Nuclear magnetic image data cleaning method and device, electronic equipment and readable storage medium

Abstract

The invention discloses a data cleaning method, and particularly relates to a nuclear magnetic image data cleaning method and device, electronic equipment and a readable storage medium, and the method comprises the steps: obtaining a to-be-processed nuclear magnetic image and a label data set of the to-be-processed nuclear magnetic image, the label data set comprises a scanning part label, a sequence label and a section label; under the condition that the label data set is not empty, obtaining scanning part information, sequence information and section information of the nuclear magnetic image to be processed based on the label data set; and under the condition that target information acquisition fails, inputting the to-be-processed nuclear magnetic image into a pre-trained deep learning model for classification processing to obtain target information, the target information including at least one of the scanned part information, the sequence information and the section information. According to the method provided by the invention, the data cleaning efficiency of the nuclear magnetic image data can be improved.

Description

Technical Field

[0001] The present invention relates to a data cleaning method, and in particular to a nuclear magnetic resonance imaging data cleaning method, device, electronic equipment and readable storage medium. Background Art

[0002] Magnetic resonance imaging data is commonly used in medical diagnosis. However, due to the lack of unified information system technical specifications, lack of medical data industry standards, incomplete data model description documents, complex internal information systems of medical institutions, the inability to interconnect, recognize and share across systems, unclear data associations between systems, uneven medical data quality leading to incomplete data, breakpoints in medical records, medical images from different equipment and manufacturers, inconsistent implementation of the DICOM international standard, and the non-standardization of DICOM image storage, it often brings great inconvenience to the retrieval, organization and use of image data.

[0003] For example, when scanning cardiac MRI images, due to non-standard processes such as equipment preset configuration, protocol parameter configuration, technician selection or self-setting, Dicom images often contain missing part information, missing sequence information, inconsistent and non-standard sequence names, missing section information, non-standard and unified section names, and missing important tag information. At the same time, the reporting system cannot be effectively aligned, resulting in a large number of cardiac MRI images being unable to be effectively retrieved through parts, sequences, sections, etc., or to distinguish cardiac MRI images from a large amount of MRI data of unknown parts and sequences, and further complete sequence classification and section classification.

[0004] In order to overcome the above problems, the method often used in the prior art is to manually check and screen to achieve the detection of cardiac magnetic resonance data, classification of sequences, and classification of sections. However, this method is time-consuming and labor-intensive, resulting in low efficiency in data cleaning. Summary of the invention

[0005] The technical problem to be solved by the present invention is that the method often adopted in the prior art is to detect cardiac magnetic resonance data, classify sequences, and classify sections through manual review and screening, but this method is time-consuming and labor-intensive, resulting in low data cleaning efficiency. In order to solve the above problems, the present invention provides a magnetic resonance imaging data cleaning method, device, electronic device and readable storage medium.

[0006] The content of the present invention includes:

[0007] In a first aspect, an embodiment of the present invention provides a method for cleaning nuclear magnetic resonance imaging data, comprising:

[0008] Obtaining the MRI images to be processed;

[0009] Acquire a label data set of the to-be-processed nuclear magnetic resonance image, wherein the label data set includes a scanning part label, a sequence label, and a section label;

[0010] When the label data set is not empty, the scanning part information, sequence information and section information of the nuclear magnetic resonance image to be processed are obtained based on the label data set; when the target information acquisition fails, the nuclear magnetic resonance image to be processed is input into a pre-trained deep learning model for classification processing to obtain the target information, and the target information includes at least one of the scanning part information, the sequence information and the section information.

[0011] Optionally, when the label data set is not empty, determining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set includes:

[0012] When the scanning part label is not empty, information matching is performed using a first regular expression, and when the matching is successful, the scanning part information is determined according to the matching result;

[0013] When the sequence tag is not empty, matching the information by using a second regular expression, and determining the sequence information according to the matching result when the match succeeds;

[0014] When the slice label is not empty, information matching is performed using a third regular expression, and when matching is successful, the slice information is determined according to the matching result.

[0015] Optionally, when the target information includes the scanning part information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0016] The nuclear magnetic resonance image to be processed is input into a pre-trained first deep learning model for binary classification processing to obtain the scanning part information, and the first deep learning model is constructed based on ResNet.

