Intelligent Grading and Determination Method and System for Blood Chylomicron Degree Based on Image Processing

Abstract

The present invention discloses an intelligent grading determination method and system for blood chylomicron degree based on image processing. The method includes the following steps: S1. Establish a blood chylomicron degree grading model: S1-1. Construct a training set; S1-2. Use a deep neural network as the basic network model and train the basic network model through the training set to obtain a blood chylomicron degree grading model; S2. Use the blood chylomicron degree grading model to automatically grade the blood chylomicron degree. The present invention provides an intelligent grading determination method and system for blood chylomicron degree based on image processing, which can, without taking out the blood in the blood bag, collect the blood bag image and automatically grade the plasma chylomicron degree based on the deep neural network and image processing technology; the present invention can improve the judgment efficiency and accuracy, reduce the workload of blood collection personnel, and is beneficial to improving the reliability of blood primary screening, improving the blood quality, and ensuring the safety of clinical blood transfusion.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to an intelligent grading and determination method and system for blood chylomicron degree based on image processing. Background Art

[0002] For front-line blood collection personnel, chylous blood is the most difficult primary screening index to control. Traditional methods are divided into two categories: instrument determination and manual determination.

[0003] Instrument determination method: Traditional instrument determination methods include large optical instruments such as enzyme-linked immunosorbent assay (ELISA) readers, fully automatic enzyme immunoassay analyzers, biochemical analyzers, and ultraviolet-visible spectrophotometers. Due to the particularity of the requirements for time and venue at the blood collection site, the above-mentioned large optical instruments are not suitable for blood collection personnel to use on mobile blood collection vehicles because of their high prices and inconvenience in carrying. At the same time, the above instruments need to take out a certain amount of blood from the blood bag for detection, so the workload of the primary screening work in the blood station is increased.

[0004] Manual determination category: The first method is the chylous blood atlas method, that is, the blood collection personnel manually compare the collected blood with the existing chylous blood atlas; the second method is the experience method, that is, the blood collection personnel observe the color of the plasma with the naked eye and experience to judge whether the blood is available, whether there is chylomicron, and what the chylomicron degree is. However, the above methods are all subjective judgment methods, with personal subjective differences and a relatively high error rate.

[0005] Therefore, a more reliable solution is needed now. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide an intelligent grading and determination method and system for blood chylomicron degree based on image processing in view of the above deficiencies in the prior art.

[0007] To solve the above technical problem, the technical solution adopted by the present invention is: An intelligent grading and determination method for blood chylomicron degree based on image processing, comprising the following steps:

[0008] S1. Establish a grading model for blood chylomicron degree:

[0009] S1-1. Construct a training set: Collect pictures of several blood bags containing plasma, manually classify the chylomicron degree of the plasma in each blood bag, label the classification labels on the corresponding pictures, and then preprocess the images to construct a training set;

[0010] S1-2. Use a deep neural network as the basic network model, and train the basic network model through the training set to obtain a grading model for blood chylomicron degree;

[0011] S2. Take pictures of the blood to be graded together with the packaged blood bag, and input the image into the blood chylomicron degree grading model, which outputs the grading result of the blood chylomicron degree.

[0012] Preferably, in step S1-1, the chief technician in the blood station manually classifies the chylomicron degree of the plasma into five grades from 1 to 5. Among them, grade 1 is normal plasma, grade 2 is slightly chylomicron, grade 3 is moderately slightly chylomicron, grade 4 is moderately severely chylomicron, and grade 5 is severely chylomicron.

[0013] Preferably, in step S1-1, the preprocessing steps of the image include: image normalization and image segmentation.

[0014] Preferably, the image segmentation method is: first cut the image into pictures with a size of 1920×1080, and then select the noise-free area and cut it into several pictures with a size of 100×100 with labels.

[0015] Preferably, in step S1-1, the method for collecting blood bag pictures is: in a darkroom environment, place the blood bag containing plasma between the camera and the transparent background board, and set an LED white light board on the side of the transparent background board far from the blood bag, and take the blood bag pictures through the camera.

[0016] Preferably, in step S1-1, multiple transparent background boards with different patterns are used to collect multiple pictures of each blood bag.

[0017] Preferably, the basic network model is an improved deep residual network ResNet-50 model.

[0018] Preferably, in the improved deep residual network ResNet-50 model, the 7×7 convolutional kernel in the first convolutional layer of the ResNet-50 model is replaced with a 5×5 convolutional kernel.

[0019] The present invention also provides an intelligent grading determination system for blood chylomicron degree based on image processing, which uses the above method to grade the blood chylomicron degree. The system includes:

[0020] An image acquisition module, which is used to acquire pictures of blood bags containing plasma;

[0021] And a blood chylomicron degree grading module, which is used to grade the chylomicron degree of the plasma according to the blood bag pictures.

[0022] Preferably, the image acquisition module includes a light-shielding cover, and an LED white light board, a transparent background board, and a camera that are arranged in the light-shielding cover in sequence along the optical path direction. A blood bag containing plasma is placed between the transparent background board and the camera. A picture of the blood bag is taken and collected by the camera and input into the blood chyle degree grading module to obtain a grading result of the chyle degree of the plasma.

