CN113592842B - A method and device for identifying sample serum quality based on deep learning - Google Patents

Abstract

The application discloses a sample serum quality identification method and identification equipment based on deep learning, wherein the method comprises the following steps: acquiring a pretreated biochemical sample image dataset; constructing a deep convolutional neural network model frame, and performing learning training on the deep convolutional neural network model frame based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model; acquiring a biochemical sample image to be identified, inputting the biochemical sample image to be identified into a deep convolutional neural network model, and acquiring the probability of hemolysis, jaundice and lipidemia of a first biochemical sample corresponding to the biochemical sample image to be identified; based on the probability of hemolysis, jaundice and lipidemia of the first biochemical sample, determining the corresponding judgment conditions of hemolysis, jaundice and lipidemia of the first biochemical sample. The method can improve the sensitivity and specificity of recognizing the serum quality, can obtain good anti-interference capability, and can improve the recognition effect of the serum quality.

Description

A method and device for identifying sample serum quality based on deep learning

Technical Field

The present application relates to the field of medical testing technology, and in particular to a sample serum quality identification method and identification device based on deep learning.

Background technique

Poor sample quality is one of the main causes of clinical laboratory errors. Unqualified samples account for 60% of pre-analytical errors. Therefore, it is necessary to attach great importance to how to identify unqualified samples.

In serum quality identification, it is necessary to pay great attention to the hemolysis, jaundice, blood lipids and other conditions of the sample. At present, visual evaluation of sample quality has been widely used in clinical laboratories, but due to the influence of environmental and physiological factors (such as Dalton's disease), there are large differences in visual evaluation results between different individuals, which can easily lead to low accuracy. At present, some sample pre-processing equipment uses traditional image segmentation algorithms combined with color difference models to identify serum quality.

However, when using traditional image segmentation algorithms combined with color difference models to identify serum quality, there are problems such as poor specificity and weak anti-interference ability, resulting in poor recognition of serum quality.

Summary of the invention

Based on this, the present application provides a sample serum quality identification method and identification device based on deep learning, which are used to improve the identification effect of serum quality.

In a first aspect, the present application provides a method for identifying sample serum quality based on deep learning, comprising:

Acquire a preprocessed biochemical sample image data set, wherein the preprocessed biochemical sample image data set includes: a plurality of biochemical sample images and serum indexes corresponding to the plurality of biochemical sample images;

Constructing a deep convolutional neural network model framework, and performing learning and training on the deep convolutional neural network model framework based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model;

Acquire a biochemical sample image to be identified, and input the biochemical sample image to be identified into the deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the biochemical sample image to be identified;

Based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample, determining the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample;

Based on the judgment, it is determined that the first biochemical sample is a qualified sample or an unqualified sample.

In a possible design, the deep convolutional neural network model framework is trained based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model, including:

Randomly dividing the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio;

Training the deep convolutional neural network model framework based on the biochemical sample training data set to obtain an initial deep convolutional neural network model;

Inputting the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and performing probability accuracy verification on the initial deep convolutional neural network model;

If the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model.

In a possible design, the deep convolutional neural network model framework is trained based on the biochemical sample training data set to obtain an initial deep convolutional neural network model, including:

Set system parameters of the tensorfow system, wherein the system parameters include an initial learning rate and an iteration parameter of the initial learning rate;

Performing image enhancement processing on the biochemical sample training data set to obtain a processed biochemical sample training data set;

The deep convolutional neural network model framework is trained on the tensorflow system based on the processed biochemical sample training data set to obtain the initial deep convolutional neural network model.

In a possible design, the deep convolutional neural network model framework is trained on the tensorflow system based on the processed biochemical sample training data set to obtain the initial deep convolutional neural network model, including:

Based on the processed biochemical sample training data set, the deep convolutional neural network model framework is trained on the tensorflow system to obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on a preset classification network, determining the sum of probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on the sum of the probabilities, the model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia are determined to obtain the initial deep convolutional neural network model.