[0017] Optionally, when the target information includes the sequence information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0018] The nuclear magnetic resonance image to be processed is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information, and the second deep learning model is constructed based on DenseNet.

[0019] Optionally, when the target information includes the section information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0020] The to-be-processed nuclear magnetic resonance image is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information, and the third deep learning model is constructed based on ResNet, DenseNet or UNet.

[0021] Optionally, when the target information includes at least two of the scanning site information, the sequence information and the section information, inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information comprises:

[0022] The nuclear magnetic resonance image to be processed is input into a pre-trained fourth deep learning model for classification processing to obtain the target information.

[0023] Optionally, when the label data set is not empty, the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image are obtained based on the label data set; when the target information acquisition fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained deep learning model for classification processing to obtain the target information, including:

[0024] When the scan part label is not empty, information is matched by using a first regular expression, and when the match is successful, the scan part information is determined according to the matching result; when the match fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained first deep learning model for binary classification processing to obtain the scan part information;

[0025] When the scanned part information is a preset part and the sequence label is not empty, information matching is performed using a second regular expression, and when the matching is successful, the sequence information is determined according to the matching result, and when the matching fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information;

[0026] When the sequence information is a preset sequence and the section label is not empty, the information is matched using a third regular expression. When the match is successful, the section information is determined based on the matching result. When the match fails, the MRI image to be processed is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information.

[0027] In a second aspect, an embodiment of the present invention provides a nuclear magnetic resonance image data cleaning device, comprising:

[0028] A first acquisition module is used to acquire the nuclear magnetic resonance image to be processed;

[0029] A second acquisition module is used to acquire a label data set of the to-be-processed nuclear magnetic resonance image, wherein the label data set includes a scanning part label, a sequence label, and a section label;

[0030] A classification module is used to obtain the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set when the label data set is not empty; when the target information acquisition fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained deep learning model for classification processing to obtain the target information, and the target information includes at least one of the scanning part information, the sequence information and the section information.

[0031] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a program stored in the memory and executable on the processor; the processor is used to read the program in the memory to implement the steps in the method for cleaning nuclear magnetic resonance image data as described in the first aspect.

[0032] In a fourth aspect, an embodiment of the present invention provides a readable storage medium for storing a program, which, when executed by a processor, implements the steps in the method for cleaning nuclear magnetic resonance imaging data as described in the first aspect.

[0033] In an embodiment of the present invention, a nuclear magnetic resonance image to be processed and a label data set of the nuclear magnetic resonance image to be processed are obtained, and the label data set includes a scanning part label, a sequence label, and a section label; when the label data set is not empty, the scanning part information, sequence information, and section information of the nuclear magnetic resonance image to be processed are obtained based on the label data set; when the target information acquisition fails, the nuclear magnetic resonance image to be processed is input into a pre-trained deep learning model for classification processing to obtain target information, and the target information includes at least one of the scanning part information, sequence information, and section information. Through the nuclear magnetic resonance image data cleaning method provided in this embodiment, a complete nuclear magnetic resonance data cleaning and sorting process system is constructed, which can be applied to the cleaning of nuclear magnetic resonance image data of different parts. For example, through the nuclear magnetic resonance image data cleaning method provided in this embodiment, the automatic detection, sequence classification, and section classification of cardiac nuclear magnetic resonance data can be realized, providing an efficient, reliable, and standardized data solution for the subsequent implementation of an artificial intelligence diagnosis and treatment system for heart disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Attached Figure 1 A flow chart of a method for cleaning nuclear magnetic resonance imaging data provided by an embodiment of the present invention;

[0035] Attached Figure 2A schematic diagram of a nuclear magnetic resonance imaging data cleaning device provided by an embodiment of the present invention;

[0036] Attached Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In the embodiments of the present application, the term "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. In the embodiments of the present application, the term "multiple" refers to two or more, and other quantifiers are similar. The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first" and "second" are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more.

