Medical image quality management method for rehabilitation monitoring

Abstract

The invention discloses a medical image quality management method for rehabilitation monitoring, and relates to the technical field of medical image quality management, and the method comprises the steps: carrying out the quality scoring of a medical image according to a quality coefficient, and carrying out the marking of the medical image through an obtained quality score; constructing an influence coefficient according to the influence degree of the medical image quality on the diagnosis accuracy, if the influence coefficient exceeds the expectation, constraining an image quality threshold value according to the scene complexity, training the marked medical image data to obtain an image quality evaluation model, and performing quality screening on the medical image by the trained image quality evaluation model to obtain a medical image quality evaluation result. And if the aggregation degree of the low-quality images in the image quality monitoring period exceeds the expectation, giving a corresponding restoration scheme for the low-quality medical images through the medical image optimization knowledge graph, and restoring the low-quality medical images. The low-quality image is specifically repaired, the image quality of the current stage is improved, and the effect is better when the medical image is managed.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image quality management, and in particular to a medical image quality management system for rehabilitation monitoring. Background Art

[0002] Medical imaging technology plays an important role in rehabilitation medicine, mainly in the following aspects:

[0003] Diagnosis and assessment of injuries: Through medical imaging technologies such as X-rays, CT, and MRI, we can clearly understand the location, degree, and scope of the patient's injury, providing an important basis for formulating a rehabilitation plan. For example, in sports injuries such as fractures, dislocations, and muscle injuries, medical imaging technology can help rehabilitation therapists understand the specific circumstances of the injury and formulate targeted rehabilitation plans. Understanding the progression of the disease: During the rehabilitation process, regular medical imaging examinations can intuitively observe changes in the disease, evaluate the rehabilitation effect, and adjust the rehabilitation plan accordingly.

[0004] Personalized rehabilitation plan: Based on the results of medical imaging, rehabilitation therapists can understand the patient's sports physiological characteristics, such as joint range of motion, muscle strength, etc., and thus develop personalized rehabilitation training plans. Monitoring training effects: During rehabilitation training, medical imaging technology can also be used to monitor the patient's training effects to ensure the effective implementation of the rehabilitation plan. Discover potential risks: For sports enthusiasts or professional athletes, medical imaging technology can help them understand their own sports physiological characteristics, discover potential sports injury risks, and take corresponding preventive measures.

[0005] Medical image quality management is a series of activities and processes to ensure that the quality of medical images (such as X-rays, CT, MRI, etc.) meets the needs of clinical diagnosis. It covers many aspects, including equipment testing, image quality standard formulation, quality control effect evaluation personnel training and assessment, etc., aiming to obtain high-quality images with the lowest radiation dose and medical expenses, and provide reliable diagnostic basis for clinicians. By establishing a sound quality assurance and quality control system, the image quality, equipment quality, radiation protection quality and personnel work quality can be fully optimized.

[0006] In the Chinese invention patent with application publication number CN118352034A, a data-based medical image data management method is disclosed. The method includes the following steps: obtaining a multimodal medical image data set, and performing image structure preprocessing on the multimodal medical image data set to obtain a multimodal medical image standard data set; annotating the image lesion area of ​​the multimodal medical image standard data set to obtain a medical lesion area annotated image data set; performing real-time storage flow monitoring and storage flow event classification on the medical lesion area annotated image data set to obtain medical annotated image storage flow event classification data; performing corresponding storage management strategy analysis and data storage management processing on the medical annotated image storage flow event classification data to obtain the medical annotated image data storage management result.

[0007] Combined with the above application and the contents of the prior art:

[0008] When the patient is in the recovery stage, after completing the initial physical examination and obtaining the corresponding medical imaging data, when diagnosing the patient based on the acquired medical imaging data, different diagnostic items have different requirements for image quality, and when the image quality is low, the risk of misdiagnosis is also high. Therefore, medical imaging quality management is required after applying the medical imaging data.

[0009] The existing quality management method usually determines whether the medical quality meets the usage requirements after obtaining medical imaging data. If it does not meet the usage expectations, it will be optimized in a targeted manner to ensure the quality of medical imaging data. However, considering that there may be large differences between different usage scenarios after obtaining medical imaging data, if the quality of medical images is still managed with consistent quality management evaluation standards, it may have a negative impact on subsequent diagnostic analysis, and thus affect the patient's recovery status.

