Method and apparatus for adjusting magnetic resonance scan parameters, scanning device

Abstract

The application relates to the technical field of magnetic resonance scanning, and discloses a method and device for adjusting magnetic resonance scanning parameters and a scanning device, wherein the method comprises the following steps: acquiring configuration data of magnetic resonance scanning and multi-source clinical evaluation data; the configuration data comprises initial magnetic resonance scanning parameters, and the multi-source clinical evaluation data is associated with the configuration data; inputting the configuration data and the multi-source clinical evaluation data into an image quality prediction model, predicting image quality corresponding to the initial magnetic resonance scanning parameters, and obtaining an initial image quality prediction result; and according to the initial image quality prediction result, performing magnetic resonance scanning or generating a target optimization suggestion of the initial magnetic resonance scanning parameters. According to the initial image quality prediction result, the initial magnetic resonance scanning parameters can be adjusted, the influence of the parameters on image quality and scanning time can be predicted without waiting for actual scanning to be completed, the parameter confirmation time is greatly shortened, and the adjustment efficiency of the magnetic resonance scanning parameters is improved.

Description

Technical Field

[0001] This application relates to the field of magnetic resonance scanning technology, and for example to a method and apparatus for adjusting magnetic resonance scanning parameters, and scanning equipment. Background Technology

[0002] Currently, magnetic resonance imaging (MRI) is one of the main imaging modalities in modern medical imaging, offering advantages such as high soft tissue resolution, no radiation exposure, and diverse imaging parameters, making it widely used in clinical diagnosis. The basic working principle of MRI is based on the magnetic resonance phenomenon. It uses radio frequency pulses to excite hydrogen nuclei in the human body, employs gradient fields for spatial encoding, receives the electromagnetic signals using receiving coils, and finally reconstructs the MRI image using Fourier transform. The selection of conventional MRI scan parameters is a key factor affecting MRI image quality, including repetition time (TR), echo time (TE), flip angle, field of view (FOV), matrix size, slice thickness, number of excitations (NEX), and receiver bandwidth (BW). A proper configuration of these scanning parameters requires a trade-off between image quality, scan time, and signal-to-noise ratio.

[0003] The related technology discloses a personalized magnetic resonance imaging method based on artificial intelligence. Before the formal scan, a pre-scan is performed to obtain a weighted map of physiological parameters and a map of magnetic field inhomogeneity. The pre-scan data is input into a deep neural network to quantify the physiological parameters. Combined with a magnetic resonance signal simulator, iterative or traversal optimization is performed in the solution space to calculate the quality score of the simulated image, thereby outputting the optimal parameters.

[0004] In the process of implementing the embodiments of this disclosure, at least the following problems were found in the related art: Related technologies optimize parameters using pre-scanned data, but cannot intuitively predict the specific impact of optimal parameters on the final image quality. This may require multiple image scans and parameter adjustments, resulting in poor parameter adjustment efficiency.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0007] This disclosure provides a method, apparatus, and scanning device for adjusting magnetic resonance scanning parameters to improve the efficiency of magnetic resonance scanning parameter adjustment.

[0008] In some embodiments, the method for adjusting magnetic resonance scanning parameters includes: acquiring configuration data and multi-source clinical evaluation data for magnetic resonance scanning; wherein the configuration data includes initial magnetic resonance scanning parameters, multi-source clinical evaluation data, and the configuration data are associated; inputting the configuration data and multi-source clinical evaluation data into an image quality prediction model to predict the image quality corresponding to the initial magnetic resonance scanning parameters, thereby obtaining an initial image quality prediction result; and performing a magnetic resonance scan or generating a target optimization suggestion for the initial magnetic resonance scanning parameters based on the initial image quality prediction result.

[0009] Optionally, the image quality prediction model is trained as follows: Multiple first historical scan samples are acquired, each including magnetic resonance imaging (MRI) scan configuration data, multi-source clinical evaluation data, and multi-dimensional data labels; wherein, the multi-dimensional data labels include image quality score labels and multiple risk index labels; feature extraction is performed on each first historical scan sample to obtain the first multimodal features corresponding to each first historical scan sample; the first multimodal features include the scan parameter features of the configuration data and the text features of the multi-source clinical evaluation data; the model is trained based on the first multimodal features and multi-dimensional data labels of each first historical scan sample to obtain the trained image quality prediction model; wherein, the output of the image quality prediction model is the predicted image quality score and the predicted values ​​of multiple risk indices.

[0010] Optionally, the method for adjusting magnetic resonance scanning parameters further includes: acquiring a current scan image; acquiring the current scan image by performing a magnetic resonance scan, or by performing a magnetic resonance scan after adjusting the initial magnetic resonance scanning parameters according to the target optimization suggestion; analyzing the configuration data, multi-source clinical evaluation data, and the current scan image based on the parameter optimization model to obtain a first optimization suggestion for the current magnetic resonance scanning parameters; responding to the user adjusting the current magnetic resonance scanning parameters according to the first optimization suggestion, integrating the adjusted current magnetic resonance scanning parameters with the configuration data, and inputting them together with the multi-source clinical evaluation data into an image quality prediction model to perform image quality prediction and obtain a target image quality prediction result; and generating a second optimization suggestion for the current magnetic resonance scanning parameters based on the target image quality prediction result.

[0011] Optionally, the parameter optimization model is trained as follows: multiple second historical scan samples are acquired; each second historical scan sample includes magnetic resonance scan configuration data, multi-source clinical evaluation data and scan images, as well as optimal scan parameters; feature extraction is performed on each second historical scan sample to obtain the second multimodal features corresponding to each second historical scan sample; the second multimodal features include scan parameter features of the configuration data, text features of the multi-source clinical evaluation data, and visual features of the scan images; the model is trained based on the second multimodal features of each second historical scan sample and the optimal scan parameters to obtain the trained parameter optimization model, the output of which includes at least one of suggested parameters, optimization schemes for specific lesions, and performance prediction curves.

[0012] Optionally, the scanning parameter characteristics include the values ​​of the magnetic resonance scanning parameters and the coupling relationships between the magnetic resonance scanning parameters.

[0013] Optionally, the configuration data for the magnetic resonance scan may also include magnetic resonance auxiliary parameters, and the first and second multimodal features may also include auxiliary parameter features.

