Visual system for rapid assessment of high altitude red blood cell storage damage and applications thereof

Abstract

The present application relates to the technical field of red blood cell quality evaluation, in particular to a visual system for rapidly evaluating highland red blood cell storage damage and application thereof. The present application divides hypoxic stored red blood cells into normal group and damage group, analyzes different cell morphologies thereof, adopts Lasso regression for variable screening, optimizes risk factor parameter dimension, and determines red blood cell morphological damage predictors (SDC, CDC and CSD). Further, a binary classification logistic regression prediction model is constructed, a new red blood cell storage damage early warning model based on morphological characteristics is established, the identification success rate of "high-risk" red blood cells with damage is controlled to be more than 75%, thereby providing a new tool for comprehensive evaluation of highland stored red blood cell damage, and having important significance for rapid early warning and evaluation of highland stored red blood cell quality change.

Description

A visualization system for rapidly assessing erythrocyte storage damage at high altitudes and its applications Technical Field

[0001] This invention relates to the field of red blood cell quality assessment technology, and in particular to a visualization system for rapidly assessing red blood cell storage damage at high altitudes and its application. Background Technology

[0002] Currently, the most in-demand blood product is erythrocyte units, with over 100 million units collected globally annually. During storage, erythrocytes undergo varying degrees of change in morphology, structure, function, and metabolism. Besides the depletion of adenosine triphosphate (ATP), 2,3-diphosphoglycerate (2,3-DPG), and nitric oxide (NO), the morphology of erythrocytes gradually changes from a biconcave disc shape to an acanthosis, eventually becoming an irreversibly smooth spherical erythrocyte. During this process, the diameter of erythrocytes decreases from 7.80±0.49 μm to 6.02±0.27 μm. This phenomenon is called storage lesion. Storage lesion has been shown to have a significant impact on recipient prognosis, with symptoms including hyperkalemia and poor vasodilation.

[0003] Red blood cells stored at high altitudes have a unique storage environment that differs from that at low altitudes, such as hypoxia and low pressure, which all have a series of effects on the quality of red blood cells stored at high altitudes. According to D'Alessandro et al., a hypoxic environment can reduce oxidative damage and alleviate the damage to hypoxic red blood cells. At the same time, it can promote glycolytic metabolism in hypoxic red blood cells by upregulating the secretion level of phosphosphingosine (S1P) and increasing the product 2,3-DPG in the RL shunt, thereby promoting oxygen unloading from red blood cells and alleviating tissue hypoxia [Hypoxic storage of red blood cells improves metabolism and post-transfusion recovery. Transfusion, 2020 April ; 60(4): 786–798. doi:10.1111 / trf.15730]. At the same time, low pressure can also cause oxygen depletion to varying degrees, forming hypobaric hypoxia. He et al.'s research found that under the low-pressure, low-oxygen storage environment of high altitudes, the phosphorylation level of band 3 protein at position Y21 of erythrocytes is significantly increased, releasing glycolytic enzymes from the band 3 cytoplasmic domain and activating glycolysis [Tyrosine phosphorylation of band 3 impairs the storage quality of suspended red blood cells in the Tibetan high-altitude polycythemia population. Doi:10.1186 / s12967-023-04428-5]. Although most current studies support that hypoxia can improve the quality of stored red blood cells to some extent, some reports indicate that hypoxic, especially anaerobic, storage conditions may increase the hemolysis rate of stored red blood cells, leading to poor recovery after transfusion. Therefore, a systematic evaluation of the quality of stored red blood cells at high altitudes is essential. Currently, relevant evaluation methods are lacking. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a visualization system for rapidly assessing red blood cell storage damage at high altitudes and its application. The system provided by this invention enables rapid assessment of red blood cell quality, allowing for rapid early warning and evaluation of the quality of red blood cells stored and transported at high altitudes. This improves the rapid detection level of red blood cell quality under the special conditions of high altitudes and enhances the capacity for blood transfusion and treatment at high altitudes.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention provides a system for assessing erythrocyte storage impairment, comprising:

[0007] The morphology detection module is used to detect the morphology of red blood cells in the sample to be tested; the detection indicators include: the number of smooth disc-shaped red blood cells, the number of serrated disc-shaped red blood cells, and the number of serrated red blood cells;

