A method, system, and storage medium for image data processing based on MRI technology

Abstract

This invention discloses an image data processing method, system, and storage medium based on MRI technology, relating to the field of medical data processing technology. The image data processing method includes: scanning and processing a phantom using several MRI imaging devices to obtain quantitative parameters of the phantom output by each MRI imaging device, constructing a first quantitative parameter vector; using the first quantitative parameter vector to construct a calibration benchmark value, and establishing a mapping relationship between the device deviation vector of each MRI imaging device and the calibration benchmark value; associating each mapping relationship with an identity tag of the MRI imaging device; reading MRI image data obtained after scanning a patient with the current MRI imaging device, extracting the identity tag and quantitative parameters, converting the quantitative parameters into a second quantitative parameter vector, and, based on the identity tag, calling the corresponding mapping relationship to standardize and output the second quantitative parameter vector. This method can standardize image data output by different imaging devices.

Description

Technical Field

[0001] This invention belongs to the field of medical data processing technology, specifically relating to an image data processing method, system, and storage medium based on MRI technology. Background Technology

[0002] Multiparameter functional magnetic resonance imaging (MRI) technology has advantages such as being non-invasive, radiation-free, multi-sequence, multi-parameter, and rich in functional sequences. It can provide more quantitative information related to cellular function and metabolism while reflecting qualitative imaging characteristics. Therefore, it is one of the main means of non-invasively obtaining chemical substances or other medical data in the body.

[0003] Autoimmune liver disease (AILD) is a group of hepatobiliary inflammatory diseases mediated by abnormal autoimmunity. It manifests as autoimmune damage to liver tissue, leading to liver fibrosis and cirrhosis, ultimately resulting in parenchymal liver damage and liver failure. The pathogenesis of AILD is not yet fully understood, and patients often lack specific clinical symptoms in the early stages, making misdiagnosis and missed diagnosis highly likely. Without timely and effective treatment, the disease rapidly progresses to liver fibrosis, eventually leading to cirrhosis and end-stage liver disease, resulting in high medical costs and a high mortality rate.

[0004] Currently, clinical assessment of AILD mainly relies on liver biopsy pathology results, but this procedure is invasive and has limitations, and cannot fully reflect the pathophysiological changes of the entire liver caused by immune damage.

[0005] Based on this, the inventors applied multiparameter functional magnetic resonance imaging (fMRI) in AILD to design a non-invasive, comprehensive, and quantitative assessment method for evaluating the pathological and physiological changes in liver tissue during the development of liver fibrosis and cirrhosis in AILD patients. Therefore, the inventors needed to obtain a large amount of comprehensive imaging data of AILD patients based on multiparameter fMRI, extract quantitative data on various chemical substances characterizing AILD, and construct an individualized, early, non-invasive, precise, stratified prognostic prediction model for AILD based on this data.

[0006] Currently, the inventors have encountered the following problems in this process: It is necessary to collect a large amount of comprehensive imaging data of AILD patients. Even if the imaging data is collected in the same hospital, the same hospital cannot guarantee that each MRI imaging device is the same model and has the same parameters. This leads to deviations in the quantitative analysis of chemical substances in the collected imaging data. These differences in data will lead to insufficient accuracy of the mathematical model when used to establish a personalized early non-invasive precise stratified prognostic prediction model for AILD.

[0007] Currently, to ensure the standardization of data from different imaging devices, phantom calibration is generally used. The phantom calibration method involves scanning the same phantom on different devices, comparing the measurement results from each device with the phantom's standard values, and constructing a calibration curve based on the difference to quantify the error of each imaging device. Subsequent image data acquired on that device is then processed using the calibration curve to obtain quantitative parameters that more closely approximate the actual data. This method aims to make all quantitative parameters as close as possible to the true data. In summary, the phantom method is primarily used to make the measured values ​​as close as possible to the true values.

[0008] However, there are certain difficulties in using the phantom method for the standardized processing of comprehensive liver MRI imaging data. This is because an ideal liver phantom needs to simultaneously simulate multiple tissue characteristics under normal or pathological conditions. Currently, there is no mature, dedicated single liver phantom that can provide simulation of multiple quantitative parameters. Summary of the Invention

[0009] The purpose of this application is to provide an image data processing method, system and storage medium based on MRI technology, which uses a phantom to collect quantitative parameter data of the liver to construct a mapping relationship between calibration benchmark values ​​and equipment deviation, thereby using this relationship to calibrate human liver image data.