[0014] Optionally, the magnetic resonance scanning parameters include a set of associated parameters. The set of associated parameters consists of at least two parameters with preset linkage constraints. During the adjustment of the magnetic resonance scanning parameters, after one parameter in the set of associated parameters is adjusted, the other parameters in the set of associated parameters are updated synchronously based on the preset linkage constraints.

[0015] Optionally, the image quality prediction results include image quality score prediction values ​​and risk index prediction values, generating optimization suggestions for magnetic resonance scanning parameters, including: comparing the image quality score prediction values ​​and risk index prediction values ​​with their respective corresponding preset thresholds to determine the dimensions that do not meet the standards in the current combination of scanning parameters; querying a preset parameter adjustment rule mapping table according to the determined types of non-compliant dimensions, and matching the corresponding adjustment strategies; wherein, the image quality prediction results include initial image quality prediction results and target image quality prediction results, the magnetic resonance scanning parameters include initial magnetic resonance scanning parameters and current magnetic resonance scanning parameters, and the optimization suggestions include target optimization suggestions and second optimization suggestions, and the mapping table has a preset correspondence between different types of non-compliant dimensions and parameter adjustment directions, adjustment steps, and adjustment reasons.

[0016] Optionally, the method for adjusting magnetic resonance scanning parameters further includes: online learning and updating of the image quality prediction model and parameter optimization model based on the user's adoption of optimization suggestions and the actual scanning results obtained after adjustment based on the optimization suggestions; the optimization suggestions include target optimization suggestions and second optimization suggestions.

[0017] In some embodiments, the apparatus for adjusting magnetic resonance scan parameters includes a processor and a memory storing program instructions, the processor being configured to, when the program instructions are executed, perform the method for adjusting magnetic resonance scan parameters as described above.

[0018] In some embodiments, the scanning device includes: a scanning device body; and the means for adjusting magnetic resonance scanning parameters as described above, which is mounted on the scanning device body.

[0019] The method, apparatus, and scanning device for adjusting magnetic resonance scanning parameters provided in this disclosure can achieve the following technical effects: In this embodiment, by introducing configuration data for magnetic resonance imaging (MRI) scans and multi-source clinical evaluation data, clinical diagnostic needs can be aligned, the standardization of scans can be improved, and the final clinical diagnostic quality requirements can be more closely met. By setting up an image quality prediction step based on an image quality prediction model, the image quality corresponding to the initial MRI scan parameters can be predicted before parameter adjustment, outputting the initial image quality prediction result. Based on the initial image quality prediction result, MRI scanning can be performed or target optimization suggestions for the initial MRI scan parameters can be generated. This allows for prediction of the impact of parameters on image quality and scan duration without waiting for the actual scan to complete, significantly reducing the probability of repeated scans, greatly shortening parameter confirmation time, and improving the efficiency of MRI scan parameter adjustment.

[0020] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0021] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein: Figure 1 This is a schematic diagram of a method for adjusting magnetic resonance scanning parameters provided in an embodiment of this disclosure; Figure 2 This is a flowchart illustrating the training process of an image quality prediction model provided in an embodiment of this disclosure. Figure 3 This is a schematic diagram of another method for adjusting magnetic resonance scanning parameters provided in an embodiment of this disclosure; Figure 4 This is a schematic diagram of an apparatus for adjusting magnetic resonance scanning parameters provided in an embodiment of this disclosure. Detailed Implementation

[0022] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0023] The terms "first," "second," etc., used in the technical solutions described in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0024] Unless otherwise stated, the term "multiple" means two or more.

[0025] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0026] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0027] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0028] Combination Figure 1 As shown, this disclosure provides a method for adjusting magnetic resonance scanning parameters. The execution subject of this method may be a processor, and the method includes: S101, the processor acquires the configuration data of the magnetic resonance scan and the multi-source clinical evaluation data; wherein, the configuration data includes the initial magnetic resonance scan parameters, the multi-source clinical evaluation data and the configuration data are associated.

[0029] S102, the processor inputs the configuration data and multi-source clinical evaluation data into the image quality prediction model, predicts the image quality corresponding to the initial magnetic resonance scan parameters, and obtains the initial image quality prediction result.

[0030] S103, the processor performs a magnetic resonance scan or generates target optimization suggestions for the initial magnetic resonance scan parameters based on the initial image quality prediction results.

[0031] In this embodiment, by introducing configuration data for magnetic resonance imaging (MRI) scans and multi-source clinical evaluation data, clinical diagnostic needs can be aligned, the standardization of scans can be improved, and the final clinical diagnostic quality requirements can be more closely met. By setting up an image quality prediction step based on an image quality prediction model, the image quality corresponding to the initial MRI scan parameters can be predicted before parameter adjustment, outputting the initial image quality prediction result. Based on the initial image quality prediction result, MRI scanning can be performed or target optimization suggestions for the initial MRI scan parameters can be generated. This allows for prediction of the impact of parameters on image quality and scan duration without waiting for the actual scan to complete, significantly reducing the probability of repeated scans, greatly shortening parameter confirmation time, and improving the efficiency of MRI scan parameter adjustment.

[0032] The configuration data for magnetic resonance scanning includes magnetic resonance scanning parameters and magnetic resonance auxiliary parameters.

[0033] For example, the configuration data for magnetic resonance imaging (MRI) scanning is obtained through the MRI scanning protocol file. The protocol file records the MRI scanning parameters and MRI auxiliary parameters. Optionally, the MRI scanning parameters include: conventional parameters such as scan sequence type, repetition time (TR), echo time (TE), field of view (FOV), matrix size, and slice thickness, as well as parallel acquisition factor (affecting the balance between signal-to-noise ratio and time), echo train length (affecting image contrast and blur), and specific absorption rate limits (ensuring that optimization suggestions do not exceed safe limits).

[0034] Magnetic resonance auxiliary parameters include the scanning location, coil type, magnet, gradient, factory-fixed parameters of the radio frequency system, and upper and lower limit thresholds of the parameters. In other words, magnetic resonance auxiliary parameters include scanning location information and inherent hardware parameter information.

[0035] In this embodiment, the configuration data for magnetic resonance scanning can be obtained through the software interface of the magnetic resonance scanning system. Furthermore, after the user adjusts the scanning parameters, the adjusted scanning parameters can also be obtained in real time through this software interface.