In a possible design, the preset classification network is a binary classification network with a Sigmoid activation function.

In one possible design, a deep convolutional neural network model framework is constructed, including:

A 1*1 convolution kernel combined with a residual network was used to construct an initial deep convolutional neural network model framework with 782 layers;

A global average pooling 2D layer and a final output layer are added after the initial convolutional neural network model framework to construct the deep convolutional neural network model framework; wherein,

The global average pooling 2D layer is used to output a feature map of the input biochemical sample image of the deep convolutional neural network model, and the final output layer is used to output the probabilities of hemolysis, jaundice and lipemia of the biochemical sample.

In a possible design, a pre-processed biochemical sample image dataset is obtained, including:

Obtaining original biochemical sample image datasets;

By manually marking the biochemical sample images with interference information in the serum part of the original biochemical sample image data set, a biochemical sample image data set after the interference information is marked is obtained;

The image resolution of the biochemical sample images in the biochemical sample image data set after the interference information is marked is set to N*M*Z, where N, M, and Z are integers greater than 1, to obtain the preprocessed biochemical sample image data set.

In a second aspect, an embodiment of the present application provides an identification device, including:

A processing unit is used to: obtain a preprocessed biochemical sample image data set, wherein the preprocessed biochemical sample image data set includes: a plurality of biochemical sample images and serum indexes corresponding to the plurality of biochemical sample images;

Constructing a deep convolutional neural network model framework, and performing learning and training on the deep convolutional neural network model framework based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model;

The judgment unit is used to: obtain a biochemical sample image to be identified, and input the biochemical sample image to be identified into the deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of a first biochemical sample corresponding to the biochemical sample image to be identified; based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample, determine the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample; based on the judgment status, determine whether the first biochemical sample is a qualified sample or an unqualified sample.

In a possible design, the processing unit is specifically used for:

Randomly dividing the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio;

Training the deep convolutional neural network model framework based on the biochemical sample training data set to obtain an initial deep convolutional neural network model;

Inputting the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and performing probability accuracy verification on the initial deep convolutional neural network model;

If the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model.

In a possible design, the processing unit is specifically used for:

Set system parameters of the tensorfow system, wherein the system parameters include an initial learning rate and an iteration parameter of the initial learning rate;

Performing image enhancement processing on the biochemical sample training data set to obtain a processed biochemical sample training data set;

The deep convolutional neural network model framework is trained on the tensorflow system based on the processed biochemical sample training data set to obtain the initial deep convolutional neural network model.

In a possible design, the processing unit is specifically used for:

Based on the processed biochemical sample training data set, the deep convolutional neural network model framework is trained on the tensorflow system to obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on a preset classification network, determining the sum of probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on the sum of the probabilities, the model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia are determined to obtain the initial deep convolutional neural network model.

In a possible design, the preset classification network is a binary classification network with a Sigmoid activation function.

In a possible design, the processing unit is specifically used for:

A 1*1 convolution kernel combined with a residual network was used to construct an initial deep convolutional neural network model framework with 782 layers;

A global average pooling 2D layer and a final output layer are added after the initial convolutional neural network model framework to construct the deep convolutional neural network model framework; wherein,

The global average pooling 2D layer is used to output a feature map of the input biochemical sample image of the deep convolutional neural network model, and the final output layer is used to output the probabilities of hemolysis, jaundice and lipemia of the biochemical sample.

In a possible design, the processing unit is specifically used for:

Obtaining original biochemical sample image datasets;

By manually marking the biochemical sample images with interference information in the serum part of the original biochemical sample image data set, a biochemical sample image data set after the interference information is marked is obtained;

The image resolution of the biochemical sample images in the biochemical sample image data set after the interference information is marked is set to N*M*Z, where N, M, and Z are integers greater than 1, to obtain the preprocessed biochemical sample image data set.