[0010] To achieve the above objectives, the solution adopted in this application is as follows:

[0011] An image data processing method based on MRI technology is implemented as follows:

[0012] Several MRI imaging devices were used to scan and process images of a simulated liver phantom, resulting in quantitative parameters of the simulated liver phantom output by each MRI imaging device. Each MRI imaging device performed the following scans on the simulated liver phantom: ADC value obtained by DWI imaging; f value, D* value, and D value obtained by T1 mapping imaging; R2* value and PDFF value obtained by IDEAL-IQ imaging; and MRE value obtained by MRE imaging. Quantitative parameters included ADC value, f value, D* value, D value, R2* value, PDFF value, and MRE value.

[0013] Quantitative parameters of a simulated liver phantom output from several MRI imaging devices are obtained, and the quantitative parameters of the phantom are constructed into a first quantitative parameter vector;

[0014] The calibration baseline value is constructed using the first quantitative parameter vector, and a mapping relationship is established between the device deviation vector of each MRI imaging device and the calibration baseline value; each mapping relationship is associated with the identity label of the MRI imaging device.

[0015] Read the MRI image data of the liver obtained after scanning the human body with the current MRI imaging equipment, extract the identity tag and the quantitative parameters of the human liver, construct the quantitative parameters of the human liver into a second quantitative parameter vector, and output the second quantitative parameter vector after standardizing and calibrating it based on the identity tag by calling the corresponding mapping relationship.

[0016] MRI imaging data includes the identification label of the MRI imaging device, quantitative parameters, and a list of quantitative parameter types.

[0017] As one specific implementation scheme, there are several phantoms simulating the liver, each providing different quantitative parameters during MRI imaging. The first quantitative parameter vector includes the quantitative parameters that the phantom can provide; the vector structure of the equipment deviation includes all the quantitative parameters that need to be collected.

[0018] As a specific implementation scheme, the process of constructing calibration reference values ​​using a first quantitative parameter vector and establishing the mapping relationship between the equipment deviation vector of each MRI imaging device and the calibration reference values ​​is as follows:

[0019] Step S10: Establish a generation model for the first quantitative parameter vector, which reflects the relationship between the quantitative parameters and the true values, equipment deviation, batch drift, and random errors;

[0020] Step S20: Construct the objective function, obtain the estimated value of the equipment deviation vector and the estimated value of the true value through inversion calculation, and use the estimated value of the true value obtained by inversion calculation as the calibration benchmark value;

[0021] Step S30: Solve the objective function iteratively by alternating coordinate descent until it converges, and obtain the mapping relationship between the equipment deviation vector and the calibration reference value.

[0022] As a specific implementation scheme, in step S10, the first quantitative parameter vector The generative model is as follows:

[0023] ;

[0024] ;

[0025] In the formula, i represents the i-th MRI imaging device. ;j represents the j-th scan, ;t represents the t-th type of motif; Represents the actual value; Indicates batch drift; This represents the random error of a single scan; The selection matrix is ​​represented by multiplication from the device deviation vector. Extract the quantitative parameter components corresponding to the t-th type of phantom. ; N represents the dimension of all quantitative parameters to be collected; dt represents the dt quantitative parameters of phantom t.

[0026] Represents the device deviation vector It follows a mean of 0 and a prior covariance of . The normal distribution; Represents batch drift vector It follows a mean of 0 and a drift covariance of . The normal distribution; This indicates that the true value follows the mean. Covariance is The normal distribution; Represents random noise It follows a mean of 0 and a noise covariance of . It follows a normal distribution.

[0027] As a specific implementation scheme, the objective function in step S20 is as follows:

[0028] ;

[0029] In the formula, Indicates the total type of the motif; For likelihood terms; and It is a priori; Indicates the calibration reference value. This represents the estimated value of the deviation vector for each device.

[0030] As a specific implementation scheme, the process of solving the objective function through alternating coordinate descent iteration in step S30 is as follows:

[0031] Fix equipment deviation and update calibration reference value:

[0032] ;

[0033] Fix the calibration reference value and update the equipment deviation:

[0034] ;

[0035] In the formula, This represents the estimated batch drift value; Indicates the total number of imaging devices; This indicates the number of scans performed on the t-type motif.

[0036] As a specific implementation plan, the process of standardizing and transforming the second quantitative parameter vector is as follows:

[0037] ;

[0038] In the formula, Represents the second quantitative parameter vector. This represents a dynamic selection matrix. This represents the second quantitative parameter vector after standardization.