[0036] Multi-source clinical evaluation data includes outcome assessment descriptions; these descriptions include image comments and the manufacturer's internal clinical research and development evaluation reports. The image comments are derived from radiologists' diagnostic descriptions of the scanned images, while the manufacturer's internal clinical research and development evaluation reports contain professional assessments of the physical performance of the magnetic resonance imaging (MRI). In this embodiment, the multi-source clinical evaluation data can be obtained via the PACS protocol, such as... Figure 2Image reports (location, indication, comparison, findings, impressions) are generated. Specifically, historical reports and request forms of the current patient in the PACS system are retrieved, and keywords are extracted from multi-source clinical evaluation data, such as diagnostic intent keywords like "metal implants," "edema signal tendency," and "suspected microfractures," as well as negative feedback on past image quality, such as "heavy image graininess" or "excessive inter-layer spacing leading to missed diagnosis."

[0037] There is an inherent correlation between multi-source clinical evaluation data and magnetic resonance imaging (MRI) scan configuration data based on the physical imaging mechanism of MRI. Specifically, semantic features involving anatomical structures or lesion types in multi-source clinical evaluation data are associated with spatial resolution-related configuration parameters (including slice thickness, matrix size, and field of view (FOV)) in the configuration data; semantic features involving tissue contrast or signal characteristics in multi-source clinical evaluation data are associated with contrast-related configuration parameters (including repetition time (TR), echo time (TE), and flip angle) in the configuration data; and semantic features involving imaging quality defects in multi-source clinical evaluation data are associated with signal-to-noise ratio (SNR) and artifact suppression-related configuration parameters (including excitation times (NEX), receiver bandwidth (BW), parallel acquisition factor, and phase encoding direction) in the configuration data.

[0038] In this embodiment, clinical evaluation text can be mapped to standard physical index vectors through a medical knowledge graph, and the association weights between text semantics and scanning parameters can be automatically learned through a cross-modal attention mechanism. For example, “microfracture” is associated with slice thickness and matrix, “edema signal” is associated with TR and TE, and “pulsation artifact” is associated with phase encoding direction. According to the above mapping relationship and association weights, the configuration parameters and clinical evaluation parameters can be effectively associated.

[0039] Building upon the aforementioned relationships, using MRI configuration parameters and multi-source clinical evaluation data as input to the image quality prediction model ensures that image quality prediction and parameter adjustment remain aligned with the actual scanning environment. This allows for decision-making based on a complete context, significantly improving the engineering practicality and clinical applicability of image quality prediction results. The same clinical evaluation may yield drastically different optimal scanning parameters depending on the context of different auxiliary parameters. For example, the optimal slice thickness and matrix configuration differ when using a head coil versus a knee-specific coil. Through the dual correlation of initial MRI scan parameters, the abstract "clinical evaluation" can be interpreted within the specific context of "auxiliary parameters." For instance, when the auxiliary parameter indicates "using a high-channel array coil," it can be more accurately understood that "focusing on small fractures" implies the use of higher spatial resolution. This contextual disambiguation based on auxiliary parameters enables truly personalized scanning. Furthermore, auxiliary parameters define the physical boundaries of the scan. By associating scan parameters with auxiliary parameters, these boundary conditions can be referenced in real-time during parameter adjustments, preventing the generation of parameter combinations that exceed the equipment's safety limits or the patient's tolerance limits in pursuit of image quality, thus improving the safety and compliance of the scan.

[0040] This embodiment of the disclosure introduces PACS image comments and manufacturer's internal clinical research and development evaluation reports into the model training closed loop. By using natural language processing technology to extract the diagnostic intent and evaluation semantics of radiologists, the parameter optimization goal is transformed from simply "improving physical quality" to "maximizing clinical diagnostic value". This makes the image features obtained after parameter optimization more in line with the reading habits of clinicians and the observation needs of specific lesions.

[0041] Optionally, the image quality prediction model is trained as follows: Multiple first historical scan samples are acquired, each including magnetic resonance imaging (MRI) scan configuration data, multi-source clinical evaluation data, and multi-dimensional data labels; wherein, the multi-dimensional data labels include image quality score labels and multiple risk index labels; feature extraction is performed on each first historical scan sample to obtain the first multimodal features corresponding to each first historical scan sample; the first multimodal features include the scan parameter features of the configuration data and the text features of the multi-source clinical evaluation data; the model is trained based on the first multimodal features and multi-dimensional data labels of each first historical scan sample to obtain the trained image quality prediction model; wherein, the output of the image quality prediction model is the predicted image quality score and the predicted values ​​of multiple risk indices.

[0042] Optionally, the risk index includes scan duration assessment, signal-to-noise ratio assessment, and artifact severity assessment; the image quality prediction model is configured with a multi-task learning network, which includes a shared feature extraction layer and multiple task-specific output layers; the shared feature extraction layer is used to extract general feature representations from the input data; the multiple task-specific output layers are used to output image quality score predictions and risk index predictions (scan duration prediction, signal-to-noise ratio prediction, and artifact severity prediction); the first multimodal feature also includes auxiliary parameter features.

[0043] In this embodiment, by using the first historical scan samples, the model can learn the mapping relationship between scan parameters and physical indicators such as signal-to-noise ratio / resolution, as well as the doctor's diagnostic evaluation of the image and the professional physical defect judgment from the clinical research side. This ensures that the final image quality prediction result reflects both objective imaging quality and aligns with clinical diagnostic requirements. The training process simultaneously extracts and fuses scan parameter features, auxiliary parameter features, and clinical text features, which can improve prediction accuracy and robustness to complex scenarios.

[0044] Optionally, feature extraction is performed on each first historical scan sample to obtain a first multimodal feature corresponding to each first historical scan sample. This includes: performing data encoding processing on the configuration data in the first historical scan sample, extracting the values ​​of magnetic resonance scanning parameters and the coupling relationship between magnetic resonance scanning parameters as scan parameter features, and extracting the values ​​of scanning site identifiers, coil identifiers, and inherent hardware parameter information in the magnetic resonance auxiliary parameters as auxiliary parameter features; performing semantic feature processing on the multi-source clinical evaluation data in the first historical scan sample, and extracting the correspondence between text keywords and physical parameters as text features. It is understood that the scan parameter features include the values ​​of magnetic resonance scanning parameters and the coupling relationship between magnetic resonance scanning parameters.

