective serum images containing normal serum, hemolyzed serum, lipemic serum, and jaundice serum, extract serum pixels, and store the R, G, and B data of serum pixels. Establish a comprehensive serum pixel database and four sub-databases containing pixel data of normal serum, hemolyzed serum, lipemic serum, and jaundice serum. Extract the R, G, and B data of pixels from the four sub-databases, calculate the target value and standard deviation of the R, G, and B data, and obtain the intervals of the R, G, and B data. Optimize the range of pixels contained in the database into three intervals. The database forms a color cube in three dimensions. S2: Serum Image Analysis: Take multiple images of the serum to be tested, compare the R, G, and B data of each serum image pixel with the comprehensive serum pixel database, record the number of matching serum pixels, select the image with the most matching pixels as the optimal image of the serum window, output it and use it for serum quality analysis. S3: Serum quality analysis: The R, G, and B data of the optimal image pixel in the serum window are compared with the R, G, and B data of the four sub-databases. The serum quality corresponding to the sub-database with the most matching pixels is the quality output result of the serum to be tested.
[0008] According to the above-described serum quality detection method, preferably, in step S3, when the R data, G data, and B data of the pixels on the optimal image of the serum window are all within the range of R data, G data, and B data of a certain database, then it matches the database.
[0009] According to the above-described serum quality detection method, preferably, when performing serum image analysis in step S2, if the number of serum pixels in the optimal image of the serum window output by the serum image auxiliary analysis module is greater than 200, it is considered a valid marker.
[0010] According to the above-mentioned serum quality detection method, preferably, step S2 uses a serum image-assisted analysis module to perform serum image analysis. The development process of the serum image-assisted analysis module is as follows: based on the comparison of the pixels of four given serum images with the serum pixel database of step S1, the K-NN algorithm is used to develop the module. The K-NN algorithm is the K-nearest neighbor algorithm, specifically a "majority voting K-nearest neighbor algorithm".
[0011] According to the above serum quality detection method, preferably, step S3 uses a serum quality auxiliary analysis module to perform serum quality analysis. The development process of the serum quality auxiliary analysis module is as follows: given a new set of serum images, the optimal image of the serum window is selected based on the serum image auxiliary analysis module, the pixels of the selected optimal image of the serum window are compared with the four sub-databases of the serum pixel database, and the K-NN algorithm is used to develop the module.
[0012] According to the serum quality detection method described above, preferably, in step S1, the type of serum to which the serum image belongs is determined by clinical judgment, and it is determined to be normal serum, hemolyzed serum, lipemic serum, or jaundice serum.
[0013] According to the serum quality detection method described above, preferably, the serum images used to establish the serum pixel database in step S1 are all manually separated from the serum window; the serum images are pre-processed images taken by TLA.
[0014] A second aspect of the present invention provides a serum quality detection system, including a TLA imaging system, a serum image auxiliary analysis module, and a serum quality auxiliary analysis module; The TLA imaging system is used to photograph the serum to be tested and obtain images of the serum. The serum image-assisted analysis module takes serum images captured by the TLA imaging system as input, compares the R, G, and B pixel data of the serum images with a comprehensive serum pixel database, and outputs the optimal image for the serum window. The serum quality auxiliary analysis module takes the optimal image of the serum window as input, compares the pixel R, G, and B data of the optimal image of the serum window with four sub-databases, and outputs the serum quality results.
[0015] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the detection method as described in the first aspect above.
[0016] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the detection method described in the first aspect above.
[0017] Compared with the prior art, the present invention has the following advantages: (1) The serum quality detection method of the present invention optimizes the range of pixels contained in the database into three intervals, forming a color cube in three dimensions, which can contain all pixels in the interval, avoiding the problem of collecting all pixels. At the same time, it optimizes the database, greatly reduces the time required for each pixel to be compared with the database, and improves the analysis efficiency.
[0018] (2) The present invention connects the serum image auxiliary analysis module with the photography system of the TLA system, takes the serum image output by the TLA system as the input value, compares the number of pixels of the four serum images, and selects the serum image with excellent serum window, thereby improving the probability of the serum window being correctly extracted from the image in the process before analysis. The success rate of selecting images with excellent serum window is between 94% and 98%.