[0038] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0040] The present application provides a method, device, electronic device and readable storage medium for cleaning nuclear magnetic resonance image data, aiming to improve the efficiency of cleaning nuclear magnetic resonance image data. Figure 1 , Figure 1 A schematic diagram of a flow chart of a method for cleaning MRI image data provided in an embodiment of the present invention.

[0041] like Figure 1 As shown, the method specifically comprises the following steps:

[0042] Step 101, obtaining a nuclear magnetic resonance image to be processed.

[0043] Step 102: Acquire a label data set of the MRI image to be processed, wherein the label data set includes a scanning part label, a sequence label, and a section label.

[0044] Step 103, when the label data set is not empty, the scanning part information, sequence information and section information of the MRI image to be processed are obtained based on the label data set; when the target information acquisition fails, the MRI image to be processed is input into a pre-trained deep learning model for classification processing to obtain the target information, and the target information includes at least one of the scanning part information, the sequence information and the section information.

[0045] In the embodiment of the present application, the MRI image to be processed is a DICOM (Digital Imaging and Communications in Medicine) image. DICOM is an international standard for storing, transmitting and processing medical images. It defines a medical image format that can be used for data exchange and has quality that meets clinical needs. It ensures compatibility between different devices and systems, allowing medical images to be seamlessly shared within a hospital and between different hospitals.

[0046] DICOM data sets (i.e., tagged data sets) consist of multiple data elements, each of which contains specific information. A data element consists of four parts: a tag (Tag) used to identify the element type, a value representation (VR) used to describe the data type, length information used to characterize the byte length of the data element, and the actual data, i.e., the value field.

[0047] In step 101 and step 102, the nuclear magnetic resonance image to be processed and the data set of the nuclear magnetic resonance image to be processed are obtained, and the label data set of the nuclear magnetic resonance image to be processed can be obtained from the data set of the nuclear magnetic resonance image to be processed, which includes the scanning part label, sequence label and section label. As a specific embodiment, the scanning part label is Dicom Tag Bodypart_exam, and SeriesDescription refers to the image description information, which generally contains series information. In cardiac images, section information (such as 2CH, SHORT) is often included, and in images of other parts, anatomical plane information (such as ax (cross-section), cor (coronal plane), sag (sagittal plane)) is often included; the sequence and section can be obtained from the SeriesDescription description information through regular matching analysis.

[0048] In the specific implementation, the modalities in DICOM images are generally available because they are scanned by different devices, such as Magnetic Resonance (MR), Computed Tomography (CT), Ultrasonography (US), etc. Therefore, this part does not need to be classified, and the MR data can be directly filtered according to the specified Tag Modality.

[0049] In step 103, when the label data set is not empty, the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image are acquired based on the label data set.

[0050] After obtaining the MRI image to be processed, the scanning part information, sequence information and section information can be first obtained according to the tag information in the DICOM data set. In the case of missing part information, missing sequence information, inconsistent and irregular sequence names, missing section information, inconsistent and irregular section names, or missing some important tag information in the DICOM image, it is impossible to directly obtain the required information according to the tag information in the DICOM data set.

[0051] Optionally, in some embodiments, when the label data set is not empty, determining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set includes:

[0052] When the scanning part label is not empty, information matching is performed using a first regular expression, and when the matching is successful, the scanning part information is determined according to the matching result;

[0053] When the sequence tag is not empty, matching the information by using a second regular expression, and determining the sequence information according to the matching result when the match succeeds;

[0054] When the slice label is not empty, information matching is performed using a third regular expression, and when matching is successful, the slice information is determined according to the matching result.

[0055] In practical applications, when the classification result of the scanned part is non-heart, the classification results of the sequence and the section cannot be obtained. As an optional implementation, the classification of the sequence and the section can be performed after determining that the scanned part is heart.

[0056] In the case where the scanning part tag is empty or the first regular expression is not matched successfully, the target information includes the scanning part information. Optionally, in some embodiments,

[0057] In the case where the target information includes the scanning part information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0058] The nuclear magnetic resonance image to be processed is input into a pre-trained first deep learning model for binary classification processing to obtain the scanning part information, and the first deep learning model is constructed based on ResNet.