[0039] As one specific implementation scheme, the process of generating the dynamic selection matrix is ​​as follows:

[0040] A list of quantitative parameter types for identifying MRI image data;

[0041] Generate a zero matrix with the same number of columns as the total number of dimensions of all quantitative parameters to be collected and the same number of rows as the number of quantitative parameter types in the quantitative parameter type list;

[0042] For the r-th quantitative parameter in the list of quantitative parameter types, query its predefined index column, modify the element 0 in the r-th row and corresponding index column of the all-zero matrix to 1, and obtain the dynamic selection matrix.

[0043] A system for implementing the above method includes a mapping relationship construction module, a relational database, a data reading module, a standardization processing module, and a dynamic selection matrix generation module;

[0044] The mapping relationship construction module constructs the quantitative parameters of the phantom into a first quantitative parameter vector; it uses the first quantitative parameter vector to construct calibration reference values ​​and establishes a mapping relationship between the equipment deviation vector of each MRI imaging device and the calibration reference values;

[0045] The relational database stores the mapping relationship between the equipment deviation vectors of each imaging device and the calibration reference values;

[0046] The data reading module reads MRI image data, including MRI image device identification tags, quantitative parameters, and a list of quantitative parameter types; and constructs the quantitative parameters into a second quantitative parameter vector.

[0047] The dynamic selection matrix generation module generates a dynamic selection matrix based on the quantitative parameter types in the quantitative parameter type list and their corresponding predefined index columns.

[0048] The standardization processing module uses the identification tag of the MRI imaging device to call up the mapping relationship between the device deviation vector and the calibration benchmark value to calibrate the second quantitative parameter vector.

[0049] A storage medium having a computer program stored thereon, which, when executed by a processor, implements the image data processing method described above.

[0050] The technical solution of this application has at least the following advantages and beneficial effects:

[0051] In this invention, the processed liver imaging data is primarily used for machine learning to construct an AILD hierarchical predictive prognostic model. For machine learning, the consistency of data distribution is more important than data accuracy. The model learns the mapping relationship from input to output; if the input scale is inconsistent, the model will confuse differences between imaging devices with disease signals. Therefore, although a phantom is introduced in the standardization process, this invention does not pursue absolute approximation to the true value. Instead, it aims for relative consistency between devices, inverting to obtain a true value estimate with consensus across all phantoms. This estimate serves as a calibration benchmark, which quantifies device errors, thereby enabling the standardization of MRI imaging data from several different devices. Furthermore, calibration is performed using variance shifting, ensuring data scale uniformity while preserving the individual biological heterogeneity in the patient's scanned imaging data.

[0052] The interference in the quantitative parameters obtained from phantom scanning is not limited to equipment deviation, but also includes batch offset and random error of the phantom. In this invention, batch offset noise interference and random error of the phantom are removed, and the equipment deviation is accurately extracted. (See attached figures for details.)

[0053] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0054] Figure 1 This is a flowchart of the method described in this invention;

[0055] Figure 2 This is a system module diagram of the system described in this invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] According to an embodiment of the present invention, an image data processing method based on MRI technology is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0058] This embodiment provides an image data processing method based on MRI technology. Figure 1 This is a flowchart of an MRI-based image data acquisition method according to an embodiment of the present invention. Its core technical means is: to construct calibration reference values ​​based on the scanning phantoms of each MRI imaging device, and to establish a mapping relationship between the device deviation of each MRI imaging device and the calibration reference values, and to standardize the patient data acquired by the MRI imaging device based on the mapping relationship.

[0059] This method is currently being used at Hospital A, which has MRI imaging equipment at both its main campus and branch campuses. Some of these devices differ in model, specifications, and parameters. Hardware data for each MRI imaging device can be directly obtained from Hospital A.

[0060] Specifically, the MRI imaging techniques used for each AILD patient will be explained as follows:

[0061] 1. Diffusion-weighted imaging (DWI):

[0062] DWI Technology Principle: DWI is an MRI technique that displays the microscopic movement of water molecules within tissues. The imaging principle of DWI is based on the Brownian motion of water molecules. Specifically, under normal circumstances, water molecules in the body diffuse freely and randomly; however, in certain pathological conditions, such as cell swelling or tumor growth, water molecule diffusion is restricted. By applying a specific gradient magnetic field pulse sequence, these minute differences in motion are detected, and images are generated accordingly, reflecting the speed and extent of water molecule diffusion in different regions.

[0063] Scanning Methods: (1) Pre-scan preparation: Patients need to prepare according to standard procedures, such as removing metal objects. (2) Localization: Use conventional T1 or T2 weighted sequences to determine the scanning range and region of interest. (3) Data acquisition: Select appropriate b-values. For the liver, at least two different b-values ​​are usually used (e.g., 0 s / mm² and 500 s / mm²) to obtain sufficient contrast and calculate the ADC map. Use single-shot spin-echo (EPI) technology or other fast imaging techniques for data acquisition. (4) Image post-processing: Use software to convert the collected information into visualized DWI images and corresponding ADC maps.