[0019] (3) The serum quality auxiliary analysis module of the present invention can automatically, efficiently, and without consumption detect serum quality and output detection results, optimizing the existing SIS method. The sensitivity of this method for detecting hemolysis, lipemia, and jaundice is 91.5%, 100.0%, and 61.2%, respectively, and the specificity is 99.5%, 60.7%, and 100.0%, respectively. The AUC of the method of the present invention for identifying lipemia serum is 0.845, the AUC for identifying jaundice serum is 0.612, the AUC for identifying hemolytic serum is 0.860, and the AUC for identifying abnormal serum is 0.810. The method of the present invention has a good ability to distinguish hemolytic serum and lipemia serum, and can distinguish abnormal serum well, but its ability to distinguish jaundice serum is weak. Attached Figure Description
[0020] Figure 1 This is a flowchart of the serum quality detection process of the present invention; Figure 2 This is an image showing the normal distribution of pixels in normal serum. Figure 3 This is an image showing the normal distribution of pixels in jaundice serum. Figure 4 This is an image showing the normal distribution of pixels in hemolyzed serum. Figure 5 This is an image showing the normal distribution of lipid serum pixels. Figure 6 Workflow diagram for serum image-assisted analysis module; Figure 7 Flowchart of the serum quality auxiliary analysis module; Figure 8 The success rate of selecting excellent serum window images for the serum image auxiliary analysis module; Figure 9 The effectiveness of serum quality testing for the serum quality auxiliary analysis module; Figure 10 The ROC curve for identifying lipemic serum using this method when the results of commercial reagent kits are used as the standard; Figure 11 The ROC curve for identifying jaundice serum using this method is based on the results of commercial reagent kits. Figure 12 This method identifies the ROC curve of hemolyzed serum when using commercial reagent kit results as a standard; Figure 13 The ROC curve for identifying abnormal serum using this method when the results of commercial reagent kits are used as the standard; Figure 14 ROC curves for identifying lipemia plotted based on human judgment (left: commercial kit, right: this method); Figure 15 ROC curves for identifying jaundice, plotted based on human judgment (left: commercial kit, right: this method). Figure 16 ROC curves for identifying hemolysis were plotted based on human judgment (left: commercial kit, right: this method). Figure 17 ROC curves for identifying abnormal serum were plotted based on the results of human judgment (left: commercial kit, right: this method). Figure 18 Ten hemolyzed samples with different concentration gradients were used. Figure 19 Seven jaundice samples with different concentration gradients; Figure 20 Seven lipid samples with different concentration gradients were collected. Figure 21 The regression curve for predicting ALT concentration using the SIS index is shown. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention is further described in detail below through embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0022] Example 1: A method for serum quality detection by image parameterization and database comparison, the process is as follows: Figure 1 As shown, the specific steps are as follows: S1: Pre-processed images taken by the laboratory TLA machine. 100 serum images were randomly sampled, and serum windows were manually separated. Serum pixels were extracted and stored as R, G, and B data, resulting in a total of 8085 valid pixels. Subsequently, it was found that the collection of abnormal serum images (jaundice, hemolysis, lipemia) was insufficient, so 20 more abnormal serum quality images were collected, extracting 5035 abnormal serum pixels. After deduplication, 2038 pixels were obtained, which were then added to the original 8085 serum pixels, resulting in 9736 pixels after deduplication. Additionally, 41 normal serum images and 32 abnormal serum images were randomly sampled from the pre-processed images. Pixel collection was performed, resulting in 2370 normal pixels, 1287 jaundice pixels, 1448 hemolysis pixels, and 1636 lipemia pixels. The 6741 collected pixels were summarized into the original 9736 serum pixels, and after deduplication, a total of 15289 pixels were obtained. The final results showed 15289 serum pixels, 3625 normal serum pixels, 4141 hemolyzed pixels, 5517 lipemic pixels, and 2006 jaundice pixels. Based on clinical judgment, the type of serum image used was determined as normal or abnormal, and whether it belonged to the normal, hemolyzed, jaundice, or lipemic serum image category. Four sub-databases were established: a comprehensive serum pixel database and sub-databases for normal serum, hemolyzed serum, lipemic serum, and jaundice serum. The R, G, and B data of the pixels in each of the four sub-databases were extracted, and the target values and standard deviations of the R, G, and B data were calculated. The range of pixels contained in the database was optimized into three intervals, and the database formed a three-dimensional color cube. No outliers were found in the serum image data in the database. SPSS 17 was used to perform a normality test (Kolmogorov-Smirnov) on the serum pixel reference value data. P>0.05 indicated