Image DR examination result evaluation method and device, electronic equipment and medium

Abstract

The invention provides an image DR examination result evaluation method and device, electronic equipment and a medium, medical image DR examination results and corresponding labels and images can be obtained from a DICOM file, the labels comprise multiple data types, and a total label score is determined according to scores of the labels of different data types; and determining an image score based on the key points in the medical image DR examination result, and determining the evaluation quality of the medical image DR examination result according to the total tag score and the image score. According to the method, quality evaluation is carried out from multiple aspects of the medical image DR examination result and the corresponding label, corresponding data can be automatically obtained, the evaluation quality data of the medical image DR examination result can be output, the accuracy, stability and efficiency of the evaluation quality of the medical image DR examination result are improved, the difference and subjectivity of manual evaluation are reduced, and the evaluation efficiency of the medical image DR examination result is improved. The method is beneficial to improving the homogenization management capability of the image examination process and details, guarantees the effective implementation of the mutual recognition of examination results, and reduces the medical risk of a patient in the treatment process.

Description

Technical Field

[0001] The present invention belongs to the technical field of data sharing processing and analysis for the mutual recognition of digital medical imaging examination results (images and information data), and particularly aims at the homogeneous intelligent quality control detection and evaluation of the results of digital DR examination items. Specifically, the present invention relates to a method, device, electronic device and medium for evaluating the results of imaging DR examination. Background Art

[0002] Digital Radiography (DR) examination has been widely used in various clinical diagnoses and treatments. The main advantages of DR examination results include low radiation, high resolution, fast imaging time, high examination efficiency, etc., making it popular in the medical field.

[0003] With the digitization, networking and platformization of medical images and the aggregation of data, it has become inevitable to promote the mutual recognition and sharing of imaging examination results among medical institutions. The requirement for homogenization of DR examination results has gradually become an important issue. Although the medical community has recognized the importance of quality control, the existing methods still have many limitations. At present, the quality control strategy mainly relies on formulating work standards, standardizing operations, manually reviewing and sampling to evaluate the value of examination results, judging the implementation of process details, and then continuously improving through the PDAC cycle (Plan: plan, Do: execute, Check: check, and Action: action). There are subjectivity, non-uniformity and time delay, which exacerbate the complexity of the quality control method, resulting in the homogenization of examination results being affected and restricting the development of mutual recognition and sharing.

[0004] In the context of the rapid development of current artificial intelligence technology, many studies are exploring intelligent evaluation of disease diagnosis and data reconstruction and reorganization in medical image data. Some are extracting single imaging images from DICOM files for quality control standard research. In the actual application of medical imaging examination scenarios, using only imaging images for quality control has defects such as the lack of monitoring of key information in the actual operation links of the imaging examination path, the inability to judge the execution interval of actual operation standard points, and insufficient data dimensions. Although the data of imaging examination result materials are stored and transmitted in the form of DICOM files, integrating patient information, image information, image acquisition parameters, image processing and display parameters related to the images, their intelligent application and exploration are insufficient, and there is an urgent need for a more systematic and scientific imaging quality control method to solve this problem.

[0005] Therefore, how to evaluate the quality of DR medical images from multiple aspects and make the corresponding evaluation results more objective is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0006] In view of the above-mentioned disadvantages of the prior art, the present invention provides a method, device, electronic device and medium for evaluating the results of imaging DR examinations to solve the above technical problems.

[0007] To achieve the above object and other related objects, the technical solutions provided in this application are as follows.

[0008] The present invention provides a method for evaluating the results of imaging DR examinations, including:

[0009] Obtain DR medical images and corresponding labels, where the labels include multiple data types;

[0010] Determine the total label score based on the scores of labels of different data types;

[0011] Determine the image score based on the key points in the DR medical image;

[0012] Determine the quality evaluation result of the DR medical image according to the total label score and the image score.

[0013] In the technical solution provided by the embodiments of this application, the determining the total label score based on the scores of labels of different data types includes: classifying the labels according to the data type, and determining the score of the label based on the classified labels and the corresponding preset label thresholds; determining the total label score based on the scores of the labels and the corresponding weights.

[0014] In the technical solution provided by the embodiments of this application, the labels of different data types include qualitative labels and quantitative labels. The determining the score of the label based on the classified labels and the corresponding preset label thresholds includes: when the label is a qualitative label, if the qualitative label is equal to the preset qualitative label value, then determine that the score of the label is the preset qualitative score; if the qualitative label is not equal to the preset qualitative label value, then determine that the score of the label is zero; when the label is a quantitative label, if the quantitative label meets the preset quantitative label value range, then determine that the score of the label is the preset quantitative score; if the quantitative label does not meet the preset quantitative label value range, then determine that the score of the label is zero; where the qualitative label is a descriptive feature of the DR medical image, and the quantitative label is a quantitative value of the DR medical image.