[0023] The beneficial effects of the present invention are as follows: The present invention provides an intelligent grading and determination method and system for blood chyle degree based on image processing, which can automatically grade the chyle degree of plasma by collecting blood bag images based on deep neural network and image processing technology without taking out the blood in the blood bag; the present invention can improve the judgment efficiency and accuracy, reduce the workload of blood collection personnel, and is conducive to improving the reliability of blood primary screening, improving blood quality, and ensuring clinical blood transfusion safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flowchart of the intelligent grading and determination method for blood chyle degree based on image processing of the present invention;

[0025] Figure 2 is a schematic diagram of the image acquisition device adopted in the present invention;

[0026] Figure 3 is a schematic diagram of image segmentation of the present invention;

[0027] Figure 4 is a schematic diagram of the structure of the improved deep residual network ResNet-50 model of the present invention;

[0028] Figure 5 is a diagram of the Loss curve (a) and accuracy curve (b) of the method of the present invention in an embodiment;

[0029] Figure 6 is the test result in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The following further describes the present invention in detail with reference to embodiments, so that those skilled in the art can implement it according to the description in the specification.

[0031] It should be understood that the terms such as "having", "including", and "comprising" used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0032] Embodiment 1

[0033] Referring to Figure 1 , this embodiment provides an intelligent grading and determination method for blood chyle degree based on image processing, including the following steps:

[0034] S1. Establish a blood chylomicron degree classification model:

[0035] S1-1. Construct a training set:

[0036] (1) Image acquisition and annotation: Collect pictures of several blood bags containing plasma. The chief technician in the blood station manually classifies the chylomicron degree of the plasma in each blood bag into five grades from 1 to 5. Among them, grade 1 is normal plasma, grade 2 is slightly chylomicron, grade 3 is moderately slightly chylomicron, grade 4 is moderately severely chylomicron, and grade 5 is severely chylomicron, and mark the classification labels on the corresponding pictures.

[0037] Refer to Figure 2 , in an embodiment, the device used for image acquisition includes: a light-shielding cover, and an LED white light board, a transparent background board, and a camera arranged in sequence along the optical path direction inside the light-shielding cover. During acquisition, place the blood bag containing plasma between the transparent background board and the camera, and take pictures of the blood bag through the camera. Among them, the transparent background board includes multiple sheets with different background patterns and can be switched into the optical path. Each blood bag can obtain multiple pictures through multiple different transparent background boards, thereby increasing the number of pictures.

[0038] (2) Then preprocess the images: image normalization and image segmentation to construct a training set.

[0039] Among them, refer to Figure 3 , the image segmentation method is: first cut the image into pictures with a size of 1920×1080, and then select the noise-free area and cut it into several pictures with a size of 100×100 with labels.

[0040] S1-2. Use a deep neural network as the basic network model, and train this basic network model through the training set to obtain a blood chylomicron degree classification model;

[0041] In this embodiment, the basic network model is an improved deep residual network ResNet-50 model. Specifically, in the improved deep residual network ResNet-50 model in this embodiment, the 7×7 convolutional kernel in the first convolutional layer of the ResNet-50 model is replaced with a 5×5 convolutional kernel. The structure of the ResNet-50 model is as Figure 4As shown in the figure, the 7×7 convolutional kernel is recognized as a relatively large convolutional kernel, aiming to extract information from a larger neighborhood range of the input image. However, in the recognition of plasma chylomicron degree in the present invention, the difference in the clarity of the background board and the clarity of the image edge is relatively subtle, and the feature differences between different categories are not obvious. In view of the above phenomenon, it is necessary to extract more subtle features from the chylous blood image to achieve a more accurate chylomicron degree image classification effect. By replacing the 7×7 convolutional kernel in the first convolutional layer of the ResNet-50 model with a 5×5 convolutional kernel, more subtle features in the chylous blood image can be obtained, improving the classification effect. Further, the 7×7 convolutional kernel has 7×7×channels = 49×channels parameters, and the number of parameters of the improved three 5×5 convolutional kernels is 5×5×channels = 25×channels, reducing the number of parameters by about 44.5%, greatly reducing the computational amount, and at the same time improving the network's ability to distinguish image details.

[0042] S2. Take a photo of the blood to be graded together with the packaged blood bag, and input the image into the blood chylomicron degree grading model. The blood chylomicron degree grading model outputs the blood chylomicron degree grading result.

[0043] Refer to Figure 5 , which is the Loss curve (a) and accuracy curve (b) diagram for the method of the present invention. In this embodiment, the Loss curve and Accuracy (ACC) curve results of the improved ResNet-50 model are analyzed. From the Loss curve, it can be seen that the Loss curve of the training set and the Loss curve of the validation set are almost parallel at the end, and the difference between the two lines is small, indicating that the improved ResNet-50 model finally tends to a stable state. The Acc curve obtained for the training set and the Acc curve obtained for the validation set indicate that the model has not reached the overfitting state and has a good classification effect, and the average accuracy of the validation set curve reaches about 0.96.