In a third aspect, an embodiment of the present application provides an identification device, the identification device comprising: at least one memory and at least one processor;

The at least one memory is used to store one or more programs;

When the one or more programs are executed by the at least one processor, the method involved in any possible design of the first aspect described above is implemented.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one program; when the at least one program is executed by a processor, the method involved in any possible design of the above-mentioned first aspect is implemented.

The beneficial effects of this application are as follows:

In the technical solution provided in the present application, a preprocessed biochemical sample image data set is obtained, and the preprocessed biochemical sample image data set includes: multiple biochemical sample images and serum indexes corresponding to the multiple biochemical sample images; a deep convolutional neural network model framework is constructed, and the deep convolutional neural network model framework is trained based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model; the biochemical sample image to be identified is obtained, and the biochemical sample image to be identified is input into the deep convolutional neural network model to obtain the probability of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the biochemical sample image to be identified; based on the probability of hemolysis, jaundice and lipemia of the first biochemical sample, the judgment of hemolysis, jaundice and lipemia corresponding to the first biochemical sample is determined. In this way, when the serum quality of the biochemical sample image is identified, the sensitivity and specificity of the serum quality can be improved, so that a good anti-interference ability of the serum quality can be obtained, and then the recognition effect of the serum quality can be improved.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a flow chart of a method for identifying sample serum quality based on deep learning provided in an embodiment of the present application;

FIG2 is a schematic diagram of a process for obtaining a preprocessed biochemical sample image data set provided in an embodiment of the present application;

FIG3 is a schematic diagram of a learning and training process for a deep convolutional neural network model provided in an embodiment of the present application;

FIG4 is a schematic diagram of a qualified sample and an unqualified sample provided in an embodiment of the present application;

FIG5 is a schematic diagram of classification of unqualified biochemical samples provided in an embodiment of the present application;

FIG6a is a schematic diagram of a ROC curve corresponding to a hemolysis classification provided in an embodiment of the present application;

FIG6b is a schematic diagram of a ROC curve corresponding to a jaundice classification provided in an embodiment of the present application;

FIG6c is a schematic diagram of a ROC curve corresponding to a lipemia classification provided in an embodiment of the present application;

FIG6d is a schematic diagram of a ROC curve corresponding to the calculation of the sum of the classification probabilities of hemolysis, jaundice, and lipemia provided in an embodiment of the present application;

FIG7 is a schematic diagram of the structure of an identification device provided in an embodiment of the present application;

FIG8 is a schematic diagram of the structure of an identification device provided in an embodiment of the present application.

Detailed ways

To facilitate understanding of the technical solution provided by the embodiments of the present application, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of methods consistent with some aspects of the present application as detailed in the appended claims.

Before introducing the embodiments of the present application, some terms in the present application are first explained to facilitate understanding by those skilled in the art.

The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and/or" used in this article refers to and includes any or all possible combinations of one or more associated listed items. It should also be understood that the term "at least one" used in this article includes one or more, "multiple" includes two and more, and "multiple" includes two and more.

Unless otherwise specified, the ordinal numbers such as "first" to "third" mentioned in the embodiments of the present application are used to distinguish multiple objects and are not used to limit the order, timing, priority or importance of the multiple objects.

Any biochemical sample involved in this article includes two parts: plasma and serum. Exemplarily, referring to FIG4 , the biochemical sample may include two parts: plasma and serum.

The sample serum quality identification method based on deep learning provided in the embodiment of the present application will be specifically described below in conjunction with Figures 1 to 6d.

Please refer to FIG1, which is a schematic diagram of a method for identifying sample serum quality based on deep learning provided in an embodiment of the present application. The execution subject of the method flow shown in FIG1 is an identification device. As shown in FIG1, the method flow may include the following steps:

S101, obtaining a preprocessed biochemical sample image dataset.