[0064] Collected parameters: (1) b value, which is a parameter for measuring diffusion sensitivity. A higher b value means that it is more sensitive to the diffusion of water molecules. (2) Apparent Diffusion Coefficient (ADC): It is calculated from data obtained from multiple different b values ​​and directly reflects the size of the water diffusion capacity in the tissue. The lower the ADC value, the more severe the diffusion restriction.

[0065] Measurement methods: Qualitative analysis: Observe the location, morphology, and other characteristics of high-signal areas on DWI images. Quantitative analysis: Select regions of interest (ROIs) on the ADC map and measure their average ADC values. The measurement of ROIs depends on the study, and ROIs are placed in different liver segments, i.e., eight ROIs (liver segments S1-S8) are placed on the obtained MRI images; their average ADC values ​​are calculated, and the difference in ADC values ​​between them and normal liver tissue is compared.

[0066] 2. T1 mapping imaging:

[0067] T1 mapping technology principle: Liver T1 mapping is a non-invasive MRI technique that quantitatively measures the longitudinal relaxation time (T1) of tissues. It can indirectly reflect changes in the microstructure and composition of the liver and is of great value in assessing liver fibrosis, cirrhosis, and hepatic steatosis.

[0068] Scanning methods: (1) Preparation stage: Ensure the patient is prepared according to the standard procedure and all metal objects are removed. (2) Localization scan: Use conventional T1-weighted or T2-weighted sequences to determine the location and extent of the liver. (3) Select an appropriate T1-mapping sequence for scanning, and adjust the scanning parameters according to the patient's tolerance, respiratory rate, etc. Use respiratory gating technology or have the patient hold their breath to reduce respiratory artifacts and ensure that the acquired patient images are short in time and of high quality. (4) Image post-processing: Use specialized software to analyze the acquired data and generate a T1 map. The T1 map displays the T1 value corresponding to each pixel and is usually color-coded.

[0069] Parameter acquisition and measurement methods: Regions of interest (ROIs) are selected on the T1 map, typically lesion areas and normal liver parenchyma as controls. T1 values ​​are measured and the average T1 value within the selected ROI is recorded. The measured T1 values ​​are compared with reference ranges for normal liver or specific pathological conditions to assess the liver's condition. Regions of interest (ROIs) are selected on the color map, and their average T1 values ​​are measured. Specifically, four ROIs (left lateral lobe, left medial lobe, right anterior lobe, and right posterior lobe) are placed on the same level of the MRI image obtained from the scan. Their average T1 values ​​are calculated and compared with those of normal liver tissue.

[0070] 3. Intravoxel Incoherent Motion (IVIM):

[0071] IVIM Technology Principle: IVIM technology is an extension of DWI imaging, capable of simultaneously providing information on water molecule diffusion and microvascular perfusion in tissues. The application of IVIM in liver imaging can help assess the microcirculatory status of the liver parenchyma and detect early changes in liver disease.

[0072] Scanning Method: (1) Pre-scan preparation: Ensure the patient is not carrying any metal objects and prepare according to the standard MRI procedure. (2) Localization scan: Use conventional T1 or T2 weighted sequences to determine the location and extent of the liver. (3) Select an appropriate IVIM sequence. (4) Perform multiple scans using multiple b-values. The b-values ​​to be acquired are: 25, 50, 75, 100, 150, 200, 400, 600, 800, and 1000. The number of repeated excitations are: 1, 1, 1, 1, 1, 2, 2, 4, 6, and 8. Use respiratory gating technology or have the patient hold their breath during the scan to reduce respiratory motion artifacts.

[0073] Data acquisition parameters: The acquired data were analyzed using specialized software to fit the above-mentioned double exponential model; the f, D*, and D values ​​of each voxel were calculated.

[0074] Measurement methods: (1) ROI selection: Select an ROI on the generated IVIM parameter map, such as the normal liver parenchyma area or the lesion area; (2) Quantitative measurement: Record the average f, D* and D values ​​within the selected ROI. That is, place eight ROIs (liver S1-S8 segments) on the obtained IVIM image; calculate their average parameter values ​​and compare their differences with normal liver tissue.