[0059] In the case where the sequence tag is empty or the second regular expression is unsuccessful in matching, the target information includes the sequence information. Optionally, in some embodiments, when the target information includes the sequence information, inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0060] The nuclear magnetic resonance image to be processed is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information, and the second deep learning model is constructed based on DenseNet.

[0061] In the case where the slice label is empty or the third regular expression is unsuccessful in matching, the target information includes the slice information. Optionally, in some embodiments, when the target information includes the slice information, inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0062] The to-be-processed nuclear magnetic resonance image is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information, and the third deep learning model is constructed based on ResNet, DenseNet or UNet.

[0063] The first deep learning model is used to perform binary classification processing to complete the classification of MRI image parts. For example, MRI data can be classified as cardiac or non-cardiac. In specific implementation, binary classification can be performed based on ResNet.

[0064] The second deep learning model is used to perform multi-classification processing, complete the sequence classification of nuclear magnetic resonance images, and obtain sequence information. In specific implementation, multi-classification can be performed based on DenseNet.

[0065] The third deep learning model is used to perform multi-classification processing or atlas segmentation processing to complete the cross-section classification of the MRI images and obtain cross-section information. In the specific implementation, multi-classification or atlas segmentation can be performed based on ResNet, DenseNet or UNet.

[0066] Optionally, in some embodiments, the MRI image to be processed is a cardiac MRI image, and cardiac atlas segmentation is performed through a third deep learning model to achieve regional segmentation of the left atrium, left ventricle, right atrium, and right ventricle, and section classification of the cardiac MRI data is achieved based on the relative position relationship of each region (including whether a certain region(s) exists in the regional segmentation results).

[0067] In some embodiments, the third deep learning model may also include multiple different sub-models, each sub-model corresponding to a sequence information, and used to perform spectrum segmentation on the data of the sequence. For example, for LGE and CINE, the third deep learning model includes two sub-models, which are used to perform spectrum segmentation on the data of the LGE sequence and the data of the CINE sequence, respectively.

[0068] In the above embodiment, three different deep learning models are used to classify the scanned parts, sequences and sections. By training the first deep learning model, the second deep learning model and the third deep learning model respectively, a better classification effect can be obtained, thereby improving the accuracy of the classification results of the scanned parts, sequences and sections.

[0069] Optionally, in some embodiments, when the target information includes at least two of the scanning site information, the sequence information and the section information, the inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0070] The nuclear magnetic resonance image to be processed is input into a pre-trained fourth deep learning model for classification processing to obtain the target information.

[0071] In this embodiment, the same model can be used to classify any two of the scan parts, sequences, and sections at the same time. Exemplarily, as an optional implementation, inputting the cardiac nuclear magnetic resonance images into the fourth deep learning model can directly classify the scan parts and sequences, and the output classification results are represented as heart-sequence. As another optional implementation, inputting the cardiac nuclear magnetic resonance images into the fourth deep learning model can directly classify the sequences and sections, and the output classification results are represented as sequence-section.

[0072] In this embodiment, the same model can be used to classify the scan part, sequence and section at the same time. For example, the cardiac MRI image can be input into the fifth deep learning model to directly classify the scan part, sequence and section, and the output classification result is represented as scan part-sequence-section (e.g., non-cardiac, cardiac-LGE-SHORT, cardiac-CINE-LONG (2CH), cardiac-CINE-LONG (4CH), etc.).

[0073] According to the foregoing content, as an optional implementation, the first deep learning model, the second deep learning model, and the third deep learning model are used in series to obtain scanning part information, sequence information, and section information in sequence. As another optional implementation, the first deep learning model and the fourth deep learning model are implemented in series, first obtaining the scanning part information, and then obtaining the sequence information and section information at one time. As another optional implementation, the scanning part information and sequence information are first obtained through the fourth deep learning model, and then the section information is obtained through the third deep learning model. As another optional implementation, the scanning part information, sequence information, and section information are directly obtained through the fifth deep learning model. In the specific implementation, the appropriate deep learning model can be pre-trained according to the actual situation. Its number and type are not limited here, and the specific details are not repeated here.