In some embodiments, the preprocessed biochemical sample image data set may include: multiple biochemical sample images and serum indexes corresponding to the multiple biochemical sample images, wherein the serum indexes corresponding to the multiple biochemical sample images can be used to characterize the serum quality of the biochemical samples corresponding to the multiple biochemical sample images.

In some embodiments, as shown in FIG. 2 , executing step S101 may specifically include the following process steps:

S201. Obtain an original biochemical sample image dataset.

In some embodiments, multiple biochemical sample images of a hospital or multiple hospitals in a certain period of time and the serum indexes corresponding to the multiple biochemical sample images can be collected as the original biochemical sample image dataset. For example, 10,667 biochemical sample images of an outpatient department of a hospital within three months and the serum indexes corresponding to the 10,667 biochemical sample images can be collected as the original biochemical sample image dataset.

S202, manually marking the biochemical sample images with interference information in the serum part of the original biochemical sample image data set, to obtain a biochemical sample image data set after the interference information is marked.

In some embodiments, the interference information may include, but is not limited to, labels, handwriting, stickers, and other information.

In some embodiments, the biochemical sample images with interference information in the serum part of the original biochemical sample image data set can be manually marked, so that when the deep convolutional neural network model is subsequently trained based on the obtained biochemical sample image data set with the marked interference information, the influence of the interference information on the sensitivity of the deep convolutional neural network model and the degree of enhancement of the anti-interference ability of the deep convolutional neural network model can be determined.

S203, setting the image resolution of the biochemical sample images in the biochemical sample image data set after the interference information is marked to N*M*Z, where N, M, and Z are integers greater than 1, to obtain a preprocessed biochemical sample image data set.

Exemplarily, N*M*Z may be set to 120*500*32.

In the embodiment of the present application, by setting the image resolution of the biochemical sample images in the biochemical sample image data set after marking the interference information to N*M*Z, it can facilitate the subsequent training of the deep convolutional neural network model.

S102, constructing a deep convolutional neural network model framework, and performing learning and training on the deep convolutional neural network model framework based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model.

In some embodiments, a 1*1 convolution kernel can be used in combination with a residual network to construct an initial deep convolutional neural network model framework of 782 layers, and then a global average pooling 2D layer and a final output layer are added after the initial convolutional neural network model framework to construct a deep convolutional neural network model framework. Among them, the global average pooling 2D layer can be used to output the feature map of the input biochemical sample image of the deep convolutional neural network model. The final output layer can be used to output the probability of hemolysis, jaundice and lipemia of the biochemical sample.

In some embodiments, as shown in FIG3 , after the deep convolutional neural network model framework is constructed, the learning and training process of the deep convolutional neural network model framework may include the following steps:

S301 . Randomly divide the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio.

In some embodiments, the preset ratio may be 8:2.

In some embodiments, Keras (an open source artificial neural network library written in Python) can be used to automatically and randomly divide the pre-processed biochemical sample image dataset into a biochemical sample training dataset and a biochemical sample verification dataset according to a preset ratio.

S302: Train the deep convolutional neural network model framework based on the biochemical sample training data set to obtain an initial deep convolutional neural network model.

In some embodiments, a tensorflow system (a symbolic mathematical system based on data flow programming) can be used as the backend of a deep convolutional neural network model framework for learning and training.

In the specific implementation process, the system parameters of the tensorfow system can be set. For example, the initial learning rate of the tensorfow system can be set to 0.0001, and the iteration parameter of the initial learning rate can be set to iterate 1/2 every 10 cycles (epochs). Then, the training time of the deep convolutional neural network model framework can be 120 cycles, and the algorithm iteration time is short.