[0010] To this end, the present invention provides a medical image quality management system for rehabilitation monitoring. Summary of the invention

[0011] 1. Technical issues to be resolved

[0012] In view of the deficiencies of the prior art, the present invention provides a medical image quality management system for rehabilitation monitoring, which constructs an influence coefficient based on the degree of influence of medical image quality on diagnostic accuracy. If the influence coefficient exceeds expectations, the image quality threshold is constrained according to the complexity of the scene, and the image quality assessment model is obtained by training with labeled medical image data. The trained image quality assessment model is used to screen the quality of medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph is used to provide corresponding repair solutions for low-quality medical images, and the low-quality medical images are repaired. Targeted repair of low-quality images is achieved to improve the image quality at the current stage, thereby solving the technical problems raised in the background technology.

[0013] (II) Technical solution

[0014] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0015] A medical image quality management method for rehabilitation monitoring, including: performing similarity analysis on the use scenarios of medical images, aggregating the collected scene similarity data to generate a scene data set, generating a scene complexity Fot from the scene data set, and issuing a labeling instruction to the outside if the scene complexity Fot exceeds expectations;

[0016] Identify the quality index of the medical image, construct the quality coefficient Qts of the medical image based on the acquired quality index data, score the quality of the medical image based on the quality coefficient Qts, and annotate the medical image with the acquired quality score;

[0017] The influence coefficient Yts is constructed according to the influence of medical image quality on diagnostic accuracy. If the influence coefficient Yts exceeds expectations, the image quality threshold is constrained according to the scene complexity Fot. If the medical image quality is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image.

[0018] The image quality assessment model is obtained by training the labeled medical image data. The trained image quality assessment model scores the quality of the medical image data, performs error analysis on the quality score and generates a corresponding error coefficient Cts. If the error coefficient Cts is higher than the error threshold, a model optimization instruction is issued to the outside.

[0019] The trained image quality assessment model is used to screen the quality of medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph will provide corresponding repair solutions for the low-quality medical images and repair the low-quality medical images.

[0020] Furthermore, the usage scenario data of medical images are collected, scenario descriptions are added to the usage scenarios, the scene similarity between different scenario descriptions is obtained through similarity analysis, and the collected scene similarity data are summarized to generate a scenario data set; the scenario complexity Fot is generated from the usage scenario data in the scenario data set.

[0021] Furthermore, the similarity Xs between the two scene descriptions is linearly normalized, and the corresponding data values ​​are mapped to the interval [0,1], and the scene complexity Fot is generated as follows:

[0022]

[0023] Where n is the number of usage scenarios, Xs ij is the similarity between the i-th usage scenario and the j-th usage scenario, Xs a is the average similarity; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1.

[0024] Further, after receiving the labeling instruction, the medical images in the current recognition cycle are collected, the quality index of the collected medical images is identified and the corresponding quality index data is obtained, and the image quality data set is generated by aggregation;

[0025] The quality coefficient Qts of medical images is constructed from the image quality data set, where the spatial resolution Fv and the artifact rate Yv are treated as dimensionless, and the corresponding weight coefficients are determined by referring to the hierarchical analysis method, and then according to the following method:

[0026]

[0027] The weight coefficient is, 0≤A f ≤1,0≤A y ≤1, C x is a constant correction factor.

[0028] Furthermore, correlation analysis was performed in different usage scenarios to obtain the correlation coefficient between the quality of medical images and the diagnostic accuracy in different usage scenarios, and the correlation coefficient was used as the influence degree;

[0029] After obtaining several influence degrees, an influence coefficient Yts is constructed. If the influence coefficient Yts exceeds a preset influence threshold, a threshold constraint instruction is issued.

[0030] Furthermore, the influence degree Po is linearly normalized, the corresponding data value is mapped to the interval [0,1], and the influence coefficient Yts is constructed as follows:

[0031]

[0032] Where n is the number of usage scenarios, Po i is the correlation coefficient in the i-th usage scenario, Po a is the average value of the correlation coefficient, Po b It is the qualified target value of the correlation coefficient; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1.

[0033] Furthermore, after receiving the threshold constraint instruction, the scene complexity Fot constrains the image quality threshold in the following way:

[0034]

[0035] Where n is the number of usage scenarios, T ij is the difference between the image quality thresholds of the i-th usage scenario and the j-th usage scenario, T a is the average value of the difference, and Fot is the scene complexity.

[0036] Furthermore, the labeled medical imaging data is used as sample data, and the sample data is divided into a training set, a validation set, and a test set. The neural convolutional network is trained with the training set and verified on the validation set to obtain a trained image quality assessment model.

[0037] The trained image quality assessment model is tested on the test set, and the corresponding test indicators are obtained. The obtained test indicators are summarized to obtain the model test indicator set.

[0038] Furthermore, the performance coefficient is generated from the model test indicator set, wherein the obtained usable precision Dy and recall rate Zy are linearly normalized and then calculated as follows:

[0039]

[0040] Weight coefficient, 0≤α≤1, 0≤β≤1; if the performance coefficient Xts exceeds the performance threshold, an error judgment instruction is issued to the outside; if the opposite is true, a model optimization instruction is issued to the outside.