[0074] The fourth and fifth deep learning models are easy to use, but the training difficulty increases. The first, second, and third deep learning models are time-consuming to use in series, but the models are flexible to update and can be focused on separately. For example, the first deep learning model can focus on improving the detection accuracy of cardiac data, that is, giving priority to ensuring a low false positive rate, that is, all data classified as 1 (heart) have extremely high accuracy, and data classified as 0 (non-heart) are acceptable even if there are some false negatives.

[0075] As a specific embodiment, step 103 includes:

[0076] When the scan part label is not empty, information is matched by using a first regular expression, and when the match is successful, the scan part information is determined according to the matching result; when the match fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained first deep learning model for binary classification processing to obtain the scan part information;

[0077] When the scanned part information is a preset part and the sequence label is not empty, information matching is performed using a second regular expression, and when the matching is successful, the sequence information is determined according to the matching result, and when the matching fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information;

[0078] When the sequence information is a preset sequence and the section label is not empty, the information is matched using a third regular expression. When the match is successful, the section information is determined based on the matching result. When the match fails, the MRI image to be processed is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information.

[0079] For ease of understanding, the following description is made by taking the to-be-processed nuclear magnetic resonance image as a cardiac nuclear magnetic resonance image as an example.

[0080] As a specific embodiment, first check whether Dicom Tag Bodypart_exam (i.e., the scan part tag) is not empty. If it is not empty, the information is matched through the first regular expression. If successful, the scan part information can be obtained to distinguish non-cardiac and cardiac. If Dicom Tag Bodypart_exam is empty or the regular expression fails, the cardiac nuclear magnetic resonance image needs to be input into the pre-trained first deep learning model for classification processing to obtain the scan part information. If the classification result is the heart (i.e., the preset part), continue to the next step of the inspection.

[0081] Check whether the Dicom Tag SeriesDescription (i.e., sequence tag) is not empty. If it is not empty, match the information through the second regular expression. If successful, the sequence information can be obtained to distinguish the sequence classification of the Dicom image: for example, LGE, CINE, ..., unknown. If the Dicom Tag SeriesDescription is empty or the regular expression fails (including unknown), the cardiac MRI image needs to be input into the pre-trained second deep learning model for classification processing to obtain the sequence information, which is divided into LGE, CINE, ..., and others. If it is classified as a preset sequence (if the subsequent business requires LGE and / or CINE sequences), continue to the next step of the inspection.

[0082] Check whether the Dicom Tag SeriesDescription (i.e., the slice label) is non-empty. If it is non-empty, match the information through the third regular expression. If successful, the slice information can be obtained to distinguish the slice classification of the Dicom image: for example, SHORT, LONG (2CH), LONG (3CH), LONG (4CH), ..., unknown. If the Dicom Tag SeriesDescription is empty or the regular expression fails (including unknown), as an optional implementation, the cardiac MRI image is input into a pre-trained third deep learning model for classification processing to obtain the slice information, which is divided into SHORT, LONG (2CH), LONG (3CH), LONG (4CH), ..., and others. As another optional implementation, the cardiac MRI image is input into a pre-trained third deep learning model for atlas segmentation. Different sections have different angles and positions, so there are typical differences in the cardiac atlas segmentation results. Some sections may not even contain certain regions. According to the clinical characteristics, we summarize and construct the section classification logic based on the relative position relationship of cardiac atlas segmentation (including whether certain regions exist in the regional segmentation results), which are divided into SHORT, LONG (2CH), LONG (3CH), LONG (4CH), ..., and others.

[0083] Through the nuclear magnetic resonance imaging data cleaning method provided in this embodiment, a complete nuclear magnetic resonance data cleaning and sorting process system is constructed, which can be applied to the cleaning of nuclear magnetic resonance imaging data of different parts. For example, through the nuclear magnetic resonance imaging data cleaning method provided in this embodiment, the automatic detection, sequence classification, and section classification of cardiac nuclear magnetic resonance data can be realized, providing an efficient, reliable, and standardized data solution for the subsequent realization of an artificial intelligence diagnosis and treatment system for heart disease.