In a specific implementation process, the biochemical sample training data set can be subjected to image enhancement processing to obtain the processed biochemical sample training data set. For example, a geometric transformation (including translation, flipping, rotation, scaling, etc.) method can be used to perform image enhancement processing on the biochemical sample training data set to obtain the processed biochemical sample training data set. Among them, in the geometric transformation method, one or more combinations of image enhancement methods such as rotation/reflection transformation, flipping transformation, scaling transformation, translation transformation, scale transformation, contrast transformation, noise disturbance (salt and pepper noise and Gaussian noise can be used), and staggered transformation can be used. Of course, other image enhancement methods can also be used to perform image enhancement processing on the biochemical sample training data set, which is not limited by the embodiment of the present application. Among them, other image enhancement processing methods can include but are not limited to: random brightness adjustment method, random contrast adjustment, etc. In a specific implementation process, the ImageDataGenerator function in the open source code library Keras can be used to implement the use of corresponding image enhancement methods to perform image enhancement processing on the biochemical sample training data set. It should be understood that the number of biochemical sample images included in the processed biochemical sample training data set is greater than the number of biochemical sample images included in the biochemical sample training data set.

In the specific implementation process, the deep convolutional neural network model framework can be trained on the tensorflow system based on the processed biochemical sample training data set to obtain the initial deep convolutional neural network model.

For example, as shown in Figure 5, when the biochemical samples are classified into hemolysis, jaundice and lipemia, there may be overlapping parts, for example, a certain biochemical sample may be a hemolysis combined with jaundice sample. In the embodiment of the present application, the deep convolutional neural network model framework can be trained on the tensorfow system based on the processed biochemical sample training data set to obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set. Then, the probability sum between the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set can be determined based on a preset classification network (such as a two-classification network of a Sigmoid activation function). Afterwards, based on the probability sum, the model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia can be determined, and the initial deep convolutional neural network model can be obtained, so that the sensitivity of the deep convolutional neural network model can be improved and the anti-interference ability of the deep convolutional neural network model can be enhanced.

Exemplarily, in the process of learning and training the deep convolutional neural network model framework provided by the embodiment of the present application using the 10,667 biochemical sample images in the above example, the ROC curve corresponding to the probability of hemolysis can be shown in Figure 6a, the ROC curve corresponding to the probability of jaundice can be shown in Figure 6b, and the ROC curve corresponding to the probability of lipemia can be shown in Figure 6c. When calculating the sum of the probabilities of hemolysis, jaundice, and lipemia, the corresponding ROC curve can be shown in Figure 6d.

As shown in Figures 6a-6d, when the deep convolutional neural network model provided in the embodiment of the present application is used to identify the serum quality of biochemical sample images, although there is interference information in the biochemical sample image data set for learning and training the deep convolutional neural network model, a higher sensitivity and specificity can be obtained, indicating that the interference information has little effect on the sensitivity of the deep convolutional neural network model and the anti-interference ability of the enhanced deep convolutional neural network model, and the deep convolutional neural network model has good anti-interference ability.

S303, inputting the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and performing probability accuracy verification on the initial deep convolutional neural network model.

In the specific implementation process, the probability accuracy threshold of the initial deep convolutional neural network model can be set. When the biochemical sample images in the biochemical sample verification data set are respectively input into the initial deep convolutional neural network model, the probabilities of hemolysis, jaundice and lipemia corresponding to the second biochemical sample corresponding to the biochemical sample image in the biochemical sample verification data set can be predicted. Then, the probabilities of hemolysis, jaundice and lipemia corresponding to the second biochemical sample actually judged based on the serum index of hemolysis, jaundice and lipemia corresponding to the second biochemical sample are compared with the predicted probabilities of hemolysis, jaundice and lipemia corresponding to the second biochemical sample, and the probability accuracy of the initial deep convolutional neural network model can be obtained. Afterwards, the probability accuracy of the initial deep convolutional neural network model is compared with the probability accuracy threshold, and it can be verified whether the probability accuracy of the initial deep convolutional neural network model meets the preset requirements.

S304: If the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model.

In the specific implementation process, the permissible error range of the probability accuracy can be set. If the error between the probability accuracy of the deep convolutional neural network model and the probability accuracy threshold is within the permissible error range, it can be determined that the probability accuracy of the initial deep convolutional neural network has passed the verification and meets the preset requirements.

