CN110852384B - Medical image quality detection method, device and storage medium - Google Patents
Medical image quality detection method, device and storage medium Download PDFInfo
- Publication number
- CN110852384B CN110852384B CN201911101597.8A CN201911101597A CN110852384B CN 110852384 B CN110852384 B CN 110852384B CN 201911101597 A CN201911101597 A CN 201911101597A CN 110852384 B CN110852384 B CN 110852384B
- Authority
- CN
- China
- Prior art keywords
- medical image
- data
- image data
- abnormal
- original
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000002159 abnormal effect Effects 0.000 claims description 141
- 238000001514 detection method Methods 0.000 claims description 47
- 238000013145 classification model Methods 0.000 claims description 40
- 238000012216 screening Methods 0.000 claims description 29
- 238000012549 training Methods 0.000 claims description 24
- 238000004458 analytical method Methods 0.000 claims description 22
- 238000004590 computer program Methods 0.000 claims description 22
- 238000000034 method Methods 0.000 claims description 22
- 238000007621 cluster analysis Methods 0.000 claims description 18
- 238000010219 correlation analysis Methods 0.000 claims description 10
- 239000000203 mixture Substances 0.000 claims description 6
- 238000003066 decision tree Methods 0.000 claims description 3
- 238000007477 logistic regression Methods 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 description 11
- 238000004891 communication Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 3
- 238000004140 cleaning Methods 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000005311 nuclear magnetism Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The application relates to a medical image quality detection method, a medical image quality detection device and a storage medium. The method comprises the following steps: acquiring original medical image data; screening abnormal data in the original medical image data to obtain abnormal-free medical image data; removing data with correlation coefficients not meeting the requirements from the non-abnormal medical image data to obtain medical image data to be detected; and classifying the quality of the medical image data to be detected by using a preset classification model, and determining the medical image meeting the quality requirement. By adopting the method, the labor cost can be reduced.
Description
Technical Field
The present disclosure relates to the field of medical data processing technologies, and in particular, to a method and apparatus for detecting medical image quality, and a storage medium.
Background
With the development of medical technology, the variety of medical devices is becoming more and more popular. The scanning technology of different medical equipment is also various, and different medical images can be obtained by different scanning technologies and imaging modes, so that massive image data are generated. Meanwhile, for image diagnosis, the quality is the sum of properties of the image itself or the property of the examination itself, which determines whether clinical diagnosis can be satisfied or not, which is the subject of evaluation. The quality of the image affects the diagnostic value of the doctor. Therefore, obtaining a high-quality image from a large amount of image data is a problem to be solved.
However, the existing low quality or broken images are usually found by the user only when called for the film reading, and then notify the examinee to re-scan and re-read the film, resulting in an increase in labor cost.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a medical image quality detection method, apparatus, and storage medium that can reduce labor costs.
A medical image quality detection method, the method comprising:
acquiring original medical image data;
screening out abnormal data in the original medical image data to obtain abnormal-free medical image data;
removing data with correlation coefficients not meeting requirements from the non-abnormal medical image data to obtain medical image data to be detected;
and classifying the quality of the medical image data to be detected by using a preset classification model, and determining the medical image meeting the quality requirement.
In one embodiment, the screening out abnormal data in the original medical image data to obtain non-abnormal medical image data includes:
and carrying out cluster analysis on the original medical image data, and screening out abnormal data according to an analysis result to obtain abnormal-free medical image data.
In one embodiment, the clustering analysis is performed on the original medical image data, and abnormal data is screened out according to an analysis result to obtain abnormal-free medical image data, which includes:
acquiring the number of clusters, and selecting data with the same number as the number of clusters from the original medical image data as a cluster center;
determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data;
or (b)
Acquiring minimum neighborhood point number and neighborhood radius, traversing the original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius;
and performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
In one embodiment, the screening out abnormal data in the original medical image data to obtain non-abnormal medical image data includes:
performing outlier factor detection on the original medical image data to obtain outlier factors corresponding to the original medical image data;
and screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data.
In one embodiment, the removing the data whose correlation coefficient does not meet the requirement from the non-abnormal medical image data to obtain medical image data to be inspected includes:
any two pieces of non-abnormal medical image data are formed into a data pair;
calculating correlation coefficients of two non-abnormal medical image data in any data pair;
and deleting any one data in the data pair to obtain medical image data to be detected when the correlation coefficient determines that the two non-abnormal medical image data in the data pair have correlation.
In one embodiment, the deleting any one of the data pairs includes:
and deleting the same non-abnormal medical image data when any at least two groups of data pairs have correlation and contain the same non-abnormal medical image data.
