A smart control method and system for magnetic resonance imaging equipment

Abstract

This invention discloses an intelligent control method and system for magnetic resonance imaging (MRI) equipment, relating to the field of MRI control technology. This system reduces the need for manual intervention through a fully automated parameter adjustment and quality assessment mechanism. This not only improves the efficiency of the scanning process but also ensures that the system can automatically optimize acquisition parameters Po, including scan time To, resolution Ro, and signal intensity So, based on actual conditions under different patients and scanning scenarios, thereby improving image quality and diagnostic accuracy. The automation process simplifies operation and reduces human error. In addition to focusing on traditional image quality standards, it also performs quality assessments using noise level Nr and sharpness Cr, ensuring more comprehensive and detailed image quality. The system combines a feedback mechanism with a continuous optimization module, continuously acquiring image quality assessment results and adjusting acquisition parameters accordingly to gradually optimize the scanning process.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging control technology, specifically to an intelligent control method and system for magnetic resonance imaging equipment. Background Technology

[0002] Magnetic resonance imaging (MRI) is a non-invasive, radiation-free medical imaging technique widely used in clinical diagnosis, playing a crucial role in the early screening and precise localization of diseases of the nervous, musculoskeletal, and cardiovascular systems. With the development of medical imaging technology, MRI technology is gradually moving towards higher resolution, faster imaging, and higher quality images. However, traditional MRI control systems primarily rely on preset scanning parameters, such as scan time, resolution, and signal intensity, for image acquisition.

[0003] Currently, most MRI control systems still rely on fixed parameter settings for scanning. While these parameters can provide basic image quality under standardized conditions, suboptimal image quality often occurs when dealing with different patients, physical conditions, and scanning sites. For example, excessively long scan times can cause patient discomfort, especially when prolonged stillness is required, leading to anxiety and unease, and potentially causing patients to interrupt the examination due to lack of cooperation. Insufficient resolution can result in unclear image details, affecting early detection and diagnosis of diseases. Insufficient signal strength can lead to excessive image noise, resulting in poor image quality and impacting the accuracy of diagnosis. Therefore, how to adjust acquisition parameters in real time according to actual conditions to ensure optimal image quality has become a major challenge in the application of current MRI technology.

[0004] This phenomenon primarily stems from the limitations of traditional MRI control systems, which cannot dynamically adjust scanning parameters based on real-time image quality feedback. Because parameter settings are typically based on fixed standards or empirical values, they cannot effectively address individual patient differences and variations in the scanning environment during actual operation. As a result, patients may experience an uncomfortable scanning process, or even misdiagnosis or missed diagnosis due to image quality issues. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent control method and system for magnetic resonance imaging equipment, solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: an intelligent control system for a magnetic resonance imaging device, comprising a data acquisition and preliminary analysis module, a deep learning model analysis module, a parameter adaptive adjustment module, an image resampling module, an image quality verification and quality assessment module, and a feedback mechanism and continuous optimization module;

[0007] The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr.

[0008] The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po.

[0009] The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon.

[0010] The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro.

[0011] The image quality verification and quality assessment module verifies the quality of image data Iro and obtains the quality assessment result Qo.

[0012] The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

[0013] Preferably, the data acquisition and preliminary analysis module includes an image data acquisition unit and a preliminary quality assessment unit;

[0014] The image data acquisition unit is responsible for acquiring raw image data Ir from the magnetic resonance imaging equipment;

[0015] The preliminary quality assessment unit performs preliminary noise and sharpness assessments on the acquired raw image data Ir, obtaining the noise level Nr and sharpness Cr;

[0016] The noise level Nr is obtained by the following formula: By calculating the noise level of an image, the influence of external interference and hardware problems on the image during the acquisition process can be reflected; the noise level Nr represents the noise level in the image; noise is usually reflected in the difference between pixel values. By calculating the difference between an image pixel and its neighboring pixels and averaging them, the noise index of the image can be obtained; if the noise level is high, it indicates that the image may be interfered with and needs to be optimized in subsequent acquisition processes.

[0017] ;

[0018] In the formula, n represents the number of pixels in the original image data Ir, Iri represents the value of the i-th pixel in the original image data Ir, and vIri represents the average value of the neighboring pixels of the i-th pixel in the original image data Ir.

[0019] Sharpness Cr is obtained using the following formula: Sharpness Cr calculates the sharpness of an image; the sharper the image, the greater the difference between the pixel value and the surrounding pixels, resulting in a larger numerator and a smaller denominator. The higher the sharpness Cr value, the better the image sharpness.

[0020] ;

[0021] In the formula, μ represents the average value of all pixel values ​​in the original image data Ir.

[0022] Preferably, the deep learning model analysis module includes a feature extraction and analysis unit and a parameter inference unit;

[0023] The feature extraction and analysis unit inputs the acquired raw image data Ir, noise level Nr, and sharpness Cr into the convolutional neural network model, and extracts features from the raw image data Ir through the convolutional layers in the convolutional neural network model, and calculates the image quality score Qe and the probability of potential problems PI.

[0024] The extracted features include edge features, texture features, and shape features;

[0025] The steps to obtain the image quality score Qe are as follows:

[0026] S1. Input the original image data Ir, noise level Nr, and sharpness Cr into the convolutional layer of the convolutional neural network model and extract features. Let the output of the p-th convolutional layer be F(p), and the formula is as follows:

[0027] ;

[0028] In the formula, Conv represents the convolutional layer operation function, W(p) represents the convolutional kernel of the p-th layer, b(p) represents the bias term, and F(p) represents the feature map of the p-th layer.

[0029] S2. Perform dimensionality reduction on the feature map F(p) of the p-th layer using a pooling layer to obtain the pooled feature map P(p); the pooled feature map P(p) of the p-th layer is obtained using the following formula:

[0030] ;

[0031] In the formula, Pool represents the pooling layer operation function;

[0032] S3. Combine the extracted features through a fully connected layer to calculate and obtain the image quality score Qe;

[0033] Image quality score Qe is obtained using the following formula:

[0034] ;

[0035] In the formula, f represents the ReLU activation function, W(fc) represents the weights of the fully connected layer, b(fc) represents the bias term of the fully connected layer, and fc represents the fully connected layer in the convolutional neural network model.

[0036] The probability PI of a potential problem occurring is obtained using the following formula:

[0037] ;

[0038] In the formula, This represents the Sigmoid activation function. PI represents the probability of a potential problem occurring, indicating the probability of the presence of potential problems in the image, including artifacts and noise, with values ​​between [0,1].

[0039] Preferably, the parameter estimation unit calculates and obtains the acquisition parameter Po based on the acquired image quality score Qe and the probability of occurrence of potential problems PI, wherein the acquisition parameter Po includes the scan time To, resolution Ro and signal strength So;

[0040] The scan time To is obtained using the following formula:

[0041] ;

[0042] In the formula, PI represents the time adjustment factor; PI represents the probability of potential problems occurring. If the system identifies serious motion artifacts or hardware problems, the probability of potential problems occurring, PI, will be larger. If the image quality is poor or there are problems, the system will automatically increase the scanning time.

[0043] Resolution Ro is obtained using the following formula:

[0044] ;

[0045] In the formula, This indicates the resolution adjustment factor; if the image sharpness is low, or the impact of potential problems is minor, the system will automatically increase the resolution.

[0046] Signal strength So is obtained using the following formula:

[0047] ;

[0048] In the formula, This represents the signal adjustment factor; if the noise level is high, the system will infer to increase the signal strength to improve image quality.

[0049] Preferably, the parameter adaptive adjustment module adjusts the acquisition parameters Po of the magnetic resonance imaging device and combines them with the image quality score Qe(t) at time t and the probability of occurrence of potential problems PI(t) at time t to calculate and obtain the adjusted acquisition parameters Pon, including the adjusted scan time Ton, the adjusted resolution Ron and the adjusted signal strength Son.

[0050] The adjusted scan time Ton is obtained using the following formula:

[0051] ;

[0052] In the formula, ΔT represents the time adjustment amount;

[0053] The time adjustment amount ΔT is obtained using the following formula;

[0054] ;

[0055] In the formula, The preset weight value represents the probability of the potential problem occurring at time t;

[0056] The adjusted resolution Ron is obtained using the following formula:

[0057] ;

[0058] In the formula, ΔR represents the image quality adjustment amount;

[0059] The image quality adjustment amount ΔR is obtained using the following formula;

[0060] ;

[0061] In the formula, C represents a constant;

[0062] The adjusted signal strength Son is obtained using the following formula:

[0063] ;

[0064] In the formula, ΔS represents the signal strength adjustment amount;

[0065] The signal strength adjustment amount ΔS is obtained using the following formula;

[0066] ;

[0067] In the formula, B1 represents a constant, and B2 represents a real number. represents the weights of the influence of noise level and potential problems on signal strength.

[0068] Preferably, the image resampling module includes an image interpolation unit and a new image acquisition unit;

[0069] The image interpolation unit adjusts the size and quality of the original image data Ir using interpolation techniques based on the adjusted scan time Ton and the adjusted resolution Ron. The purpose of the interpolation operation is to generate a new image based on the original image, according to the adjusted scan time Ton and the adjusted resolution Ron, in order to improve the accuracy and quality of the image.

[0070] The size of the original image data Ir is adjusted by adjusting the resolution Ron. The original image data Ir has a size of (Wr, Hr). By adjusting the resolution Ron, a new image size of (Wro, Hro) is obtained.

[0071] Where Wro = Ron * Wr, Hro = Ron * Hr;

[0072] For the pixels in the original image data Ir, the adjusted image data Irn(x,y) is obtained by calculating using bicubic interpolation.

[0073] The details of the adjusted image data Irn(x,y) are adjusted by adjusting the scan time Ton. Specifically, the details of the interpolated adjusted image data Irn(x,y) are enhanced by a filter, and the enhancement formula is as follows:

[0074] The scanning time can be adjusted to control the reconstruction of details during the interpolation of image data; a longer scanning time usually means more data acquisition points, resulting in higher image accuracy; for this effect, the details of the interpolated image can be smoothed and enhanced.

[0075] ;

[0076] In the formula, G represents the enhancement coefficient, and Filter represents the smoothing filter.

[0077] Preferably, the new image acquisition unit adjusts the brightness and contrast of the adjusted image data Irn(x,y) using the adjusted signal strength Son, and acquires image data Iro.

[0078] Changes in signal strength directly affect the brightness and contrast of an image; when the signal strength Son increases, the brightness of the image will increase; while when the signal strength decreases, the image may need contrast enhancement to compensate for the low signal conditions.

[0079] Image data Iro is obtained using the following formula:

[0080] ;

[0081] In the formula, k1 represents the brightness gain factor, and k2 represents the contrast adjustment parameter; usually, the contrast is increased when the signal strength Son is low after adjustment in order to maintain the visibility of image details.

[0082] Preferably, the image quality verification and quality assessment module analyzes the noise characteristics, sharpness characteristics, and contrast characteristics of image data Iro to obtain noise assessment SNR (Iro), sharpness assessment C (Iro), and contrast assessment CD (Iro), thereby assessing the quality of image data Iro and obtaining the quality assessment result Qo.

[0083] The noise assessment SNR (Iro) is obtained using the following formula:

[0084] ;

[0085] In the formula, μIro represents the mean value of the Iro pixel values ​​in the image data, and σIro represents the standard deviation of the Iro pixel values ​​in the image data.

[0086] The sharpness assessment C (Iro) is obtained using the following formula:

[0087] ;

[0088] In the formula, N represents the total number of pixels in the image data Iro. This represents the gradient value at pixel (i, j);

[0089] Contrast ratio (CD) (Iro) is obtained using the following formula:

[0090] ;

[0091] In the formula, max(Iro) represents the peak value of grayscale in image data Iro, and min(Iro) represents the valley value of grayscale in image data Iro;

[0092] The quality assessment result Qo is obtained using the following formula:

[0093] ;

[0094] In the formula, These represent the preset weight values ​​for noise evaluation SNR (Iro), sharpness evaluation C (Iro), and contrast evaluation CD (Iro), respectively. .

[0095] Preferably, the feedback mechanism and continuous optimization module compares the obtained quality assessment result Qo with the preset quality standard threshold TQo to judge the image quality and obtain the image quality status.

[0096] Image quality status is obtained using the following formula:

[0097] When the quality assessment result Qo ≥ the quality standard threshold TQo, the image quality meets the standard and no adjustment of the acquisition parameters is required.

[0098] When the quality assessment result Qo < the quality standard threshold TQo, the image quality does not meet the standard, and the acquisition parameters need to be adjusted to generate a parameter adjustment feedback signal.

[0099] The feedback signal brings the quality assessment result Qo back to the deep learning model analysis module; the collected parameter Po is adjusted through the convolutional neural network model.

[0100] A method for intelligent control of a magnetic resonance imaging (MRI) device includes the following steps:

[0101] Step 1: The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs a preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr.

[0102] Step 2: The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po;

[0103] Step 3: The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon.

[0104] Step 4: The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro;

[0105] Step 5: The image quality verification and quality assessment module performs quality verification on the image data Iro and obtains the quality assessment result Qo;

[0106] Step Six: The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

[0107] This invention provides an intelligent control method and system for magnetic resonance imaging equipment, which has the following beneficial effects:

[0108] (1) During system operation, the acquisition parameters Po, including scan time To, resolution Ro, and signal strength So, are automatically adjusted based on the real-time acquired image data and its quality assessment indicators to ensure that the image quality is maximized for each scan. This intelligent dynamic adjustment not only eliminates the need for manual intervention but also optimizes parameter settings for different patients and scan sites, thereby effectively improving diagnostic results.

[0109] Traditional MRI systems, due to their use of fixed scanning parameters, often result in excessively long scan times or poor image quality, causing unnecessary physical discomfort and psychological stress for patients. However, by automatically adjusting parameters such as scan time and resolution, the system can optimize the scanning process according to actual needs, reducing unnecessary scan time, alleviating patient discomfort, and reducing patient anxiety.

[0110] (2) By introducing a data acquisition and preliminary analysis module, the system can perform a preliminary assessment of the noise level and sharpness of the raw image data, ensuring real-time monitoring of image quality during the acquisition process. This real-time image quality assessment can provide more accurate data support for subsequent deep learning model analysis, thereby enabling precise adjustment of acquisition parameters to ensure optimal image quality. This innovation effectively avoids the problem that fixed parameter settings cannot adapt to different patients and scanning environments.

[0111] The deep learning model analysis module extracts features from image data using convolutional neural networks, including edge features, texture features, and shape features, which are crucial for evaluating image quality. Convolutional layers automatically extract detailed image features, and through a combination of feature maps and pooling layers, image quality is accurately analyzed. This efficient feature extraction allows the system to more comprehensively assess image quality and further improve the accuracy of image diagnosis.

[0112] (3) The system reduces the need for manual intervention through a fully automated parameter adjustment and quality assessment mechanism. This not only improves the efficiency of the scanning process but also ensures that the system can automatically optimize the acquisition parameters Po, such as scan time To, resolution Ro, and signal intensity So, according to different patients and scanning conditions, thereby improving image quality and diagnostic accuracy. The automation process makes operation simpler and reduces the occurrence of human error. In addition to focusing on traditional image quality standards, the system also conducts quality assessments through noise level Nr and sharpness Cr to ensure more comprehensive and detailed image quality. The system combines a feedback mechanism with a continuous optimization module to continuously acquire image quality assessment results and adjust the acquisition parameters accordingly, thereby gradually optimizing the scanning process.

[0113] (4) The system intelligently infers the acquisition parameter Po by combining the image quality score Qe and the probability of potential problems PI. This parameter inference mechanism based on real-time image quality feedback can accurately and dynamically adjust the acquisition parameters according to the specific scanning environment and the patient's needs, thereby avoiding the problem of fixed parameters that cannot be adjusted in real time in traditional MRI systems.

[0114] The adaptive parameter adjustment module ensures optimal image quality for each scan by fine-tuning the acquisition parameter Po. Through flexible adjustments to scan time To, resolution Ro, and signal intensity So, the system can adapt to different clinical needs, ensuring clear and noise-free diagnostic images while avoiding patient discomfort caused by excessively long scan times.

[0115] In practice, the system dynamically adjusts the scan time To, resolution Ro, and signal strength So based on changes in the probability of a potential problem (PI) and the image quality score (Qe). For example, when the probability of a potential problem (PI) is high, the system automatically increases the scan time To and resolution Ro to ensure image quality; conversely, it reduces the scan time To to optimize the efficiency of the diagnostic process. Attached Figure Description

[0116] Figure 1 This is a schematic flowchart of the intelligent control system of a magnetic resonance imaging device according to the present invention;

[0117] Figure 2 This is a schematic diagram of the steps of an intelligent control method for a magnetic resonance imaging device according to the present invention;

[0118] Figure 3 This is a schematic diagram of the process for judging the quality assessment results of the present invention;

[0119] Figure 4 This is a schematic diagram illustrating the steps for obtaining image quality scores according to the present invention.

[0120] Figure 5 This is a line graph diagram of the statistical data chart of the present invention. Detailed Implementation

[0121] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. Example 1

[0122] This invention provides an intelligent control system for a magnetic resonance imaging (MRI) device. Please refer to [link / reference]. Figure 1It includes a data acquisition and preliminary analysis module, a deep learning model analysis module, a parameter adaptive adjustment module, an image resampling module, an image quality verification and quality assessment module, and a feedback mechanism and continuous optimization module;

[0123] The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr.

[0124] The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po.

[0125] The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon.

[0126] The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro.

[0127] The image quality verification and quality assessment module verifies the quality of image data Iro and obtains the quality assessment result Qo.

[0128] The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

[0129] In this embodiment, by introducing a deep learning model, the system can automatically adjust the acquisition parameters Po, including scan time To, resolution Ro, and signal intensity So, based on real-time acquired image data and its quality assessment indicators, ensuring maximum image quality for each scan. This intelligent dynamic adjustment not only eliminates the need for manual intervention but also optimizes parameter settings for different patients and scan sites, thereby effectively improving diagnostic results.

[0130] Traditional MRI systems, due to their use of fixed scanning parameters, often result in excessively long scan times or poor image quality, causing unnecessary physical discomfort and psychological stress for patients. However, by automatically adjusting parameters such as scan time and resolution, the system can optimize the scanning process according to actual needs, reducing unnecessary scan time, alleviating patient discomfort, and reducing patient anxiety.

[0131] The system analyzes raw image data in real time, combining indicators such as noise level and image sharpness to predict potential image quality problems and automatically adjust the acquisition parameters Po. This intelligent optimization effectively avoids image quality issues caused by fixed parameter settings, such as blurry images and excessive noise, improving image quality and thus enhancing doctors' ability to diagnose and accurately locate diseases in their early stages. By adjusting scanning parameters in real time, the intelligent control system ensures optimal image quality for each acquisition, reducing re-examinations caused by substandard image quality and lowering medical costs.

[0132] This system incorporates a feedback mechanism and a continuous optimization module. It continuously acquires image quality assessment results and optimizes acquisition parameters based on these results. This self-learning and continuous optimization capability allows the system to gradually adapt to the needs of different patients and scanning scenarios over long-term use, thereby providing personalized diagnostic support and further improving the system's intelligence and adaptability. Through automated and intelligent control, the system reduces intervention by doctors or operators in parameter settings, saving time and effort on manual adjustments. Doctors can focus more on diagnostic work, while the system ensures efficiency and image quality for each scan, improving the overall efficiency of the MRI examination process. Example 2

[0133] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 and Figure 4 Specifically: the data acquisition and preliminary analysis module includes an image data acquisition unit and a preliminary quality assessment unit;

[0134] The image data acquisition unit is responsible for acquiring raw image data Ir from the magnetic resonance imaging equipment;

[0135] The preliminary quality assessment unit performs preliminary noise and sharpness assessments on the acquired raw image data Ir, obtaining the noise level Nr and sharpness Cr;

[0136] The noise level Nr is obtained using the following formula:

[0137] ;

[0138] In the formula, n represents the number of pixels in the original image data Ir, Iri represents the value of the i-th pixel in the original image data Ir, and vIri represents the average value of the neighboring pixels of the i-th pixel in the original image data Ir.

[0139] Cr resolution is obtained using the following formula:

[0140] ;

[0141] In the formula, μ represents the average value of all pixel values ​​in the original image data Ir.

[0142] The deep learning model analysis module includes a feature extraction and analysis unit and a parameter inference unit;

[0143] The feature extraction and analysis unit inputs the acquired raw image data Ir, noise level Nr, and sharpness Cr into the convolutional neural network model, and extracts features from the raw image data Ir through the convolutional layers in the convolutional neural network model, and calculates the image quality score Qe and the probability of potential problems PI.

[0144] The extracted features include edge features, texture features, and shape features;

[0145] The steps to obtain the image quality score Qe are as follows:

[0146] S1. Input the original image data Ir, noise level Nr, and sharpness Cr into the convolutional layer of the convolutional neural network model and extract features. Let the output of the p-th convolutional layer be F(p), and the formula is as follows:

[0147] ;

[0148] In the formula, Conv represents the convolutional layer operation function, W(p) represents the convolutional kernel of the p-th layer, b(p) represents the bias term, and F(p) represents the feature map of the p-th layer.

[0149] S2. Perform dimensionality reduction on the feature map F(p) of the p-th layer using a pooling layer to obtain the pooled feature map P(p); the pooled feature map P(p) of the p-th layer is obtained using the following formula:

[0150] ;

[0151] In the formula, Pool represents the pooling layer operation function;

[0152] S3. Combine the extracted features through a fully connected layer to calculate and obtain the image quality score Qe;

[0153] Image quality score Qe is obtained using the following formula:

[0154] ;

[0155] In the formula, f represents the ReLU activation function, W(fc) represents the weights of the fully connected layer, b(fc) represents the bias term of the fully connected layer, and fc represents the fully connected layer in the convolutional neural network model.

[0156] The probability PI of a potential problem occurring is obtained using the following formula:

[0157] ;

[0158] In the formula, This represents the Sigmoid activation function.

[0159] In this embodiment, by introducing a data acquisition and preliminary analysis module, the system can perform a preliminary assessment of the noise level and sharpness of the raw image data, ensuring real-time monitoring of image quality during the acquisition process. This real-time image quality assessment provides more accurate data support for subsequent deep learning model analysis, thereby enabling precise adjustment of acquisition parameters to ensure optimal image quality. This innovation effectively avoids the problem that fixed parameter settings cannot adapt to different patients and scanning environments.

[0160] The deep learning model analysis module extracts features from image data using convolutional neural networks, including edge features, texture features, and shape features, which are crucial for evaluating image quality. Convolutional layers automatically extract detailed image features, and through a combination of feature maps and pooling layers, image quality is accurately analyzed. This efficient feature extraction allows the system to more comprehensively assess image quality and further improve the accuracy of image diagnosis.

[0161] The system extracts and analyzes features from the original image using a convolutional neural network, and calculates an image quality score (Qe) by combining indicators such as noise level and sharpness, thereby achieving a comprehensive assessment of image quality. This score calculation process utilizes convolutional and pooling layers, and combines features through fully connected layers, enabling precise quantification of image quality. This mechanism allows image quality to be optimized according to specific diagnostic needs, avoiding the instability and reliance on manual adjustments inherent in traditional MRI systems.

[0162] By calculating the probability PI of potential problems occurring through a deep learning model, the system can identify potential issues in image data in real time, such as excessive noise and image blurring, and issue early warnings. Through this intelligent prediction mechanism, the system can promptly detect and intervene in potential problems during image acquisition, avoiding misdiagnosis or missed diagnosis due to substandard image quality.

[0163] The system reduces the need for manual intervention through fully automated parameter adjustment and quality assessment mechanisms. This not only improves the efficiency of the scanning process but also ensures that the system can automatically optimize acquisition parameters Po, such as scan time To, resolution Ro, and signal intensity So, based on different patients and scanning conditions, thereby improving image quality and diagnostic accuracy. Automation simplifies operation and reduces the occurrence of human error.

[0164] The system not only focuses on traditional image quality standards but also assesses quality through noise level (Nr) and sharpness (Cr) to ensure more comprehensive and detailed image quality. Combining a feedback mechanism and a continuous optimization module, the system continuously acquires image quality assessment results and adjusts acquisition parameters accordingly, gradually optimizing the scanning process. This approach, based on real-time feedback and continuous optimization, enables the MRI scanning system to learn and improve itself, continuously enhancing performance over long-term use and ensuring maximum optimization of the results for each scan. Example 3

[0165] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 and Figure 3 Specifically: the parameter prediction unit calculates and obtains the acquisition parameters Po based on the acquired image quality score Qe and the probability of occurrence of potential problems PI, wherein the acquisition parameters Po include the scan time To, resolution Ro and signal strength So;

[0166] The scan time To is obtained using the following formula:

[0167] ;

[0168] In the formula, Indicates the time adjustment factor;

[0169] Resolution Ro is obtained using the following formula:

[0170] ;

[0171] In the formula, Indicates the resolving adjustment factor;

[0172] Signal strength So is obtained using the following formula:

[0173] ;

[0174] In the formula, This represents the signal adjustment factor.

[0175] The parameter adaptive adjustment module adjusts the acquisition parameters Po of the magnetic resonance imaging device and combines them with the image quality score Qe(t) at time t and the probability of occurrence of potential problems PI(t) at time t to calculate and obtain the adjusted acquisition parameters Pon, including the adjusted scan time Ton, the adjusted resolution Ron, and the adjusted signal strength Son.

[0176] The adjusted scan time Ton is obtained using the following formula:

[0177] ;

[0178] In the formula, ΔT represents the time adjustment amount;

[0179] The time adjustment amount ΔT is obtained using the following formula;

[0180] ;

[0181] In the formula, The preset weight value represents the probability of the potential problem occurring at time t;

[0182] The adjusted resolution Ron is obtained using the following formula:

[0183] ;

[0184] In the formula, ΔR represents the image quality adjustment amount;

[0185] The image quality adjustment amount ΔR is obtained using the following formula;

[0186] ;

[0187] In the formula, C represents a constant;

[0188] The adjusted signal strength Son is obtained using the following formula:

[0189] ;

[0190] In the formula, ΔS represents the signal strength adjustment amount;

[0191] The signal strength adjustment amount ΔS is obtained using the following formula;

[0192] ;

[0193] In the formula, B1 represents a constant and B2 represents a real number.

[0194] In this embodiment, the system intelligently infers the acquisition parameter Po by combining the image quality score Qe and the probability of potential problems occurring PI. This parameter inference mechanism based on real-time image quality feedback can accurately and dynamically adjust the acquisition parameters according to the specific scanning environment and the patient's needs, thereby avoiding the problem of fixed parameters that cannot be adjusted in real time in traditional MRI systems.

[0195] The adaptive parameter adjustment module ensures optimal image quality for each scan by fine-tuning the acquisition parameter Po. Through flexible adjustments to scan time To, resolution Ro, and signal intensity So, the system can adapt to different clinical needs, ensuring clear and noise-free diagnostic images while avoiding patient discomfort caused by excessively long scan times.

[0196] In practice, the system dynamically adjusts the scan time To, resolution Ro, and signal strength So based on changes in the probability of a potential problem (PI) and the image quality score (Qe). For example, when the probability of a potential problem (PI) is high, the system automatically increases the scan time To and resolution Ro to ensure image quality; conversely, it reduces the scan time To to optimize the efficiency of the diagnostic process.

[0197] By fine-tuning the time To, resolution Ro, and signal strength So, the system can significantly reduce unnecessary waiting time during the scanning process while ensuring image quality. Excessive scanning time not only increases the physical burden on patients but can also affect the efficiency of the overall scanning process. Optimizing the system can effectively improve scanning efficiency and reduce patient waiting time and operator workload.

[0198] The system precisely controls the signal strength So and resolution Ro, ensuring that each image acquisition is performed at the optimal signal and resolution. This avoids noise issues caused by excessively low signal strength and overexposure issues caused by excessively high signal strength. Simultaneously, adjusting the resolution Ro allows for optimization based on the required level of detail, ensuring the image possesses high-resolution detail without wasting scanning time due to excessively high resolution.

[0199] The system eliminates the need for manual intervention by automatically adjusting and predicting the acquisition parameters Po. Operators no longer need to manually adjust acquisition parameters under different patient and scanning conditions, which not only improves operational efficiency but also reduces the incidence of human error. The system's adaptive adjustment is based on the actual image quality score Qe and the probability of potential problems PI, and can be fine-tuned based on real-time feedback. Each adjustment, including time adjustment ΔT, resolution adjustment ΔR, and signal intensity adjustment ΔS, is calculated based on precise algorithms to ensure that each adjustment effectively improves image quality and avoids quality degradation or resource waste caused by blind adjustments. Example 4

[0200] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 , Figure 3 and Figure 5 Specifically: the image resampling module includes an image interpolation unit and a new image acquisition unit;

[0201] The image interpolation unit adjusts the size and quality of the original image data Ir using interpolation techniques based on the adjusted scan time Ton and the adjusted resolution Ron.

[0202] The size of the original image data Ir is adjusted by adjusting the resolution Ron. The original image data Ir has a size of (Wr, Hr). By adjusting the resolution Ron, a new image size of (Wro, Hro) is obtained.

[0203] Where Wro = Ron * Wr, Hro = Ron * Hr;

[0204] For the pixels in the original image data Ir, the adjusted image data Irn(x,y) is obtained by calculating using bicubic interpolation.

[0205] The details of the adjusted image data Irn(x,y) are adjusted by adjusting the scan time Ton. Specifically, the details of the interpolated adjusted image data Irn(x,y) are enhanced by a filter, and the enhancement formula is as follows:

[0206] ;

[0207] In the formula, G represents the enhancement coefficient, and Filter represents the smoothing filter.

[0208] The new image acquisition unit adjusts the brightness and contrast of the adjusted image data Irn(x,y) using the adjusted signal strength Son, and acquires the image data Iro.

[0209] Image data Iro is obtained using the following formula:

[0210] ;

[0211] In the formula, k1 represents the brightness gain factor, and k2 represents the contrast adjustment parameter.

[0212] The image quality verification and quality assessment module analyzes the noise characteristics, sharpness characteristics, and contrast characteristics of image data Iro to obtain noise assessment SNR (Iro), sharpness assessment C (Iro), and contrast assessment CD (Iro), thereby assessing the quality of image data Iro and obtaining the quality assessment result Qo.

[0213] The noise assessment SNR (Iro) is obtained using the following formula:

[0214] ;

[0215] In the formula, μIro represents the mean value of the Iro pixel values ​​in the image data, and σIro represents the standard deviation of the Iro pixel values ​​in the image data.

[0216] The sharpness assessment C (Iro) is obtained using the following formula:

[0217] ;

[0218] In the formula, N represents the total number of pixels in the image data Iro. This represents the gradient value at pixel (i, j);

[0219] Contrast ratio (CD) (Iro) is obtained using the following formula:

[0220] ;

[0221] In the formula, max(Iro) represents the peak value of grayscale in image data Iro, and min(Iro) represents the valley value of grayscale in image data Iro;

[0222] The quality assessment result Qo is obtained using the following formula:

[0223] ;

[0224] In the formula, These represent the preset weight values ​​for noise evaluation SNR (Iro), sharpness evaluation C (Iro), and contrast evaluation CD (Iro), respectively. .

[0225] The feedback mechanism and continuous optimization module compare the obtained quality assessment result Qo with the preset quality standard threshold TQo to judge the image quality and obtain the image quality status.

[0226] Image quality status is obtained using the following formula:

[0227] When the quality assessment result Qo ≥ the quality standard threshold TQo, the image quality meets the standard and no adjustment of the acquisition parameters is required.

[0228] When the quality assessment result Qo < the quality standard threshold TQo, the image quality does not meet the standard, and the acquisition parameters need to be adjusted to generate a parameter adjustment feedback signal.

[0229] The feedback signal brings the quality assessment result Qo back to the deep learning model analysis module; the collected parameter Po is adjusted through the convolutional neural network model.

[0230] Specific examples:

[0231] The preset quality standard threshold TQo = 10.5;

[0232] The noise assessment SNR (Iro) was obtained as 8.00;

[0233] Sharpness assessment C(Iro) = 25;

[0234] Contrast ratio assessment CD (Iro) = 0.80;

[0235] Preset They are 0.4, 0.3, and 0.3 respectively;

[0236] Calculate the quality assessment result Qo:

[0237] ;

[0238] The quality assessment result Qo = 10.94 > the quality standard threshold TQo = 10.5, indicating that the image quality meets the standard.

[0239] Table 1

[0240] ;

[0241] In this embodiment, the image resampling module uses the adjusted scan time Ton and adjusted resolution Ron to adaptively interpolate and adjust the original image data Ir. By adjusting the resolution to modify the image size, and combining this with bicubic interpolation for detail optimization, the image size and quality can be precisely matched to the current acquisition requirements. This process allows for flexible adjustment of the image size while ensuring image quality, meeting the needs of different clinical scenarios.

[0242] Following image interpolation, image detail enhancement uses a smoothing filter and enhancement coefficient G to enhance image details. This detail enhancement process optimizes image quality, allowing image sharpness to be maintained or even improved while adjusting image size. This optimization is particularly important for certain clinical applications with high detail requirements, providing more refined and clearer image data.

[0243] Image brightness and contrast are dynamically adjusted via the adjusted signal strength (Son). The brightness gain factor (k1) and contrast adjustment parameter (k2) control the image display effect, ensuring that both brightness and contrast match the actual scanning conditions, avoiding image distortion or failure to clearly display important information due to unsuitable brightness or contrast. The image quality verification and assessment module evaluates image quality from multiple dimensions using noise assessment (SNR) (Iro), sharpness assessment (C) (Iro), and contrast assessment (CD) (Iro). This comprehensive quality assessment method can comprehensively measure various image quality indicators, ensuring accurate and reliable evaluation results. This is crucial for clinical diagnosis, helping doctors quickly assess whether images meet standards and make better judgments accordingly.

[0244] In image quality assessment, noise and sharpness are accurately evaluated using the pixel mean, standard deviation, and gradient values ​​in a formula. This allows the system to accurately identify noise levels in images and automatically adjust parameters to improve sharpness, avoiding low-quality images that could negatively impact diagnosis. In this way, image quality improvement relies not only on parameter tuning but also on automatic optimization through intelligent algorithms. The feedback mechanism and continuous optimization module automatically determine whether the image meets the standard by comparing the quality assessment results with a preset quality standard threshold TQo and generates a feedback signal based on the assessment results. If the image quality does not meet the standard, the system automatically returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po. This continuous optimization feedback mechanism ensures that optimal image quality can be achieved through parameter fine-tuning after each scan.

[0245] The system can optimize image quality based on real-time feedback, ensuring that the final image always meets the preset quality standard threshold TQo. This process helps eliminate image quality problems caused by suboptimal scanning conditions, guarantees that the final image meets high standards for clinical diagnosis, and avoids misdiagnosis due to quality issues.

[0246] Precise image acquisition and quality adjustment not only improve diagnostic accuracy but also significantly increase diagnostic efficiency by reducing repetitive scanning time. For patients, this translates to a more efficient and comfortable diagnostic experience, avoiding repeated scanning processes caused by substandard image quality. Through automated image acquisition optimization and continuous feedback mechanisms, the system can automatically adjust parameters based on real-time image quality, avoiding unnecessary repetitive scans, maximizing the utilization efficiency of MRI equipment, preventing waste of equipment resources, and thus improving hospital operational efficiency. Example 5

[0247] A smart control method for a magnetic resonance imaging device, please refer to Figure 2 Specifically, it includes the following steps:

[0248] Step 1: The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs a preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr.

[0249] Step 2: The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po;

[0250] Step 3: The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon.

[0251] Step 4: The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro;

[0252] Step 5: The image quality verification and quality assessment module performs quality verification on the image data Iro and obtains the quality assessment result Qo;

[0253] Step Six: The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

[0254] In this embodiment, an automated quality assessment and acquisition parameter adjustment process is implemented. After each acquisition, the system can analyze the image quality in real time and automatically adjust the acquisition parameters Po, including scan time To, resolution Ro, and signal intensity So, based on the assessment results. This reduces the need for manual intervention and optimizes the operation process. By integrating deep learning algorithms, the system can perform real-time quality assessment of the raw image data, automatically predict the probability PI of potential problems, and infer the acquisition parameters. This allows the MRI equipment to dynamically adjust according to the actual image quality during each scan, ensuring that the final image meets the standards and avoiding misdiagnosis or rescanning due to low-quality images.

[0255] The image resampling module re-acquires images based on the real-time adjusted acquisition parameters Pon, ensuring that image size, quality, and detail are effectively adjusted. During the adjustment process, deep learning models intelligently predict and dynamically optimize parameters to maximize image sharpness and detail, reducing the complexity of manual adjustments and enhancing the diagnostic value of the images.

[0256] Continuous optimization and feedback mechanism: The method's feedback mechanism and continuous optimization function enable the system to adjust acquisition parameters based on quality assessment results after each image acquisition. This adaptive optimization method ensures quality improvement after each image acquisition, avoiding repeated scanning due to inappropriate parameters. This not only improves image quality but also reduces the time and equipment resource waste required for repeated scanning due to substandard image quality.

[0257] The data acquisition and preliminary analysis module evaluates the noise level (Nr) and sharpness (Cr) to ensure that the deep learning model analysis module is based on accurate preliminary data. This step helps the system to analyze noise and sharpness issues in images in detail and dynamically adjust parameters based on the probability of these issues occurring, further optimizing image quality. Through deep learning analysis and adaptive adjustment, the system can automatically optimize acquisition parameters based on real-time feedback without manual intervention. This eliminates the need for time-consuming adjustments to scan settings after each scan and avoids rescanning due to substandard image quality, improving overall diagnostic efficiency and reducing patient waiting time.

[0258] Through intelligent control and automatic optimization, the scanning efficiency of the equipment has been significantly improved. Repeated scans caused by improper parameter adjustments are avoided, increasing equipment utilization and efficiency. This not only saves time but also increases equipment productivity, thereby reducing operating costs. The design of the feedback mechanism and continuous optimization module enables this intelligent control method to continuously self-optimize. The quality assessment results after each scan provide new feedback to the system, which dynamically adjusts the acquisition parameters based on this feedback, thereby gradually improving image quality and ensuring that each scan achieves optimal results.

[0259] This intelligent control method allows magnetic resonance imaging (MRI) equipment to more accurately adapt to the needs of different patients and different scan sites, optimizing image quality in real time. This provides doctors with higher-quality imaging data, thereby improving the accuracy and reliability of clinical diagnosis and enhancing support for early disease detection and diagnosis.

[0260] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent control system for a magnetic resonance imaging device, characterized in that: It includes modules for data acquisition and preliminary analysis, deep learning model analysis, adaptive parameter adjustment, image resampling, image quality verification and quality assessment, and feedback mechanism and continuous optimization. The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr. The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po. The deep learning model analysis module includes a feature extraction and analysis unit and a parameter inference unit; The feature extraction and analysis unit inputs the acquired raw image data Ir, noise level Nr, and sharpness Cr into the convolutional neural network model, and extracts features from the raw image data Ir through the convolutional layers in the convolutional neural network model, and calculates the image quality score Qe and the probability of potential problems PI. The extracted features include edge features, texture features, and shape features; The steps to obtain the image quality score Qe are as follows: S1. Input the original image data Ir, noise level Nr, and sharpness Cr into the convolutional layer of the convolutional neural network model and extract features. Let the output of the p-th convolutional layer be F(p), and the formula is as follows: ； In the formula, Conv represents the convolutional layer operation function, W(p) represents the convolutional kernel of the p-th layer, b(p) represents the bias term, and F(p) represents the feature map of the p-th layer. S2. Perform dimensionality reduction on the feature map F(p) of the p-th layer using a pooling layer to obtain the pooled feature map P(p); the pooled feature map P(p) of the p-th layer is obtained using the following formula: ； In the formula, Pool represents the pooling layer operation function; S3. Combine the extracted features through a fully connected layer to calculate and obtain the image quality score Qe; Image quality score Qe is obtained using the following formula: ； In the formula, f represents the ReLU activation function, W(fc) represents the weights of the fully connected layer, b(fc) represents the bias term of the fully connected layer, and fc represents the fully connected layer in the convolutional neural network model. The probability PI of a potential problem occurring is obtained using the following formula: ； In the formula, This represents the Sigmoid activation function; The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon. The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro. The image quality verification and quality assessment module verifies the quality of image data Iro and obtains the quality assessment result Qo. The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

2. The intelligent control system for a magnetic resonance imaging device according to claim 1, characterized in that: The data acquisition and preliminary analysis module includes an image data acquisition unit and a preliminary quality assessment unit; The image data acquisition unit is responsible for acquiring raw image data Ir from the magnetic resonance imaging equipment; The preliminary quality assessment unit performs preliminary noise and sharpness assessments on the acquired raw image data Ir, obtaining the noise level Nr and sharpness Cr; The noise level Nr is obtained using the following formula: ； In the formula, n represents the number of pixels in the original image data Ir, Iri represents the value of the i-th pixel in the original image data Ir, and vIri represents the average value of the neighboring pixels of the i-th pixel in the original image data Ir. Cr resolution is obtained using the following formula: ； In the formula, μ represents the average value of all pixel values ​​in the original image data Ir.

3. The intelligent control system for a magnetic resonance imaging device according to claim 2, characterized in that: The parameter inference unit calculates and obtains the acquisition parameters Po based on the acquired image quality score Qe and the probability of occurrence of potential problems PI, wherein the acquisition parameters Po include the scan time To, resolution Ro and signal strength So; The scan time To is obtained using the following formula: ； In the formula, Indicates the time adjustment factor; Resolution Ro is obtained using the following formula: ； In the formula, Indicates the resolving adjustment factor; Signal strength So is obtained using the following formula: ； In the formula, This represents the signal adjustment factor.

4. The intelligent control system for a magnetic resonance imaging device according to claim 3, characterized in that: The parameter adaptive adjustment module adjusts the acquisition parameters Po of the magnetic resonance imaging device and combines them with the image quality score Qe(t) at time t and the probability of occurrence of potential problems PI(t) at time t to calculate and obtain the adjusted acquisition parameters Pon, including the adjusted scan time Ton, the adjusted resolution Ron, and the adjusted signal strength Son. The adjusted scan time Ton is obtained using the following formula: ； In the formula, ΔT represents the time adjustment amount; The time adjustment amount ΔT is obtained using the following formula; ； In the formula, The preset weight value represents the probability of the potential problem occurring at time t; The adjusted resolution Ron is obtained using the following formula: ； In the formula, ΔR represents the image quality adjustment amount; The image quality adjustment amount ΔR is obtained using the following formula; ； In the formula, C represents a constant; The adjusted signal strength Son is obtained using the following formula: ； In the formula, ΔS represents the signal strength adjustment amount; The signal strength adjustment amount ΔS is obtained using the following formula; ； In the formula, B1 represents a constant and B2 represents a real number.