[0015] In the technical solution provided by the embodiments of this application, the determining the total label score based on the scores of the labels and the corresponding weights includes: assigning weights to the preset qualitative score and the preset quantitative score; performing a weighted calculation on the preset qualitative score and the preset quantitative score to obtain the total label score.

[0016] In the technical solution provided by the embodiments of the present application, determining the image score based on the key points in the DR medical image includes: providing an image quality evaluation model for characterizing the correspondence between the key points in the image and the score; inputting the DR medical image into the image quality evaluation model to score the key points of the DR medical image based on the correspondence to obtain the image score.

[0017] In the technical solution provided by the embodiments of the present application, the steps for establishing the image quality evaluation model include: obtaining sample DR medical images; performing key point positioning and recognition on the sample DR medical images to obtain the key points of the sample DR medical images; inputting the key points of the sample DR medical images into a neural network model for feature extraction and score prediction to obtain a sample correspondence; constructing a loss function based on the sample correspondence and the actual correspondence; updating the parameters of the neural network model according to the loss value of the loss function until the loss value is within a preset loss threshold range to obtain the image quality evaluation model.

[0018] In the technical solution provided by the embodiments of the present application, the loss value includes the loss of the target detection region and the loss of the target point, and the determination method of the loss of the target detection region is as follows:

[0019] Loss box [[ID=J11]]=λ coord [(x pred -x true ) 2 +(y pred -y true ) 2 +(w pred -w true )2 +(h pred -h true ) 2

[0020] In the above expression, Loss box is the loss of the target detection region, x pred represents the predicted X coordinate of the center of the target detection region, x true represents the actual X coordinate of the center of the target detection region, y pred represents the predicted y coordinate of the center of the target detection region, y true represents the actual y coordinate of the center of the target detection region, w pred represents the predicted width of the target detection region, w true represents the actual width of the target detection region, h pred represents the predicted height of the target detection region, h true ​Represents the actual height of the target detection area, λ coord is a hyperparameter, that is, a preset parameter before starting training;

[0021] The loss of the target point is determined as follows:

[0022]

[0023] In the above expression, Loss keypoints is the loss of the target point, n represents the number of target points in the sample DR medical image, x pred，i , y pred，i represents the predicted coordinates of the i-th target point; x true，i , y true，i represents the true coordinates of the i-th target point.

[0024] According to another aspect of the embodiments of the present application, there is provided an evaluation device for the results of an imaging DR examination, including: an acquisition module for acquiring a DR medical image and a corresponding label, the label including multiple data types; a label evaluation module for determining a total label score based on the scores of labels of different data types; an image evaluation module for determining an image score based on key points in the DR medical image; and a comprehensive evaluation module for determining a quality evaluation result of the DR medical image according to the total label score and the image score.

[0025] According to another aspect of the embodiments of the present application, there is provided an electronic device, including: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, enabling the electronic device to implement the evaluation method for the results of an imaging DR examination as described above.

[0026] According to another aspect of the embodiments of the present application, there is provided a computer-readable storage medium, on which computer-readable instructions are stored, when the computer-readable instructions are executed by a processor of a computer, enabling the computer to execute the evaluation method for the results of an imaging DR examination as described above.

[0027] The present application provides an evaluation method, device, electronic device and medium for the results of imaging DR examinations. DR medical images and corresponding labels are obtained. The labels include multiple data types, and the total label score is determined according to the scores of labels of different data types. The image score is determined based on key points in the DR medical images, and the quality evaluation result of the DR medical images is determined according to the total label score and the image score. The present application conducts quality evaluation on DR medical images from multiple aspects of DR medical images and corresponding labels, increases the reference materials for the quality evaluation of DR medical images, can automatically perform quality evaluation on the quality of DR medical images, improves the accuracy and efficiency of the quality evaluation of DR medical images, reduces the differences and subjectivity of manual evaluation, is beneficial to enhancing the homogeneous management ability of the imaging examination process and details, and ensures the effective implementation of the mutual recognition of examination results, reducing the medical risks during patients' visits. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flowchart of the evaluation method for the results of imaging DR examinations shown in an exemplary embodiment of the present invention;

[0029] Figure 2 is a schematic diagram showing the determination of the total label score shown in an exemplary embodiment of the present invention;

[0030] Figure 3 is a schematic diagram showing the determination of the image score shown in an exemplary embodiment of the present invention;