[0075] 4. IDEAL-IQ Imaging:

[0076] IDEAL-IQ Technology Principle: IDEAL-IQ can simultaneously separate and quantify the water, fat, and mineral (iron) components in tissues. This technology is particularly suitable for liver imaging, providing accurate information about the fat and iron content within the liver. This is crucial for diagnosing non-alcoholic fatty liver disease (AILD), hepatitis B-related liver fibrosis, evaluating liver transplant donors, and monitoring disease progression. Studies have shown that AILD patients, lacking timely and effective treatment or with poor treatment response, rapidly develop liver fibrosis and cirrhosis, a process also accompanied by changes in the chemical composition of liver tissue. IDEAL-IQ is based on multi-echo gradient echo sequences and utilizes echo signals from different phases to distinguish water, fat, and iron signals. This technology uses six different echo time points to acquire data, allowing for the acquisition of pure water images, pure fat images, and other relevant quantitative parameters such as R2* (reflecting iron deposition) and proton density fat fraction (PDFF) in a single scan.

[0077] Acquisition parameters: (1) Water-lipid separation: By analyzing the signal changes under different TE (Echo Time), IDEAL-IQ can effectively distinguish water and fat signals. (2) Fat quantification: By calculating the proton density fat fraction (PDFF), the percentage of liver fat content can be directly obtained. (3) Iron quantification: R2* spectrum can be used to assess the iron deposition in the liver.

[0078] Scanning Method: (1) Pre-scan preparation: Ensure the patient is not carrying any metal objects and prepare according to the standard MRI examination procedure. (2) Localization scan: Use conventional T1 or T2 weighted sequences to determine the location and extent of the liver. (3) Select a specific IDEAL-IQ scan sequence, and combine it with respiratory gating or have the patient hold their breath to reduce motion artifacts. Finally, perform the scan under the set parameters to obtain raw image data corresponding to multiple TE values. (4) Post-processing: Use specialized software to process the acquired data to generate water images, fat images, PDFF images, and R2* images, etc.

[0079] Measurement methods: (1) ROI selection: Select ROIs on the generated IDEAL-IQ image. In this invention, eight ROIs (segments S1-S8 of the liver) are placed on the scanned IVIM image. Calculate their average parameter values ​​and compare their differences with normal liver tissue. (2) Quantitative analysis: Record the average PDFF value within the selected ROI, which is a key indicator for measuring liver fat content; record the R2* value to assess iron deposition, which is a key indicator for measuring intrahepatic iron content. (3) Compare the differences between different regions, such as the difference between healthy tissue and diseased tissue.

[0080] (5) Magnetic Resonance Elastography (MRE):

[0081] MRE (Mechanical Resonance Imaging) Technology Principle: The basic principle of MRE is to introduce mechanical vibration waves into the body and use MRI technology to detect the propagation of these waves within the tissue. Because different tissues have different elastic properties, the propagation speed of the vibration waves will also vary. By analyzing these wave patterns, the shear modulus of the tissue can be calculated, thus indirectly reflecting the tissue stiffness. This technology combines MRI and mechanical principles to measure the stiffness or elasticity of liver tissue. The liver stiffness value measured by this technology can assess the grade of liver fibrosis and provides more accurate and direct information than traditional serological markers.

[0082] Scanning Methods: (1) Pre-scan preparation: Patients need to prepare according to standard MRI examination procedures, including removing all metal objects. Sometimes fasting may be required to reduce the impact of gastrointestinal gas on image quality. (2) Localization scan: Use conventional T1 or T2 weighted sequences to determine the location and extent of the liver; determine the optimal vibration transmission path, usually selecting a flat area of ​​the abdomen with thin muscles. (3) Vibrator installation: Fix the vibrator in the selected skin position to ensure that the vibration can be effectively transmitted to the liver. Post-processing: Use specialized software to process the acquired data, generate shear modulus maps or elasticity maps, and analyze the obtained images to quantify the stiffness of the liver. Measurement methods:

[0083] (1) ROI selection: Select ROIs on the generated elastogram and place eight ROIs (segments S1-S8 of the liver) on the scanned MRE image; calculate their average parameter values ​​and compare their differences with normal liver tissue. (2) Quantitative measurement: Record the average shear modulus value within the selected ROI and compare the differences between different regions, such as the difference between healthy tissue and diseased tissue. (3) According to the clinical guidelines and laboratory standards for liver fibrosis, compare the measured shear modulus value MRE with a specific threshold to assess the degree of fibrosis.

[0084] Example

[0085] The first aspect of the present invention provides an image data processing method based on MRI technology, the implementation process of which is as follows:

[0086] Step S100: The phantom is scanned and image processed using several MRI imaging devices to obtain quantitative parameters of the phantom output by several MRI imaging devices, and the quantitative parameters are used to construct a first quantitative parameter vector.