[0045] In this embodiment, in magnetic resonance imaging, parameters such as TR, TE, NEX, and matrix are not independent but jointly determine imaging results such as signal-to-noise ratio, contrast, and scan time. By extracting the coupling relationship between parameters (such as the combined effect of TR×NEX / matrix), the extracted scan parameter features can reflect the true physical imaging laws, avoiding image quality prediction deviations caused by ignoring the interaction between parameters, thereby improving the accuracy of subsequent parameter adjustments. Furthermore, unstructured and subjective clinical descriptions such as "blurred edges," "obvious graininess," and "pulsation artifacts" are transformed into structured features associated with specific physical parameters (such as slice thickness, NEX, and phase encoding direction), eliminating the ambiguity of natural language descriptions. This allows the model to uniformly understand the essence of clinical needs under different physicians and different expressions, providing reliable input for subsequent parameter optimization based on clinical needs. Scan parameter features and text features can be extracted within the same "physical parameter semantic space," and the two are naturally aligned. This aligned multimodal feature makes it easier for the model to learn "what kind of clinical evaluation semantics corresponds to what kind of parameter adjustment direction", laying a high-quality data foundation for subsequent cross-modal attention mechanisms, feature fusion and image quality prediction, and significantly improving the model's convergence speed and prediction accuracy.

[0046] Optionally, the model is trained based on the first multimodal features of each first historical scan sample and the real image quality label to obtain a trained image quality prediction model. This includes: fusing the scan parameter features, auxiliary parameter features, and text features of the first multimodal features to obtain a first fused feature; inputting the first fused feature into the image prediction quality model for model training; the model output is a predicted image quality score and a predicted risk index; the training objective is to minimize the difference between the predicted image quality score and the predicted risk index and the multidimensional data label; and the trained model is evaluated to obtain an image quality prediction model that passes the evaluation.

[0047] In this embodiment, the training process integrates scanning parameter features, auxiliary parameter features, and text features from multi-source clinical evaluation data. This aligns with clinical diagnostic needs, ensuring that the prediction results balance physical quality and diagnostic value. The training objective is set to minimize the differences between the predicted image quality score and risk index values ​​and the multidimensional data labels, ensuring stable quantitative comparability of the model's prediction results. After training, an accuracy evaluation is used to select usable models, avoiding the problem of models only fitting historical samples and failing to adapt to new patients or new equipment conditions. This ensures that the model maintains predictive stability across different medical institutions and scanning scenarios.

[0048] Optionally, the image quality rating labels use a Likert scale of 5 and are labeled by a senior radiologist.

[0049] Optionally, combined Figure 2As shown, the scanning parameters of the configuration data include: TR, TE, flip angle, FOV, matrix, layer thickness, excitation times NEX, and receive bandwidth BW.

[0050] Optionally, auxiliary parameter features of the configuration data include scanning location identifier, coil type identifier, system hardware parameters, and upper and lower limit thresholds.

[0051] Optionally, textual features of multi-source clinical evaluation data can be obtained as follows: Natural language processing techniques are used to extract keywords and semantic features from the multi-source clinical evaluation data. Combined with... Figure 2 As shown, text features include: location, indication, comparison, discovery, and impression.

[0052] In this embodiment, colloquial descriptions in doctors' comments (such as "graininess," "blur," and "ghosting") are mapped to standard physical indicator vectors (such as SNR defect coefficient, contrast-to-noise ratio deviation, and artifact level predictor) through a medical knowledge graph. A pre-trained medical language model (such as Med-BERT) can be used to encode these features into high-dimensional hidden states. A cross-modal attention mechanism is then employed to calculate the correlation weights between "image defect descriptions" and "scanning protocol parameters" in the text features (e.g., identifying the high correlation between the description of "pulsation artifacts" and the "phase encoding direction" parameter), thereby capturing the physical imaging improvement needs hidden behind the text.

[0053] Optionally, combined Figure 2 As shown, feature fusion is achieved through a pre-trained feature extraction module.

[0054] Optionally, feature splicing, attention mechanisms, and other methods can be used to achieve the fusion of scanning parameter features and text features.

[0055] It should be noted that the use of auxiliary parameter features in model training in the above embodiments is only one optional implementation. In other embodiments, the model can be trained solely based on magnetic resonance imaging (MRI) scan parameter features and textual features from multi-source clinical evaluation data, without extracting auxiliary parameter features. In one optional approach, auxiliary parameters do not directly participate in model training and inference, but can be used as post-validation conditions during the parameter adjustment phase to verify whether the current optimization recommendations exceed the safety boundaries of the device or the patient. In a specific embodiment, combined with Figure 2As shown, multiple first historical scan samples are first acquired. Each first historical scan sample includes a protocol document, a result evaluation description, and corresponding image quality labels. For the protocol document, not only are numerical values ​​extracted, but also the coupling relationships between parameters; for example, the system calculates the composite feature affecting the theoretical signal-to-noise ratio of the scan: TR × excitation / Matrix. For the result evaluation description, image comments from the PACS layer need to be extracted, such as diagnostic feedback like "obvious graininess," "blurred edges," and "pulsation artifacts"; and clinical research reports from manufacturers during the R&D period need to be extracted to obtain more professional physical evaluations, such as "geometric distortion caused by B0 inhomogeneity" and "inter-layer crosstalk." For each first historical scan sample, the scan parameter features of the protocol document and the text features of the result evaluation description are extracted. During feature extraction, a tabular data encoder can be used to process the protocol parameters, and a pre-trained Medical-BERT can be used to process the Chinese clinical text, uniformly mapping it to a 512-dimensional latent vector space. An attention mechanism can be used to learn the correspondence between text keywords and physical parameters. For example, when the text shows "metallic artifacts," the model automatically increases the weight given to the parameters "bandwidth (BW)" and "echo spacing." After setting training parameters such as the number of rounds, learning rate, and accuracy, the scan parameter features and text features are fused and input into the image quality prediction model for training. The model can use a multilayer perceptron (MLP), convolutional neural network (CNN), or Transformer architecture. The training objective is the mean squared error between the predicted image quality score and the predicted risk index and the multidimensional data labels. Finally, the model performance is evaluated on the validation set, calculating metrics such as accuracy, recall, and F1 score to ensure the model has good generalization ability. The trained image quality prediction model can provide predicted image quality scores and predicted risk index values ​​(scan time, signal-to-noise ratio, etc.) for the current parameter combination.