In the specific implementation process, if the error between the probability accuracy of the deep convolutional neural network model and the probability accuracy threshold is not within the allowable error range, it can be determined that the probability accuracy of the initial deep convolutional neural network has not passed the verification and does not meet the preset requirements. When the probability accuracy of the initial deep convolutional neural network has not passed the verification, it can return to step S101 until the probability accuracy of the initial deep convolutional neural network has passed the verification.

S103, obtaining a biochemical sample image to be identified.

It should be noted that, in the specific implementation process, the embodiment of the present application does not limit the execution order between steps S101-S102 and step S103. For example, the identification device may first execute steps S101-S102 and then execute step S103, or may first execute step S103 and then execute steps S101-S102, or may execute steps S101-S102 and step S103 at the same time.

S104: Input the image of the biochemical sample to be identified into a deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the image of the biochemical sample to be identified.

S105 . Determine the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample.

In some embodiments, the judgment may include whether the first biochemical sample is a qualified biochemical sample or an unqualified biochemical sample, and if it is an unqualified biochemical sample, whether it is a biochemical sample that is a combination of one or more of hemolysis, icterus and lipemia.

In a specific implementation process, the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample can be compared with the preset probability thresholds of hemolysis, jaundice and lipemia respectively. When it is determined that one or more of the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample is greater than or equal to the corresponding probability threshold, it can be determined that the first biochemical sample is a biochemical sample that is a combination of one or more of hemolysis, jaundice and lipemia. When it is determined that any one of the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample is less than the corresponding probability threshold, it can be determined that the first biochemical sample is a biochemical sample that is a combination of one or more of hemolysis, jaundice and lipemia.

It can be seen from the above description that in the technical solution provided in the embodiment of the present application, a preprocessed biochemical sample image data set is obtained, and the preprocessed biochemical sample image data set includes: multiple biochemical sample images and serum indexes corresponding to each of the multiple biochemical sample images; a deep convolutional neural network model framework is constructed, and the deep convolutional neural network model framework is trained based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model; the biochemical sample image to be identified is obtained, and the biochemical sample image to be identified is input into the deep convolutional neural network model to obtain the probability of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the biochemical sample image to be identified; based on the probability of hemolysis, jaundice and lipemia of the first biochemical sample, the judgment of hemolysis, jaundice and lipemia corresponding to the first biochemical sample is determined. In this way, when the serum quality of the biochemical sample image is identified, the sensitivity and specificity of identifying the serum quality can be improved, so that good anti-interference ability can be obtained, and then the recognition effect of serum quality can be improved.

Based on the same inventive concept, the embodiment of the present application further provides an identification device. As shown in FIG7 , the identification device 600 may include:

The processing unit 601 is used to: obtain a preprocessed biochemical sample image data set, wherein the preprocessed biochemical sample image data set includes: a plurality of biochemical sample images and serum indexes corresponding to the plurality of biochemical sample images;

Constructing a deep convolutional neural network model framework, and performing learning and training on the deep convolutional neural network model framework based on the preprocessed biochemical sample image data set to obtain a deep convolutional neural network model;

The judgment unit 602 is used to: obtain a biochemical sample image to be identified, and input the biochemical sample image to be identified into the deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the biochemical sample image to be identified; based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample, determine the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample; based on the judgment status, determine whether the first biochemical sample is a qualified sample or an unqualified sample.

In a possible design, the processing unit 601 is specifically configured to:

Randomly dividing the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio;

Training the deep convolutional neural network model framework based on the biochemical sample training data set to obtain an initial deep convolutional neural network model;

Inputting the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and performing probability accuracy verification on the initial deep convolutional neural network model;

If the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model.

In a possible design, the processing unit 601 is specifically configured to:

Set system parameters of the tensorfow system, wherein the system parameters include an initial learning rate and an iteration parameter of the initial learning rate;

Performing image enhancement processing on the biochemical sample training data set to obtain a processed biochemical sample training data set;

The deep convolutional neural network model framework is trained on the tensorflow system based on the processed biochemical sample training data set to obtain the initial deep convolutional neural network model.

