Multi-modal medical image quality inspection system based on deep learning

Abstract

The invention discloses a multi-modal medical image quality inspection system based on deep learning, which belongs to the technical field of medical image processing quality inspection and comprises a data acquisition and extraction module, an evaluation calculation module and a result display module. After receiving the first modal image data, the first texture feature, the second modal image data, the second texture feature, the third modal image data and the third texture feature, the evaluation calculation module obtains a first modal image quality score, a second modal image quality score and a third modal image quality score; the method comprises the steps of obtaining a bimodal image quality score according to a single-modal image quality score and vector conversion of bimodal related features, and finally obtaining a multi-modal stable feature based on the bimodal image quality score and repeatedly fusing the extracted fused texture features. According to the method, the correlation between multiple factors and double modes and the fusion stability of a single mode are comprehensively considered, so that the overall quality of the multi-mode image is comprehensively evaluated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical image processing quality inspection, and particularly relates to a multi-modal medical image quality inspection system based on deep learning. BACKGROUND

[0002] In the field of medical diagnosis, the application of multi-modal medical image technology is increasingly widespread. Different modalities of medical images can provide information about different aspects of the human body, among which CT is good at showing the anatomical structure of the human body, MRI has strong resolution of soft tissue, and PET can reflect the metabolic function of the human body.

[0003] The prior art only focuses on the indicators of a single modality when performing medical image processing quality inspection, resulting in an incomplete evaluation of image quality. With the development of medical images, some existing technologies have systems that fuse and inspect multiple modalities of medical images. However, in some existing technologies, even if multiple modal indicators are considered, different clinical application scenarios have different requirements for clarity, noise level, and contrast, but some existing technologies still use fixed weights, which cannot be adjusted according to specific circumstances.

[0004] Some existing technologies do not fully consider the correlation between two modal images when evaluating the fusion quality of dual-modality images. The images of different modalities differ in feature space, and if this difference is not considered, the fused images will have information loss and redundancy problems.

[0005] In addition, some existing technologies also ignore the stability of the fusion results when performing final quality inspection on multi-modal images. The multi-modal image fusion process is affected by various factors, and if stability is not considered, the quality inspection results will fluctuate greatly, affecting the accuracy of diagnosis. SUMMARY

[0006] The technical problem to be solved by the present application is that the prior art lacks comprehensive consideration, ignores the correlation between multi-modalities, and does not consider the stability of fusion. Therefore, we propose a multi-modal medical image quality inspection system based on deep learning.

[0007] The technical solution is mainly as follows: a multi-modal medical image quality inspection system based on deep learning, comprising a data acquisition and extraction module, an evaluation calculation module, and a result display module. The output ends of a computed tomography device, a magnetic resonance imaging device, and a positron emission tomography device are connected to the input end of the data acquisition and extraction module, facilitating the simultaneous acquisition of input data from each device forming multiple modalities. The data acquisition and extraction module acquires first modality image data transmitted by the computed tomography device, second modality image data transmitted by the magnetic resonance imaging device, and third modality image data transmitted by the positron emission tomography device. The first texture feature in the first modality image data, the second texture feature of the second modality image data, and the third texture feature of the third modality image data are extracted. After the evaluation calculation module receives the first modality image data, the first texture feature, the second modality image data, the second texture feature, the third modality image data, and the third texture feature, the following operations are performed: After the quality evaluation of the weight allocation of the first modality image data, the second modality image data, and the third modality image data, respectively, the first modality image quality score, the second modality image quality score, and the third modality image quality score are obtained. The first texture feature, the second texture feature, and the third texture feature are converted and evaluated in the form of a vector of any two modality combinations to obtain a double modality related feature of any two modality combinations. According to the first modality image quality score, the second modality image quality score, the third modality image quality score, and the double modality related feature, a double modality image fusion of any two modality combinations is performed to obtain a double modality image quality score of any two modality combinations. After the first modality image data, the second modality image data, and the third modality image data are repeatedly fused m times, the fused texture features are extracted, converted, and evaluated in the form of a vector to obtain a multi-modality stable feature. Based on the double modality related feature and the multi-modality stable feature, a multi-modality image fusion is performed to obtain a multi-modality quality inspection score. The result display module is responsible for displaying the result of the multi-modality quality inspection score.