[0084] The detection of cardiac MRI data, the classification of sequences, and the classification of sections are the basis for the development of subsequent tasks, such as the segmentation of cardiac maps, the calculation of cardiac parameters, and the diagnosis of diseases, which all need to be performed on specified sections. The method provided in the embodiment of the present application overcomes the problem of missing important tag information on MRI imaging data and being unable to complete data cleaning and sorting. It can extract data from unknown parts and complete the classification of multiple sequences and multiple sections. It is particularly suitable for cleaning cardiac MRI data and can achieve better results.

[0085] like Figure 2 As shown, an embodiment of the present invention provides a nuclear magnetic resonance image data cleaning device 200, comprising:

[0086] The first acquisition module 201 is used to acquire the nuclear magnetic resonance image to be processed;

[0087] The second acquisition module 202 is used to acquire a label data set of the to-be-processed nuclear magnetic resonance image, wherein the label data set includes a scanning part label, a sequence label, and a section label;

[0088] The classification module 203 is used to obtain the scanning part information, sequence information and section information of the nuclear magnetic resonance image to be processed based on the label data set when the label data set is not empty; when the target information acquisition fails, the nuclear magnetic resonance image to be processed is input into a pre-trained deep learning model for classification processing to obtain the target information, and the target information includes at least one of the scanning part information, the sequence information and the section information.

[0089] Optionally, when the label data set is not empty, determining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set includes:

[0090] When the scanning part label is not empty, information matching is performed using a first regular expression, and when the matching is successful, the scanning part information is determined according to the matching result;

[0091] When the sequence tag is not empty, matching the information by using a second regular expression, and determining the sequence information according to the matching result when the match succeeds;

[0092] When the slice label is not empty, information matching is performed using a third regular expression, and when matching is successful, the slice information is determined according to the matching result.

[0093] Optionally, when the target information includes the scanning part information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0094] The nuclear magnetic resonance image to be processed is input into a pre-trained first deep learning model for binary classification processing to obtain the scanning part information, and the first deep learning model is constructed based on ResNet.

[0095] Optionally, when the target information includes the sequence information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0096] The nuclear magnetic resonance image to be processed is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information, and the second deep learning model is constructed based on DenseNet.

[0097] Optionally, when the target information includes the section information, the step of inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes:

[0098] The to-be-processed nuclear magnetic resonance image is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information, and the third deep learning model is constructed based on ResNet, DenseNet or UNet.

[0099] Optionally, when the target information includes at least two of the scanning site information, the sequence information and the section information, inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information comprises:

[0100] The nuclear magnetic resonance image to be processed is input into a pre-trained fourth deep learning model for classification processing to obtain the target information.

[0101] Optionally, when the label data set is not empty, the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image are obtained based on the label data set; when the target information acquisition fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained deep learning model for classification processing to obtain the target information, including:

[0102] When the scan part label is not empty, information is matched by using a first regular expression, and when the match is successful, the scan part information is determined according to the matching result; when the match fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained first deep learning model for binary classification processing to obtain the scan part information;

[0103] When the scanned part information is a preset part and the sequence label is not empty, information matching is performed using a second regular expression, and when the matching is successful, the sequence information is determined according to the matching result, and when the matching fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information;

[0104] When the sequence information is a preset sequence and the section label is not empty, the information is matched using a third regular expression. When the match is successful, the section information is determined based on the matching result. When the match fails, the MRI image to be processed is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information.

[0105] The nuclear magnetic resonance image data cleaning device 200 provided in the embodiment of the present application can execute the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated in this embodiment.

[0106] It should be noted that the division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0107] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0108] like Figure 3 As shown, an embodiment of the present application provides an electronic device 300, including: a memory 302, a processor 301, and a program stored in the memory 302 and executable on the processor 301; the processor 301 is used to read the program in the memory 302 to implement the steps in the magnetic resonance imaging data cleaning method as described above.

[0109] The embodiment of the present application also provides a readable storage medium, on which a program is stored. When the program is executed by a processor, the various processes of the above-mentioned nuclear magnetic resonance image data cleaning method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here. Among them, the readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical storage (such as compact disk (CD), digital video disc (DVD), Blu-ray Disc (BD), high-definition versatile disc (HVD), etc.), and semiconductor memory (such as read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read only memory (EEPROM), non-volatile memory (NAND FLASH), solid-state drive (Solid State Disk or Solid State Drive, SSD)), etc.