[0008] An evaluation module, connected to the morphological detection module, is used to evaluate the probability of erythrocyte storage damage in the sample to be tested; the evaluation includes: obtaining the probability of erythrocyte storage damage in the sample to be tested using Formula I and Formula II;

[0009] logitP = -0.304 - 0.268 × SDC + 1.365 × CDC + 0.603 × CSD (Equation I)

[0010] p = 1 / (1+e) -logitP Formula II;

[0011] Where SDC represents the number of smooth discoid red blood cells, CDC represents the number of serrated discoid red blood cells, and CSD represents the number of serrated red blood cells; p represents the probability of red blood cell storage damage in the sample to be tested.

[0012] Preferably, the detection includes: detecting red blood cells using a YOLOV5s network with an attention mechanism.

[0013] Preferably, the system further includes a visualization module connected to the evaluation module, used to visualize the probability of red blood cell storage damage in the test sample.

[0014] Preferably, the visualization module includes: using visualization software R and the function package rms to visualize the results of the evaluation module.

[0015] Preferably, the visualization software R is R4.4.2.

[0016] The present invention provides the application of the system described above in assessing red blood cell damage.

[0017] Preferably, the assessment of red blood cell damage includes assessing the probability of red blood cell damage in the sample to be tested.

[0018] Preferably, the erythrocyte damage includes high-altitude erythrocyte damage.

[0019] Preferably, the high-altitude erythrocyte damage includes high-altitude erythrocyte damage caused by one or more of the following reasons: storage, shaking, turbulence, and irradiation.

[0020] Beneficial effects:

[0021] The system provided by this invention enables rapid assessment of red blood cell quality. This invention collects red blood cell samples under different storage conditions and for different durations, performs microscopic imaging on each sample, and classifies red blood cell morphology into nine categories. Hypoxic stored red blood cells are divided into normal and damaged groups, and their different cell morphologies are analyzed. Lasso regression is used for variable screening to optimize the dimensions of risk factor parameters and determine morphological damage predictors (SDC, CDC, and CSD). Furthermore, a binary logistic regression prediction model is constructed to establish a novel early warning model for red blood cell storage damage based on morphological characteristics. This model ensures that the success rate of identifying damaged "high-risk" red blood cells is controlled above 75%, thus providing a new tool for comprehensive assessment of red blood cell damage during high-altitude storage and transportation. This is of great significance for rapid early warning and assessment of changes in the quality of red blood cells stored at high altitudes. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the embodiments will be briefly described below.

[0023] Figure 1 shows the variable selection results based on Lasso regression;

[0024] Figure 2 shows the forest plot of the multivariate logistic regression analysis;

[0025] Figure 3 is a line chart based on SDC, CDC, and CSD;

[0026] Figure 4 shows the model evaluation analysis; where A represents the analysis results of the ROC curve; and B represents the visualization results of the HL test and calibration curve. Detailed Implementation

[0027] This invention provides a system for assessing erythrocyte storage impairment, comprising:

[0028] The morphology detection module is used to detect the morphology of red blood cells in the sample to be tested; the detection indicators include: the number of smooth disc-shaped red blood cells, the number of serrated disc-shaped red blood cells, and the number of serrated red blood cells;

[0029] An evaluation module, connected to the morphological detection module, is used to evaluate the probability of erythrocyte storage damage in the sample to be tested; the evaluation includes: obtaining the probability of erythrocyte storage damage in the sample to be tested using Formula I and Formula II;

[0030] logitP = -0.304 - 0.268 × SDC + 1.365 × CDC + 0.603 × CSD (Equation I)

[0031] p = 1 / (1+e) -logitP Formula II;

[0032] Where SDC represents the number of smooth discoid red blood cells, CDC represents the number of serrated discoid red blood cells, and CSD represents the number of serrated red blood cells; p represents the probability of red blood cell storage damage in the sample to be tested.

