CN-122017260-A - Blood type auxiliary studying and judging method

CN122017260ACN 122017260 ACN122017260 ACN 122017260ACN-122017260-A

Abstract

The invention relates to a blood type auxiliary research and judgment method, which is used for carrying out sedimentation on a sample after mixing, centrifuging and reacting treatment in blood type detection operation on a weakly aggregated and classified blood sample, obtaining the light transmittance variance of the sample, the reactivity and the reaction acceleration rate of the reaction sample through light transmittance analysis, combining a light transmittance change upper limit calibration threshold value, a light transmittance change acceleration upper limit calibration threshold value and a local turbidity gradient variance upper limit calibration threshold value under the dynamic adjustment of a temperature adjustment factor reflected by real-time temperature, carrying out threshold value comparison and judgment, further determining the optimal sedimentation duration of the sample, ensuring that the sample can fully react under the optimal sedimentation duration while effectively controlling the overlarge sedimentation duration, further accurately controlling the blood detection flow, and assisting the blood type research and judgment.

Inventors

ZHANG LIBO
HAN WENPING
PANG RONGRONG
CHENG YANG
LIU YANG
HUANG MIN
DONG RUIPING
LI YAN
Hu Polu
BAO JINGJING

Assignees

南京红十字血液中心
连云港市红十字中心血站

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. A blood type auxiliary studying and judging method is characterized in that based on a preset number of blood samples belonging to weak agglutination classification, an agglutination sedimentation time period model is obtained according to the following steps A to B, and then blood type detection is realized by executing the following step D aiming at a blood sample to be detected; Step A, based on each time-space combination between each preset temperature interval and each preset sample amount, aiming at the samples subjected to mixing, centrifugation and reaction treatment in blood group detection operation under each time-space combination, performing sedimentation and determining the optimal sedimentation duration, further constructing each sample, forming a sample library, and then entering the step B; Step B, training a target network to be trained based on each sample in a sample library by taking a temperature interval in the sample and the sample quantity of a blood sample as input and taking a preset sedimentation duration interval corresponding to the optimal sedimentation duration in the sample as output to obtain an agglutination sedimentation duration model; And D, according to blood type detection operation, obtaining a target sample after mixing, centrifuging and reacting treatment of the blood sample to be detected, according to an agglutination sedimentation time model, performing sedimentation treatment on the target sample, and then performing centrifuging treatment to realize blood type detection of the blood sample to be detected.
2. The method for blood group assisted studying and judging according to claim 1, wherein in the step A, based on each time-space combination between preset temperature intervals and preset sample amounts, the following steps A1 to A5 are executed for each time-space combination and each blood sample under the time-space combination respectively, based on the environment of the temperature interval in the time-space combination, so as to construct samples; Step A1, mixing blood typing reagent with corresponding dosage in blood type detection operation by a test tube method for blood samples with sample quantity in space-time combination, centrifuging and reacting the treated sample, precipitating the sample, and initializing And (C) entering a step A2; a2, detecting and obtaining the upper, middle and lower three positions of the region where the sample is located in the test tube to correspond to the first position Transmittance at each time 、、 And form the sample corresponding to the first Overall light transmittance at each instant And obtain 、、 Variance corresponding to the three Then enter step A3; Step A3. Press Obtaining the corresponding first sample Reactivity at each moment And wait for Not less than a preset first time number When the method is used, the following formula is adopted: ; ; Obtaining sample No. 1 Degree of reaction at each moment Corresponding reaction rate And obtaining the sample No Degree of reaction at each moment Corresponding reaction acceleration rate , wherein, Indicating a preset number of second moments in time, Indicating that the sample corresponds to the first The degree of reaction at each moment in time, Representation of The length of time that the time passes, Indicating sample No. 1 Degree of reaction at each moment The corresponding reaction rate is used for the preparation of the catalyst, Representation of The length of time that the time passes, Indicating the overall light transmittance of the sample at the 0 th moment, and then entering the step A4; Step A4. According to the current ambient temperature Pressing down Obtaining a temperature adjustment factor And is expressed as follows: ; ; ; Obtaining a light transmittance upper limit calibration threshold Light transmission change acceleration upper limit calibration threshold Local turbidity gradient variance upper limit calibration threshold , wherein, Represents a preset upper threshold value of the light transmittance, Indicating a preset upper limit threshold value of the light transmission change acceleration, Representing a preset local turbidity gradient variance upper threshold, Indicating the preset temperature compensation coefficient of the temperature sensor, Representing a preset reference temperature, and then entering a step A5; step A5, judging whether the following conditions 1, 2 and 3 are satisfied simultaneously, if yes, obtaining from the 0 th moment to the 0 th moment Duration of time elapsed from moment to moment And press Obtaining the optimal sedimentation time length Combining the optimal precipitation time length with the temperature interval and the sample size in the space-time combination Constructing a sample, wherein, Indicating a preset duration of the security reservation, Indicating a preset minimum settling period of time, Indicating the preset maximum sedimentation time length, otherwise, according to the preset first adjacent time interval time length, waiting to enter the next time, aiming at Adding 1 for updating, and returning to the step A2; condition 1 based on The absolute value of the reaction rate of the sample corresponding to each time is smaller than that of the sample from each time to the continuous preset number of time in the historical time direction ; Condition 2. ; Condition 3. 。