In one embodiment, the method further comprises:
acquiring training sample data, wherein the training sample data comprises negative sample data and positive sample data;
adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data;
and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
In one embodiment, the preset classification model includes, but is not limited to, any one or more of a decision tree model, a logistic regression model, a neural network model, and a support vector machine model.
A medical image quality detection apparatus, the apparatus comprising:
the acquisition module is used for acquiring original medical image data;
the abnormality analysis module is used for screening abnormal data in the original medical image data to obtain abnormal-free medical image data;
the correlation analysis module is used for removing data with correlation coefficients not meeting requirements from the non-abnormal medical image data to obtain medical image data to be detected;
the detection module is used for carrying out quality classification on the medical image data to be detected by utilizing a preset detection model, and determining medical images meeting quality requirements.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the medical image quality detection method of any of the above.
According to the medical image quality detection method, the medical image quality detection device and the storage medium, after the original medical image data are acquired, abnormal data are screened out, data with correlation coefficients not meeting the requirements are removed, the data which can be detected are obtained, and then the data to be detected are subjected to quality classification by using a preset detection model. According to the method, abnormal data and data which do not meet the requirements are removed, noise data is prevented from affecting the detection quality, the burden of subsequent detection work is reduced, and further, high-quality images are automatically screened from original medical image data by combining medical image data with data analysis, so that the labor cost is reduced.
Drawings
FIG. 1 is a diagram of an application environment of a method for detecting quality of a medical image according to an embodiment;
FIG. 2 is a flow chart of a method for detecting quality of a medical image according to an embodiment;
FIG. 3 is a flowchart illustrating a step of removing data with correlation coefficients not meeting the requirements from non-abnormal medical image data to obtain medical image data to be inspected in one embodiment;
FIG. 4 is a block diagram showing a structure of a device for detecting quality of medical images according to an embodiment;
fig. 5 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The medical image quality detection method provided by the application can be applied to an application environment shown in fig. 1. The application environment involves a terminal 102 and a server 104, the terminal 102 communicating with the server 104 over a network. After the terminal 102 acquires the original medical image data, the above medical image quality detection method can be implemented by the terminal 102 alone. The terminal 102 may also transmit the original medical image data to the server 104, and the server 104 may implement the medical image quality detection method. Specifically, taking the terminal 102 as an example, after the terminal 102 acquires the original medical image data, screening out abnormal data in the original medical image data to obtain abnormal-free medical image data; the terminal 102 removes data with correlation coefficients not meeting the requirements from the non-abnormal medical image data to obtain medical image data to be detected; the terminal 102 performs quality classification on the medical image data to be detected by using a preset classification model, and determines the medical image meeting the quality requirement. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smartphones, tablet computers, and portable wearable devices, and the server 104 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers.
In one embodiment, as shown in fig. 2, a medical image quality detection method is provided, and the method is applied to the terminal in fig. 1 for illustration, and includes the following steps:
step S202, acquiring original medical image data.
The original medical image data refers to medical image data which is not subjected to medical image quality detection, and can be medical image data acquired by a terminal from a PACS system (Picture Archiving and Communication Systems, image archiving and communication system), medical image data generated by a medical device in real time or medical image data uploaded by a user in real time. The medical image data in the PACS system are data stored in a digital mode through various interfaces of various medical images generated in daily life. For example, medical images generated by medical equipment such as nuclear magnetism, magnetic resonance, CT (Computed Tomography ), ultrasound, various X-ray machines, various infrared instruments, microscopes and the like are digitally stored through corresponding interfaces.
Step S204, the abnormal data in the original medical image data is screened out, and the abnormal-free medical image data is obtained.
The abnormal data refers to noise data, that is, data in the original medical image data, which deviates from an expected value or has a large difference from most of the data. The abnormal-free medical image number is the original medical image data remained after the abnormal data is removed.
Specifically, after the original medical image data is acquired, the terminal performs cluster analysis on the original medical image data, and determines abnormal data according to a cluster analysis result. And then, removing the determined abnormal data from the original medical image data to obtain the abnormal-free medical image data. The clustering analysis refers to a process of classifying data into different classes or clusters, objects classified into the same cluster have great similarity, and objects among different clusters have great dissimilarity. Thus, in the present embodiment, the clearly abnormal data can be determined by cluster analysis. For example, data that is not allocated, or the amount of data in a certain cluster is much smaller than the amount of data in other clusters, can be determined to be abnormal data.
Step S206, data with correlation coefficients not meeting the requirements are removed from the abnormal medical image data, and medical image data to be detected is obtained.
The correlation coefficient refers to a value obtained by performing correlation analysis based on statistics, and is used for representing the correlation between two objects. The medical image data to be detected is the abnormal-free medical image data which are remained after the data of which the correlation coefficient does not meet the requirement are removed.