[0008] Preferably, the first modality image data includes computed tomography definition, computed tomography noise level, and computed tomography contrast transmitted by the computed tomography device. The second modality image data includes magnetic resonance imaging definition, magnetic resonance imaging noise level, and magnetic resonance imaging contrast transmitted by the magnetic resonance imaging device. The third modality image data includes positron emission tomography definition, positron emission tomography noise level, and positron emission tomography contrast transmitted by the positron emission tomography device.

[0009] Preferably, the evaluation calculation module comprises a single modality evaluation unit, a dual modality evaluation unit and a multi-modality evaluation unit.

[0010] Preferably, the single modality evaluation unit adds the weight distribution of the computed tomography definition, the computed tomography noise level and the computed tomography contrast to obtain the first modality image quality score; add the weight distribution of the magnetic resonance imaging definition, the magnetic resonance imaging noise level and the magnetic resonance imaging contrast to obtain the second modality image quality score; add the weight distribution of the positron emission tomography definition, the positron emission tomography noise level and the positron emission tomography contrast to obtain the third modality image quality score; Wherein, the first modality image quality score, the second modality image quality score and the third modality image quality score are each weight distribution total value of 1, and are respectively weight distributed according to the emphasis of the magnetic resonance imaging definition, the magnetic resonance imaging noise level and the magnetic resonance imaging contrast, the emphasis of the magnetic resonance imaging definition, the magnetic resonance imaging noise level and the magnetic resonance imaging contrast, and the emphasis of the positron emission tomography definition, the positron emission tomography noise level and the positron emission tomography contrast; The single modality evaluation unit can obtain flexible weight distribution of the computed tomography device, the magnetic resonance imaging device and the positron emission tomography device for the respective definition, noise level and contrast by respectively weight distributing the first modality image data, the second modality image data and the third modality image data transmitted by the computed tomography device, the magnetic resonance imaging device and the positron emission tomography device, so as to achieve the purpose of flexible adjustment of emphasis.

[0011] Preferably, the arbitrary two modality combination comprises a first modality combination, a second modality combination and a third modality combination; The first modality combination: the computed tomography device and the magnetic resonance imaging device; The second modality combination: the computed tomography device and the positron emission tomography device; The third modality combination: the magnetic resonance imaging device and the positron emission tomography device; The dual modality evaluation unit obtains the dual modality image quality score of the arbitrary two modality combination based on the first modality combination, the second modality combination and the third modality combination: According to the first texture feature and the second texture feature, the double-modal correlation feature of the first modality combination is obtained, and according to the first modality image quality score and the second modality image quality score, the double-modal image quality score of the first modality combination is obtained; According to the first texture feature and the third texture feature, the double-modal correlation feature of the second modality combination is obtained, and according to the first modality image quality score and the third modality image quality score, the double-modal image quality score of the second modality combination is obtained; According to the second texture feature and the third texture feature, the double-modal correlation feature of the third modality combination is obtained, and according to the first modality image quality score and the second modality image quality score, the double-modal image quality score of the third modality combination is obtained; The double-modal evaluation unit can realize the fusion of the correlation between all double modalities in the multi-modality through the acquisition of the double-modal image quality score of any two modality combinations.

[0012] Preferably, the steps of vector form conversion and evaluation of the double-modal correlation feature are as follows: S1: The first texture feature, the second texture feature and the third texture feature are extracted according to the any two modality combinations; S2: After extraction, the vector form is converted, that is, v i → (x1, x2,... x n ) and v i+1 → (y1, y2,... y n ); S3: The vector dot product and vector norm of the vector form v i → (x1, x2,... x n ) and v i+1 → (y1, y2,... y n ) are obtained; S4: According to the vector dot product divided by the vector norm, the double-modal correlation feature of the any two modality combinations is obtained.

[0013] Preferably, the multi-modality evaluation unit obtains an average quality feature according to the average degree of the double-modal image quality score of the first modality combination, the double-modal image quality score of the second modality combination and the double-modal image quality score of the third modality combination; The average quality score of the dual-modal image is multiplied by the multimodal stable features to obtain the stable average quality features. The multimodal stable features are obtained by repeatedly fusing the first modal image data, the second modal image data, and the third modal image data input from the computed tomography (CT) scanner, magnetic resonance imaging (MRI) scanner, and positron emission tomography (PET) scanner, based on repeated fusion experiments. This ensures the accuracy of the stable average quality features during quality inspection. Based on the average degree of difference between the bimodal image quality scores of the first modal combination, the second modal combination, and the third modal combination and the average bimodal image quality score, quality discrete features are obtained. The stable average quality feature is added to the discrete quality feature to obtain the multimodal quality inspection score.