[0041] Furthermore, after receiving the error judgment instruction, several medical imaging data are reselected as the judgment sample set, and the trained image quality assessment model performs quality scoring on the medical imaging data in the judgment sample set, which is compared with the professional assessment results of the radiologist, and the degree of error between the two is used as the error ratio.

[0042] Furthermore, after continuously acquiring the error ratios Cb of several medical images, a corresponding error coefficient Cts is constructed, wherein the error ratio Cb is linearly normalized, and then in the following manner:

[0043]

[0044] Weight coefficient: 0≤S1≤1, 0≤S2≤1; the weight coefficient can be consistent with the previous value, Cb i is the error ratio of the i-th medical image, Cb a is the mean of the error ratio, Cb b is the acceptable value of the error ratio, and p is the number of medical images.

[0045] Furthermore, medical images are acquired within a preset image quality monitoring cycle, and the trained image quality assessment model is used to score the quality of the medical images. If the acquired quality score is lower than a preset image quality threshold, an alarm instruction is issued to the outside, and the time node of each alarm instruction is recorded;

[0046] If the alarm concentration Mop exceeds the concentration threshold, an image repair instruction is issued to the outside; after receiving the image repair instruction, low-quality medical images are screened out and feature extraction is performed to obtain corresponding optimization features; medical image optimization is used as the target word, and a medical image optimization knowledge graph is pre-constructed.

[0047] Furthermore, after receiving several alarm instructions within the image quality monitoring period, the alarm concentration Mop is generated according to the density of the alarm instructions, in the following manner:

[0048]

[0049] Weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1; n is the number of alarm instructions, P ij is the time interval between the ith alarm instruction and the jth alarm instruction, P a is the average value of the time interval, P b Acceptable values ​​for the time interval.

[0050] Medical image quality management system for rehabilitation monitoring, including,

[0051] The scene analysis unit performs similarity analysis on the use scenarios of medical images, aggregates the collected scene similarity data to generate a scene data set, and generates a scene complexity Fot from the scene data set. If the scene complexity Fot exceeds expectations, a labeling instruction is issued to the outside.

[0052] An image annotation unit identifies quality indicators of medical images, constructs a quality coefficient Qts of the medical images from the acquired quality indicator data, and scores the quality of the medical images based on the quality coefficient Qts, and annotates the medical images with the acquired quality scores;

[0053] The correlation analysis unit constructs the influence coefficient Yts according to the influence of the medical image quality on the diagnostic accuracy. If the influence coefficient Yts exceeds the expectation, the image quality threshold is constrained according to the scene complexity Fot. If the medical image quality is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image.

[0054] The model performance evaluation unit uses the labeled medical image data to train an image quality evaluation model, uses the trained image quality evaluation model to score the quality of the medical image data, performs error analysis on the quality score and generates a corresponding error coefficient Xts. If the error coefficient Cts is higher than the error threshold, a model optimization instruction is issued to the outside.

[0055] The quality analysis and restoration unit uses a trained image quality assessment model to perform quality screening on medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph will provide corresponding restoration solutions for the low-quality medical images and restore the low-quality medical images.

[0056] (III) Beneficial effects

[0057] The present invention provides a medical image quality management system for rehabilitation monitoring, which has the following beneficial effects:

[0058] 1. Judge the complexity of medical image usage scenarios based on the scene complexity Fot, determine the current management expectations for medical image quality, and improve the efficiency and effectiveness of subsequent image management.

[0059] 2. The quality coefficient Qts is constructed from the acquired quality index data. After each image is scored according to the quality coefficient Qts, the medical image is labeled according to the acquired quality score. Compared with labeling the quality of medical images based on manual experience, the labeling has higher accuracy and higher reliability in subsequent quality assessment of medical images.

[0060] 3. Determine the degree of impact of medical image quality on specific diagnosis through correlation analysis. If the impact is high in complex usage scenarios, constrain the image quality of medical images to ensure subsequent usage results.

[0061] 4. By constraining the image quality threshold through the complexity of the scene used by medical images, Fot, the dynamic adjustment of the medical image quality threshold can be achieved, which is more targeted when screening low-quality medical images.

[0062] 5. Evaluate the performance of the image quality assessment model obtained through training based on the performance coefficient Xts, and then determine whether its performance can meet the usage requirements. Otherwise, continue to optimize it to ensure the quality assessment effect of the image quality assessment model on medical images.