Specifically, after obtaining the abnormality-free medical image data, the abnormality-free medical image data is subjected to correlation detection by pearson correlation detection. And then determining data with stronger correlation according to the correlation coefficient obtained by correlation detection, and removing the non-abnormal medical image data with stronger correlation to obtain medical image data to be detected. Determining if the correlation is strong may result from comparing the correlation coefficient to a coefficient threshold. For example, a correlation coefficient greater than a coefficient threshold indicates a stronger correlation, and the coefficient threshold and specifically the comparison mode may be set according to the actual situation.
Step S208, quality classification is carried out on the medical image data to be detected by using a preset classification model, and medical images meeting the quality requirements are determined.
The classification model is a model which is obtained by training a large number of medical image data samples in advance, and the trained classification model is used for classifying the medical image data in quality. Classification models include, but are not limited to, any one or more of decision tree models, logistic regression models, neural network models, support vector machine models.
Specifically, after obtaining the medical image data to be inspected, invoking a pre-trained classification model. Inputting the medical image data to be detected into the called classification model, and classifying the medical image data to be detected in quality through the classification model. And then, determining whether the quality of the medical image data to be detected input into the classification model meets the requirements or not according to the classification result output by the classification model. When the classification result is that the quality of the medical image data to be detected is not satisfactory, the user can be further notified to perform scanning again.
According to the medical image quality detection method, after the original medical image data are acquired, abnormal data are screened out, data with correlation coefficients not meeting the requirements are removed, the data which can be detected are obtained, and then the data to be detected are subjected to quality classification by using a preset detection model. According to the method, abnormal data and data which do not meet the requirements are removed, noise data is prevented from affecting the detection quality, the burden of subsequent detection work is reduced, and further, high-quality images are automatically screened from original medical image data by combining medical image data with data analysis, so that the labor cost is reduced.
In one embodiment, performing cluster analysis on the original medical image data, and screening out abnormal data according to an analysis result to obtain non-abnormal medical image data specifically includes: acquiring the number of clusters, and selecting data with the same number as the number of clusters from the original medical image data as a cluster center; performing data distribution based on the distance between each clustering center and the rest of original medical image data; and determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data.
The number of clusters is a value preset for defining the number of clusters, for example, when the number of clusters is 4, namely, 4 clusters are clustered. A cluster center may be understood as a representation of a randomly selected class of data.
Specifically, when cluster analysis is performed, the number of clusters is first acquired. And then randomly selecting the data with the same quantity as the clustering number from the original medical image data, wherein the selected data is the clustering center. For example, when the number of clusters is 4, namely, 4 data are randomly selected from the original medical image data as cluster centers. And then, taking the cluster center as a standard, and calculating the Euclidean distance between the cluster center and other residual original medical image data. Based on the Euclidean distance, the rest original medical image data is distributed to the cluster center with the smallest Euclidean distance, and the data distributed to the same cluster center is expressed as one type of data. For example, if the Euclidean distance between the data A and the first cluster center is the smallest, the data A is distributed to the first cluster center. Finally, after the data allocation is completed, a cluster of data with the smallest data amount may be deleted as the abnormal data. The abnormal data may also be determined according to the difference between the data amounts of the various types, for example, when there is an excessive difference between the data amount of one cluster of data and the data amount of the other cluster, the data having the excessive difference of the cluster may be directly determined. And when the difference in the data amount between the clusters is not large, it is determined that there is no abnormal data. Wherein whether the difference is excessive may be determined by comparing the difference in the data amount of each cluster with a difference threshold.
In addition, when the difference of the data amount between the clusters is not large, the increment can be performed based on the existing number of clusters. And (3) re-clustering based on the number of clusters after increasing, and determining the abnormal data according to the new round of clustering results until the abnormal data can be determined. For example, the number of clusters originally is increased from 4 to 5,5 to 6, and so on. And then randomly determining a clustering center from the original medical image data according to the new clustering number to perform data distribution. If the clustering is repeated in multiple increments, and abnormal data with excessive data difference cannot be obtained, the fact that the abnormal data does not exist can be determined. In this embodiment, the obvious noise abnormal data is removed through cluster analysis, so that the quality of the data is improved.