that the data were normally distributed. Data that follow a normal distribution are described as mean ± standard deviation (x±s). The 2.5% and 97.5% distribution sites of each group were calculated after removing outliers. The reference intervals for each group were established using the percentage method (P2.5-P97.5). S2: Use the TLA system to take four images of the serum to be tested. Compare the R, G, and B data of the pixels in each serum image with the comprehensive serum pixel database, record the number of matching serum pixels, and select the image with the most matching pixels as the optimal image for the serum window: First, iterate through the serum images and select the serum image with the most pixels as the output; otherwise, select the last serum image as the output. Furthermore, if the number of serum pixels in the output serum image is greater than 200, it is considered a valid label, output, and used for serum quality analysis; otherwise, the labeling fails. The development process of the serum image-assisted analysis module in step S2 is as follows: Based on the comparison of the pixels of four given serum images with the serum pixel database from step S1, the K-NN algorithm is used to develop the module. The K-NN algorithm is also known as the K-Nearest Neighbors algorithm, specifically a "majority-voting K-Nearest Neighbors algorithm".
[0023] S3: Compare the R, G, and B data of the pixels in the optimal image of the serum window with the R, G, and B data of the four sub-databases: First, if the RGB data of the serum pixels in the optimal image of the serum window are within the range of the sub-database, proceed to the next matching step; otherwise, mark it as failed. Furthermore, if the R, G, and B data of the pixels in the optimal image of the serum window are all within the range of the R, G, and B data of a certain database, then match with that database. The serum quality corresponding to the sub-database with the most matching pixels of that image is the quality output result of the serum to be tested. We performed normality tests on the R, G, and B data of pixels from normal serum (denoted as norm), jaundice serum, hemolyzed serum, and lipemia serum, respectively, and found that they conformed to a normal distribution. Figures 2-5 ), take 95% of the interval as the interval for R data, G data, and B data.
[0024] The development process of the serum quality auxiliary analysis module in step S3 is as follows: Step S3 uses the serum quality auxiliary analysis module to perform serum quality analysis. The development process of the serum quality auxiliary analysis module is as follows: Given a new set of serum images, the optimal image of the serum window is selected based on the serum image auxiliary analysis module. The pixels of the selected optimal image of the serum window are compared with the four sub-databases of the serum pixel database. The serum quality corresponding to the sub-database with the most matching numbers is output as the output result. The K-NN algorithm is used to develop the module.
[0025] Example 2: A serum quality detection system, comprising a TLA imaging system, a serum image auxiliary analysis module, and a serum quality auxiliary analysis module; The TLA imaging system is used to photograph the serum to be tested and obtain images of the serum. The serum image-assisted analysis module takes the pre-processed raw serum images captured by the TLA imaging system as input, compares the R, G, and B pixel data of the serum images with a comprehensive serum pixel database, and outputs the image with the highest number of matches as the optimal image for the serum window to the LIS system. The workflow of the serum image-assisted analysis module is as follows: Figure 6 As shown; The serum quality auxiliary analysis module takes the optimal image of the serum window to be analyzed from the serum image auxiliary analysis module as input, identifies the serum window, and compares the pixel R, G, and B data of the optimal image with four sub-databases. The database with the highest number of matches is output as the serum quality result. The workflow of the serum quality auxiliary analysis module is as follows: Figure 7 As shown.
[0026] Example 3: An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the detection method as described in Embodiment 1.
[0027] Example 4: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the detection method as described in Example 1.
[0028] The serum quality detection method described in Embodiment 1, which involves image parameterization and database comparison, can be implemented as a computer software program. For example, Embodiment 1 of this disclosure includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the method shown in the serum quality detection procedure.
[0029] It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0030] Example 5: Based on the serum quality detection system of Example 2, performance tests for serum quality detection were conducted: (1) Success rate of selecting excellent serum windows by the serum image-assisted analysis module 1) Sample source Serum samples collected from patients who underwent examinations at the First Affiliated Hospital of Zhengzhou University between January 19 and January 29, 2022, are photographed using a Roche cobas 8000 C701 fully automated biochemical analyzer.