[0087] Step S200: Construct a mapping relationship between the calibration benchmark value and the equipment deviation vector of each MRI imaging device using the first quantitative parameter vector, forming a relationship database; each mapping relationship is associated with the identity tag of the MRI imaging device;

[0088] Step S300: Read the MRI image data obtained after scanning the patient with the current MRI imaging equipment, extract the identity label and quantitative parameters, construct the quantitative parameters into a second quantitative parameter vector, and based on the identity label, call the corresponding mapping relationship to standardize and output the second quantitative parameter vector;

[0089] MRI image data includes the identification label of the MRI imaging device that outputs the data, quantitative parameters, and a list of quantitative parameter types.

[0090] This embodiment provides information on phantoms used in specific application scenarios: a CaliberMRI phantom is used to scan and detect T1, T2, and ADC values; D, D*, and f values ​​can be obtained by fitting the CaliberMRI DWI signal; a CIRS MRE phantom is used to scan and detect MRE values; and a self-made phantom is used to measure PDFF and R2* values. For PDFF value measurement, a self-made fat emulsion-agarose gel is used as the phantom; for R2* value measurement, a self-made ferric chloride-agarose gel is used as the phantom. In addition to these phantoms, other phantoms can be selected and used as phantom fabrication technology develops.

[0091] Ideally, quantitative parameters obtained from phantom scanning are used as measurements to quantify the equipment bias of different MRI imaging devices. When acquiring image data, the obtained equipment bias is used to eliminate equipment noise interference, so that the data finally used for training the AILD prediction model can be scaled uniformly.

[0092] In practice, the interference in quantitative parameters obtained from phantom scanning is not limited to equipment bias; it also includes batch offset and random errors of the phantom. Therefore, it is necessary to remove batch offset noise and random errors from the phantom and accurately extract the equipment bias. Furthermore, no existing technology possesses a phantom that can fully characterize all quantitative parameters of the liver. Therefore, in real-world scenarios, multiple phantoms are scanned using MRI imaging equipment to obtain different quantitative parameters. This makes it impossible to achieve an absolute approximation to the true values ​​compared to traditional phantom methods. Moreover, the equipment bias of each MRI imaging device remains consistent when scanning different phantoms.

[0093] Based on this, in step S200, the mapping relationship between the calibration reference value and the equipment deviation vector of the MRI imaging equipment is constructed. The specific implementation process is as follows:

[0094] Step S210: Establish the first quantitative parameter vector The generative model reflects the relationship between quantitative parameters and true values, equipment deviations, batch drift, and random errors.

[0095] ;

[0096] ;

[0097] ;

[0098] ;

[0099] ;

[0100] Among them, the first quantitative parameter vector The structure is: ;

[0101] In the formula, i represents the i-th MRI imaging device. ;j represents the j-th scan of the t-th motif, ; ; The selection matrix is ​​derived from the device deviation vector. Select the quantitative parameter component corresponding to the t-th type phantom. , The total number of dimensions of the quantitative parameters to be collected is 8, where dt represents the number of dimensions of the quantitative parameters that can be collected by the phantom (<8).

[0102] Represents the actual value. ; Indicates batch drift. ; This represents the random error of a single scan. ; Represents the device deviation vector It follows a mean of 0 and a prior covariance of . The normal distribution; Represents batch drift vector It follows a mean of 0 and a drift covariance of . The normal distribution; This indicates that the calibration reference value follows the mean. The covariance is The normal distribution For reference only; Represents random noise It follows a mean of 0 and a noise covariance of . The normal distribution;

[0103] This represents the mean value of the scanning phantom t by imaging device i; Indicates the number of iterations; The variance representing the estimated deviation of device i; Represents the cross product of deviations; The hyperparameter representing the batch offset is an empirical value; Represents the identity matrix; The number of quantitative parameters (or dimensions) provided by the t-th type of motif.

[0104] In this embodiment, since factors such as the non-uniformity of the device gradient magnetic field affect all sequences simultaneously, the device deviation should cover each phantom. However, the quantitative parameters obtained by scanning and acquiring different phantoms are different. Therefore, the corresponding quantitative parameter components are selected from the device deviation vector by selecting a matrix.

[0105] Step S220: Based on the generative model of the first quantitative parameter vector, construct the objective function through maximum a posteriori estimation. :

[0106] ;

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, Indicates the total type of the motif; For likelihood terms; and It is a priori; Indicates the calibration reference value. This represents the estimated deviation of each device; const represents a constant term that does not depend on the optimization variables.