[0033] As one implementation method, the detection includes: detecting red blood cells using a YOLOV5s network with an attention mechanism, the specific detection method of which can be found in the literature [Zhang C, Wang J, Sun W, et al. Assaying storage lesion of irradiated red blood cells by deep learning with attention mechanism[J]. Optics and lasers in enginering, 2024, 181(000):7.DOI:10.1016 / j.optlaseng.2024.108409.]. Previously, this invention proposed a simple, efficient, and low-cost method for red blood cell morphology recognition and red blood cell quality assessment based on deep learning technology. This method eliminates the need for staining red blood cells, avoiding the influence of dyes on red blood cells. An improved YOLOV5s network with an attention mechanism is used to detect red blood cells; finally, red blood cell quality is quantified by the proportion of morphology index (MI) or smooth disc cell (SDC).

[0034] In one implementation, the system further includes a visualization module connected to the evaluation module, used to visualize the probability of erythrocyte storage damage in the test sample. In one implementation, the visualization module includes: using visualization software R and the function package rms to visualize the results of the evaluation module. In one implementation, the visualization software R is R4.4.2.

[0035] This invention collected red blood cell samples under different storage conditions and for different durations, performed microscopic imaging on each sample, and classified red blood cell morphology into nine categories. Hypoxic stored red blood cells were divided into normal and damaged groups, and their different cell morphologies were analyzed. Lasso regression was used for variable screening to optimize the dimensions of risk factor parameters and determine morphological damage predictors, i.e., morphological predictors affecting the degree of red blood cell damage at high altitudes. Furthermore, a binary logistic regression prediction model was constructed to establish a novel early warning model for red blood cell storage damage based on morphological characteristics. This model achieves a success rate of over 75% in identifying damaged "high-risk" red blood cells, thus providing a new tool for comprehensive assessment of red blood cell damage during high-altitude storage and transportation. This is of great significance for rapid early warning and assessment of changes in the quality of red blood cells stored at high altitudes.

[0036] Based on the above advantages, the present invention provides the application of the system described in the above technical solution in assessing red blood cell damage.

[0037] As one implementation, the assessment of red blood cell damage includes assessing the probability of red blood cell damage in the sample to be tested.

[0038] In one implementation, the erythrocyte damage includes high-altitude erythrocyte damage.

[0039] As one implementation, the high-altitude erythrocyte damage includes high-altitude erythrocyte damage caused by one or more of the following reasons: storage, shaking, turbulence, and irradiation.

[0040] As one implementation method, the presence of (1) and / or (2) in the sample to be tested is considered as red blood cell damage:

[0041] (1) Hemolysis rate ≥ 0.8%;

[0042] (2) Two or more of the following are outside the reference range: phosphatidylserine, adenosine triphosphate, 2,3-diphosphoglyceric acid and pH. The reference range for phosphatidylserine is 0 – 0.030, the reference range for adenosine triphosphate is 4.73 – 28.93 nmol / mg, the reference range for 2,3-diphosphoglyceric acid is 183.89 – 870.96 nmol / L, and the reference range for pH is 6.31 – 7.21.

[0043] To further illustrate the present invention, the following detailed description, in conjunction with embodiments and accompanying drawings, provides a visualization system for rapidly assessing high-altitude erythrocyte storage damage and its applications, but these descriptions should not be construed as limiting the scope of protection of the present invention.

[0044] Example 1

[0045] 1. A rapid method for detecting red blood cell quality based on morphological recognition

[0046] During blood transfusion, storage lesions of red blood cells have a significant impact on patient prognosis. Red blood cell morphology and function are closely related, and morphological changes during storage can reflect red blood cell quality to some extent. Traditional manual classification methods are heavily influenced by subjective factors. This invention initially developed a rapid red blood cell morphology recognition technology based on YOLOv5s image recognition technology, as described in the reference [Zhang C, Wang J, Sun W, et al. Assaying storage lesion of irradiated red blood cells by deep learning with attention mechanism[J]. Optics and lasers in enginering, 2024, 181(000):7.DOI:10.1016 / j.optlaseng.2024.108409.]. This technology includes: using microscopic imaging to establish a nine-category dataset of red blood cells, as detailed below:

[0047] STC, target / oral type cells; SDC, smooth discoid; CDC, serrated discoid; CDD, serrated discoid; CSD, serrated; CSE, serrated spheroid; SSE, smooth spheroid; SV, lateral view cells; AC, abnormal cells.

[0048] By differentiating the weights of different erythrocyte morphologies (SV and AC are not weighted), and weighting based on the proportion of the nine-category morphology classification, the percentages of morphology index (MI) and smooth discoid cells (SDCs) are obtained to reflect the degree of erythrocyte storage damage.