[0056] Optionally, the image quality prediction result includes an image quality score prediction value and a risk index prediction value. Based on the image quality prediction result, optimization suggestions for magnetic resonance scanning parameters are generated, including: comparing the image quality score prediction value and the risk index prediction value with their respective preset thresholds to determine the dimensions that do not meet the standards in the current combination of scanning parameters; querying a preset parameter adjustment rule mapping table according to the determined types of non-compliant dimensions to match the corresponding adjustment strategies; wherein, the image quality prediction result includes an initial image quality prediction result and a target image quality prediction result, the magnetic resonance scanning parameters include initial magnetic resonance scanning parameters and current magnetic resonance scanning parameters, the optimization suggestions include target optimization suggestions and second optimization suggestions, and the non-compliant dimensions include at least one of the following: overall quality non-compliant, signal-to-noise ratio non-compliant, scanning time exceeding expectations, insufficient resolution, and artifact degree exceeding the standard, and the mapping table has a preset correspondence between different types of non-compliant dimensions and parameter adjustment directions, adjustment steps, and adjustment reasons.

[0057] In this embodiment, the adjustment of initial MRI scan parameters supports both global optimization to improve the overall score and local correction to specifically repair abnormal indicators such as artifacts and excessively long scan times, and can be flexibly switched according to actual clinical needs. For example, in emergency scenarios, priority can be given to correcting the abnormal indicator of "excessively long scan time," while in research scans, the goal of maximizing the overall score can be prioritized, significantly improving the adaptability and practicality of parameter optimization. When the signal-to-noise ratio assessment value is lower than a preset threshold, the strategy of "increasing the number of excitations (NEX)" or "increasing the repetition time (TR)" is matched; when the scan duration assessment value exceeds the preset threshold, the strategy of "increasing the parallel acquisition factor" or "reducing the matrix size" is matched. By generating a structurally complete adjustment strategy, it is easy for physicians / technicians to quickly understand and implement, reducing the risk of misoperation.

[0058] Optionally, the method for adjusting magnetic resonance scanning parameters further includes: acquiring a current scan image; acquiring the current scan image by performing a magnetic resonance scan, or by performing a magnetic resonance scan after adjusting the initial magnetic resonance scanning parameters according to the target optimization suggestion; analyzing the configuration data, multi-source clinical evaluation data, and the current scan image based on the parameter optimization model to obtain a first optimization suggestion for the current magnetic resonance scanning parameters; responding to the user adjusting the current magnetic resonance scanning parameters according to the first optimization suggestion, integrating the adjusted current magnetic resonance scanning parameters with the configuration data, and inputting them together with the multi-source clinical evaluation data into an image quality prediction model to perform image quality prediction and obtain a target image quality prediction result; and generating a second optimization suggestion for the current magnetic resonance scanning parameters based on the target image quality prediction result.

[0059] Combination Figure 3 As shown in the embodiments of this disclosure, another method for adjusting magnetic resonance scanning parameters is provided, including: S401, the processor acquires configuration data for the magnetic resonance scan and multi-source clinical evaluation data; wherein, the configuration data includes initial magnetic resonance scan parameters, multi-source clinical evaluation data, and configuration data associated with each other.

[0060] S402, the processor inputs the configuration data and multi-source clinical evaluation data into the image quality prediction model, predicts the image quality corresponding to the initial magnetic resonance scan parameters, and obtains the initial image quality prediction result.

[0061] S403, the processor performs a magnetic resonance scan or generates target optimization suggestions for the initial magnetic resonance scan parameters based on the initial image quality prediction results.

[0062] S404, The processor acquires the current scan image; the current scan image is acquired by performing a magnetic resonance scan, or by performing a magnetic resonance scan after adjusting the initial magnetic resonance scan parameters according to the target optimization suggestions.

[0063] The S405 processor analyzes configuration data, multi-source clinical evaluation data, and current scan images based on a parameter optimization model to obtain the first optimization suggestion for the current magnetic resonance scan parameters.

[0064] S406, the processor responds to the user adjusting the current magnetic resonance scanning parameters according to the first optimization suggestion, integrates the adjusted current magnetic resonance scanning parameters with the configuration data, and inputs them together with the multi-source clinical evaluation data into the image quality prediction model to perform image quality prediction and obtain the target image quality prediction result.

[0065] S407, the processor generates a second optimization suggestion for the current magnetic resonance scan parameters based on the target image quality prediction results.

[0066] In this embodiment, the image quality prediction model incorporates two image quality prediction steps: an initial image quality prediction result can be output based on the initial configuration before parameter adjustment, and a target image quality prediction result can be output in real time after the user adjusts the scanning parameters. This allows for prediction of the impact of parameter adjustments on image quality and scan duration without waiting for the actual scan to complete, significantly reducing the probability of repeated scans, greatly shortening parameter confirmation time, and improving parameter adjustment efficiency. Furthermore, the parameter optimization model simultaneously inputs configuration data, multi-source clinical evaluation data, and the current scan image. Through multimodal feature fusion, it can simultaneously understand the parameter configuration logic, clinical needs, and visual defects of the current image, outputting initial optimization suggestions that are more relevant to the actual scenario and adaptable to differentiated situations. It should be noted that in step S407, the mapping table method used in the initial MRI scan parameter adjustment process described above can be used to provide a second optimization suggestion.

[0067] Optionally, the parameter optimization model can be trained as follows: Multiple second historical scan samples are acquired; each second historical scan sample includes configuration data for the magnetic resonance imaging (MRI) scan, multi-source clinical evaluation data, and scan images, as well as the corresponding optimal scan parameters; feature extraction is performed on each second historical scan sample to obtain the second multimodal features corresponding to each second historical scan sample; the second multimodal features include scan parameter features of the configuration data, text features of the multi-source clinical evaluation data, and visual features of the scan images; the model is trained based on the second multimodal features of each second historical scan sample and the optimal scan parameters to obtain the trained parameter optimization model. The output of the parameter optimization model includes at least one of suggested parameters, optimization schemes for specific lesions, and performance prediction curves. It is understood that the second multimodal features may also include auxiliary parameter features.

[0068] In this embodiment, the second historical scan sample enables the model to learn the mapping relationship between the current multimodal state and expert-level parameter adjustments. The scan parameter features correspond to the existing configuration baseline, the text features carry the diagnostic priority, and the visual features can identify the current noise distribution, artifact morphology, and anatomical structure clarity. This achieves a "visual-text-parameter" cross-attention fusion logic, which allows the model to output compensatory adjustment suggestions that are more in line with the current actual scenario.