In another embodiment, performing cluster analysis on the original medical image data, and screening out abnormal data according to the analysis result to obtain non-abnormal medical image data specifically includes: acquiring minimum neighborhood point number and neighborhood radius, traversing original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius; and performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
Among them, the cluster analysis used in the present embodiment is a DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise, density-based cluster algorithm). The DBSCAN algorithm is a clustering method that defines clusters as the largest set of densely connected points, can divide a region having a sufficiently high density into clusters, and can find arbitrary shapes in noisy data. The minimum domain number and the neighborhood radius are parameters in the DBSCAN algorithm, the neighborhood radius refers to a neighborhood distance threshold of a certain sample, and the minimum domain number is a threshold describing the number of samples in a neighborhood with a certain sample distance as the domain radius. Since the DBSCAN algorithm describes the compactness of a sample set based on a set of neighbors, its parameters minimum neighborhood number and neighborhood radius are used to describe the sample distribution compactness of the neighborhood. Therefore, by giving the minimum neighborhood point number and the neighborhood radius of the parameters, the DBSCAN algorithm can perform cluster analysis based on the minimum neighborhood point number and the neighborhood radius.
Specifically, a preset minimum neighborhood point number and a preset neighborhood radius are obtained, and each piece of original medical image data is traversed. When it is determined that the original medical image data at least contains the minimum domain point number within the domain radius of a certain original medical image data, the original medical image data is determined as core data and a cluster is established for the core data. Then, all objects in the neighborhood radius of the core data are put into a candidate set, and original medical image data which do not belong to other clusters in the candidate set are put into the established clusters. In this process, the judged data in the candidate set can be regarded as traversed data, and the data amount of the original medical image data in the data neighborhood radius needs to be further checked. And when the data quantity is at least the minimum neighborhood point number, putting the original medical image data in the neighborhood radius of the data into the candidate set. And continuing to put the data in the candidate set into the cluster until the data in the candidate set is traversed, thereby obtaining a complete cluster.
If the next cluster is needed to be obtained by clustering, the core data can be randomly determined again in the rest data, the operation is repeated, the principle is the same and is not repeated here until all the original medical image data are traversed. And finally, determining the original medical image data which are free from the clusters as abnormal data, and deleting the abnormal data to obtain abnormal-free original medical image data. In this embodiment, since the accuracy of model classification is generally reduced due to the noise data, the medical image data is subjected to data cleaning by the cluster analysis, so that the obvious abnormal medical image data can be removed, and the quality of the medical image data is improved. And clean data can be provided for subsequent correlation analysis by cleaning the medical image data, so that the accuracy of the correlation analysis is improved, and the accuracy of classification of the medical image data by the model is further improved.
In one embodiment, screening out abnormal data in the original medical image data to obtain non-abnormal medical image data includes: performing outlier factor detection on the original medical image data to obtain outlier factors corresponding to the original medical image data; and screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data.
Specifically, the original medical image data is detected through an outlier factor detection algorithm LOF (Local Outlier Factor, local anomaly factor algorithm) to obtain outlier factors, wherein the outlier factors are abnormal original medical image data. And then deleting the original medical image data represented by the outlier factors to obtain the non-abnormal medical image data.
In addition, in one embodiment, when abnormal data in the original medical image data is screened out to obtain abnormal-free medical image data, the abnormal data may be determined by adopting a clustering analysis method. When the clustering analysis determines that the original medical image data does not have abnormal data, or when the clustering analysis cannot determine the abnormal data even through clustering is performed by increasing the clustering number for a plurality of times, the outlier factor detection is performed through the LOF algorithm, and the abnormal data is obtained. In this embodiment, when the cluster analysis algorithm cannot clearly remove the abnormal data, the outlier factor detection algorithm is further used to detect and remove the abnormal data, so that the abnormal data can be accurately and effectively removed when various different types of medical image data are faced.
In one embodiment, as shown in fig. 3, the method for removing data with correlation coefficients not meeting the requirement from the non-abnormal medical image data to obtain medical image data to be inspected includes the following steps:
Step S302, any two pieces of non-abnormal medical image data are combined into a data pair.
Wherein, the data pair refers to a set comprising two different non-abnormal medical image data. Since correlation coefficients between the respective non-abnormal medical image data need to be calculated, data pairs need to be composed between every two of all the non-abnormal medical image data. Thus, the number of data pairs finally obtained is determined by the number of non-abnormal medical image data.
Specifically, after obtaining the non-abnormal medical image data, performing traversal combination on each non-abnormal medical image data to obtain corresponding data pairs. The combination of traversal is that every two of the data need to be combined, for example, assuming that there are 4 pieces of non-abnormal medical image data a, B, C and D, the pair of data obtained after the combination of traversal is (a, B) (a, C) (a, D) (B, C) (B, D) (C, D).
Step S304, calculating the correlation coefficient of two non-abnormal medical image data in any data pair.
Specifically, after the data pair is obtained, the correlation coefficient between the two non-abnormal medical image data in the data pair is calculated by using a calculation formula of the pearson correlation coefficient. For example, there are 6 data pairs such as (a, B) (a, C) (a, D) (B, C) (B, D) (C, D), and the like, and the non-abnormal medical image data in the 6 data pairs are calculated to obtain 6 corresponding correlation coefficients.