[0014] Preferably, the steps for acquiring the multimodal stability features are as follows: S1: The first modal image data, the second modal image data, and the third modal image data are repeatedly fused m times to obtain m fusion results f1, f2, ... f m ; S2: For m fusion results f1, f2, ... f m The fused texture features are extracted to obtain the feature vector f1. → f2 → , ...f m → ; S3: Obtain the feature vector f1 → f2 → , ...f m → The mean and variance of the eigenvectors; S4: Utilize The multimodal stability features are obtained.

[0015] The technical effects and advantages of this invention are as follows: In this invention, the data acquisition and extraction module comprehensively and objectively evaluates the quality of a single modality image by acquiring and extracting the first modality image data, second modality image data, and third modality image data from the computed tomography (CT) scanner, magnetic resonance imaging (MRI) scanner, and positron emission tomography (PET) scanner, respectively, as well as the weight allocation. Furthermore, the weight allocation can be adjusted according to different clinical application scenarios and needs.

[0016] In the dual-mode evaluation unit, a dual-mode correlation feature of any two modal combinations is introduced, and the dual-mode correlation feature is calculated by the cosine similarity method, so as to accurately measure the correlation between two modal images. In the evaluation of the fusion quality of the dual-mode image, the fusion quality is adjusted in combination with the modal correlation, thereby improving the accuracy of the evaluation.

[0017] In addition, the multi-modal evaluation unit introduces a multi-modal stability feature. The multi-modal stability feature is calculated by calculating the variance of multiple repeated fusion experiments to evaluate the stability of the fusion result. In the final quality inspection score, the average quality is adjusted in combination with the multi-modal stability feature, which can comprehensively evaluate the overall quality of the multi-modal image. This comprehensive evaluation method can more accurately reflect the actual quality level of the multi-modal image than the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The method flowchart of the multi-modal medical image quality inspection system. DETAILED DESCRIPTION

[0019] The application will now be further described in detail in conjunction with the accompanying drawings and preferred embodiments.

[0020] Referring to Figure 1 The application provides a technical solution: a multi-modal medical image quality inspection system based on deep learning, which includes a data acquisition and extraction module, an evaluation and calculation module, and a result display module. The output ends of a computer tomography device, a magnetic resonance imaging device, and a positron emission tomography device are respectively connected to the input end of the data acquisition and extraction module. The data acquisition and extraction module acquires first modal image data transmitted by the computer tomography device, second modal image data transmitted by the magnetic resonance imaging device, and third modal image data transmitted by the positron emission tomography device. The first modal image data includes computer tomography definition, computer tomography noise level, and computer tomography contrast transmitted by the computer tomography device. The second modal image data includes magnetic resonance imaging definition, magnetic resonance imaging noise level, and magnetic resonance imaging contrast transmitted by the magnetic resonance imaging device. The third modal image data includes positron emission tomography definition, positron emission tomography noise level, and positron emission tomography contrast transmitted by the positron emission tomography device. The first texture feature in the first modal image data, the second texture feature of the second modal image data, and the third texture feature of the third modal image data are extracted. The evaluation calculation module receives the first modality image data, the first texture feature, the second modality image data, the second texture feature, the third modality image data, and the third texture feature, and performs the following operations: After quality evaluation of weight distribution of the first modality image data, the second modality image data, and the third modality image data, first modality image quality scores, second modality image quality scores, and third modality image quality scores are obtained. The first texture feature, the second texture feature, and the third texture feature are converted and evaluated in the form of vectors of any two modality combinations to obtain double-modality related features of any two modality combinations. According to the first modality image quality scores, the second modality image quality scores, the third modality image quality scores, and the double-modality related features, double-modality image fusion of any two modality combinations is performed to obtain double-modality image quality scores of any two modality combinations. After m times of repeated fusion of the first modality image data, the second modality image data, and the third modality image data, the fused texture features are extracted, converted, and evaluated to obtain multi-modality stable features. Based on the double-modality related features and the multi-modality stable features, multi-modality image fusion is performed to obtain multi-modality quality inspection scores. The evaluation calculation module includes a single-modality evaluation unit, a double-modality evaluation unit, and a multi-modality evaluation unit. The result display module is responsible for displaying the results of the multi-modality quality inspection scores.