[0063] 6. Based on the error coefficient Cts, the performance of the image quality assessment model can be evaluated again, and the performance of the trained image quality assessment model can be judged and evaluated from different angles to ensure the quality of medical images.

[0064] 7. The trained image quality assessment model is used to score the quality of medical images. When applying medical image data, the quality of the images can be automatically evaluated. Combined with the alarm instructions issued, targeted processing can be achieved after low-quality images are generated.

[0065] 8. After screening out low-quality images, the corresponding repair solutions are given for the low-quality medical images through the medical image optimization knowledge graph, which can achieve targeted repair of low-quality images, improve the image quality at the current stage, achieve better results in medical image management, and also guarantee the subsequent application effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of the medical image quality management method of the present invention;

[0067] Figure 2 It is a schematic diagram of the structure of the medical image quality management system of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] See also Figure 1 The present invention provides a medical image quality management method for rehabilitation monitoring, comprising:

[0070] Step 1: Perform similarity analysis on the usage scenarios of medical images, aggregate the collected scene similarity data to generate a scene data set, and generate scene complexity Fot from the scene data set. If the scene complexity Fot exceeds expectations, issue a labeling instruction to the outside world;

[0071] The step 1 includes the following contents:

[0072] Step 101: collect usage scenario data of medical images. For example, for usage scenarios of long-term tracking of patient conditions, in telemedicine services, doctors obtain patient images remotely for diagnosis, etc., add scenario descriptions to the usage scenarios, obtain scenario similarities between different scenario descriptions through similarity analysis, and aggregate the collected scenario similarity data to generate a scenario data set;

[0073] Step 102: Generate scene complexity Fot from the scene data in the scene data set, wherein the similarity Xs between two scene descriptions is linearly normalized and the corresponding data values ​​are mapped to the interval [0,1] in the following manner:

[0074]

[0075] Where n is the number of usage scenarios, Xs ij is the similarity between the i-th usage scenario and the j-th usage scenario, Xs a is the average similarity; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1;

[0076] Based on historical data and expectations for the use and management of medical images, the scene threshold is pre-set; if the scene complexity Fot exceeds the scene threshold, it means that the current use scenario of medical images is relatively complex, and the requirements for image quality in complex application scenarios are high, requiring strict quality management. At this time, a labeling instruction is issued to the outside;

[0077] When using, combine the contents in steps 101 and 102:

[0078] When medical imaging data is needed, similarity analysis is performed on different usage scenarios, and the scenario complexity Fot is obtained based on the similarity analysis between different usage scenarios. The complexity of the medical imaging usage scenario is judged based on the scenario complexity Fot. The more complex the usage scenario, the higher the quality requirements for medical imaging. By establishing the scenario complexity, the current management expectations for medical imaging quality can be determined, and the efficiency and effectiveness of subsequent imaging management can be improved.

[0079] The existing quality management method usually determines whether the medical quality meets the usage requirements after obtaining medical imaging data. If it does not meet the usage expectations, it will be optimized in a targeted manner to ensure the quality of medical imaging data. However, considering that there may be large differences between different usage scenarios after obtaining medical imaging data, if the quality of medical images is still managed with consistent quality management evaluation standards, it may have a negative impact on subsequent diagnostic analysis.

[0080] Step 2: Identify the quality index of the medical image, construct the quality coefficient Qts of the medical image from the acquired quality index data, and score the quality of the medical image based on the quality coefficient Qts, and mark the medical image with the acquired quality score;

[0081] The step 2 includes the following contents:

[0082] Step 201: after receiving the labeling instruction, collect the medical images in the current recognition cycle, identify the quality index of the collected medical images and obtain the corresponding quality index data, including contrast, spatial resolution, artifact rate, distortion, peak signal-to-noise ratio, structural similarity, etc., and summarize and generate an image quality data set;

[0083] Step 202: construct a quality coefficient Qts of the medical image from the image quality data set, wherein the spatial resolution Fv and the artifact rate Yv are treated as dimensionless, and the corresponding weight coefficients are determined by referring to the hierarchical analysis method, and then according to the following method:

[0084]

[0085] The weight coefficient is, 0≤A f ≤1,0≤A y ≤1, C x is a constant correction coefficient, and its value falls within [-0.15, 0.15];

[0086] Based on historical data and management expectations for image quality, each image is scored according to the quality coefficient and the corresponding quality score is obtained. The medical images are marked with the quality score; and relevant parameters when obtaining the image, such as equipment type, exposure time, and patient characteristics, are recorded;

[0087] When using, combine the contents in steps 201 and 202:

[0088] After quality testing of medical images and obtaining corresponding quality indicators, the quality coefficient Qts is constructed based on the obtained quality indicator data. After each image is scored based on the quality coefficient Qts, the medical images are labeled based on the obtained quality scores. Compared with labeling the quality of medical images based on manual experience, the labeling has higher accuracy and higher reliability in subsequent quality assessment of medical images.