Step S306, when it is determined that the two non-abnormal medical image data in the data pair have correlation according to the correlation coefficient, deleting any one data in the data pair to obtain the medical image data to be detected.
Specifically, the correlation coefficient of the data pair is compared with a coefficient threshold value, and the data pair having correlation is determined. For example, when the coefficient threshold is 0.8, it may be determined that two non-abnormal medical image data in the data pair having the correlation equal to or greater than 0.8 have correlation. When the coefficient threshold is 0.6, it can be determined that two non-abnormal medical image data with a correlation coefficient greater than or equal to 0.6 have a correlation coefficient. It should be understood that the setting of the coefficient threshold may be set according to practical situations, and is not limited herein. Then, when the correlation between the two non-abnormal medical image data in the data pair is determined according to the coefficient threshold value, any one non-abnormal medical image data in the data pair can be deleted, and the other non-abnormal medical image data is reserved. And finally, the preserved non-abnormal medical image data is the medical image data to be detected. For example, data B is removed from the pair of data (A, B), and data A is retained. In this embodiment, since medical imaging is typically stored in PACS (Picture Archiving and Communications System medical image management and communication system) in the form of DICOM (Digital Imaging and Communications in Medicine, digital imaging and communication in medicine) image standard, the docm file in turn relies on various tag values (tags) to identify unique medical image data and maintain associations between various medical image data. Therefore, if medical image data is classified directly in the presence of various tags, the classification efficiency of the model is lowered. Therefore, before classifying the medical image data, the data with high correlation is deleted by performing correlation analysis on the data, so that the data with redundant attributes is reduced, and the burden of model classification is reduced to improve the accuracy of classification.
In addition, in one embodiment, when any at least two sets of data pairs have a correlation and contain the same non-abnormal medical image data, the same non-abnormal medical image data is deleted.
Specifically, when two sets of data pairs each have a correlation and contain identical non-abnormal medical image data, and further two sets of data that are not identical do not have a correlation, the identical non-abnormal medical image data is deleted. For example, assume that data pair (a, B) and data pair (a, C) both have a correlation, and data a is contained at the same time, while data B and data C have no correlation. To guarantee the diversity of data, data a may be deleted. However, when data B and data C have a correlation at the same time, after data a is removed, one data deletion should be arbitrarily selected between B and C. Similarly, when the pair of data (a, B), the pair of data (a, C) and the pair of data (a, D) all have correlation, and the data B, C and D do not have correlation, the data a is selected to be deleted. And the data BCD are deleted according to the correlation coefficient between them. In this embodiment, when two or more sets of data pairs have correlation and include the same medical image data, the same medical image data is directly removed, so that not only can attribute redundancy data be reduced to reduce classification burden for the model, but also the diversity of the data can be further ensured on the basis of removing the redundancy data.
In one embodiment, the medical image quality detection further comprises the step of training a classification model, the training of which should be done before the quality classification using the classification model. For example, the classification model may be trained prior to acquiring the raw medical image data.
The training classification model specifically comprises: acquiring training sample data, wherein the training sample data comprises negative sample data and positive sample data; adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data; and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
Specifically, when the classification model is trained, firstly, historical medical image data is acquired, wherein the historical medical image data is real medical image data acquired through historical film shooting. And deleting abnormal data and data with higher correlation in the historical medical image data to obtain positive sample data. Negative sample data is invalid medical image data uploaded by a user, and the negative sample data is real data obtained by manually generating data instead of shooting. Positive sample data may be considered high quality medical image data and negative sample data may be considered low quality medical image data.
The type of sample data is marked by adding a corresponding tag to the sample data. That is, a positive sample tag is added to positive sample data to identify it as high quality sample data, and a negative sample tag is added to negative sample data to identify it as low quality sample data. Positive and negative sample tags may be represented by tags that add boolean attributes to the sample data. For example, data with an attribute value of 1 is added as high-quality positive-sample data, and data with an attribute value of 0 is added as low-quality negative-sample data. After the label is added, mixing the positive sample data and the negative sample data according to a certain proportion. The invalid unbalanced data may be simulated as positive sample data 10: the ratio of negative sample data 1 was mixed. Finally, a certain proportion, for example, 70% of the mixed data is selected as training data to be input into the classification model, and the classification model is trained, so that the model has the capability of distinguishing high-quality medical image data from low-quality medical image data. And the rest 30% of data can be used for testing the classification effect of the classification model after the classification model is trained, and parameters of the classification model are further adjusted according to the tested classification result, so that the accuracy of the classification model is improved.
It should be understood that, although the steps in the flowcharts of fig. 2-3 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 2-3 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily occur sequentially, but may be performed alternately or alternately with at least a portion of the sub-steps or stages of other steps or steps.