[0021] In this embodiment, the close cooperation of the data acquisition and extraction module, the evaluation calculation module, the result display module, and the units brings significant benefits to the multi-modality medical image quality inspection system based on deep learning. The single-modality evaluation unit in the evaluation calculation module obtains multi-modality image data with the help of the data acquisition and extraction module, and obtains single-modality feature data after feature extraction. The double-modality evaluation unit optimizes the fusion quality evaluation and improves the system's ability to judge the image fusion effect by considering the modality correlation coefficient. The multi-modality evaluation unit considers the fusion stability and accurately calculates the multi-modality quality inspection scores.

[0022] Referring to Figure 1 In this embodiment, the single-modality evaluation unit obtains the first modality image quality scores by adding the weight distribution of the computed tomography definition, the computed tomography noise level, and the computed tomography contrast. The second modality image quality scores are obtained by adding the weight distribution of the magnetic resonance imaging definition, the magnetic resonance imaging noise level, and the magnetic resonance imaging contrast. The third modality image quality score is obtained by summing the weighted values ​​of positron emission tomography (PET) sharpness, PET noise level, and PET contrast. In the weighting of the first modality image quality score, the second modality image quality score, and the third modality image quality score, the total weighting value is 1. The weighting is then assigned separately based on the emphasis on the sharpness, noise level, and contrast of magnetic resonance imaging (MRI), the emphasis on the sharpness, noise level, and contrast of MRI, and the emphasis on the sharpness, noise level, and contrast of positron emission tomography (PET).

[0023] Let computed tomography (CT) be labeled a, magnetic resonance imaging (MRI) be labeled b, and positron emission tomography (PET) be labeled c. The specific calculation formula for single-modal image quality scoring is as follows: ; in: DF i The single-modal image quality score is given for any one of the following devices: computed tomography, magnetic resonance imaging, and positron emission tomography, where i = a, b, c; Q i The resolution of any one of the following devices: computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). DF i The single-modal image quality score is given for any one of the following devices: computed tomography, magnetic resonance imaging, and positron emission tomography, where i = a, b, c; Q i The resolution of any one of the following devices: computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). ; in: RF 2i Let 2i be the bimodal image quality score for any two-modal combination, and 2i = (a, b) / (a, c) / (c, b); DF i and DF i+1 These are the single-modal image quality scores for any two-modal combination, i.e., i+1=i=a, b, c; g is the modal correlation coefficient reflecting the bimodal correlation characteristics. The steps for vector form transformation and evaluation of the bimodal correlation characteristics are as follows: S1: Extract the first texture feature, the second texture feature, and the third texture feature based on any two modal combinations; S2: After extraction, convert it into vector form, i.e., v i → (x1, x2, ... x n ) and v i+1 → (y1, y2, ... y n ); S3: Obtain the vector form v i → (x1, x2, ... x n ) and v i+1 → (y1, y2, ... y n The vector dot product and vector modulus are as follows: ; ; S4: Obtain the bimodal correlation features of any two-mode combination by dividing the vector dot product by the vector modulus, as follows: .

[0024] In this embodiment, it is worth noting that a modal correlation coefficient g is introduced. This coefficient, calculated using the cosine similarity method, accurately measures the similarity and correlation between two modalities of images. In multimodal medical image fusion, the information contained in different modalities is both complementary and correlated. The adjustment of fusion quality through the modal correlation coefficient g makes the evaluation of fusion quality more consistent with reality, thereby improving the effect and quality of bimodal image fusion. Furthermore, a geometric mean is employed in this embodiment. and arithmetic mean Combination method, geometric mean It places greater emphasis on the overall quality of the two single-modal images, while the arithmetic mean... This reflects their average quality, and combining the two can evaluate the quality of bimodal image fusion from different perspectives.