[0089] Step 3: Construct the influence coefficient Yts according to the influence of medical image quality on diagnostic accuracy. If the influence coefficient Yts exceeds expectations, constrain the image quality threshold according to the scene complexity Fot. If the medical image quality is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image.

[0090] The step three includes the following contents:

[0091] Step 301: perform correlation analysis in different usage scenarios to obtain correlation coefficients between the quality of medical images and the diagnostic accuracy in different usage scenarios, and use the correlation coefficients as the influence degree;

[0092] After obtaining several influences, the influence coefficient Yts is constructed, wherein the influence Po is linearly normalized and the corresponding data value is mapped to the interval [0,1[, according to the following method:

[0093]

[0094] Where n is the number of usage scenarios, Po i is the correlation coefficient in the i-th usage scenario, Po a is the average value of the correlation coefficient, Po b is the qualified target value of the correlation coefficient; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1, the value of the weight coefficient is consistent with the previous value;

[0095] If the influence coefficient Yts exceeds the preset influence threshold, a threshold constraint instruction is issued;

[0096] When in use, correlation analysis is performed to determine the degree of impact of the image quality of medical images on specific diagnosis. If the degree of impact is high in complex usage scenarios, higher quality requirements for medical images will be imposed. Therefore, it is necessary to constrain the image quality of medical images to ensure subsequent use results.

[0097] Step 302: After receiving the threshold constraint instruction, constrain the image quality threshold according to the scene complexity Fot. The constraint method is as follows:

[0098]

[0099] Where n is the number of usage scenarios, T ij is the difference between the image quality thresholds of the i-th usage scenario and the j-th usage scenario, T a is the average value of the difference, Fot is the scene complexity;

[0100] If the quality of the medical image is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image;

[0101] When used, combine the contents in steps 301 and 302:

[0102] When screening low-quality medical images, the image quality threshold is constrained by the scene complexity Fot used by the medical images, so as to achieve dynamic adjustment of the medical image quality threshold, which is more targeted when screening low-quality medical images.

[0103] Step 4: train the labeled medical image data to obtain an image quality assessment model, use the trained image quality assessment model to score the quality of the medical image data, perform error analysis on the quality score and generate a corresponding error coefficient Xts. If the error coefficient Cts is higher than the error threshold, issue a model optimization instruction to the outside.

[0104] The step 4 includes the following contents:

[0105] Step 401: Use the annotated medical image data as sample data, divide the sample data into a training set, a validation set, and a test set, train a neural convolutional network with the training set and verify it on the validation set, and then obtain a trained image quality assessment model;

[0106] Test the trained image quality assessment model on the test set and obtain the corresponding test indicators, including usable precision, recall rate, and F1-score, etc. The obtained test indicators are summarized to obtain the model test indicator set;

[0107] Step 402: Generate a performance coefficient from the model test indicators in the model test indicator set, wherein the obtained usable precision Dy and recall rate Zy are linearly normalized, and then in the following manner:

[0108]

[0109] Weight coefficient, 0≤α≤1, 0≤β≤1; the weight coefficient can be obtained by referring to the hierarchical analysis method;

[0110] Pre-set performance thresholds based on historical data and performance management expectations for image quality assessment models;

[0111] If the performance coefficient Xts exceeds the performance threshold, it means that the current model performance can basically meet the current usage needs, and an error judgment instruction is issued to the outside; if the opposite is true, it is necessary to continue to optimize the trained image quality assessment model, and issue a model optimization instruction to the outside;

[0112] When in use, after completing the construction of the image quality assessment model and testing it, the performance coefficient Xts is generated by the acquired test data, and the performance of the image quality assessment model acquired through training is evaluated based on the performance coefficient Xts to determine whether its performance can meet the use requirements. Otherwise, it is further optimized to ensure the quality assessment effect of the image quality assessment model on medical images;

[0113] Step 403: after receiving the error judgment instruction, reselect a number of medical image data as the judgment sample set, use the trained image quality assessment model to perform quality scoring on the medical image data in the judgment sample set, compare it with the professional assessment result of the radiologist, and use the error degree between the two as the error ratio;

[0114] Step 404: after continuously acquiring the error ratios Cb of a number of medical images, construct the corresponding error coefficient Cts, wherein the error ratio Cb is linearly normalized, and then in the following manner:

[0115]

[0116] Weight coefficient: 0≤S1≤1, 0≤S2≤1; the weight coefficient can be consistent with the previous value, Cb i is the error ratio of the i-th medical image, Cb a is the mean of the error ratio, Cb b is the acceptable value of the error ratio, and p is the number of medical images;

[0117] The error threshold is set in advance based on historical data and management expectations of model performance; if the error coefficient Cts is higher than the error threshold, it means that the trained image quality assessment model can meet the actual usage requirements. Otherwise, the image quality assessment model needs to be optimized. At this time, a model optimization instruction is issued to the outside.