[0111] In this embodiment, for the true values ​​that cannot be accurately obtained, a true value estimate with consensus across all phantoms is obtained through statistical Bayesian inversion, which serves as a calibration benchmark value, aiming at relative consistency between devices. The calibration benchmark value is used to quantify device errors, thereby enabling the standardized processing of MRI image data from several different devices. After being scaled uniformly, it is used for training the AILD individualized early non-invasive precise hierarchical prognostic prediction model.

[0112] Step S230: Solve the objective function by alternating coordinate descent twice until it converges, and then obtain the mapping relationship between the equipment deviation vector and the calibration reference value;

[0113] Fix equipment deviation and update calibration reference value:

[0114] ;

[0115] Fix the calibration reference value and update the equipment deviation:

[0116] ;

[0117] In the formula, This represents an estimate of batch drift.

[0118] Furthermore, in step S300, the standardization transformation process for the second quantitative parameter vector is as follows:

[0119] ;

[0120] In the formula, This represents the second quantitative parameter vector acquired by the MRI imaging equipment. This represents a dynamic selection matrix. This represents the second quantitative parameter vector after standardization.

[0121] Furthermore, the generation process of the dynamic selection matrix is ​​as follows:

[0122] S310: List of quantitative parameter types for identifying MRI image data;

[0123] S320: Generate a zero matrix with the same number of columns as the number of dimensions of all quantitative parameters to be collected and the same number of rows as the number of quantitative parameter types in the quantitative parameter type list;

[0124] S330: For the r-th quantitative parameter in the list of quantitative parameter types, query its predefined index column, modify the element 0 in the r-th row and corresponding index column of the all-zero matrix to 1, and obtain the dynamic selection matrix.

[0125] In this embodiment, a predefined index table clearly defines the index of each quantitative parameter. For example, the T1 value is in column 1, and the ADC value is in column 2. The total number of quantitative parameters to be collected in this embodiment is 8.

[0126] In some feasible embodiments, device bias is used as the weight of the training samples, so that the training strategy can be adjusted according to the weight in subsequent training.

[0127] like Figure 2 As shown, a second aspect of the present invention provides an image data processing system based on MRI technology, including a mapping relationship construction module, a relationship database, a data reading module, a standardization processing module, and a dynamic selection matrix generation module;

[0128] The mapping relationship construction module constructs the quantitative parameters of the phantom into a first quantitative parameter vector; it uses the first quantitative parameter vector to construct calibration reference values ​​and establishes a mapping relationship between the equipment deviation vector of each MRI imaging device and the calibration reference values;

[0129] The relational database stores the mapping relationship between the equipment deviation vectors of each imaging device and the calibration reference values;

[0130] The data reading module reads MRI image data, including MRI image device identification tags, quantitative parameters, and a list of quantitative parameter types; and constructs the quantitative parameters into a second quantitative parameter vector.

[0131] The dynamic selection matrix generation module generates a dynamic selection matrix based on the quantitative parameter types in the quantitative parameter type list and their corresponding predefined index columns.

[0132] The standardization processing module uses the identification tag of the MRI imaging device to call up the mapping relationship between the device deviation vector and the calibration benchmark value to calibrate the second quantitative parameter vector.

[0133] A third aspect of the present invention provides a computer storage medium storing computer-executable instructions and the aforementioned image data processing method, serving as its hardware carrier. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above-mentioned types of memory.

[0134] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0135] The various embodiments of the present invention have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. An image data processing method based on MRI technology, characterized in that, The implementation process is as follows: Several MRI imaging devices were used to scan and process images of a simulated liver phantom, resulting in quantitative parameters of the simulated liver phantom output by each MRI imaging device. Each MRI imaging device performed the following scans on the simulated liver phantom: ADC value obtained by DWI imaging; f value, D* value, and D value obtained by T1 mapping imaging; R2* value and PDFF value obtained by IDEAL-IQ imaging; and MRE value obtained by MRE imaging. Quantitative parameters included ADC value, f value, D* value, D value, R2* value, PDFF value, and MRE value. Quantitative parameters of a simulated liver phantom output from several MRI imaging devices are obtained, and the quantitative parameters of the phantom are constructed into a first quantitative parameter vector; The calibration baseline value is constructed using the first quantitative parameter vector, and a mapping relationship is established between the device deviation vector of each MRI imaging device and the calibration baseline value; each mapping relationship is associated with the identity label of the MRI imaging device. Read the MRI image data of the liver obtained after scanning the human body with the current MRI imaging equipment, extract the identity tag and the quantitative parameters of the human liver, construct the quantitative parameters of the human liver into a second quantitative parameter vector, and output the second quantitative parameter vector after standardizing and calibrating it based on the identity tag by calling the corresponding mapping relationship. MRI imaging data includes the identification label of the MRI imaging device, quantitative parameters, and a list of quantitative parameter types; There are several phantoms simulating the liver, each providing different quantitative parameters during MRI imaging. The first quantitative parameter vector includes the quantitative parameters that the phantom can provide; the vector structure of the equipment deviation includes all the quantitative parameters that need to be acquired. The process of constructing calibration reference values ​​using the first quantitative parameter vector and establishing the mapping relationship between the equipment deviation vector of each MRI imaging device and the calibration reference values ​​is as follows: Step S10: Establish a generation model for the first quantitative parameter vector, which reflects the relationship between the quantitative parameters and the true values, equipment deviation, batch drift, and random errors; Step S20: Construct the objective function, obtain the estimated value of the equipment deviation vector and the estimated value of the true value through inversion calculation, and use the estimated value of the true value obtained by inversion calculation as the calibration benchmark value; Step S30: Solve the objective function iteratively by alternating coordinate descent until it converges, and obtain the mapping relationship between the equipment deviation vector and the calibration reference value.