3. The method of claim 2, wherein in step A5, it is determined whether condition 1, condition 2, and condition 3 are not satisfied, if so, the time interval is set according to the preset first adjacent time, and the next time is reached Adding 1 for updating, returning to the step A2, otherwise further judging whether the following conditions 1, 2 and 3 are satisfied at the same time, if so, obtaining the optimal sedimentation time And constructing a sample, otherwise, according to the preset second adjacent time interval duration, waiting for the next time, aiming at the next time And (3) updating by adding 1, and returning to the step (A2), wherein the preset second adjacent time interval duration is smaller than the preset first adjacent time interval duration.
4. The method for assisting in studying and judging blood type according to claim 2, wherein in the step A5, if it is judged that the conditions 1,2 and 3 are satisfied at the same time, the following step A6 is entered; step A6, executing centrifugal rotation under a preset centrifugal force on the sample for a preset period of time, wherein the centrifugal rotation under the preset centrifugal force does not damage the aggregated part in the sample, and then entering step A7; A7, obtaining a top view of the sample, judging whether the ratio of the dark red area extending outwards from the central position is larger than the ratio of the preset minimum agglutination area, and if so, obtaining the optimal sedimentation time And constructing a sample, otherwise, failing to construct the sample of the blood sample under the space-time combination.
5. The method for blood group assisted study according to claim 1, wherein the step D comprises steps D-1 to D-5; d-1, mixing blood typing reagents with corresponding doses according to blood type detection operation, and performing centrifugation and reaction treatment to obtain a corresponding target sample, wherein if the target sample is strongly agglutinated, the step D-5 is performed, and if the target sample is not strongly agglutinated, the step D-2 is performed; D-2, executing an agglutination and precipitation time period model according to a temperature interval corresponding to the environmental temperature of the target sample and the blood sample quantity of the blood sample to be detected, obtaining a corresponding precipitation time period interval, forming an optimal precipitation time period according to the upper limit time of the precipitation time period interval, precipitating the optimal precipitation time period aiming at the target sample, and then entering the step D-3; D-3, executing centrifugal rotation under a preset centrifugal force on the target sample for a preset period of time, wherein the centrifugal rotation under the preset centrifugal force does not damage the aggregated part in the target sample, and then entering step D-4; D-4, obtaining a top view of the target sample, judging whether the pixel value change from deep to shallow exists in the outward extension of the central position, if so, judging that the target sample is weakly agglutinated, and entering the step D-5, otherwise, judging that the target sample is unagglutinated, and entering the step D-5; step D-5, further determining the blood type of the blood sample to be detected according to whether the target sample is strong agglutination, weak agglutination or non-agglutination.
6. The method for assisting in studying and judging blood type according to claim 1, further comprising the step F of, And F, aiming at the target sample judged to be weakly aggregated in the step D, obtaining the quantity of the blood sample corresponding to the target sample to be detected and a temperature interval corresponding to the ambient temperature where the step D is executed, constructing a corresponding space-time combination, judging whether the same space-time combination exists in a sample library, if so, executing the step A aiming at the blood sample to be detected, constructing a corresponding sample, and adding the sample library, otherwise, not carrying out further processing.
7. The method for assisting in studying and judging blood type according to claim 1, wherein the target network to be trained in the step B comprises an initial feature extraction module, a feature compression and attention mechanism module, a multi-scale cavity space pyramid pooling module, a multi-scale branch module, a feature pyramid fusion module, a regularization module, an addition fusion module, a global context extraction module and a classification head module; The input end of the initial feature extraction module forms the input end of the target network to be trained, the output end of the initial feature extraction module is connected with the input end of the feature compression and attention mechanism module, the output end of the feature compression and attention mechanism module is connected with the input end of the multi-scale cavity space pyramid pooling module, the output end of the multi-scale cavity space pyramid pooling module is respectively connected with the input end of the regularization module and the two input ends of the multi-scale branch module, the two output ends of the multi-scale branch module are respectively connected with the two input ends of the feature pyramid fusion module, the output end of the feature pyramid fusion module is connected with the two input ends of the regularization module in a butt joint mode, the output end of the addition fusion module is connected with the input end of the