[0118] When used, combine the contents in steps 401 to 404:

[0119] Different from the performance test on the test set, the image quality is evaluated by the trained image quality assessment model, and compared with the scores of professionals, the error between the two is determined, and then the corresponding error coefficient Cts is obtained. According to the error coefficient Cts, the performance of the image quality assessment model can be evaluated again, thereby judging and evaluating the performance of the trained image quality assessment model from two different angles, thereby ensuring the quality of medical images.

[0120] Step 5: Use the trained image quality assessment model to screen the quality of medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph will provide corresponding repair solutions for the low-quality medical images and repair the low-quality medical images.

[0121] The step five includes the following contents:

[0122] Step 501: obtain medical images within a preset image quality monitoring period, and use the trained image quality assessment model to score the quality of the medical images. If the obtained quality score is lower than the preset image quality threshold, issue an alarm instruction to the outside, and record the time node of each alarm instruction;

[0123] When in use, after completing the setting of the image quality threshold, the trained image quality assessment model will score the quality of the medical image. When applying medical image data, it can automatically evaluate its quality, and combined with the alarm instructions issued, it can achieve targeted processing after low-quality images are generated;

[0124] Step 502: after receiving several alarm instructions within the image quality monitoring period, generate the alarm concentration Mop according to the density of the alarm instructions, in the following manner:

[0125]

[0126] Weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1; the weight coefficient value is consistent with the previous value; n is the number of alarm instructions, P ij is the time interval between the ith alarm instruction and the jth alarm instruction, P a is the average value of the time interval, P b is the acceptable value for the time interval;

[0127] Based on historical data and management expectations for image quality, a concentration threshold is pre-set; if the alarm concentration Mop exceeds the concentration threshold, it means that there are many low-quality medical images in the current stage and the overall image quality is low. Under the current usage conditions, it will have a negative impact on subsequent diagnostic analysis and other activities. At this time, an image repair instruction is issued to the outside;

[0128] When in use, after acquiring several low-quality images in succession, if the concentration of low-quality images is high, it means that the image quality in the current batch is generally low, which may have a certain negative impact on the subsequent diagnosis. Therefore, by repairing the images in the current batch, the effect of subsequent diagnostic analysis can be guaranteed in complex usage scenarios;

[0129] Step 503: After receiving the image restoration instruction, low-quality medical images are screened out, corresponding image quality data, as well as device type, exposure time, and patient data are identified, and feature extraction is performed after taking them as image-related data to obtain corresponding optimization features;

[0130] Taking medical image optimization as the target word, after deep retrieval and entity relationship building, a medical image optimization knowledge graph is pre-built. Based on the correspondence between the optimization features of low-quality medical images and the optimization scheme, the medical image optimization knowledge graph provides corresponding repair schemes for low-quality medical images; the repair schemes are executed to repair low-quality medical images;

[0131] When using, combine the contents in steps 501 to 503:

[0132] After screening out low-quality images, the corresponding repair solutions are given for the low-quality medical images through the medical image optimization knowledge graph, which can achieve targeted repair of low-quality images, improve the image quality at the current stage, achieve better results in medical image management, and also guarantee the subsequent application effects.

[0133] The construction method of medical image optimization knowledge graph can refer to the following:

[0134] First, we need to clarify the construction goals of the medical imaging optimization knowledge graph, which usually includes improving the accuracy and efficiency of medical imaging diagnosis and promoting the scientific research progress of medical imaging.

[0135] Data source selection: Select high-quality medical imaging data as the data source, which can come from hospitals, medical imaging centers, scientific research institutions, etc. At the same time, it is also necessary to collect patient information, diagnostic reports, treatment records, etc. related to the imaging.

[0136] Data preprocessing: The collected medical imaging data is preprocessed, including denoising, enhancement, segmentation and other steps to improve the quality and usability of the data. In addition, the text data needs to be cleaned, segmented, annotated and other processes to facilitate subsequent knowledge extraction.

[0137] Entity recognition: Identify key entities such as lesions, organs, diseases, symptoms, etc. from pre-processed medical images and text data. This usually requires the use of machine learning or deep learning algorithms such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs).

[0138] Relationship extraction: Based on the identification of key entities, further extract the relationship between entities, such as the relationship between lesions and diseases, the relationship between diseases and symptoms, etc. Relationship extraction can be based on rule-based methods or deep learning-based methods.