[0031] 2) Experimental methods Method before use: The serum image selection method built into the Roche cobas 8000 C701 was used as a control. It identifies the barcode on the serum sample tube and rotates it at a fixed angle to find a good image of the serum window that is not obscured by the barcode. This method requires a very high degree of standard in the barcode affixing by the blood collector, is greatly affected by human factors, and has a low success rate in selecting images with good serum windows.
[0032] Post-use method: Connect the serum image auxiliary analysis module of this invention to the imaging system of the TLA system, take the serum image output by the TLA system as the input value, the serum image auxiliary analysis module extracts pixels from the images taken from the four directions of the sample tube, and finds the serum image with the most serum pixels by comparing with the serum pixel database. Select the image with excellent serum window. Compare the number of pixels in the four serum images and select the serum image with excellent serum window.
[0033] 3) Experimental Results The success rate of selecting excellent serum window images before and after use, such as Figure 8 As shown, the success rate before use was 23%–38%. Using the serum image-assisted analysis module of this invention, the success rate of selecting images with excellent serum windows is between 93% and 98%. This invention effectively solves the problem of barcodes obscuring serum and hindering subsequent serum quality testing.
[0034] (2) Sensitivity and specificity of serum quality auxiliary analysis module 1) Sample source Serum samples collected from patients undergoing examinations at the First Affiliated Hospital of Zhengzhou University, randomly selected between June 6 and June 16, 2022, were photographed by a pre-processing machine of the same model (Roche C701) in the laboratory. The serum images were then selected by the serum image analysis module in this patent, resulting in 1,101 specimen images.
[0035] 2) Experimental methods Doctors manually judged the accuracy of images based on clinical review, eliminating erroneous images. The remaining 1000 serum images were then judged by both humans and a serum quality analysis module to determine whether they indicated hemolysis, jaundice, lipemia, or normality. The results of the human judgments were used as the true values, and the results from the serum quality analysis module were used as the predicted values. A confusion matrix was constructed and listed, yielding the following results. Figure 9 Experimental data.
[0036] When there are enough serum pixels in the serum window (>200), the serum quality auxiliary analysis module can automatically, efficiently, and without resource consumption detect serum quality based on the serum image output by the serum image auxiliary analysis module, output the detection results, and optimize the existing SIS method.
[0037] 3) Experimental Results Based on physicians' manual judgment through clinical review, the detection sensitivity and specificity of this method are as follows: Figure 9As shown, the sensitivity of this method for detecting hemolysis, lipemia, jaundice, and normal serum was 91.5%, 61.2%, 60.7%, and 90.9%, respectively, and the specificity was 100%, 99.5%, 100.0%, and 98.4%, respectively. We found that the results for lipemia and jaundice produced by this method had a high probability of being false positives, but the sensitivity and specificity for the results for normal serum and hemolysis were very high. In other words, when the output result indicates that the serum is normal serum, or when the output result indicates that the serum quality is hemolyzed, it has sufficiently high reliability, and this method can effectively distinguish between normal and abnormal serum.
[0038] This invention effectively enhances the ability of SIS (Self-Assessment System) to identify unsuitable samples, significantly optimizing SIS. It can also assist, and even replace, clinical laboratory technicians in making correct decisions regarding the normality of serum samples when the output result indicates hemolysis. However, we remain concerned about the possibility of a serum sample being in a state of lipemia or jaundice, which could be misclassified as normal serum, leading to pre-analysis errors. Laboratory staff can reject samples they deem lipemia / jaundice but which the method classifies as normal serum based on their own judgment. Therefore, this method can achieve fully automated, zero-cost, and highly efficient serum quality testing, but it still relies to some extent on manual judgment when the result is lipemia / jaundice.
[0039] (3) ROC curve of serum quality test 1) Sample source The serum samples collected from patients undergoing examinations at the First Affiliated Hospital of Zhengzhou University were photographed using the same model (Roche C701) pretreatment machine in the laboratory. After manually removing erroneous samples, 569 valid serum images were obtained.
[0040] 2) Experimental methods To further evaluate the efficacy of this method, experiments were conducted using the results of serum quality testing from commercial kits as the standard and the serum quality testing results from the same tube of serum produced by this method as a prediction.