In one embodiment, as shown in fig. 4, there is provided a medical image quality detecting apparatus, comprising: an acquisition module 402, an anomaly analysis module 404, a correlation analysis module 406, and a detection module 408, wherein:
an acquisition module 402 is configured to acquire raw medical image data.
The anomaly analysis module 404 is configured to screen out anomaly data in the original medical image data to obtain anomaly-free medical image data.
The correlation analysis module 406 is configured to remove data with correlation coefficients that do not meet the requirements from the non-abnormal medical image data, and obtain medical image data to be inspected.
The detection module 408 is configured to perform quality classification on the medical image data to be detected by using a preset classification model, and determine a medical image that meets the quality requirement.
In one embodiment, the anomaly analysis module 404 is further configured to perform cluster analysis on the raw medical image data, and screen out the anomaly data according to the analysis result to obtain anomaly-free medical image data.
In one embodiment, the anomaly analysis module 404 is further configured to obtain the number of clusters, and select, from the raw medical image data, data that is the same as the number of clusters as a cluster center; determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data; or (b)
Acquiring minimum neighborhood point number and neighborhood radius, traversing original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius; and performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
In one embodiment, the anomaly analysis module 404 is further configured to perform outlier factor detection on the raw medical image data to obtain outliers corresponding to each raw medical image data; and screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data.
In one embodiment, the correlation analysis module 406 is further configured to combine any two non-anomaly medical image data into a data pair; calculating correlation coefficients of two non-abnormal medical image data in any data pair; when the correlation coefficient is used for determining that the two non-abnormal medical image data in the data pair have correlation, deleting any one data in the data pair to obtain the medical image data to be detected.
In one embodiment, the correlation analysis module 406 is further configured to delete the same non-abnormal medical image data when any at least two data pairs have correlation and contain the same non-abnormal medical image data.
In one embodiment, the medical image quality detection apparatus further comprises a training module for acquiring training sample data, the training sample data comprising negative sample data and positive sample data; adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data; and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
For specific limitations of the medical image quality detection apparatus, reference may be made to the above limitations of the medical image quality detection method, and no further description is given here. The modules in the medical image quality detection device can be realized in whole or in part by software, hardware and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing raw medical image data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a medical image quality detection method.
It will be appreciated by those skilled in the art that the structure shown in fig. 5 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of:
acquiring original medical image data;
screening abnormal data in the original medical image data to obtain abnormal-free medical image data;
removing data with correlation coefficients not meeting the requirements from the non-abnormal medical image data to obtain medical image data to be detected;
and classifying the quality of the medical image data to be detected by using a preset classification model, and determining the medical image meeting the quality requirement.
In one embodiment, the processor when executing the computer program further performs the steps of:
and carrying out cluster analysis on the original medical image data, and screening out abnormal data according to an analysis result to obtain abnormal-free medical image data.
In one embodiment, the processor when executing the computer program further performs the steps of:
acquiring the number of clusters, and selecting data with the same number as the number of clusters from the original medical image data as a cluster center; determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data; or (b)
Acquiring minimum neighborhood point number and neighborhood radius, traversing original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius; and performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
In one embodiment, the processor when executing the computer program further performs the steps of:
performing outlier factor detection on the original medical image data to obtain outlier factors corresponding to the original medical image data; and screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data.
In one embodiment, the processor when executing the computer program further performs the steps of:
combining any two non-abnormal medical image data into a data pair; calculating correlation coefficients of two non-abnormal medical image data in any data pair; when the correlation coefficient is used for determining that the two non-abnormal medical image data in the data pair have correlation, deleting any one data in the data pair to obtain the medical image data to be detected.
In one embodiment, the processor when executing the computer program further performs the steps of:
when any at least two data pairs have correlation and contain the same non-abnormal medical image data, deleting the same non-abnormal medical image data.
In one embodiment, the processor when executing the computer program further performs the steps of:
acquiring training sample data, wherein the training sample data comprises negative sample data and positive sample data; adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data; and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring original medical image data;
screening abnormal data in the original medical image data to obtain abnormal-free medical image data;
removing data with correlation coefficients not meeting the requirements from the non-abnormal medical image data to obtain medical image data to be detected;
And classifying the quality of the medical image data to be detected by using a preset classification model, and determining the medical image meeting the quality requirement.