[0049] 2. Rapid prediction of high-altitude erythrocyte injury based on the Nomogram prediction model

[0050] Morphology-based rapid red blood cell quality testing technologies utilize portable instruments, require small sample volumes, and have low operator skill requirements, making them a promising means for rapid quality assessment of stored red blood cells in high-altitude regions. However, the intuitiveness with which this technology reflects red blood cell damage still needs further improvement.

[0051] This invention collected high-altitude red blood cell samples under different storage conditions and for different storage times, and detected indicators reflecting changes in red blood cell quality, including hemolysis rate, pH, ATP, 2,3-DPG, and phosphatidylserine (PS). A comprehensive statistical algorithm was used to determine the normal reference ranges for these indicators affecting the quality of high-altitude red blood cells. Details are as follows:

[0052] Hemolysis rate: 0-0.8%; pH: 6.31-7.21; ATP: 4.73-28.93 nmol / mg; 2,3-DPG: 183.89-870.96 nmol / L; PS: 0-0.030. Red blood cell damage is determined to have occurred if the test results of high-altitude red blood cell samples meet (1) and / or (2).

[0053] (1) Hemolysis rate ≥ 0.8%;

[0054] (2) Two or more of the four items PS, ATP, 2,3-DPG and pH exceed the reference range (below the minimum or above the maximum).

[0055] Based on this, the present invention establishes a nomogram prediction model to quickly determine the probability of damage occurrence by observing the morphology of red blood cells in high-altitude areas.

[0056] First, morphological differences between different groups were analyzed. Based on the definition of high-altitude erythrocyte damage, the stored erythrocytes in the screening dataset were divided into a normal group and a damaged group. Different cell morphologies were analyzed, and the results are shown in Table 1. The results showed that the damaged group had significant differences from the normal values ​​in AC, STC, SDC, CDC, CSD, and SSE, and these differences were statistically significant (P<0.05). However, there were no statistically significant differences between the two groups in SV, CDD, and CSE (P>0.05). To further screen variables, all variables will be included to comprehensively select the most suitable variables.

[0057] Table 1. Morphological differences M (Q1, Q3) between the damaged group and the normal group.

[0058]

[0059] Note: The values ​​outside the parentheses in the table are the medians, and the values ​​inside the parentheses are the interquartile ranges, i.e., the values ​​at the 25th and 75th percentiles; the data used for subsequent modeling is the original data corresponding to each sample, where the original data for each type of red blood cell is the number of cells of the corresponding class in an image.

[0060] To further screen high-quality risk factors and optimize risk factor parameters, Lasso regression was used for variable screening before Logistic regression analysis. The results are shown in Figure 1. Figure A shows the coefficient variation characteristics of the nine variables; Figure B illustrates the process of selecting the optimal value of parameter λ in the Lasso regression model using cross-validation. Two vertical dashed lines were drawn at the optimal λ value using 10-fold cross-validation, the minimum standard, and the 1-Se standard. Lasso regression uses the adjustment parameter lambda to regulate the correlation between the screened variables. The larger the lambda value, the stronger the degree of variable screening, and consequently, the coefficients of non-characteristic variables are reduced to zero.

[0061] In this study, nine red blood cell morphology categories were used as variables in the dataset: SV, AC, STC, SDC, CDC, CDD, CSD, CSE, and SSE. The glmnet package in R (version 4.4.2) was used, and Lasso variable selection was employed to analyze the dataset (N=64) to identify predictive factors related to red blood cell morphology. Ultimately, three variables were selected: SDC, CDC, and CSD.

[0062] Based on the selected variables, univariate and multivariate logistic regression analyses were performed. With the occurrence of erythrocyte storage damage as the dependent variable and SDC, CDC, and CSD as independent variables, a binary logistic regression prediction model was constructed, as detailed below:

[0063] logitP= -0.304-0.268×SDC+1.365×CDC+0.603×CSD;

[0064] Where SDC, CDC, and CSD correspond to the specific values ​​of the tested sample; logitP represents ln(p / (1-p), which can be used to deduce the probability value (p) of damage to a sample based on the specific SDC, CDC, and CSD data, i.e., p=1 / (1+e -logitP ).

[0065] The results are shown in Table 2 and Figure 2.