[0041] 3) Experimental Results ROC curves were plotted based on the samples. Figures 10-13 The AUC values obtained by the method of this invention for identifying lipemic serum were 0.845, for jaundice serum were 0.612, for hemolyzed serum were 0.860, and for abnormal serum were 0.810. This indicates that, compared with the test results of commercial kits, this method has a good ability to distinguish hemolyzed serum, a good ability to distinguish lipemic serum, and can distinguish abnormal serum relatively well. However, the AUC value for identifying jaundice suggests that the method's ability to distinguish jaundice serum is relatively weak.
[0042] To more comprehensively evaluate the efficacy of this method, we further used the results of physicians' manual judgment based on clinical review to conduct serum quality checks on these 569 samples as a standard, and plotted ROC curves for the detection results of commercial kits and the method of this invention respectively. Figures 14-17 The commercial kits showed an AUC of 0.640 for identifying lipemic serum, 0.655 for hemolyzed serum, 0.868 for jaundice serum, and 0.809 for abnormal serum. The method of this invention showed an AUC of 0.894 for identifying lipemic serum, 0.829 for jaundice serum, 0.885 for hemolyzed serum, and 0.939 for abnormal serum. Therefore, if human judgment is used as the standard, this method can excellently identify various abnormal serums and outperforms commercial kits in all aspects of serum quality testing, especially in the identification of lipemic serum.
[0043] (4) Thresholds for concentration changes of hemolysis, lipemia, and jaundice during serum quality testing More broadly, serum quality testing also includes the detection of the concentration of various clinical test items in serum. The method of this invention can be further extended to the evaluation and correction of serum clinical test items, explore the correlation between common clinical biochemical items and the SIS index obtained by this method, and quantify the concentration change threshold corresponding to the "acceptable image quality range". This facilitates clinical laboratory technicians to make a unified and objective judgment when reviewing specimens, thereby reducing human error in judging whether serum specimens are acceptable. At the same time, it can also correct the test values of clinical test items affected by hemolysis / lipemia / jaundice, so as to provide a reliable value to the testers who need the true value of the test item.
[0044] Through experiments, we identified clinical tests where, all other things being equal, the SIS index increased with the severity of hemolysis / lipemia / jaundice, reaching a threshold. These tests were deemed "unacceptable" due to the influence of hemolysis / lipemia / jaundice. SPSS analysis was used to determine correlations and establish regression curves to predict the measured values (X) of these tests based on their SIS hemolysis / lipemia / jaundice indices. Specifically, when the SIS hemolysis / lipemia / jaundice index exceeded the threshold corresponding to the clinical test, the measured concentration was deemed "unacceptable" due to the influence of hemolysis / lipemia / jaundice; conversely, it was "acceptable." This quantifies the threshold for acceptable image quality ranges corresponding to the concentration changes of the test. On the other hand, this method can correct the test values of clinical test items that are "unacceptable" due to the influence of hemolysis / lipemia / jaundice. The difference between the estimated value of the clinical test item in the sample and the corresponding estimated value when the SIS hemolysis / lipemia / jaundice index is zero is the concentration value that theoretically changes due to the influence of hemolysis / lipemia / jaundice, which is set as the correction value (C). Thus, the test error caused by hemolysis / lipemia / jaundice is transformed into a quantifiable parameter. Subtracting the correction value (C) from the test value (X) gives the corrected concentration of the clinical test item, which is the true value (Xc) of the concentration of the clinical test item given by this method.
[0045] Experimental Design: High concentrations of hemoglobin, triglycerides, and bilirubin solutions were added in gradients to mixed serum to prepare a series of gradient samples. These samples were tested on a Roche P612 pretreatment system to evaluate the consistency of the values (hemolysis index, jaundice index, and lipemia index) provided by the SIS system. The samples were then tested on a Roche cobas 8000 C701 fully automated biochemical analyzer to assess the impact of different gradients of hemolysis, lipemia, and jaundice on the tested parameters. The indices measured by commercial reagent kits were used as the criteria for judging the severity of hemolysis, lipemia, and jaundice. A deviation of ±15% from the control sample concentration for each clinical test parameter was set as the threshold.