In one embodiment, the computer program when executed by the processor further performs the steps of:
and carrying out cluster analysis on the original medical image data, and screening out abnormal data according to an analysis result to obtain abnormal-free medical image data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring the number of clusters, and selecting data with the same number as the number of clusters from the original medical image data as a cluster center; determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data; or (b)
Acquiring minimum neighborhood point number and neighborhood radius, traversing original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius; and performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
performing outlier factor detection on the original medical image data to obtain outlier factors corresponding to the original medical image data; and screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
combining any two non-abnormal medical image data into a data pair; calculating correlation coefficients of two non-abnormal medical image data in any data pair; when the correlation coefficient is used for determining that the two non-abnormal medical image data in the data pair have correlation, deleting any one data in the data pair to obtain the medical image data to be detected.
In one embodiment, the computer program when executed by the processor further performs the steps of:
when any at least two data pairs have correlation and contain the same non-abnormal medical image data, deleting the same non-abnormal medical image data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring training sample data, wherein the training sample data comprises negative sample data and positive sample data; adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data; and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (8)
1. A medical image quality detection method, the method comprising:
acquiring original medical image data; the original medical image data are medical image data generated by medical equipment in real time or medical image data uploaded by a user in real time;
performing cluster analysis on the original medical image data, and screening out abnormal data according to an analysis result;
if no abnormal data exists, outlier factor detection is carried out on the original medical image data, and outlier factors corresponding to the original medical image data are obtained;
Screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data;
removing data with correlation coefficients not meeting requirements from the non-abnormal medical image data to obtain medical image data to be detected;
classifying the quality of the medical image data to be detected by using a preset classification model, and determining medical images meeting the quality requirement;
and when the quality classification result is that the quality of the medical image data to be detected is not in accordance with the requirement, notifying a user to rescan.
2. The method according to claim 1, wherein the performing cluster analysis on the raw medical image data, and screening out abnormal data according to the analysis result, to obtain abnormal-free medical image data, includes:
acquiring the number of clusters, and selecting data with the same number as the number of clusters from the original medical image data as a cluster center;
performing data distribution based on the distance between each clustering center and the rest of the original medical image data;
determining abnormal data according to the data quantity of each distributed clustering center, and deleting the abnormal data to obtain abnormal-free medical image data;
or (b)
Acquiring minimum neighborhood point number and neighborhood radius, traversing the original medical image data, and determining core data according to the minimum neighborhood point number and the neighborhood radius;
And performing data distribution based on the core data and the neighborhood radius, and screening out the unassigned original medical image data to obtain the non-abnormal medical image data.
3. The method according to claim 1, wherein said removing data whose correlation coefficient does not meet the requirement from the non-abnormal medical image data to obtain medical image data to be inspected comprises:
any two pieces of non-abnormal medical image data are formed into a data pair;
calculating correlation coefficients of two non-abnormal medical image data in any data pair;
and deleting any one data in the data pair to obtain medical image data to be detected when the correlation coefficient determines that the two non-abnormal medical image data in the data pair have correlation.
4. A method according to claim 3, wherein said deleting any one of said pair of data comprises:
and deleting the same non-abnormal medical image data when any at least two groups of data pairs have correlation and contain the same non-abnormal medical image data.
5. The method according to claim 1, wherein the method further comprises:
Acquiring training sample data, wherein the training sample data comprises negative sample data and positive sample data;
adding a negative sample tag to the negative sample data and a positive sample tag to the positive sample data;
and mixing the negative sample added with the negative sample label and the positive sample added with the positive sample label, inputting the mixture into a preset classification model, and training the classification model.
6. The method of claim 1 or 5, wherein the classification model pre-set includes any one or more of, but is not limited to, a decision tree model, a logistic regression model, a neural network model, and a support vector machine model.