2. The image data processing method based on MRI technology according to claim 1, characterized in that, In step S10, the first quantitative parameter vector The generative model is as follows: ； ； In the formula, i represents the i-th MRI imaging device. ;j represents the j-th scan, ;t represents the t-th type of motif; Represents the actual value; Indicates batch drift; This represents the random error of a single scan; The selection matrix is ​​represented by multiplication from the device deviation vector. Extract the quantitative parameter components corresponding to the t-th type of phantom. ; N represents the dimension of all quantitative parameters to be collected; dt represents the dt quantitative parameters of phantom t. Represents the device deviation vector It follows a mean of 0 and a prior covariance of . The normal distribution; Represents batch drift vector It follows a mean of 0 and a drift covariance of . The normal distribution; This indicates that the true value follows the mean. Covariance is The normal distribution; Represents random noise It follows a mean of 0 and a noise covariance of . It follows a normal distribution.

3. The image data processing method based on MRI technology according to claim 2, characterized in that, The objective function in step S20 is as follows: ； In the formula, Indicates the overall type of the motif; For likelihood terms; and It is a priori; Indicates the calibration reference value. This represents the estimated value of the deviation vector for each device.

4. The image data processing method based on MRI technology according to claim 3, characterized in that, In step S30, the process of solving the objective function through iterative alternating coordinate descent is as follows: Fix equipment deviation and update calibration reference value: ; Fix the calibration reference value and update the equipment deviation: ； In the formula, This represents the estimated batch drift value; Indicates the total number of imaging devices; This indicates the number of scans performed on the t-type motif.

5. The image data processing method based on MRI technology according to claim 1, characterized in that, The process of standardizing the second quantitative parameter vector is as follows: ； In the formula, This represents the second quantitative parameter vector. This represents a dynamic selection matrix. This represents the second quantitative parameter vector after standardization.

6. The image data processing method based on MRI technology according to claim 5, characterized in that, The process of generating the dynamic selection matrix is ​​as follows: A list of quantitative parameter types for identifying MRI image data; Generate a zero matrix with the same number of columns as the total number of dimensions of all quantitative parameters to be collected and the same number of rows as the number of quantitative parameter types in the quantitative parameter type list; For the r-th quantitative parameter in the list of quantitative parameter types, query its predefined index column, modify the element 0 in the r-th row and corresponding index column of the all-zero matrix to 1, and obtain the dynamic selection matrix.

7. A system for implementing the method according to any one of claims 1-6, characterized in that, It includes a mapping relationship construction module, a relational database, a data reading module, a standardization processing module, and a dynamic selection matrix generation module; The mapping relationship construction module constructs the quantitative parameters of the phantom into a first quantitative parameter vector; it uses the first quantitative parameter vector to construct a calibration reference value and establishes a mapping relationship between the equipment deviation vector of each MRI imaging device and the calibration reference value; The relational database stores the mapping relationship between the equipment deviation vectors of each imaging device and the calibration reference values; The data reading module reads MRI image data, including MRI image device identification tags, quantitative parameters, and a list of quantitative parameter types; and constructs the quantitative parameters into a second quantitative parameter vector. The dynamic selection matrix generation module generates a dynamic selection matrix based on the quantitative parameter types in the quantitative parameter type list and their corresponding predefined index columns. The standardization processing module uses the identification tag of the MRI imaging device to call up the mapping relationship between the device deviation vector and the calibration benchmark value to calibrate the second quantitative parameter vector.

8. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the image data processing method as described in any one of claims 1-6.

Images