[0046] 1) Investigate the correlation between hemolysis index and common clinical biochemical tests. Experimental Method: Take 1.5 mL of EDTA-anticoagulated whole blood sample, centrifuge to retain red blood cells, and then wash twice with physiological saline until the saline is clear. Add 0.5 mL of deionized water to the RBCs and sonicate. Confirm complete hemolysis under a microscope to obtain a high-concentration hemoglobin solution. Add this hemoglobin solution in a gradient to sufficient mixed serum at hemoglobin concentrations of 0 mg / dL, 5 mg / dL, 10 mg / dL, 20 mg / dL, 40 mg / dL, 100 mg / dL, 250 mg / dL, 500 mg / dL, 1000 mg / dL, and 1500 mg / dL, preparing 10 gradient samples to achieve visible mild, moderate, and severe hemolysis. Figure 18 A series of gradient specimens were tested for biochemical items on the Roche cobas 8000 C701 fully automated biochemical analyzer to assess the impact of hemolysis degree on different test items and correlate it with the hemolysis index value presented by SIS images.
[0047] Table 1 shows the experimental data on the effects of hemolysis on common clinical biochemical parameters. It can be observed that in experiments investigating the correlation between the hemolysis index and common clinical biochemical parameters, the SIS hemolysis index increased with increasing hemolysis degree, and the ALT concentration generally showed an upward trend. When the SIS hemolysis index reached approximately 10⁸, the ALT concentration reached the threshold. When the SIS index continued to increase, ALT exceeded the normal range. Similar changes in AST and GGT concentrations were observed, the only differences being that AST concentration generally increased with increasing SIS hemolysis index, while GGT concentration decreased, and the rate of change and threshold values differed.
[0048] Table 1. Experimental data on the effects of hemolysis on common clinical biochemical tests. 2) Investigate the correlation between lipid levels and common clinical biochemical tests. Experimental Method: Weigh 200 mg of pure triglycerides and dissolve them in DMF to prepare a triglyceride solution with a concentration of 200 mg / mL. The high-concentration triglyceride solution was then added in a gradient to the mixed serum (ensuring that the volume of DMF in each serum sample was consistent to eliminate the influence of DMF on the detection). This resulted in a series of serum samples containing different triglyceride concentrations: 0 mg / mL, 2 mg / mL, 3 mg / mL, 6 mg / mL, 10 mg / mL, 20 mg / mL, and 30 mg / mL, labeled as sample 1-sample 7 (sample 1 was the control sample, without added triglycerides but with an equal volume of DMF). The prepared serum samples were made to exhibit visible lipemia. Figure 19A series of gradient samples were tested for biochemical items on the Roche cobas 8000 C701 fully automated biochemical analyzer to evaluate the impact of different triglyceride concentrations on different test items and correlate them with the lipid index values presented in the SIS images.
[0049] Table 2 shows the experimental data on the effects of lipemia on common clinical biochemical items. It can be found that in the experiment exploring the correlation between lipemia index and common clinical biochemical items, as the degree of lipemia increases, the SIS lipemia index also increases, while the concentration of HDLC generally shows a downward trend. When the SIS lipemia index reaches 35, the concentration of HDLC reaches the threshold. The concentration changes of ALT and GLU are similar to those of HDLC, except that the trend and threshold of change are different as the SIS lipemia index increases.
[0050] Table 2. Experimental data on the effects of lipemia on common clinical biochemical tests. 3) Investigate the correlation between jaundice index and common clinical biochemical tests. Experimental Method: Weigh 20 mg of pure bilirubin and dissolve it in DMF to prepare a bilirubin solution with a concentration of 200 mg / dL. Add high concentrations of bilirubin in a gradient to the mixed serum (to ensure a consistent DMF volume in each serum sample), preparing serum samples with bilirubin concentrations of 0 mg / dL, 3.5 mg / dL, 7 mg / dL, 14 mg / dL, 28 mg / dL, 43 mg / dL, and 53 mg / dL, labeled sample 1-sample 7 respectively (sample 1 is the control sample, without added bilirubin). The prepared serum samples were designed to show visibly mild, moderate, and severe jaundice. Figure 20 A series of gradient specimens were tested for biochemical items on a Roche cobas 8000 C701 fully automated biochemical analyzer to assess the impact of different jaundice concentrations on different test items and to correlate them with the jaundice index values presented in the SIS image.
[0051] The experimental data on the impact of jaundice on common clinical biochemical items are shown in Table 3. None of the clinical biochemical items in Table 3 reached the threshold, indicating that their test values can be considered unaffected by the degree of jaundice and do not require correction.