[0025] Reference Figure 1 As shown in this implementation scheme: the multimodal evaluation unit obtains the average quality feature based on the average degree of the bimodal image quality scores of the first modality combination, the second modality combination, and the third modality combination. The average quality score of the dual-modal image is multiplied by the multimodal stable features to obtain the stable average quality features; Based on the average degree of difference between the bimodal image quality scores of the first modal combination, the second modal combination, and the third modal combination and the average bimodal image quality score, the quality discrete features are obtained. Adding the stabilized average quality feature and the quality dispersion feature obtains a multi-modal quality inspection score; The specific calculation formula for calculating the multi-modal quality inspection score is as follows: ; ZF is the multi-modal quality inspection score, n is the quantity of modal combination, The calculation result is the average quality of the double-modal image quality scores output by all arbitrary two-modal combinations; The steps for obtaining the multi-modal stabilized feature are as follows: S1: repeatedly fusing the first modal image data, the second modal image data and the third modal image data for m times to obtain m fusion results f1, f2,...fm. m ; S2: extracting the fusion post-texture feature from the m fusion results f1, f2,...fm to obtain feature vectors f1 m , f2 → ,...fm → m → ; S3: obtaining the feature vector average value and the feature vector variance of the feature vectors f1 → , f2 → ,...fm m → , which are specifically as follows: ; ; f avg → is the feature vector average value, f i → is any one of the feature vectors f1 → , f2 → ,...fm m → ; S4: obtaining the multi-modal stabilized feature by using ; w is a stability coefficient w reflecting the stability of the multi-modal stabilized feature.

[0026] In this embodiment, it is worth noting that the stability coefficient w is introduced, which is calculated by repeatedly fusing experiments for multiple times to obtain the variance, and is used to measure the stability of the multi-modal image fusion result. Due to the complexity of data and the uncertainty of algorithm, there will be certain fluctuations in the fusion result in the multi-modal image fusion process. The stability coefficient w is used to measure the average quality of the multi-modal image. ​The adjustment fully considers the influence of fusion stability on the final quality inspection result, so that the quality inspection result is more reliable, the error caused by unstable fusion is reduced, and the multi-modal quality inspection score ZF comprehensively considers multiple double-modal image quality scores RF 2i The overall quality of the multi-modal image is comprehensively evaluated by calculating the average quality and the standard deviation The average quality reflects the average level of the multi-group double-modal fusion, and the standard deviation reflects the dispersion degree of each fusion result, and this comprehensive evaluation method can grasp the quality condition of the multi-modal image from the macro and micro levels, so that the final quality inspection score can more accurately reflect the actual quality of the multi-modal image.

[0027] It should be noted that any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should also be within the protection scope of the present application.