[0044] Refer to Figure 6 , which is the test result of the blood chylomicron degree grading model constructed based on the improved ResNet-50 model in the present invention. The test set includes data of five chylomicron degree levels from 1 to 5: Grade1, Grade2, Grade3, Grade4, Grade5, and the data distributions are 700 cases, 400 cases, 400 cases, 400 cases, and 400 cases respectively. The obtained confusion matrix is as Figure 5 shown. It can be seen that the blood chylomicron degree grading model has good performance in all five levels of blood chylomicron degree, and the overall determination accuracy reaches 0.96, having good ability to intelligently determine the blood chylomicron degree.

[0045] Example 2

[0046] An intelligent grading determination system for the degree of blood chyle based on image processing, which uses the method of Embodiment 1 to grade the degree of blood chyle. The system includes:

[0047] An image acquisition module, which is used to acquire pictures of blood bags containing plasma;

[0048] And a blood chyle degree grading module, which is used to grade the chyle degree of plasma according to the blood bag pictures.

[0049] In this embodiment, the structure of the image acquisition module is the same as the device used for image acquisition in Embodiment 1, and includes a light-shielding cover and an LED white light board, a transparent background board, and a camera arranged in sequence along the optical path direction inside the light-shielding cover. The blood chyle degree grading module is a computer embedded with the blood chyle degree grading model in Embodiment 1. During operation, a blood bag containing plasma is placed between the transparent background board and the camera, and the blood bag picture is taken and collected by the camera and input into the computer embedded with the blood chyle degree grading module, and finally the grading result of the chyle degree of plasma is obtained.

[0050] Although the embodiments of the present invention have been disclosed as above, it is not limited to only the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to specific details.

Claims

1. An intelligent grading and determination method for the chylomicron degree of blood based on image processing, characterized in that, It includes the following steps: S1. Establish a blood chylomicron degree classification model: S1-1. Construct a training set: Collect pictures of several blood bags containing plasma, manually classify the chylomicron degree of the plasma in each blood bag, label the classification labels on the corresponding pictures, and then preprocess the images to construct a training set; S1-2. Use a deep neural network as the basic network model, and train the basic network model with the training set to obtain a blood chylomicron degree classification model; S2. Take pictures of the blood to be classified and the blood bag containing the blood together, input the image into the blood chylomicron degree classification model, and the blood chylomicron degree classification model outputs the blood chylomicron degree classification result; The basic network model is an improved deep residual network ResNet-50 model; In the improved deep residual network ResNet-50 model, the 7×7 convolutional kernel in the first convolutional layer of the ResNet-50 model is replaced with a 5×5 convolutional kernel.

2. The intelligent grading determination method for blood chylomicron degree based on image processing according to claim 1, wherein In the step S1-1, the chief technician in the blood station manually classifies the chylomicron degree of the plasma into five grades from 1 to 5. Among them, grade 1 is normal plasma, grade 2 is mild chylomicron, grade 3 is moderately mild chylomicron, grade 4 is moderately severe chylomicron, and grade 5 is severe chylomicron.

3. The intelligent grading determination method for blood chylomicron degree based on image processing according to claim 1, characterized in that, In the step S1-1, the image preprocessing steps include: image normalization and image segmentation.

4. The intelligent grading determination method for blood chylomicron degree based on image processing according to claim 3, wherein Among them, The image segmentation method is: First, cut the image into pictures with a size of 1920×1080, and then select the noise-free area and cut it into several 100×100 pictures with labels.

5. The intelligent grading determination method for blood chylomicron degree based on image processing according to claim 1, wherein In the step S1-1, the method for collecting blood bag pictures is: In a darkroom environment, place the blood bag containing plasma between the camera and the transparent background board, and set an LED white light board on the side of the transparent background board away from the blood bag, and take pictures of the blood bag through the camera to obtain blood bag pictures.

6. The intelligent grading determination method for blood chylomicron degree based on image processing according to claim 5, wherein In the step S1-1, multiple transparent background boards with different patterns are used to collect multiple pictures of each blood bag.

7. An intelligent grading determination system for the degree of blood chylomicron based on image processing, characterized in that, It uses the method described in any one of claims 1-6 to classify the blood chylomicron degree. The system includes: An image acquisition module, which is used to collect pictures of blood bags containing plasma; And a blood chylomicron degree classification module, which is used to classify the chylomicron degree of the plasma according to the blood bag pictures.

8. The intelligent grading determination system for blood chylomicron degree based on image processing according to claim 7, characterized in that The image acquisition module includes a light-shielding cover and an LED white light board, a transparent background board and a camera arranged in sequence along the optical path direction in the light-shielding cover. The blood bag containing plasma is placed between the transparent background board and the camera, and the blood bag pictures are collected by taking pictures through the camera and input into the blood chylomicron degree classification module to obtain the classification result of the chylomicron degree of the plasma.

Images