[0052] Table 3 Experimental data on the effects of jaundice on common clinical biochemical tests In our experiments exploring the correlation between hemolysis index and common clinical biochemical tests, we found a strong non-linear correlation between ALT concentration and SIS hemolysis index. As the amount of experimental data increased, we obtained a regression model between ALT concentration and SIS hemolysis index. Figure 21 This allows for the assessment and correction of ALT concentrations detected in clinical biochemical tests using the SIS index from serum images.
[0053] In experiments exploring the correlation between lipid levels and common clinical biochemical tests, the concentrations of HDLC, ALT, and GLU showed a strong non-linear correlation with the SIS index. As the amount of experimental data increases, regression models of the SIS index and more clinical tests will be obtained, enabling the assessment and correction of clinical tests using the SIS index from serum images.
Claims
1. A method for detecting serum quality through image parameterization and database comparison, characterized in that, Includes the following steps: S1: Randomly select effective serum images containing normal serum, hemolyzed serum, lipemic serum, and jaundice serum, extract serum pixels, and store the R, G, and B data of serum pixels. Establish a comprehensive serum pixel database and four sub-databases containing pixel data of normal serum, hemolyzed serum, lipemic serum, and jaundice serum. Extract the R, G, and B data of pixels from the four sub-databases, calculate the target value and standard deviation of the R, G, and B data, and obtain the intervals of the R, G, and B data. Optimize the range of pixels contained in the database into three intervals. The database forms a color cube in three dimensions. S2: Serum Image Analysis: Take multiple images of the serum to be tested, compare the R, G, and B data of each serum image pixel with the comprehensive serum pixel database, record the number of matching serum pixels, select the image with the most matching pixels as the optimal image of the serum window, output it and use it for serum quality analysis. S3: Serum quality analysis: The R, G, and B data of the optimal image pixel in the serum window are compared with the R, G, and B data of the four sub-databases. The serum quality corresponding to the sub-database with the most matching pixels is the quality output result of the serum to be tested.
2. The serum quality detection method according to claim 1, characterized in that, In step S3, if the R, G, and B data of the pixels in the optimal image of the serum window are all within the range of R, G, and B data in a certain database, then it is matched with that database.
3. The serum quality detection method according to claim 1, characterized in that, In step S2, when performing serum image analysis, if the number of serum pixels in the optimal image of the serum window output by the serum image auxiliary analysis module is greater than 200, it is considered a valid marker.
4. The serum quality detection method according to claim 1, characterized in that, Step S2 uses a serum image-assisted analysis module to perform serum image analysis. The development process of the serum image-assisted analysis module is as follows: Based on the comparison of the pixels of four given serum images with the serum pixel database of step S1, the K-NN algorithm is used to develop the module.
5. The serum quality detection method according to claim 4, characterized in that, Step S3 uses a serum quality auxiliary analysis module to perform serum quality analysis. The development process of the serum quality auxiliary analysis module is as follows: given a new set of serum images, the optimal image of the serum window is selected based on the serum image auxiliary analysis module. The pixels of the selected optimal image of the serum window are compared with the four sub-databases of the serum pixel database. The K-NN algorithm is used to develop the module.
6. The serum quality detection method according to claim 1, characterized in that, In step S1, the serum type to which the serum image belongs is determined through clinical judgment.
7. The serum quality detection method according to claim 6, characterized in that, In step S1, the serum images used to establish the serum pixel database are manually separated into serum windows; the serum images are pre-processed images taken by the TLA photography system.
8. A serum quality detection system, characterized in that, Includes a TLA imaging system, a serum image-assisted analysis module, and a serum quality-assisted analysis module; The TLA imaging system is used to photograph the serum to be tested and obtain images of the serum. The serum image-assisted analysis module takes serum images captured by the TLA imaging system as input, compares the R, G, and B pixel data of the serum images with a comprehensive serum pixel database, and outputs the optimal image for the serum window. The serum quality auxiliary analysis module takes the optimal image of the serum window as input, compares the pixel R, G, and B data of the optimal image of the serum window with four sub-databases, and outputs the serum quality results.
9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the processor executes the computer program, it implements the detection method as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the detection method as described in any one of claims 1 to 7.