[0110] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0111] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), and includes a number of instructions for a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0112] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for cleaning nuclear magnetic resonance imaging data, characterized in that: include: Obtaining the MRI images to be processed; Acquire a label data set of the to-be-processed nuclear magnetic resonance image, wherein the label data set includes a scanning part label, a sequence label, and a section label; When the label data set is not empty, obtaining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set; In the event that the target information cannot be obtained, the MRI image to be processed is input into a pre-trained deep learning model for classification processing to obtain the target information, which includes at least one of the scanning site information, the sequence information and the section information.

2. The method according to claim 1, characterized in that: When the label data set is not empty, determining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set includes: When the scanning part label is not empty, information matching is performed using a first regular expression, and when the matching is successful, the scanning part information is determined according to the matching result; When the sequence tag is not empty, matching the information by using a second regular expression, and determining the sequence information according to the matching result when the match succeeds; When the slice label is not empty, information matching is performed using a third regular expression, and when matching is successful, the slice information is determined according to the matching result.

3. The method according to claim 1, wherein When the target information includes the scanning part information, the inputting of the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes: The nuclear magnetic resonance image to be processed is input into a pre-trained first deep learning model for binary classification processing to obtain the scanning part information, and the first deep learning model is constructed based on ResNet.

4. The method according to claim 1, wherein When the target information includes the sequence information, the inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes: The nuclear magnetic resonance image to be processed is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information, and the second deep learning model is constructed based on DenseNet.

5. The method according to claim 1, wherein When the target information includes the section information, the inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes: The to-be-processed nuclear magnetic resonance image is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information, and the third deep learning model is constructed based on ResNet, DenseNet or UNet.

6. The method according to claim 1, wherein When the target information includes at least two of the scanning part information, the sequence information and the section information, the inputting the to-be-processed nuclear magnetic resonance image into a pre-trained deep learning model for classification processing to obtain the target information includes: The nuclear magnetic resonance image to be processed is input into a pre-trained fourth deep learning model for classification processing to obtain the target information.

7. The method according to claim 1, characterized in that: When the label data set is not empty, obtaining the scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set; In the case of failure to obtain target information, the to-be-processed nuclear magnetic resonance image is input into a pre-trained deep learning model for classification processing to obtain the target information, including: When the scan part label is not empty, information is matched by using a first regular expression, and when the match is successful, the scan part information is determined according to the matching result; when the match fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained first deep learning model for binary classification processing to obtain the scan part information; When the scanned part information is a preset part and the sequence label is not empty, information matching is performed using a second regular expression, and when the matching is successful, the sequence information is determined according to the matching result, and when the matching fails, the to-be-processed nuclear magnetic resonance image is input into a pre-trained second deep learning model for multi-classification processing to obtain the sequence information; When the sequence information is a preset sequence and the section label is not empty, the information is matched using a third regular expression. When the match is successful, the section information is determined based on the matching result. When the match fails, the MRI image to be processed is input into a pre-trained third deep learning model for multi-classification processing or atlas segmentation processing to obtain the section information.

8. A nuclear magnetic resonance imaging data cleaning device, characterized in that: include: A first acquisition module is used to acquire the nuclear magnetic resonance image to be processed; A second acquisition module is used to acquire a label data set of the to-be-processed nuclear magnetic resonance image, wherein the label data set includes a scanning part label, a sequence label, and a section label; A classification module, configured to obtain scanning part information, sequence information and section information of the to-be-processed nuclear magnetic resonance image based on the label data set when the label data set is not empty; In the event that the target information cannot be obtained, the MRI image to be processed is input into a pre-trained deep learning model for classification processing to obtain the target information, which includes at least one of the scanning site information, the sequence information and the section information.

9. An electronic device, comprising: A memory, a processor, and a program stored in the memory and executable on the processor; wherein the processor is used to read the program in the memory to implement the steps in the method for cleaning nuclear magnetic resonance imaging data as described in any one of claims 1 to 7.

10. A readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the steps in the method for cleaning nuclear magnetic resonance image data as described in any one of claims 1 to 7 are implemented.

Images