In a possible design, the processing unit 601 is specifically configured to:

Based on the processed biochemical sample training data set, the deep convolutional neural network model framework is trained on the tensorflow system to obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on a preset classification network, determining the sum of probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set;

Based on the sum of the probabilities, the model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia are determined to obtain the initial deep convolutional neural network model.

In a possible design, the preset classification network is a binary classification network with a Sigmoid activation function.

In a possible design, the processing unit 601 is specifically configured to:

A 1*1 convolution kernel combined with a residual network was used to construct an initial deep convolutional neural network model framework with 782 layers;

A global average pooling 2D layer and a final output layer are added after the initial convolutional neural network model framework to construct the deep convolutional neural network model framework; wherein,

The global average pooling 2D layer is used to output a feature map of the input biochemical sample image of the deep convolutional neural network model, and the final output layer is used to output the probabilities of hemolysis, jaundice and lipemia of the biochemical sample.

In a possible design, the processing unit 601 is specifically used for:

Obtaining original biochemical sample image datasets;

By manually marking the biochemical sample images with interference information in the serum part of the original biochemical sample image data set, a biochemical sample image data set after the interference information is marked is obtained;

The image resolution of the biochemical sample images in the biochemical sample image data set after the interference information is marked is set to N*M*Z, where N, M, and Z are integers greater than 1, to obtain the preprocessed biochemical sample image data set.

The identification device 600 in the embodiment of the present application and the sample serum quality identification method based on deep learning shown in the above-mentioned Figure 1 are inventions based on the same concept. Through the above-mentioned detailed description of the sample serum quality identification method based on deep learning, those skilled in the art can clearly understand the implementation process of the identification device 600 in this embodiment, so for the sake of brevity of the specification, it will not be repeated here.

Based on the same inventive concept, the embodiment of the present application further provides an identification device, as shown in FIG8 , the identification device 700 may include: at least one memory 701 and at least one processor 702. Among them:

The at least one memory 701 is used to store one or more programs.

When one or more programs are executed by at least one processor 702, the sample serum quality identification method based on deep learning shown in FIG. 1 is implemented.

The identification device 700 may also preferably include a communication interface (not shown in FIG. 8 ), which is used for communicating with an external device and performing data exchange transmission.

It should be noted that the memory 701 may include a high-speed RAM memory, and may also include a nonvolatile memory (nonvolatile memory), such as at least one disk memory.

In a specific implementation process, if the memory, processor and communication interface are integrated on a chip, the memory, processor and communication interface can communicate with each other through an internal interface. If the memory, processor and communication interface are implemented independently, the memory, processor and communication interface can be connected to each other through a bus and communicate with each other.

Based on the same inventive concept, an embodiment of the present application also provides a computer-readable storage medium, which can store at least one program. When the at least one program is executed by a processor, the sample serum quality identification method based on deep learning shown in Figure 1 above is implemented.

It should be understood that a computer-readable storage medium is any data storage device that can store data or programs, which can then be read by a computer system. Examples of computer-readable storage media include read-only memory, random access memory, CD-ROM, HDD, DVD, magnetic tape, and optical data storage devices, etc.

The computer readable storage medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, radio frequency (RF), etc., or any suitable combination of the above.

The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application.