[0066] Table 2. Univariate and multivariate logistic regression analyses of factors associated with erythrocyte storage impairment in the dataset.

[0067]

[0068] The results showed that the three independent variables SDC, CDC, and CSD were all statistically significant in the model (P<0.05) and were independent risk factors for erythrocyte storage damage. There was no multicollinearity among the factors.

[0069] Based on the Lasso-logistic regression of the dataset, and using three erythrocyte morphological indicators, the visualization software R (4.4.2) with the function package rms was used to visualize the results, resulting in a nomogram-based visualization prediction model for erythrocyte storage damage.

[0070] As shown in Figure 3, the specific values ​​of SDC, CDC, and CSD can be used to find the corresponding scores in the model and calculate the total score. The corresponding predicted value is the predicted probability of storage damage to the blood sample.

[0071] The model's discriminative power was evaluated using receiver operating characteristic (ROC) curves. Model discriminative power is the model's ability to correctly distinguish whether storage damage has occurred in a sample. The calculated C-index of the predictive model was 0.973 (95% CI 0.941-1.000).

[0072] An additional 500 samples were selected as the validation set, and the C-index of the model was recalculated to 0.989. The highest Youden index was selected as the cutoff point of the receiver operating characteristic (ROC) curve, with a specificity of 93.75%, sensitivity of 90.62%, positive predictive value of 93.55%, and negative predictive value of 90.91%, indicating that the model has good discrimination (Figure 4A). The model's calibration degree is the degree of consistency between the model's predictions and the actual observed values. It is mainly analyzed using the Hosmer-Lemeshow (HL) test and Bootstrap repeated sampling. The HL test result showed a χ² of 2.515 and a p-value of 0.961, indicating that the model's expected probability fits the actual probability well. The fitting results are shown in the calibration plot, showing good consistency between the predicted and observed values ​​(Figure 4B).

[0073] In summary, the red blood cell morphology recognition method provided by this invention can predict the probability of red blood cell injury at high altitudes with good accuracy. This method is expected to provide an effective means for the rapid detection of red blood cell injury at high altitudes during both peacetime and wartime conditions.

[0074] Although the above embodiments have provided a detailed description of the present invention, they are only some embodiments of the present invention, and not all embodiments. People can obtain other embodiments based on these embodiments without creative effort, and these embodiments all fall within the protection scope of the present invention.

Claims

1. A system for assessing erythrocyte storage damage, characterized in that, include: The morphology detection module is used to detect the morphology of red blood cells in the sample to be tested; The detection indicators include: the number of smooth discoid red blood cells, the number of serrated discoid red blood cells, and the number of serrated red blood cells; an evaluation module, connected to the morphology detection module, is used to evaluate the probability of red blood cell storage damage in the test sample; the evaluation includes: obtaining the probability of red blood cell storage damage in the test sample using Formula I and Formula II; logitP = -0.304 - 0.268 × SDC + 1.365 × CDC + 0.603 × CSD Formula I; p = 1 / (1 + e -logitP Formula II; where SDC is the number of smooth discoid red blood cells, CDC is the number of serrated discoid red blood cells, and CSD is the number of serrated red blood cells; p is the probability of red blood cell storage damage in the sample to be tested.

2. The system according to claim 1, characterized in that, The detection includes: detecting red blood cells using a YOLOv5s network with an attention mechanism.

3. The system according to claim 1, characterized in that, The system also includes a visualization module connected to the evaluation module, used to visualize the probability of red blood cell storage damage in the test sample.

4. The system according to claim 3, characterized in that, The visualization module includes: using visualization software R and the function package rms to visualize the results of the evaluation module.

5. The system according to claim 4, characterized in that, The visualization software R is R4.4.

2.

6. The use of the system according to any one of claims 1 to 5 in assessing erythrocyte damage.

7. The application according to claim 6, characterized in that, The assessment of red blood cell damage includes assessing the probability of red blood cell damage in the sample to be tested.

8. The application according to claim 6 or 7, characterized in that, The erythrocyte damage includes high-altitude erythrocyte damage.

9. The application according to claim 8, characterized in that, The high-altitude red blood cell damage includes high-altitude red blood cell damage caused by one or more of the following reasons: storage, shaking, turbulence, and irradiation.

Images