[0069] Optionally, a model is trained based on the second multimodal features and optimal scanning parameters of each second historical scan sample to obtain a trained parameter optimization model. This includes: fusing the scanning parameter features, text features, and visual features of the second multimodal features to obtain a second fused feature; inputting the second fused feature into the parameter optimization model for model training; the model output being suggested parameters; the training objective being to minimize the difference between the suggested parameters and the optimal scanning parameters; evaluating the trained model to obtain a parameter optimization model that passes the evaluation; and the evaluation objective being the accuracy of the parameter optimization suggestions and the image quality improvement effect. The suggested parameters can be precise values ​​that should be increased or decreased for each parameter based on the current parameters; specifically, they can be represented as parameter compensation vectors.

[0070] In this embodiment, a cross-attention mechanism is used to fuse scan parameter features, text features, and visual features. This allows for the identification of artifacts and noise distribution issues in the current image through visual features, and the capture of diagnostic priorities such as "paying attention to minor fractures" and "requiring pressure on artifacts from metal implants" through text features. Furthermore, it can output targeted compensation adjustments based on current scan parameter features, avoiding the problem of general recommendations not matching specific scenarios and improving robustness. Through image quality prediction and parameter optimization models, the experience of senior technicians and the diagnostic needs of radiologists can be digitized and modeled. By automatically outputting target optimization suggestions, even less experienced technicians can obtain expert-level scan configurations, thereby eliminating image quality fluctuations caused by differences in personnel skill levels and achieving standardization and homogenization of MRI examinations.

[0071] Optionally, the optimal scanning parameters are determined by a senior technician based on the image quality assessment results.

[0072] Optionally, visual features are extracted using pre-trained convolutional neural networks (such as ResNet, VGG, etc.), including features such as signal-to-noise ratio, contrast, spatial resolution, uniformity, and artifacts.

[0073] Optionally, a multimodal attention mechanism can be used to fuse scanning parameter features, text features, and visual features, enabling the model to fully understand the current scanning state.

[0074] Optionally, the parameter optimization model can also output optimization schemes for specific lesions. During model training, recommended optimization strategy labels for different lesion types can be added. These labels are pre-annotated by imaging experts based on clinical diagnostic needs. Furthermore, loss weights for these recommended optimization strategy labels for different lesion types are added to the damage function.

[0075] Optionally, the parameter optimization model can also output performance prediction curves, i.e., curves showing the changes in signal-to-noise ratio (SNR) and scan time as the target scan parameters (such as repetition time TR or excitation count NEX) vary within a preset range. During model training, training labels can be generated using a magnetic resonance signal simulator—using the scan parameters in each second historical scan sample as a benchmark, the values ​​of the target scan parameters are changed within a preset range at preset step sizes, while keeping other parameters constant. The simulator calculates the SNR and scan time corresponding to each value point, forming curve labels. During model training, the SNR and scan time values ​​corresponding to different adjustments around the current scan parameters can be used as annotation data for the performance prediction curves.

[0076] In one specific embodiment, multiple second historical scan samples are acquired. Each second historical scan sample includes a protocol document, a result evaluation description, a scan image, and corresponding optimal scan parameters. The protocol document defines the current baseline parameter space, the result evaluation description introduces clinical "diagnostic weights," and the scan image, combined with real-time image analysis, extracts spatial features to identify actual noise distribution, artifact morphology, and anatomical structure clarity. For example, if the evaluation emphasizes "paying attention to small blood vessels," the model automatically increases contrast and matrix weights. For each second historical scan sample, the scan parameter features from the protocol document, the text features from the result evaluation description, and the visual features from the scan image are extracted. This enables feature fusion from visual to parameter and text to visual. For example, if "phase wrapping" is detected in the image, the decision is to increase the field of view (FOV) or enable the anti-folding option; if the doctor mentions "suspected tumor enhancement," the model will focus on the enhanced region in the scan image and analyze whether its signal-to-noise ratio is sufficient to support a qualitative assessment. The final output of the parameter optimization model is suggested parameters, optimization schemes for specific lesions, and performance prediction curves.

[0077] It should be noted that the above-described implementation method for training the parameter optimization model is only one optional approach. In other implementations, while extracting scan parameter features and textual features from multi-source clinical data, the parameter optimization model can also extract auxiliary parameter features. In this case, the auxiliary parameters directly participate in the model's training and inference, used to verify whether the current optimization recommendations exceed the safety boundaries of the device or the patient.

[0078] Optionally, the adjusted magnetic resonance scanning parameters are integrated with the configuration data, including generating a new combination of scanning parameters based on the adjusted magnetic resonance scanning parameters and the configuration data.

[0079] In this embodiment, the adjusted magnetic resonance scanning parameters do not take effect in isolation. Instead, they are combined with the original magnetic resonance scanning configuration data to generate a new combination of scanning parameters. This avoids ignoring the coupling relationships in the original protocol (such as the linkage logic between TR, TE, and excitation count) when calculating and adjusting parameters separately, ensuring the consistency between the prediction results and the actual scanning effect. The new combination is generated and the prediction is completed immediately after the user modifies the parameters, without waiting for pre-scanning or formal scanning. This significantly shortens the parameter confirmation cycle and reduces duplicate scans caused by improper parameter settings.

[0080] Optionally, the magnetic resonance scanning parameters include a set of associated parameters, which consists of at least two parameters with preset linkage constraints. During the adjustment of the magnetic resonance scanning parameters, after one parameter in the set of associated parameters is adjusted, the other parameters in the set of associated parameters are updated synchronously based on the preset linkage constraints.

[0081] In this embodiment, by pre-setting linkage constraints (such as physical formulas or empirical lookup tables), when adjusting active parameters (such as increasing bandwidth), the passive parameters (such as sampling time) are forcibly updated synchronously. This ensures that the newly generated parameter combinations always satisfy physical laws and equipment hardware limitations, eliminates the generation of illegal protocols, and significantly reduces equipment error rates and scan failure rates. Furthermore, when adjusting parameters, users only need to focus on core clinical decision parameters (such as adjusting slice thickness according to diagnostic needs), and the system can automatically complete the adaptation calculations for the remaining related parameters. This simplifies complex multi-dimensional parameter adjustments into a single-dimensional decision-making problem, significantly reducing operational difficulty and enabling even non-expert personnel to quickly develop high-quality scanning plans. In addition, magnetic resonance imaging quality is the result of the combined effect of global parameters. By adjusting one parameter to trigger the linkage of the entire related group, a series of chain reactions caused by the adjustment of one parameter (such as a decrease in signal-to-noise ratio due to reduced slice thickness, which necessitates increasing NEX or reducing the matrix for compensation) are systematically coordinated. This ensures that after parameter adjustment, the overall scanning plan reaches a new, self-consistent balance point in multiple dimensions such as resolution, signal-to-noise ratio, contrast, and scan time, achieving systematic global optimization.