Claims (7)

1. A method for identifying sample serum quality based on deep learning, comprising: Acquire a preprocessed biochemical sample image data set, wherein the preprocessed biochemical sample image data set includes: a plurality of biochemical sample images and serum indexes corresponding to the plurality of biochemical sample images; Constructing a deep convolutional neural network model framework, and randomly dividing the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio; Set system parameters of the tensorfow system, wherein the system parameters include an initial learning rate and an iteration parameter of the initial learning rate; Performing image enhancement processing on the biochemical sample training data set to obtain a processed biochemical sample training data set; Based on the processed biochemical sample training data set, the deep convolutional neural network model framework is trained on the tensorflow system to obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set; Based on a preset classification network, determining the sum of probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set; Based on the sum of probabilities, determining model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia, and obtaining an initial deep convolutional neural network model; Inputting the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and performing probability accuracy verification on the initial deep convolutional neural network model; If the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model; Acquire a biochemical sample image to be identified, and input the biochemical sample image to be identified into the deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample corresponding to the biochemical sample image to be identified; Based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample, the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample is determined. 2. The method according to claim 1, characterized in that the preset classification network is a binary classification network with a Sigmoid activation function. 3. The method according to claim 1 or 2, characterized in that constructing a deep convolutional neural network model framework comprises: A 1*1 convolution kernel combined with a residual network was used to construct an initial deep convolutional neural network model framework with 782 layers; A global average pooling 2D layer and a final output layer are added after the initial convolutional neural network model framework to construct the deep convolutional neural network model framework; wherein, The global average pooling 2D layer is used to output a feature map of the input biochemical sample image of the deep convolutional neural network model, and the final output layer is used to output the probabilities of hemolysis, jaundice and lipemia of the biochemical sample. 4. The method according to claim 1 or 2, characterized in that obtaining a preprocessed biochemical sample image data set comprises: Obtaining original biochemical sample image datasets; By manually marking the biochemical sample images with interference information in the serum part of the original biochemical sample image data set, a biochemical sample image data set after the interference information is marked is obtained; The image resolution of the biochemical sample images in the biochemical sample image data set after the interference information is marked is set to N*M*Z, where N, M, and Z are integers greater than 1, to obtain the preprocessed biochemical sample image data set. 5. An identification device, comprising: The processing unit is used to: obtain a preprocessed biochemical sample image data set, wherein the preprocessed biochemical sample image data set includes: multiple biochemical sample images and serum indexes corresponding to the multiple biochemical sample images; construct a deep convolutional neural network model framework, and randomly divide the preprocessed biochemical sample image data set into a biochemical sample training data set and a biochemical sample verification data set according to a preset ratio; set system parameters of a tensorflow system, wherein the system parameters include an initial learning rate and an iteration parameter of the initial learning rate; perform image enhancement processing on the biochemical sample training data set to obtain a processed biochemical sample training data set; and train the deep convolutional neural network model framework on the tensorflow system based on the processed biochemical sample training data set. , obtain the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set; based on the preset classification network, determine the probability sum between the probabilities of hemolysis, jaundice and lipemia corresponding to the biochemical sample images in the processed biochemical sample training data set; based on the probability sum, determine the model parameters of the deep convolutional neural network model framework for judging hemolysis, jaundice and lipemia, and obtain an initial deep convolutional neural network model; input the biochemical sample images in the biochemical sample verification data set into the initial deep convolutional neural network model respectively, and verify the probability accuracy of the initial deep convolutional neural network model; if the probability accuracy of the initial deep convolutional neural network model meets the preset requirements, the initial deep convolutional neural network model is used as the deep convolutional neural network model; The judgment unit is used to: obtain a biochemical sample image to be identified, and input the biochemical sample image to be identified into the deep convolutional neural network model to obtain the probabilities of hemolysis, jaundice and lipemia of a first biochemical sample corresponding to the biochemical sample image to be identified; based on the probabilities of hemolysis, jaundice and lipemia of the first biochemical sample, determine the judgment status of hemolysis, jaundice and lipemia corresponding to the first biochemical sample; based on the judgment status, determine whether the first biochemical sample is a qualified sample or an unqualified sample. 6. An identification device, comprising: at least one memory and at least one processor; The at least one memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method according to any one of claims 1 to 4 is implemented. 7. A computer-readable storage medium, characterized in that the computer-readable storage medium stores at least one program; when the at least one program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.