7. A medical image quality detection apparatus, the apparatus comprising:
the acquisition module is used for acquiring original medical image data; the original medical image data are medical image data generated by medical equipment in real time or medical image data uploaded by a user in real time;
the anomaly analysis module is used for carrying out cluster analysis on the original medical image data and screening out anomaly data according to an analysis result; if no abnormal data exists, outlier factor detection is carried out on the original medical image data, and outlier factors corresponding to the original medical image data are obtained; screening out abnormal data according to each outlier factor to obtain abnormal-free medical image data;
The correlation analysis module is used for removing data with correlation coefficients not meeting requirements from the non-abnormal medical image data to obtain medical image data to be detected;
the detection module is used for classifying the quality of the medical image data to be detected by using a preset detection model, and determining medical images meeting the quality requirement;
the medical image quality detection device is also used for notifying a user to rescan when the quality classification result shows that the quality of the medical image data to be detected does not meet the requirement.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911101597.8A CN110852384B (en) | 2019-11-12 | 2019-11-12 | Medical image quality detection method, device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911101597.8A CN110852384B (en) | 2019-11-12 | 2019-11-12 | Medical image quality detection method, device and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110852384A CN110852384A (en) | 2020-02-28 |
CN110852384B true CN110852384B (en) | 2023-06-27 |
Family
ID=69600543
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911101597.8A Active CN110852384B (en) | 2019-11-12 | 2019-11-12 | Medical image quality detection method, device and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110852384B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113935896A (en) * | 2020-06-29 | 2022-01-14 | 深圳辉煌耀强科技有限公司 | Image stitching method, device, computer equipment and storage medium |
CN114343719B (en) * | 2022-03-17 | 2022-05-31 | 深圳华声医疗技术股份有限公司 | Ultrasonic imaging control method, ultrasonic imaging terminal, ultrasonic imaging apparatus, and medium |
CN115860579B (en) * | 2023-02-27 | 2023-05-09 | 山东金利康面粉有限公司 | Production quality monitoring system for flour processing |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109146705A (en) * | 2018-07-02 | 2019-01-04 | 昆明理工大学 | A kind of method of electricity consumption characteristic index dimensionality reduction and the progress stealing detection of extreme learning machine algorithm |
CN109740694A (en) * | 2019-01-24 | 2019-05-10 | 燕山大学 | A non-technical loss detection method for smart grid based on unsupervised learning |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB201012519D0 (en) * | 2010-07-26 | 2010-09-08 | Ucl Business Plc | Method and system for anomaly detection in data sets |
US11170545B2 (en) * | 2018-01-24 | 2021-11-09 | New York University | Systems and methods for diagnostic oriented image quality assessment |
CN109753991A (en) * | 2018-12-06 | 2019-05-14 | 中科恒运股份有限公司 | Abnormal data detection method and device |
CN109949268A (en) * | 2019-01-24 | 2019-06-28 | 郑州大学第一附属医院 | A machine learning-based method for grading the differentiation level of hepatocellular carcinoma |
CN110232719B (en) * | 2019-06-21 | 2021-10-26 | 腾讯科技(深圳)有限公司 | Medical image classification method, model training method and server |
-
2019
- 2019-11-12 CN CN201911101597.8A patent/CN110852384B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109146705A (en) * | 2018-07-02 | 2019-01-04 | 昆明理工大学 | A kind of method of electricity consumption characteristic index dimensionality reduction and the progress stealing detection of extreme learning machine algorithm |
CN109740694A (en) * | 2019-01-24 | 2019-05-10 | 燕山大学 | A non-technical loss detection method for smart grid based on unsupervised learning |
Non-Patent Citations (2)
Title |
---|
Precise Processing of Point Cloud Data in Omni-Directional Scanning Based on Three-Dimensional Laser Sensor;Ma, Dongling;《Journal of Nanoelectronics and Optoelectronics》;全文 * |
基于数据模式聚类算法的离群点检测;李永丽 等;《吉林大学学报(理学版)》;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN110852384A (en) | 2020-02-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10706328B2 (en) | Focus-weighted, machine learning disease classifier error prediction for microscope slide images | |
JP6402265B2 (en) | Method, computer device and storage device for building a decision model | |
RU2533500C2 (en) | System and method for combining clinical signs and image signs for computer-aided diagnostics | |
US20190102878A1 (en) | Method and apparatus for analyzing medical image | |
CN113052831B (en) | Brain medical image abnormality detection method, device, equipment and storage medium | |
CN110852384B (en) | Medical image quality detection method, device and storage medium | |
CN111931931A (en) | Deep neural network training method and device for pathology full-field image | |
JP2015087903A (en) | Apparatus and method for information processing | |
CN112602155B (en) | Systems and methods for processing and using trained models | |
CN108241853A (en) | A kind of video frequency monitoring method, system and terminal device | |
US20230252305A1 (en) | Training a model to perform a task on medical data | |
KR20230021665A (en) | Systems and methods for processing electronic images to determine salient information in digital pathology | |
KR102354476B1 (en) | Providing method and system for diagnosing lesions of bladder | |
CN116402739A (en) | Quality evaluation method and device for electronic endoscope detection flow | |
CN114067964B (en) | Medical image data processing method, device, computer equipment and storage medium | |
US12339932B2 (en) | Classification system | |
CN112102235A (en) | Human body part recognition method, computer device, and storage medium | |
CN113160199B (en) | Image recognition method and device, computer equipment and storage medium | |
CN114723723A (en) | Medical image processing method, computer device and storage medium | |
JP2024527831A (en) | System and method for image processing for image matching | |
CN112598630A (en) | Pulmonary nodule detection method, apparatus, medium, and computer program product | |
CN115861718B (en) | Gastric biopsy image classification method, apparatus, device, medium and program product | |
CN111292298A (en) | Breast cancer pathological typing determination method, device and storage medium | |
CN116416488A (en) | Picture sample screening method and device, electronic equipment and medium | |
CN114820576A (en) | Mammary gland feature extraction and detection model training method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |