Magnetic resonance arterial marker image quality evaluation method and system and electronic equipment

Abstract

The invention discloses a magnetic resonance arterial marker image quality evaluation method and system and electronic equipment, and belongs to the technical field of magnetic resonance. The method comprises the following steps: S1, obtaining a to-be-evaluated ASL image, and recognizing a grey matter region and a white matter region in the ASL image; s2, based on signals in a grey matter area and a white matter area in the ASL image, respectively calculating seven quantitative index values including a grey matter-white matter signal ratio, a grey matter variable coefficient, a joint variable coefficient, a contrast noise ratio, an abnormal value ratio, grey matter area connectivity and index structure similarity; and S3, normalizing the seven index values of the ASL image to be evaluated, and inputting the normalized seven index values into a pre-constructed and fitted logistic regression model to obtain an evaluation result representing the quality of the image. The 3D ASL image quality evaluation method can be directly used for quality evaluation of 3D ASL image data acquired clinically and mainstream, and a quick and effective tool is provided for clinical image screening.

Description

Technical Field

[0001] The present invention belongs to the technical field of magnetic resonance, and particularly relates to a method, a system and an electronic device for evaluating the quality of magnetic resonance arterial labeling images. Background Art

[0002] Magnetic resonance spin arterial labeling imaging (ASL) is a non-invasive and non-ionizing radiation method for quantitatively measuring regional cerebral blood flow, and is often applied to the research of brain development and brain diseases in large-scale pediatric cohorts. However, since ASL is sensitive to head movement, the images are easily interfered by artifacts. At present, optimized parameters for ASL imaging in adults have been established clinically, while the imaging parameters for children have not been optimized, and children's images usually have more quality problems, which affect subsequent diagnoses. Therefore, the evaluation of ASL image quality is very important, and this work is currently usually done manually by doctors, which is time-consuming, laborious, and subject to subjective biases. In recent years, some studies have been devoted to automated ASL quality evaluation methods. For example, the authors such as Dolui proposed a method for evaluating the quality of 2D ASL images for adults (Dolui S, Wang Z, Wolf R L, et al. Automated Quality Evaluation Index for Arterial Spin Labeling Derived Cerebral Blood Flow Maps [J]. Journal of Magnetic Resonance Imaging, 2024). This method designed 3 indicators for quality evaluation (the proportion of negative gray matter values, joint spatial variability, and structural similarity), and combined with an exponential fitting model to predict the image quality score. In addition, the authors such as Shirzadi also proposed a method for evaluating the quality of ASL images (Shirzadi Z, Stefanovic B, Chappell M A, et al. Enhancement of automated blood flow estimates (ENABLE) from arterial spin-labeled MRI [J]. Journal of Magnetic Resonance Imaging, 2018, 47(3): 647-655). This method designed 4 indicators (signal-to-noise ratio, contrast-to-noise ratio, detectability rate, and coefficient of variation), and predicted the image quality score by assigning different weights to each indicator. However, most of these methods are designed for adults and 2D ASL images, but the head movement of children is more obvious, and the change of ASL image quality is more complex. Summary of the Invention

[0003] Existing ASL image quality assessment methods mainly focus on adults, and there is still a lack of assessment methods designed for children's images. Children's images still require manual evaluation by doctors, which is not only time-consuming and laborious but may also have human biases. The purpose of the present invention is to solve the above problems existing in the prior art and provide a high-precision and objective automatic assessment method.

[0004] The specific technical solution adopted by the present invention is as follows:

[0005] In the first aspect, the present invention provides a method for assessing the quality of magnetic resonance arterial spin labeling images, which includes:

[0006] S1. Obtain the ASL image to be evaluated and identify the gray matter region and white matter region therein;

[0007] S2. Based on the signals in the gray matter region and white matter region of the ASL image, calculate 7 quantitative index values including the gray-white matter signal ratio, gray matter coefficient of variation, combined coefficient of variation, contrast-to-noise ratio, outlier ratio, gray matter region connectivity, and index structural similarity;

[0008] S3. Normalize the 7 index values of the ASL image to be evaluated and input them into a pre-constructed and fitted logistic regression model to obtain an evaluation result representing the quality of the image.

[0009] As a preference of the above first aspect, when identifying the gray matter region and white matter region in the ASL image, the ASL image needs to be registered to the corresponding T1-weighted image, and the gray and white matter regions are segmented based on the T1-weighted image. The generated gray matter region mask and white matter region mask are applied to the registered ASL image to identify the gray matter region and white matter region in the ASL image.

[0010] As a preference of the above first aspect, among the 7 quantitative index values:

[0011] The gray-white matter signal ratio is defined as the ratio of the signal mean of the gray matter region to the signal mean of the white matter region;

[0012] The gray matter coefficient of variation is defined as the ratio of the signal variance of the gray matter region to the signal mean of the gray matter region;

[0013] The combined coefficient of variation is defined as the ratio of the sum of the signal variances of the gray matter region and the white matter region to the difference between the signal means of the gray matter region and the white matter region;

[0014] The contrast-to-noise ratio is defined as the ratio of the difference between the signal means of the gray matter region and the white matter region to the signal variance of the global range of the two regions;

[0015] The outlier ratio is defined as the ratio of the abnormal pixel points in the global range of the gray matter and white matter regions calculated based on the 3σ criterion;

[0016] The connectivity of the gray matter region is defined as the number of connected regions obtained by segmenting the gray matter region with the reference gray matter signal value as the threshold;

[0017] The index structural similarity is defined as the structural similarity index SSIM between the ASL image to be evaluated and the corresponding pseudo-reference ASL image; in the pseudo-reference ASL image, all white matter regions are assigned the signal mean value of the white matter region in the ASL image to be evaluated, and all gray matter regions are assigned 2.5 times the signal mean value of the white matter region in the ASL image to be evaluated.

[0018] As a preference of the first aspect above, the evaluation result output by the logistic regression model is a binary classification label.

[0019] As a preference of the first aspect above, the logistic regression model is pre-trained on a dataset annotated by experts. For each sample in the dataset, the above-mentioned 7 quantization index values are extracted from the ASL sample image as the model input, and the true label of each sample is a 0-1 label representing the quality of the ASL image.

[0020] As a preference of the first aspect above, the ASL image to be evaluated is a 3D ASL image collected for children.

[0021] In a second aspect, the present invention provides a magnetic resonance arterial labeling image quality evaluation system, which includes:

[0022] An image acquisition module for acquiring the ASL image to be evaluated and identifying the gray matter region and white matter region therein;

[0023] An index quantization module for calculating a total of 7 quantization index values of the gray-white matter signal ratio, gray matter coefficient of variation, combined coefficient of variation, contrast-to-noise ratio, outlier ratio, gray matter region connectivity, and index structural similarity respectively based on the signals in the gray matter region and white matter region of the ASL image;

[0024] A quality evaluation module for normalizing the 7 index values of the ASL image to be evaluated and inputting them into a pre-constructed and fitted logistic regression model to obtain an evaluation result representing the quality of the image.

[0025] In a third aspect, the present invention provides a computer program product, including a computer program / instructions, which when executed by a processor, can implement the magnetic resonance arterial labeling image quality evaluation method as described in any item of the first aspect above.

[0026] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for evaluating the quality of magnetic resonance arterial labeling images as described in any one of the above first aspects can be implemented.

[0027] Fifthly, the present invention provides a computer electronic device, which includes a memory and a processor;

[0028] The memory is used for storing a computer program;

[0029] The processor is used for implementing the method for evaluating the quality of magnetic resonance arterial labeling images as described in any one of the above first aspects when executing the computer program.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention designs a method for evaluating the quality of magnetic resonance arterial labeling images. Compared with the traditional 2D method, it can be directly used for the quality evaluation of 3D ASL image data collected in the mainstream of clinical practice, and the evaluation efficiency is higher. This method is particularly suitable for children's 3D ASL images, and its evaluation accuracy for children's 3D ASL images is higher than that of the evaluation model. The present invention fills the gap in the field of automatic evaluation of the quality of children's ASL images and provides a fast and effective tool for clinical image screening. Description of the Drawings

[0032] Figure 1 It is a schematic diagram of the steps of the method for evaluating the quality of magnetic resonance arterial labeling images;

[0033] Figure 2 It is a flowchart of the method for evaluating the quality of magnetic resonance arterial labeling images;

[0034] Figure 3 It is a schematic diagram of the module composition of the system for evaluating the quality of magnetic resonance arterial labeling images;

[0035] Figure 4 It is a schematic diagram of the structure of the computer electronic device;

[0036] Figure 5 It is the specific implementation steps and comparison and verification steps of the method for evaluating the quality of ASL images in the embodiments of the present invention;

[0037] Figure 6 It is an example diagram of normal segmentation and incorrect segmentation of gray matter in T1W images: (a) Normal segmentation; (b) Incorrect segmentation due to the presence of a tumor;

[0038] Figure 7 It is an ROC curve diagram of two methods in 51 cases of data after training and then testing in the embodiments of the present invention;

[0039] Figure 8 This is an example of the classification results of two methods in cross-validation in the embodiments of the present invention. Detailed implementation manners

[0040] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention with reference to the accompanying drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below. The technical features in various embodiments of the present invention can be combined correspondingly without conflict.

[0041] As Figure 1 shown, in a preferred embodiment of the present invention, a method for evaluating the quality of magnetic resonance arterial spin labeling (ASL) images is provided, which includes the following steps:

[0042] S1. Obtain the ASL image to be evaluated, and identify the gray matter region and white matter region therein.

[0043] It should be noted that the method for identifying the gray matter region and white matter region from the ASL image can adopt any corresponding method in the existing technologies. However, since it is relatively difficult to directly and automatically identify the gray matter region and white matter region from the ASL image, when identifying the gray matter region and white matter region in the ASL image, the ASL image can be registered to the corresponding T1-weighted (T1W) image, and based on the T1-weighted image, the gray matter and white matter regions are segmented. The generated gray matter region mask and white matter region mask are applied to the registered ASL image to identify the gray matter region and white matter region in the ASL image.

[0044] Of course, in addition to identifying the gray and white matter regions based on the T1-weighted image, other medical images can also be combined for auxiliary identification, and this is not limited.

[0045] S2. Based on the signals in the gray matter region and white matter region in the ASL image, calculate a total of 7 quantitative index values, namely the gray-to-white matter signal ratio, the coefficient of variation of gray matter, the combined coefficient of variation, the contrast-to-noise ratio, the proportion of outliers, the connectivity of the gray matter region, and the structural similarity of the index.

[0046] It should be noted that among the above 7 quantitative index values, their respective definitions and calculation methods are as follows:

[0047] 1) The gray-to-white matter signal ratio is defined as the ratio of the average signal of the gray matter region to the average signal of the white matter region. The smaller the deviation between this value and the reference value (2.5:1), the higher the possible image quality.

[0048] 2) The gray matter coefficient of variation is defined as the ratio of the signal variance of the gray matter region to the signal mean of the gray matter region. The signal intensity within the same tissue of high-quality images is usually relatively uniform, while low-quality images often have an increased signal variance due to factors such as head movement and vascular artifacts. Therefore, the smaller this value, the higher the possible quality.

[0049] 3) The combined coefficient of variation is defined as the ratio of the sum of the signal variances of the gray matter region and the white matter region to the difference between the signal means of the gray matter region and the white matter region;

[0050] 4) The contrast-to-noise ratio is defined as the ratio of the difference between the signal means of the gray matter region and the white matter region to the signal variance of the global range of the two regions;

[0051] 5) The outlier ratio is defined as the ratio of the outlier pixels (i.e., pixels within the gray matter region and the white matter region that exceed the global mean ± 3 times the standard deviation) calculated based on the 3σ criterion in the global range of the gray matter and white matter regions;

[0052] 6) The connectivity of the gray matter region is defined as the number of connected regions obtained by segmenting the gray matter region using the reference gray matter signal value as the threshold. The smaller this value, the higher the possible quality.

[0053] 7) The index structural similarity is defined as the structural similarity index (Structural Similarity Index Measurement, SSIM) between the ASL image to be evaluated and the corresponding pseudo-reference ASL image; in the pseudo-reference ASL image, all white matter regions are assigned the signal mean of the white matter region in the ASL image to be evaluated, and all gray matter regions are assigned 2.5 times the signal mean of the white matter region in the ASL image to be evaluated.

[0054] S3. Normalize the 7 index values of the ASL image to be evaluated and input them into the pre-constructed and fitted logistic regression model to obtain the evaluation result representing the quality of the image.

[0055] It should be noted that the evaluation result output by the above logistic regression model is generally a binary classification label. One label represents that the quality of the ASL image to be evaluated is good and it belongs to an available image, and the other label represents that the quality of the ASL image to be evaluated is poor and it belongs to an unavailable image. Before actually using this logistic regression model for quality evaluation, it needs to be pre-trained on a dataset annotated by experts. For each sample in the dataset, the above 7 quantitative index values are extracted from the ASL sample image as the model input, and the true label of each sample is a 0-1 label representing the quality of the ASL image. The logistic regression model needs to be trained until the evaluation performance meets the corresponding requirements before it can be used for actual image quality evaluation.

[0056] For the method for evaluating the quality of magnetic resonance artery-labeled images shown in S1 to S3 above, the overall process can be referred to Figure 2 as shown. The method for evaluating the quality of magnetic resonance artery-labeled images is applicable to 2D ASL images or 3D ASL images. Relevant verification experiments show that this method is particularly applicable to 3D ASL images collected for children, and the evaluation accuracy of this method for 3D ASL images of children is higher than that of the evaluation model. Of course, the method of the present invention can be used to evaluate the quality of any ASL image, and is not limited to being only used for 3D ASL images collected for children.

[0057] Therefore, based on the same inventive concept, as Figure 3 shown, the present invention also provides a magnetic resonance artery-labeled image quality evaluation system corresponding to the magnetic resonance artery-labeled image quality evaluation method provided in the above embodiment, which includes:

[0058] An image acquisition module, configured to acquire an ASL image to be evaluated and identify the gray matter region and white matter region therein;

[0059] An index quantification module, configured to calculate a total of 7 quantified index values including the gray-white matter signal ratio, the coefficient of variation of gray matter, the combined coefficient of variation, the contrast-to-noise ratio, the proportion of outliers, the connectivity of the gray matter region, and the index structural similarity based on the signals in the gray matter region and white matter region of the ASL image;

[0060] A quality evaluation module, configured to normalize the 7 index values of the ASL image to be evaluated and input them into a pre-constructed and fitted logistic regression model to obtain an evaluation result representing the quality of the image.

[0061] It should be noted that the method steps shown in S1 to S3 above can essentially be implemented in the form of a computer program.

[0062] Therefore, based on the same inventive concept, as Figure 4 shown, the present invention also provides a computer electronic device corresponding to the magnetic resonance artery-labeled image quality evaluation method provided in the above embodiment, which includes a memory and a processor;

[0063] The memory is used to store a computer program;

[0064] The processor is configured to, when executing the computer program, implement the magnetic resonance artery-labeled image quality evaluation method as described above;

[0065] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention.

[0066] Therefore, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to a method for evaluating the quality of magnetic resonance artery labeling images. A computer program is stored on the storage medium, and when the computer program is executed by a processor, it can implement the method for evaluating the quality of magnetic resonance artery labeling images as described above.

[0067] Therefore, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, they can implement the method for evaluating the quality of magnetic resonance artery labeling images as described above.

[0068] Specifically, in the computer-readable storage media of the above three embodiments, the stored computer program is executed by a processor, and the steps of S1 to S3 described above can be executed.

[0069] It can be understood that the above storage medium may include a random access memory (RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory. At the same time, the storage medium may also be various media such as a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc that can store program codes.

[0070] It can be understood that the above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0071] In addition, it should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the embodiments provided in the present application, the division of steps or modules in the system and method is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0072] Next, the present invention will further show the detailed implementation process and technical effects of the magnetic resonance artery labeling image quality assessment method shown in the above S1-S3 steps on a specific data set through a specific embodiment, so as to facilitate understanding the essence of the present invention.

[0073] Embodiment

[0074] In this embodiment, 51 cases of pediatric 3D ASL and 3D T1W image data from the database of the Children's Hospital, Zhejiang University were used for experiments. This data set was acquired by the 3D pCASL sequence of a Philips 3T magnetic resonance scanner, and the average age of the patients was 4.34±4.08 years. The quality scores of all ASL images were independently scored by 2 ASL imaging experts with 10 years and 1 year of work experience and 1 radiologist with 4 years of work experience, and were divided into 0 points (poor quality, unacceptable images) and 1 point (good quality, acceptable images). The Fleiss kappa value among the scores of the three experts was 0.537, indicating moderate consistency. The final score of each image was determined by the majority of the three.

[0075] In this embodiment, the specific implementation steps and comparative verification steps of the magnetic resonance artery labeling image quality assessment method (denoted as QAC) are as Figure 5 shown as follows:

[0076] First, the data is preprocessed. The SPM12 in MATLAB is used to segment the T1W images in the database to obtain the gray matter probability map and the white matter probability map. Then, the ASL images in the database are registered to the T1 image space, so that the ASL images are aligned with the gray matter probability map and the white matter probability map of the T1W images, and the gray matter probability map and the white matter probability map are binarized with 0.5 as the threshold, that is, the registered ASL images and the corresponding gray and white matter binary masks are obtained. Based on the gray and white matter binary masks, the gray matter region and the white matter region can be identified from the registered ASL images.

[0077] Secondly, based on the signals of the gray matter region and the white matter region in the registered ASL images respectively, 7 quantization index values of each ASL image are calculated. The specific calculation formulas are as follows:

[0078] 1. Gray matter to white matter signal ratio: The ratio of the mean signal of the gray matter region to the mean signal of the white matter region.

[0079] 2. Coefficient of variation of gray matter: The ratio of the signal variance of the gray matter region to the mean signal of the gray matter region.

[0080] 3. Combined coefficient of variation: The ratio of the sum of the variances of gray and white matter (i.e., the signal variance of the gray matter region plus the signal variance of the white matter region) to the difference in the mean signals of gray and white matter (i.e., the mean signal of the gray matter region minus the mean signal of the white matter region).

[0081] 4. Contrast-to-noise ratio: The ratio of the difference in the mean signals of gray and white matter to the signal variance over the entire range of the two regions (i.e., the variance calculated from all the signals in the gray matter region and the white matter region).

[0082] 5. Proportion of outliers: The proportion of outlier pixels over the entire range of the gray and white matter regions calculated based on the 3σ criterion. Outlier pixels are those within the gray matter region and the white matter region that exceed the global mean signal ± 3 times the standard deviation.

[0083] 6. Connectivity of gray matter region: The gray matter region is segmented using the reference gray matter signal value as the threshold, and the number of connected regions is obtained. In this embodiment, the reference gray matter signal value is set to 15 ml / (100 g·min).

[0084] 7. Index structural similarity: Calculate the mean signal of the white matter region in the ASL image, then assign all values in the white matter region to the calculated mean signal of the white matter region, and assign the gray matter region to 2.5 times the calculated mean signal of the white matter region, thereby constructing a pseudo-reference ASL image. Then, use the structural similarity index SSIM to calculate the similarity between the original ASL image and the pseudo-reference ASL image. The calculation method of SSIM belongs to the prior art, and its formula is as follows:

[0085]

[0086] Where: X and Y represent the original ASL image and the pseudo-reference ASL image respectively, μ X and μ Y represent the mean signals in X and Y respectively, σ XY represents the covariance of the signals in X and Y, σ X and σ Y represent the variances of the signals in X and Y respectively, and C1 and C2 are constant terms.

[0087] Finally, a logistic regression model is trained based on the seven metric values of each image and the corresponding quality scores, and its accuracy is verified. Since these seven metrics have different dimensions and orders of magnitude, directly inputting them into the model may interfere with the accuracy of the results. Therefore, all metrics are normalized using the Min-Max method before input. Due to the relatively small size of this dataset, five-fold cross-validation is used to verify the two methods, and the accuracy, F1 score, and the area under the ROC curve (AUC) of the model trained on all images and then tested are used as evaluation metrics. The division ratio of the training set to the test set is 4:1. The optimal parameters of the logistic regression of the present invention are selected by the grid search method, with the regularization coefficient C = 100, the optimization algorithm selection parameter ='solver', and the penalty term penalty = l1.

[0088] In addition, to compare the evaluation method (QAC model) of the present invention with the QEI evaluation method in the prior art (i.e., the 2D ASL image quality evaluation method for adults, see Dolui S, Wang Z, Wolf R L, et al. Automated Quality Evaluation Index for Arterial Spin Labeling Derived Cerebral Blood Flow Maps[J]. Journal of Magnetic Resonance Imaging, 2024), this embodiment replicates the QEI model in this literature. The QEI model designs three metrics for quality evaluation: joint spatial variability, proportion of negative gray matter, and structural similarity, and uses an exponential fitting model to perform 2D evaluation of the image quality score. However, since this QEI model was originally only for 2D ASL images, in order to make this method applicable to 3D ASL data, two modifications are made in this embodiment: 1. Expand the metrics designed for 2D ASL images to 3D images for calculation; 2. Change the metric of the proportion of negative gray matter to calculate the proportion of gray matter with zero value to adapt to Dicom data. On this basis of the modification, the parameters of the exponential fitting of the QEI model are retrained on the 3D ASL dataset of children in this embodiment to fit the parameters of the model.

[0089] Figure 7The ROC curves and AUC values of the two methods after training and testing on 51 cases of data are shown. The AUC of this method is 0.95, while that of the QEI model is 0.84. Table 1 shows the accuracy, F1 score, and model threshold of the two methods in five-fold cross-validation. The average accuracy of the method of the present invention is 82.36%, the F1 score is 85.86%, the average accuracy of the QEI model is 72.55%, and the F1 score is 77.26%. The results show that the method of the present invention exceeds the comparative method in terms of accuracy, F1 score, and AUC, achieving better evaluation performance.

[0090] Table 1 Accuracy, F1 score, and threshold of five-fold cross-validation of two methods

[0091]

[0092] Figure 8 Some examples in the cross-validation results of the two methods are shown, where (a) represents the images judged acceptable by both methods, (b) represents the images judged unacceptable by QAC of the method of the present invention and acceptable by the QEI model, (c) represents the images judged acceptable by QAC of the method of the present invention and not acceptable by the QEI model, and (d) represents the images judged unacceptable by both methods. The experimental results show that:

[0093] 1. The method of the present invention has excellent evaluation performance on pediatric ASL images, exceeding the QEI model. The average accuracy of the method of the present invention in five-fold cross-validation is 82.36%, the F1 score is 85.86%, and the area under the ROC curve after training and testing on all data is 0.95. In contrast, the accuracy of the QEI model is 72.55%, the F1 score is 77.26%, and the area under the ROC curve on all data is 0.84.

[0094] 2. The method of the present invention points out the potential impact of the presence of lesions on quality assessment. In this embodiment, it is found that the presence of lesions will cause abnormal gray and white matter segmentation of T1W images, as Figure 6 shown. At the same time, abnormal signal values will also occur in the lesion area due to factors such as abnormal blood circulation, but these abnormal signals are not caused by poor image quality, resulting in misjudgment by the model. This also provides a direction for subsequent research and improvement.

[0095] In summary, the 3D ASL image quality assessment model proposed by the present invention has achieved a high evaluation accuracy on 51 cases of pediatric ASL datasets, and its performance exceeds that of existing representative methods. And we also found the potential impact of lesions on quality assessment in the experiment, providing ideas for subsequent research and improvement.

[0096] The embodiments described above are only some preferred implementation solutions of the present invention, but are not intended to limit the present invention. Those of ordinary skill in the relevant technical field can still make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all technical solutions obtained by means of equivalent replacement or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A method for assessing the quality of a magnetic resonance artery marking image, characterized in that: include: S1, obtaining an ASL image to be evaluated and identifying the gray matter area and white matter area therein; S2. Based on the signals in the gray matter and white matter regions in the ASL images, seven quantitative index values ​​were calculated, including gray-white matter signal ratio, gray matter coefficient of variation, joint coefficient of variation, contrast-to-noise ratio, outlier ratio, gray matter region connectivity, and index structure similarity; S3. The seven index values ​​of the ASL image to be evaluated are normalized and input into a pre-built and fitted logistic regression model to obtain an evaluation result representing the image quality.

2. The magnetic resonance artery marking image quality assessment method according to claim 1, characterized in that: When identifying gray matter and white matter areas in ASL images, the ASL images need to be registered to the corresponding T1-weighted images, and gray and white matter areas need to be segmented based on the T1-weighted images. The generated gray matter area mask and white matter area mask are applied to the registered ASL images to identify the gray matter and white matter areas in the ASL images.

3. The magnetic resonance artery marking image quality assessment method according to claim 1, characterized in that: Among the 7 quantitative index values: The gray-white matter signal ratio is defined as the ratio of the signal mean of the gray matter area to the signal mean of the white matter area; The gray matter coefficient of variation is defined as the ratio of the signal variance of the gray matter region to the signal mean of the gray matter region; The joint coefficient of variation is defined as the ratio of the sum of the signal variances of the gray matter region and the white matter region to the difference between the signal means of the gray matter region and the white matter region; The contrast-to-noise ratio is defined as the ratio of the difference between the signal means of the gray matter region and the white matter region to the signal variance of the global range of the two regions; The outlier ratio is defined as the ratio of abnormal pixels in the global range of the gray matter and white matter regions calculated based on the 3σ criterion; The gray matter region connectivity is defined as the number of connected regions obtained after segmenting the gray matter region using the reference gray matter signal value as a threshold; The indicator structural similarity is defined as the structural similarity index SSIM between the ASL image to be evaluated and the corresponding pseudo-reference ASL image; The white matter areas in the pseudo-reference ASL image are all assigned a signal mean value of the white matter areas in the ASL image to be evaluated, and the gray matter areas are all assigned a signal value that is 2.5 times the signal mean value of the white matter areas in the ASL image to be evaluated.

4. The magnetic resonance artery marking image quality assessment method according to claim 1, characterized in that: The evaluation result output by the logistic regression model is a binary classification label.

5. The magnetic resonance artery marking image quality assessment method according to claim 4, characterized in that: The logistic regression model is pre-trained on a dataset annotated by experts. Each sample in the dataset extracts the 7 quantitative index values ​​from the ASL sample image as model input, and the true value label of each sample is a 0-1 label representing the quality of the ASL image.

6. The magnetic resonance artery marking image quality assessment method according to claim 1, characterized in that: The ASL image to be evaluated is a 3D ASL image collected for children.

7. A magnetic resonance artery marking image quality assessment system, characterized in that: include: An image acquisition module, used for acquiring an ASL image to be evaluated and identifying gray matter areas and white matter areas therein; The index quantification module is used to calculate the gray-white matter signal ratio, gray matter coefficient of variation, joint coefficient of variation, contrast-to-noise ratio, outlier ratio, gray matter area connectivity, and index structure similarity, based on the signals in the gray matter area and white matter area in the ASL image. The quality assessment module is used to normalize the seven index values ​​of the ASL image to be evaluated and input them into the pre-built and fitted logistic regression model to obtain the evaluation results representing the image quality.

8. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the method for evaluating the quality of a magnetic resonance artery marking image as claimed in any one of claims 1 to 7 can be implemented.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the magnetic resonance artery marking image quality assessment method according to any one of claims 1 to 7 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is used to implement the magnetic resonance artery marking image quality assessment method according to any one of claims 1 to 7 when executing the computer program.

Images