[0082] Optionally, generating a new combination of scanning parameters includes: acquiring a set of associated parameters for magnetic resonance scanning parameters; and, in response to the adjusted magnetic resonance scanning parameters, calculating the value of at least one slave parameter associated with the adjusted scanning parameters based on a preset linkage constraint in the set of associated parameters.

[0083] In this embodiment, due to the strict physical coupling between magnetic resonance scanning parameters, such as the limitations of TR and minimum TE, the inverse relationship between receiving bandwidth and sampling time, and the logical constraints of layer thickness and interlayer spacing, the associated parameter group mechanism automatically calculates and updates its associated slave parameters (such as sampling time and chemical shift artifact size) according to physical formulas when actively adjusting a certain parameter (such as increasing the receiving bandwidth BW). This ensures that the newly generated parameter combination always meets the physical laws and device hardware limitations, and avoids the generation of illegal protocols.

[0084] Optionally, based on the target image quality prediction result, a target optimization suggestion for magnetic resonance scanning parameters is generated, including: comparing the target image quality prediction result with a preset threshold, and generating a target optimization suggestion based on the comparison result.

[0085] In this embodiment, by directly comparing the target image quality prediction result with a preset threshold to output suggestions, the "parameter quality judgment" that originally relied on the subjective experience of senior technicians is transformed into an objective and referable standard. This quickly clarifies whether the current parameters need adjustment and in what direction, effectively eliminating parameter configuration fluctuations caused by differences in personnel capabilities and achieving standardized scanning operations. The target optimization suggestion can simultaneously output the parameter adjustment direction and provide multi-dimensional information such as the expected image quality improvement, scan time changes, and signal-to-noise ratio changes. Users can intuitively weigh image quality and examination efficiency when adjusting parameters, avoiding excessively extending scan time due to blindly pursuing high image quality or sacrificing the image quality required for diagnosis to compress time, thus reducing unnecessary rescanning from the root cause.

[0086] In one specific embodiment, the user modifies scanning parameters, such as TR, TE, and slice thickness, on the MRI scanning interface. The system generates a new combination of scanning parameters based on the adjusted parameters and the current protocol file. This new combination is then input into the image quality prediction model to obtain the predicted target image quality. Based on a comparison between the predicted target image quality and a preset threshold, the system provides optimization suggestions to the user, including the direction of parameter adjustment, the expected improvement in image quality, changes in scanning time, and predicted changes in signal-to-noise ratio. For example, if the predicted target image quality is lower than the preset threshold, the system prompts, "The current parameters may lead to a decrease in image quality; it is recommended to increase the number of excitations or adjust the flip angle." If the predicted target image quality is higher than the preset threshold, the system prompts, "The current parameter combination is optimal; image quality is expected to improve; the scanning time will increase by 2 minutes." Furthermore, the system interface provides an interactive window, allowing interaction with the technician to present the prediction results and facilitate technician interaction.

[0087] Optionally, the method for adjusting magnetic resonance scanning parameters further includes: online learning and updating of the image quality prediction model and parameter optimization model based on the user's adoption of optimization suggestions and the actual scanning results obtained after adjustment based on the optimization suggestions; the optimization suggestions include target optimization suggestions and second optimization suggestions.

[0088] In this embodiment, by collecting actual suggestion adoption data and actual scan results as the basis for updates, the image quality prediction model and parameter optimization model can gradually adapt to the aging level of the current equipment, the uniformity characteristics of the magnetic field, and the physiological characteristics of the local patient population. This enables the model to dynamically adapt to specific scenarios, allowing it to continuously adapt to individual differences between different devices and patients. For example, if a user repeatedly adopts a certain type of adjustment direction and the subsequent PACS diagnostic evaluation is positive, the model will automatically strengthen the output weight of that type of scenario. If a high predicted score occurs but the actual diagnostic evaluation is poor, the model will correct the prediction bias through incremental training, avoiding the problem of "high physical image quality but not meeting diagnostic requirements."

[0089] In one specific embodiment, if the user adopts the optimization suggestions and performs a scan, the system records the actual scan results to update the image quality prediction model and parameter optimization model, achieving continuous optimization. Reinforcement learning (RL) reward functions are defined according to different protocol sets, and the "excellent, good, poor" evaluations given by physicians in the subsequent PACS system are used as reward signals to fine-tune the front-end parameter optimization model, enabling the device to become increasingly intelligent with use.

[0090] Optionally, the method for adjusting magnetic resonance scanning parameters may further include: obtaining real-time feedback information of the current scan.

[0091] In this embodiment, real-time feedback information of the current scan can be obtained through the intelligent interactive interface of the scanning system. By monitoring the message reports of the UI layer, the "intention" of the technician during operation can be captured. For example, frequent zooming / moving may indicate that the lesion location is difficult to focus. If the technician forcibly stops and rescans halfway through the scan, the system needs to record the current parameters as a "negative sample".

[0092] The method for adjusting magnetic resonance imaging (MRI) scan parameters provided in this disclosure can intelligently optimize MRI scan parameters based on an image quality prediction model and a parameter optimization model, thereby significantly improving the signal-to-noise ratio, contrast, and spatial resolution of MRI images and enhancing diagnostic accuracy. The optimized combination of scan parameters can reasonably shorten scan time and improve examination efficiency while ensuring image quality. The system can also provide professional guidance to inexperienced operators, reducing image quality problems caused by improper parameter settings. It immediately predicts image quality changes after user parameter modifications, providing intuitive optimization suggestions and improving parameter adjustment efficiency. Furthermore, through an online learning mechanism, the model can continuously optimize, adapting to individual differences in different devices and patients, thus enhancing system robustness.

[0093] Combination Figure 4 As shown, this disclosure provides an apparatus 600 for adjusting magnetic resonance imaging (MRI) scan parameters, including a processor 601 and a memory 602. Optionally, the apparatus may further include a communication interface 603 and a bus 604. The processor 601, communication interface 603, and memory 602 can communicate with each other via the bus 604. The communication interface 603 can be used for information transmission. The processor 601 can call logical instructions in the memory 602 to execute the method for adjusting MRI scan parameters described in the above embodiment.