global context extraction module, the output end of the global context extraction module is connected with the input end of the classification head module, and the output end of the classification head module forms the output end of the target network to be trained; the multi-scale branching module comprises a first branching module and a second branching module, wherein the first branching module sequentially comprises a first maximum pooling layer, a depth separable convolution layer and an SE attention module from an input end to an output end, and the pooling core of the first maximum pooling layer is as follows The step length is 2, the input end of the first maximum pooling layer forms the input end of the first branch module, the output end of the SE attention module forms the output end of the first branch module, the second branch module sequentially comprises a pool second maximum pooling layer, a depth separable convolution layer and an SE attention module from the input end to the output end, and the pool core size of the second maximum pooling layer is that The input end of the first branch module and the input end of the second branch module form two input ends of a multi-scale branch module, and the output end of the first branch module and the output end of the second branch module form two output ends of the multi-scale branch module; the feature pyramid fusion module comprises an up-sampling module, The device comprises a convolution module and an addition fusion module, wherein the input end of the up-sampling module and one of the input ends of the addition fusion module form two input ends of the feature pyramid fusion module, the input end of the up-sampling module is connected with the output end of a second branch module in the multi-scale branch module in a butt joint mode, and the input end of the addition fusion module is connected with the output end of a first branch module in the multi-scale branch module in a butt joint mode The input end of the convolution module is provided with a signal processing unit, The output end of the convolution module is butted with the other input end of the addition fusion module, and the output end of the addition fusion module forms the output end of the feature pyramid fusion module.
8. The blood type auxiliary research and judgment method according to claim 7, wherein the initial feature extraction module sequentially comprises a 2D convolution layer, a normalization layer, a ReLU activation layer, a depth separable convolution layer, a normalization layer and a ReLU activation layer from an input end to an output end, the input end of the 2D convolution layer forms the input end of the initial feature extraction module, and the output end of a sequentially second ReLU activation layer forms the output end of the initial feature extraction module; The feature compression and attention mechanism module sequentially comprises a first bottleneck residual module, an SE attention module, a second bottleneck residual module and an SE attention module from an input end to an output end, wherein the input end of the first bottleneck residual module forms the input end of the feature compression and attention mechanism module, the output end of the second SE attention module sequentially forms the output end of the feature compression and attention mechanism module, the structure of the first bottleneck residual module is the same as that of the second bottleneck residual module, and the structure of the first bottleneck residual module and the structure of the second bottleneck residual module respectively sequentially comprise from the input end to the output end A convolution module, A depth splittable convolution module, Convolution module, first in sequence The input end of the convolution module forms the input end of the bottleneck residual error module, and the sequence is second The output end of the convolution module forms the output end of the bottleneck residual error module; The multi-scale cavity space pyramid pooling module comprises a splicing layer, The convolution module, the normalization layer and four paths of pre-branches are connected in series from an input end to an output end The convolution module and the normalization layer are respectively connected in series from the input end to the output end The input ends of all the front branches are connected to form the input end of the multi-scale cavity space pyramid pooling module, the output ends of all the front branches are connected with the input ends of the splicing layer in a butt joint mode, and the output ends of the splicing layer are connected in series The convolution module is then abutted against the input end of the normalization layer, and the output end of the normalization layer forms the output end of the multi-scale cavity space pyramid pooling module.
9. The method for blood group assisted studying and judging according to claim 7, wherein the regularization module sequentially comprises DropBlock discarding layers, a Dropout random inactivation layer, a normalization layer and a ReLU activation layer from an input end to an output end, wherein the input end of the DropBlock discarding layers forms the input end of the regularization module, and the output end of the ReLU activation layer forms the output end of the regularization module.
10. The blood type auxiliary research and judgment method according to claim 7, wherein the global context extraction module sequentially comprises a global average pooling layer, a flattening operation layer, a linear layer, a ReLU activation layer and a linear layer from an input end to an output end, wherein the input end of the global average pooling layer forms an input end of the global context extraction module, and the output end of the linear layer forms an output end of the global context extraction module; The classifying head module comprises a full-connection layer and a Softmax layer which are connected in series from an input end to an output end, wherein the input end of the full-connection layer forms the input end of the classifying head module, and the output end of the Softmax layer forms the output end of the classifying head module.