Claims

1. A deep learning-based multi-modal medical image quality inspection system, characterized in that: it comprises a data acquisition and extraction module, an evaluation calculation module, and a result display module; the input end of the data acquisition and extraction module is connected with the output end of a computer tomography device, a magnetic resonance imaging device, and a positron emission tomography device; the data acquisition and extraction module acquires first modality image data transmitted by the computer tomography device, second modality image data transmitted by the magnetic resonance imaging device, and third modality image data transmitted by the positron emission tomography device; first texture features in the first modality image data, second texture features in the second modality image data, and third texture features in the third modality image data are extracted; the evaluation calculation module receives the first modality image data, the first texture features, the second modality image data, the second texture features, the third modality image data, and the third texture features, and performs the following operations: after performing weight distribution quality evaluation on the first modality image data, the second modality image data, and the third modality image data, respectively, first modality image quality scores, second modality image quality scores, and third modality image quality scores are acquired; the first texture features, the second texture features, and the third texture features are converted into vector forms of any two modality combinations, evaluated, and double modality related features of any two modality combinations are acquired; double modality image fusion of the any two modality combinations is performed according to the first modality image quality scores, the second modality image quality scores, the third modality image quality scores, and the double modality related features, and double modality image quality scores of the any two modality combinations are acquired; after performing m times of repeated fusion on the first modality image data, the second modality image data, and the third modality image data, fusion texture features are extracted, vector conversion and evaluation are performed, and multi-modal stable features are acquired; multi-modal image fusion is performed based on the double modality related features and the multi-modal stable features, and a multi-modal quality inspection score is acquired; and the result display module is responsible for displaying the result of the multi-modal quality inspection score. The first modality image data includes computer tomography definition, computer tomography noise level, and computer tomography contrast transmitted by the computer tomography device; the second modality image data includes magnetic resonance imaging definition, magnetic resonance imaging noise level, and magnetic resonance imaging contrast transmitted by the magnetic resonance imaging device; and the third modality image data includes positron emission tomography definition, positron emission tomography noise level, and positron emission tomography contrast transmitted by the positron emission tomography device. The evaluation calculation module comprises a single modality evaluation unit, a double modality evaluation unit, and a multi-modal evaluation unit. The single modality evaluation unit adds the computer tomography definition, the computer tomography noise level, and the computer tomography contrast after weight distribution to acquire the first modality image quality score. ​ ​ ​ ​ ​ ​ ​ ​ 2.The deep learning based multi-modal medical image quality inspection system according to claim 1, characterized in that: ​ ​ ​ 3.The deep learning based multi-modal medical image quality inspection system according to claim 2, characterized in that: ​ 4.The deep learning based multi-modal medical image quality inspection system according to claim 3, characterized in that: ​ According to the magnetic resonance imaging clarity, the magnetic resonance imaging noise level, the addition after the weight distribution of the magnetic resonance imaging contrast, the second modality image quality score is obtained; According to the positron emission tomography clarity, the positron emission tomography noise level, the addition after the weight distribution of the positron emission tomography contrast, the third modality image quality score is obtained; Wherein, the first modality image quality score, the second modality image quality score, the third modality image quality score in weight distribution, each weight distribution total value is 1, and according to the magnetic resonance imaging clarity, the magnetic resonance imaging noise level, the magnetic resonance imaging contrast, the magnetic resonance imaging clarity, the magnetic resonance imaging noise level, the magnetic resonance imaging contrast, the positron emission tomography clarity, the positron emission tomography noise level, the positron emission tomography contrast, the positron emission tomography clarity, the positron emission tomography noise level, the positron emission tomography contrast, the weight distribution is respectively distributed. 5.The deep learning based multi-modal medical image quality inspection system according to claim 4, characterized in that: The arbitrary two modality combination includes first modality combination, second modality combination, third modality combination; First modality combination: the computer tomography device and the magnetic resonance imaging device; Second modality combination: the computer tomography device and the positron emission tomography device; Third modality combination: the magnetic resonance imaging device and the positron emission tomography device; The double modality evaluation unit obtains the double modality image quality score of any two modality combination based on the first modality combination, the second modality combination, the third modality combination: According to the first modality combination, the double modality related feature, and the first modality image quality score and the second modality image quality score, the double modality image quality score of the first modality combination is obtained; According to the double modality related feature of the second modality combination, and the first modality image quality score and the third modality image quality score, the double modality image quality score of the second modality combination is obtained; According to the double modality related feature of the third modality combination, and the first modality image quality score and the second modality image quality score, the double modality image quality score of the third modality combination is obtained. 6.The deep learning based multi-modal medical image quality inspection system according to claim 5, characterized in that: The steps of vector form conversion and evaluation of the double modality related feature are as follows: S1: the first texture feature, the second texture feature, the third texture feature are extracted according to the arbitrary two modality combination; S2: conversion to vector form after extraction; S3: the vector dot product and the vector module of the vector form are obtained; S4: according to the vector dot product divided by the vector module, the double modality related feature of the arbitrary two modality combination is obtained. 7.The deep learning based multi-modal medical image quality inspection system according to claim 5, characterized in that: The multi-modality evaluation unit obtains the average quality feature according to the average degree of the double modality image quality score of the first modality combination, the double modality image quality score of the second modality combination, the double modality image quality score of the third modality combination. The double-mode image quality average score is multiplied by the multi-modal stability feature to obtain a stabilized average quality feature; The average degree of difference between the double-mode image quality score of the first modality combination, the double-mode image quality score of the second modality combination, the double-mode image quality score of the third modality combination and the double-mode image quality average score is obtained to obtain a quality dispersion feature; The stabilized average quality feature and the quality dispersion feature are added to obtain a multi-modal quality inspection score. 8.The deep learning based multi-modal medical image quality inspection system according to claim 7, characterized in that: The multi-modal stability feature is obtained by the following steps: S1: The first modality image data, the second modality image data and the third modality image data are repeatedly fused m times to obtain m fusion results; S2: The fusion texture features of the m fusion results are extracted to obtain feature vectors; S3: The feature vector average value and the feature vector variance of the feature vectors are obtained; S4: utilizing the multi-modal stability features.

Images