[0094] Furthermore, the logic instructions in the aforementioned memory 602 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0095] The memory 602, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 601 executes functional applications and data processing by running the program instructions / modules stored in the memory 602, that is, it implements the method for adjusting magnetic resonance scanning parameters in the above embodiments.

[0096] The memory 602 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 602 may include high-speed random access memory and may also include non-volatile memory.

[0097] This disclosure provides a scanning device, including: a scanning device body, and the aforementioned device for adjusting magnetic resonance scanning parameters. The device for adjusting magnetic resonance scanning parameters is installed in the scanning device body. The installation relationship described herein is not limited to placement within the scanning device, but also includes installation connections with other components of the scanning device, including but not limited to physical connections, electrical connections, or signal transmission connections. Those skilled in the art will understand that the device for adjusting magnetic resonance scanning parameters can be adapted to any suitable scanning device body, thereby realizing other feasible embodiments.

[0098] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to perform the above-described method for adjusting magnetic resonance scanning parameters.

[0099] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code.

[0100] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the technical solutions described herein. As used in the technical solutions described herein, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used herein refers to any and all possible combinations of one or more of the associated listed elements. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0101] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0102] The methods and products disclosed in the embodiments herein (including but not limited to devices and equipment) can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0103] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A method for adjusting magnetic resonance scanning parameters, characterized in that, include: Acquire configuration data for magnetic resonance imaging (MRI) scans and multi-source clinical evaluation data; wherein, the configuration data includes initial MRI scan parameters, multi-source clinical evaluation data, and the configuration data are correlated. The configuration data and multi-source clinical evaluation data are input into the image quality prediction model to predict the image quality corresponding to the initial magnetic resonance scan parameters, and the initial image quality prediction results are obtained. Based on the initial image quality prediction results, perform magnetic resonance scanning or generate target optimization suggestions for initial magnetic resonance scanning parameters.

2. The method according to claim 1, characterized in that, The image quality prediction model is trained as follows: Multiple first historical scan samples were acquired. Each first historical scan sample included configuration data of magnetic resonance imaging scans, multi-source clinical evaluation data, and multidimensional data labels. Among them, the multidimensional data labels included image quality score labels and multiple risk index labels. Feature extraction is performed on each first historical scan sample to obtain the first multimodal feature corresponding to each first historical scan sample; the first multimodal feature includes the scan parameter features of the configuration data and the text features of the multi-source clinical evaluation data; The model is trained based on the first multimodal features and multidimensional data labels of each first historical scan sample to obtain the trained image quality prediction model; the output of the image quality prediction model is the predicted image quality score and multiple risk index predicted values.

3. The method according to claim 1, characterized in that, Also includes: Get the currently scanned image; The current scan image is acquired by performing a magnetic resonance scan, or by performing a magnetic resonance scan after adjusting the initial magnetic resonance scan parameters according to the target optimization suggestions; Based on the parameter optimization model, configuration data, multi-source clinical evaluation data and current scan images are analyzed to obtain the first optimization suggestion for the current magnetic resonance scan parameters; In response to the user adjusting the current magnetic resonance scanning parameters according to the first optimization suggestion, the adjusted current magnetic resonance scanning parameters are integrated with the configuration data and then input together with the multi-source clinical evaluation data into the image quality prediction model to perform image quality prediction and obtain the target image quality prediction result. Based on the target image quality prediction results, a second optimization suggestion for the current magnetic resonance scanning parameters is generated.

4. The method according to claim 3, characterized in that, The optimized model is trained as follows: Acquire multiple second historical scan samples; each second historical scan sample includes magnetic resonance scan configuration data, multi-source clinical evaluation data and scan images, as well as optimal scan parameters; Feature extraction is performed on each second historical scan sample to obtain the second multimodal features corresponding to each second historical scan sample; the second multimodal features include the scan parameter features of the configuration data, the text features of the multi-source clinical evaluation data, and the visual features of the scan images; The model is trained based on the second multimodal features and optimal scanning parameters of each second historical scan sample to obtain a trained parameter optimization model. The output of the parameter optimization model includes at least one of the following: suggested parameters, optimization scheme for specific lesions, and performance prediction curve.

5. The method according to claim 2 or 4, characterized in that, The characteristics of scanning parameters include the values ​​of magnetic resonance scanning parameters and the coupling relationship between magnetic resonance scanning parameters.

6. The method according to claim 2 or 4, characterized in that, The configuration data for magnetic resonance scanning also includes magnetic resonance auxiliary parameters, and the first and second multimodal features also include auxiliary parameter features.

7. The method according to claim 3, characterized in that, The magnetic resonance scanning parameters include a set of associated parameters. The set of associated parameters consists of at least two parameters with preset linkage constraints. When adjusting the magnetic resonance scanning parameters, once one parameter in the set of associated parameters is adjusted, the other parameters in the set of associated parameters are updated synchronously based on the preset linkage constraints.

8. The method according to claim 2 or 3, characterized in that, Image quality prediction results include predicted image quality scores and risk index values, generating optimization suggestions for magnetic resonance scanning parameters, including: The predicted values ​​of image quality score and risk index are compared with their respective preset thresholds to determine the dimensions that do not meet the standards in the current combination of scanning parameters. Based on the identified non-compliant dimension type, query the pre-defined parameter adjustment rule mapping table and match the corresponding adjustment strategy; The image quality prediction results include the initial image quality prediction results and the target image quality prediction results. The magnetic resonance scanning parameters include the initial magnetic resonance scanning parameters and the current magnetic resonance scanning parameters. The optimization suggestions include the target optimization suggestions and the second optimization suggestions. The mapping table has a pre-set correspondence between different non-compliant dimension types and parameter adjustment directions, adjustment steps, and adjustment reasons.

9. The method according to claim 1 or 3, characterized in that, Also includes: Based on the user's adoption of optimization suggestions and the actual scanning results obtained after adjustments based on the optimization suggestions, the image quality prediction model and parameter optimization model are learned and updated online. The optimization suggestions include objective optimization suggestions and secondary optimization suggestions.

10. An apparatus for adjusting magnetic resonance scanning parameters, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to, when running the program instructions, perform the method for adjusting magnetic resonance scan parameters as described in any one of claims 1 to 9.

11. A scanning device, characterized in that, include: Scan the device itself; The apparatus for adjusting magnetic resonance scanning parameters as described in claim 10 is installed on the scanning device body.

Images