5. The intelligent control system for a magnetic resonance imaging device according to claim 4, characterized in that: The image resampling module includes an image interpolation unit and a new image acquisition unit; The image interpolation unit adjusts the size and quality of the original image data Ir using interpolation techniques based on the adjusted scan time Ton and the adjusted resolution Ron. The size of the original image data Ir is adjusted by adjusting the resolution Ron. The original image data Ir has a size of (Wr, Hr). By adjusting the resolution Ron, a new image size of (Wro, Hro) is obtained. Where Wro = Ron * Wr, Hro = Ron * Hr; For the pixels in the original image data Ir, the adjusted image data Irn(x,y) is obtained by calculating using bicubic interpolation. The details of the adjusted image data Irn(x,y) are adjusted by adjusting the scan time Ton. Specifically, the details of the interpolated adjusted image data Irn(x,y) are enhanced by a filter, and the enhancement formula is as follows: ； In the formula, G represents the enhancement coefficient, Filter represents the smoothing filter, and (x, y) represents the pixel coordinates.

6. The intelligent control system for a magnetic resonance imaging device according to claim 5, characterized in that: The new image acquisition unit adjusts the brightness and contrast of the adjusted image data Irn(x,y) using the adjusted signal strength Son, and acquires the image data Iro. Image data Iro is obtained using the following formula: ； In the formula, k1 represents the brightness gain factor, and k2 represents the contrast adjustment parameter.

7. The intelligent control system for a magnetic resonance imaging device according to claim 6, characterized in that: The image quality verification and quality assessment module analyzes the noise characteristics, sharpness characteristics, and contrast characteristics of image data Iro to obtain noise assessment SNR (Iro), sharpness assessment C (Iro), and contrast assessment CD (Iro), thereby assessing the quality of image data Iro and obtaining the quality assessment result Qo. The noise assessment SNR (Iro) is obtained using the following formula: ； In the formula, μIro represents the mean value of the Iro pixel values ​​in the image data, and σIro represents the standard deviation of the Iro pixel values ​​in the image data. The sharpness assessment C (Iro) is obtained using the following formula: ； In the formula, N represents the total number of pixels in the image data Iro. This represents the gradient value at pixel (i, j); Contrast ratio (CD) (Iro) is obtained using the following formula: ； In the formula, max(Iro) represents the peak value of grayscale in image data Iro, and min(Iro) represents the valley value of grayscale in image data Iro; The quality assessment result Qo is obtained using the following formula: ； In the formula, These represent the preset weight values ​​for noise evaluation SNR (Iro), sharpness evaluation C (Iro), and contrast evaluation CD (Iro), respectively. .

8. The intelligent control system for a magnetic resonance imaging device according to claim 7, characterized in that: The feedback mechanism and continuous optimization module compare the obtained quality assessment result Qo with the preset quality standard threshold TQo to judge the image quality and obtain the image quality status. Image quality status is obtained using the following formula: When the quality assessment result Qo ≥ the quality standard threshold TQo, the image quality meets the standard and no adjustment of the acquisition parameters is required. When the quality assessment result Qo < the quality standard threshold TQo, the image quality does not meet the standard, and the acquisition parameters need to be adjusted to generate a parameter adjustment feedback signal. The feedback signal brings the quality assessment result Qo back to the deep learning model analysis module; the collected parameter Po is adjusted through the convolutional neural network model.

9. A method for intelligent control of a magnetic resonance imaging (MRI) device, applied to an intelligent control system of an MRI device according to any one of claims 1 to 8, characterized in that: Includes the following steps: Step 1: The data acquisition and preliminary analysis module acquires raw image data Ir from the magnetic resonance imaging equipment and performs a preliminary quality assessment to obtain preliminary assessment indicators, including noise level Nr and sharpness Cr. Step 2: The deep learning model analysis module uses a convolutional neural network model and combines noise level Nr and sharpness Cr to perform real-time analysis on the original image data Ir, predict the image quality score Qe and the probability of potential problems PI of the original image data Ir, and infer the acquisition parameters Po; The deep learning model analysis module includes a feature extraction and analysis unit and a parameter inference unit; The feature extraction and analysis unit inputs the acquired raw image data Ir, noise level Nr, and sharpness Cr into the convolutional neural network model, and extracts features from the raw image data Ir through the convolutional layers in the convolutional neural network model, and calculates the image quality score Qe and the probability of potential problems PI. The extracted features include edge features, texture features, and shape features; The steps to obtain the image quality score Qe are as follows: S1. Input the original image data Ir, noise level Nr, and sharpness Cr into the convolutional layer of the convolutional neural network model and extract features. Let the output of the p-th convolutional layer be F(p), and the formula is as follows: ； In the formula, Conv represents the convolutional layer operation function, W(p) represents the convolutional kernel of the p-th layer, b(p) represents the bias term, and F(p) represents the feature map of the p-th layer. S2. Perform dimensionality reduction on the feature map F(p) of the p-th layer using a pooling layer to obtain the pooled feature map P(p); the pooled feature map P(p) of the p-th layer is obtained using the following formula: ； In the formula, Pool represents the pooling layer operation function; S3. Combine the extracted features through a fully connected layer to calculate and obtain the image quality score Qe; Image quality score Qe is obtained using the following formula: ； In the formula, f represents the ReLU activation function, W(fc) represents the weights of the fully connected layer, b(fc) represents the bias term of the fully connected layer, and fc represents the fully connected layer in the convolutional neural network model. The probability PI of a potential problem occurring is obtained using the following formula: ； In the formula, This represents the Sigmoid activation function; Step 3: The parameter adaptive adjustment module automatically adjusts the acquisition parameters Po of the magnetic resonance imaging device by using a convolutional neural network model, and obtains the adjusted acquisition parameters Pon. Step 4: The image resampling module re-acquires the original image data Ir according to the adjusted acquisition parameters Pon, and obtains the image data Iro; Step 5: The image quality verification and quality assessment module performs quality verification on the image data Iro and obtains the quality assessment result Qo; Step Six: The feedback mechanism and continuous optimization module judges the image quality based on the obtained quality assessment result Qo. If the image quality does not meet the standard, it returns to the deep learning model analysis module to continue adjusting the acquisition parameters Po.

Images