Description

Blood type auxiliary studying and judging method Technical Field The invention relates to a blood type auxiliary studying and judging method, and belongs to the technical field of blood type auxiliary detection. Background In the existing blood type detection technology, an automatic interpretation system generally performs result analysis according to a preset agglutination feature model, and the basic principle is that the performance of a reaction hole site is compared with a standard template, and if the reaction form of a sample meets a preset rule or template feature, the sample is determined to be agglutinated or non-agglutinated. However, in the actual detection process, the expression form of the weak agglutination sample is complex and various, and is often in an atypical or irregular form, and is difficult to accurately identify by a given rule model. The samples are extremely easy to be misjudged or missed, so that the blood typing is wrong or the accidental antibodies are missed to be detected, and the blood transfusion safety is directly affected. The blood group analysis equipment on the market currently generally adopts the interpretation logic based on rule matching, and the limitation of the blood group analysis equipment is increasingly prominent along with the continuous improvement of the clinical requirements on detection precision. In order to make up for the defect of automatic interpretation, detection personnel often need to check all reaction results by naked eyes to eliminate potential interpretation errors. However, the naked eye interpretation is not only influenced by the subjective experience difference of individuals and lacks a unified objective standard, but also the visual fatigue of the detection personnel is easy to generate in the long-time and large-batch rechecking operation, so that the rechecking accuracy and reliability are further reduced. Therefore, the existing blood group interpretation mode which is dependent on the combination of the fixed rule and the manual review has the defects of complex operation, low efficiency, difficulty in stably identifying difficult samples such as weak agglutination and the like, and becomes a key technical bottleneck for restricting the improvement of the accuracy and the efficiency of blood group detection. Disclosure of Invention The invention aims to solve the technical problem of providing a blood group auxiliary judging method, which is characterized in that the optimal sedimentation time length corresponding to a weak agglutination classification blood sample is determined by using the reaction acceleration rate of a light transmittance reaction sample and combining a change temperature to dynamically adjust a calibration threshold value, so that the sample is constructed, an agglutination sedimentation time length model is obtained by training, and the blood group judgment operation is efficiently assisted by combining blood group detection operation. The invention designs a blood type auxiliary studying and judging method, based on preset quantity of blood samples belonging to weak agglutination classification, obtaining an agglutination sedimentation time length model according to the following steps A to B, and then carrying out the following step D for realizing blood type detection aiming at a blood sample to be detected; Step A, based on each time-space combination between each preset temperature interval and each preset sample amount, aiming at the samples subjected to mixing, centrifugation and reaction treatment in blood group detection operation under each time-space combination, performing sedimentation and determining the optimal sedimentation duration, further constructing each sample, forming a sample library, and then entering the step B; Step B, training a target network to be trained based on each sample in a sample library by taking a temperature interval in the sample and the sample quantity of a blood sample as input and taking a preset sedimentation duration interval corresponding to the optimal sedimentation duration in the sample as output to obtain an agglutination sedimentation duration model; And D, according to blood type detection operation, obtaining a target sample after mixing, centrifuging and reacting treatment of the blood sample to be detected, according to an agglutination sedimentation time model, performing sedimentation treatment on the target sample, and then performing centrifuging treatment to realize blood type detection of the blood sample to be detected. In the step A, based on each time-space combination between preset temperature intervals and preset sample amounts, respectively aiming at each time-space combination and each blood sample under the time-space combination, executing the following steps A1 to A5 based on the environment of the temperature intervals in the time-space combination to construct samples; Step A1, mixing blood typing reagent with corresponding dosage