[0139] Knowledge modeling: The extracted entities and relationships are represented graphically to form a medical imaging optimization knowledge graph. This usually requires the use of a graph database or a specialized knowledge graph building tool.

[0140] Knowledge fusion: Medical imaging knowledge graphs from different sources and formats are integrated to form a more comprehensive and accurate knowledge system. This requires solving problems such as entity alignment and relationship fusion.

[0141] Knowledge update: With the continuous advancement of medical technology and the emergence of new knowledge, the medical imaging optimization knowledge graph needs to be updated regularly to maintain its timeliness and accuracy.

[0142] The Analytic Hierarchy Process (AHP) is a decision-making method that decomposes decision-related elements into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. It is particularly suitable for dealing with target systems with hierarchical and staggered evaluation indicators, and when the target value is difficult to describe quantitatively, the AHP is an effective decision-making tool.

[0143] The core of the hierarchical analysis method is to decompose the decision problem into multiple levels to form a hierarchical structure, which usually includes the target level, the criterion level, the sub-criterion level and the solution level. By solving the eigenvector of the judgment matrix, the priority weight of each element of each level to an element of the previous level is obtained, and the final weight of each alternative solution to the total goal is recursively merged by the weighted sum method, so as to find the optimal solution.

[0144] See also Figure 2 The present invention provides a medical image quality management system for rehabilitation monitoring, comprising:

[0145] The scene analysis unit performs similarity analysis on the use scenarios of medical images, aggregates the collected scene similarity data to generate a scene data set, and generates a scene complexity Fot from the scene data set. If the scene complexity Fot exceeds expectations, a labeling instruction is issued to the outside.

[0146] An image annotation unit identifies quality indicators of medical images, constructs a quality coefficient Qts of the medical images from the acquired quality indicator data, and scores the quality of the medical images based on the quality coefficient Qts, and annotates the medical images with the acquired quality scores;

[0147] The correlation analysis unit constructs the influence coefficient Yts according to the influence of the medical image quality on the diagnostic accuracy. If the influence coefficient Yts exceeds the expectation, the image quality threshold is constrained according to the scene complexity Fot. If the medical image quality is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image.

[0148] The model performance evaluation unit uses the labeled medical image data to train an image quality evaluation model, uses the trained image quality evaluation model to score the quality of the medical image data, performs error analysis on the quality score and generates a corresponding error coefficient Xts. If the error coefficient Cts is higher than the error threshold, a model optimization instruction is issued to the outside.

[0149] The quality analysis and restoration unit uses a trained image quality assessment model to perform quality screening on medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph will provide corresponding restoration solutions for the low-quality medical images and restore the low-quality medical images.

[0150] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0151] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0152] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0153] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0154] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0155] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0156] If the functions 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 application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0157] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A medical image quality management method for rehabilitation monitoring, characterized in that: include, Conduct similarity analysis on the use scenarios of medical images, aggregate the collected scene similarity data to generate a scene data set, and generate the scene complexity Fot from the scene data set. If the scene complexity Fot exceeds expectations, issue a labeling instruction to the outside world. Identify the quality index of the medical image, construct the quality coefficient Qts of the medical image based on the acquired quality index data, score the quality of the medical image based on the quality coefficient Qts, and annotate the medical image with the acquired quality score; The influence coefficient Yts is constructed according to the influence of medical image quality on diagnostic accuracy. If the influence coefficient Yts exceeds expectations, the image quality threshold is constrained according to the scene complexity Fot. If the medical image quality is lower than the image quality threshold, the corresponding medical image is regarded as a low-quality image. The image quality assessment model is obtained by training the labeled medical image data. The trained image quality assessment model scores the quality of the medical image data, performs error analysis on the quality score and generates a corresponding error coefficient Cts. If the error coefficient Cts is higher than the error threshold, a model optimization instruction is issued to the outside. The trained image quality assessment model is used to screen the quality of medical images. If the concentration of low-quality images within the image quality monitoring cycle exceeds expectations, the medical image optimization knowledge graph provides corresponding repair solutions for low-quality medical images and repairs the low-quality medical images. After receiving several alarm instructions within the image quality monitoring cycle, the alarm concentration Mop is generated according to the density of the alarm instructions in the following way: Weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1; n is the number of alarm instructions, P ij is the time interval between the ith alarm instruction and the jth alarm instruction, P a is the average value of the time interval, P b Acceptable values ​​for the time interval.

2. The medical image quality management method for rehabilitation monitoring according to claim 1, characterized in that: Collect usage scenario data of medical images, add scenario descriptions to the usage scenarios, obtain the scenario similarity between different scenario descriptions through similarity analysis, and aggregate the collected scenario similarity data to generate a scenario data set; generate the scenario complexity Fot from the usage scenario data in the scenario data set; Among them, the similarity Xs between the two scene descriptions is linearly normalized, and the corresponding data values ​​are mapped to the interval [0,1], and the scene complexity Fot is generated as follows: Where n is the number of usage scenarios, Xs ij is the similarity between the i-th usage scenario and the j-th usage scenario, Xs a is the average similarity; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1.

3. The medical image quality management method for rehabilitation monitoring according to claim 2, characterized in that: After receiving the labeling instruction, the medical images within the current recognition cycle are collected, the quality index of the collected medical images is identified and the corresponding quality index data is obtained, and the image quality data set is generated by aggregation; The quality coefficient Qts of medical images is constructed from the image quality data set, where the spatial resolution Fv and the artifact rate Yv are treated as dimensionless, and the corresponding weight coefficients are determined by referring to the hierarchical analysis method, and then according to the following method: Weight coefficient, 0≤A f ≤1,0≤A y ≤1, C x is a constant correction factor.

4. The medical image quality management method for rehabilitation monitoring according to claim 3, characterized in that: Correlation analysis is performed in different usage scenarios to obtain the correlation coefficient between the quality of medical images and the diagnostic accuracy in different usage scenarios and use it as the influence degree; after obtaining several influence degrees, the influence coefficient Yts is constructed. If the influence coefficient Yts exceeds the preset influence threshold, a threshold constraint instruction is issued.

5. The medical image quality management method for rehabilitation monitoring according to claim 4, characterized in that: Perform linear normalization on the influence degree Po, map the corresponding data value to the interval [0,1], and construct the influence coefficient Yts as follows: Where n is the number of usage scenarios, Po i is the correlation coefficient in the i-th usage scenario, Po a is the average value of the correlation coefficient, Po b It is the qualified target value of the correlation coefficient; weight coefficient: 0≤F1≤1, 0≤F2≤1, and F1+F2=1.

6. The medical image quality management method for rehabilitation monitoring according to claim 5, characterized in that: After receiving the threshold constraint instruction, the scene complexity Fot constrains the image quality threshold in the following way: Where n is the number of usage scenarios, T ij is the difference between the image quality thresholds of the i-th usage scenario and the j-th usage scenario, T a is the average value of the difference, and Fot is the scene complexity.

7. The medical image quality management method for rehabilitation monitoring according to claim 6, characterized in that: The labeled medical imaging data is used as sample data, and the sample data is divided into a training set, a validation set, and a test set. The neural convolutional network is trained with the training set and verified on the validation set to obtain the trained image quality assessment model; the trained image quality assessment model is tested on the test set, and corresponding test indicators are obtained, and the obtained test indicators are summarized to obtain a model test indicator set; The performance coefficient is generated from the model test indicator set, where the obtained usable precision Dy and recall rate Zy are linearly normalized and then calculated as follows: Weight coefficient, 0≤α≤1, 0≤β≤1; if the performance coefficient Xts exceeds the performance threshold, an error judgment instruction is issued to the outside; if the opposite is true, a model optimization instruction is issued to the outside.

8. The medical image quality management method for rehabilitation monitoring according to claim 7, characterized in that: After receiving the error judgment instruction, several medical imaging data are reselected as the judgment sample set. The trained image quality assessment model performs quality scoring on the medical imaging data in the judgment sample set, which is compared with the professional evaluation results of the radiologists, and the degree of error between the two is used as the error ratio.

9. The medical image quality management method for rehabilitation monitoring according to claim 8, characterized in that: After continuously acquiring the error ratios Cb of several medical images, the error coefficient Cts is constructed, wherein the error ratio Cb is linearly normalized and then follows the following method: Weight coefficient: 0≤S1≤1, 0≤S2≤1; the weight coefficient can be consistent with the previous value, Cb i is the error ratio of the i-th medical image, Cb a is the mean of the error ratio, Cb b is the acceptable value of the error ratio, and p is the number of medical images.

10. The medical image quality management method for rehabilitation monitoring according to claim 9, characterized in that: Acquire medical images within a preset image quality monitoring cycle, and use the trained image quality assessment model to score the quality of the medical images. If the acquired quality score is lower than the preset image quality threshold, issue an alarm command to the outside, and record the time node of each alarm command; If the alarm concentration Mop exceeds the concentration threshold, an image repair instruction is sent to the outside; After receiving the image restoration instruction, low-quality medical images are screened out and feature extraction is performed to obtain corresponding optimization features; Taking medical image optimization as the target word, a medical image optimization knowledge graph is pre-built.

Images