[0031] Figure 4 is a block diagram of the evaluation device for the results of imaging DR examinations shown in an exemplary embodiment of the present invention;

[0032] Figure 5 is a comprehensive evaluation system designed based on the evaluation device for the results of imaging DR examinations shown in an exemplary embodiment of the present invention;

[0033] Figure 6 is a schematic diagram of the structure of a computer system of an electronic device suitable for implementing the present invention shown in an exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] The following illustrates the embodiments of the present invention through specific specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0035] It should be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of the present invention. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The types, quantities, and proportions of the components in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0036] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0037] DICOM file (Digital Imaging and Communications in Medicine) is a digital image format commonly used in the medical field. It is used to store medical images and related data and can be shared and exchanged between different devices, providing important assistance for doctors in diagnosis and treatment. A DICOM file consists of two main components: metadata and pixel data.

[0038] The metadata in a DICOM file is called a "tag", and each tag contains specific information about the image. A tag consists of an identifier and a value. The identifier tells us what type of information this tag corresponds to, for example, patient name or date and time. The value provides the specific information, such as the actual name of the patient or the exact time when the image was acquired. A DICOM file contains multiple different tags, and these tags are organized into different groups for effective organization and management. The group and element numbers of the tags are used to determine the unique identifier of the tag. Pixel data is the actual medical image data, such as CT, MRI images, or DR medical images. The specific coding structure of the DICOM file format enables the metadata and pixel data to be stored in the same file, facilitating reading and understanding.

[0039] Digital Radiography, abbreviated as DR examination, has been widely used in various clinical diagnoses and treatments. The main advantages of DR examination results include low radiation, high resolution, fast imaging time, and high examination efficiency, making it popular in the medical field.

[0040] With the digitization, networking, and platformization of medical images, it has become inevitable to promote the mutual recognition and sharing of imaging examination results among medical institutions. The requirement for homogenization of DR examination results has gradually become an important issue. Although the medical community has recognized the importance of quality control, existing methods still have many limitations. Currently, quality control strategies mainly rely on formulating work standards, standardizing operations, and manually reviewing and sampling to evaluate the value of examination results, judging the implementation of process details, and then making continuous improvements through the PDAC cycle (Plan: Plan, Do: Execute, Check: Check, and Action: Act). There are subjectivity, non-uniformity, and time delays, which exacerbate the complexity of the quality control method, resulting in the homogenization of examination results being affected and restricting the development of mutual recognition and sharing.

[0041] Against the backdrop of the rapid development of current artificial intelligence technology, many studies have explored intelligent evaluation of disease diagnosis and data reconstruction and reorganization in medical image data. Some have conducted research on quality control standards by extracting single imaging images from DICOM files. In the actual application of medical imaging examination scenarios, using only imaging images for quality control has defects such as the lack of monitoring of key information in the actual operation links of the imaging examination path, the inability to determine the execution range of actual operation standard points, and insufficient data dimensions. Although the data of imaging examination result materials are stored and transmitted in the form of DICOM files, integrating patient information, image information, image acquisition parameters, image processing, and display parameters related to the images, there is insufficient intelligent application and exploration, and there is an urgent need for a more systematic and scientific imaging quality control method to solve this problem.

[0042] To solve the above problems, this application obtains DR medical images and their corresponding tags, evaluates the quality of DR medical images and different types of tags respectively, obtains the image scores corresponding to the DR medical images and the total tag scores corresponding to the tags, and determines the quality evaluation results of the DR medical images based on the image scores and the total tag scores.

[0043] Please refer to Figure 1 , which shows a flowchart of an evaluation method for imaging DR examination results shown in an exemplary embodiment of the present invention.

[0044] As Figure 1 shown, in an exemplary embodiment, providing relevant steps of an evaluation method for imaging DR examination results at least includes steps S110 to S140, which are introduced in detail as follows:

[0045] Step S110, obtain DR medical images and their corresponding tags, where the tags include multiple data types.

[0046] Specifically, the DR medical images and the corresponding tags are stored in DICOM files. The DR medical images are extracted from the pixel data of the DICOM files to obtain the DR medical images, and the tags corresponding to the DR medical images are extracted from the metadata of the DICOM files through identifiers, obtaining the tags corresponding to the DR medical images; among them, the tags include multiple data types.

[0047] Step S120, determining the total tag score based on the scores of tags of different data types.

[0048] Specifically, in an exemplary embodiment of the present application, determining the total tag score based on the scores of tags of different data types includes: classifying the tags according to the data type to determine the scores of the tags based on the classified tags and the corresponding preset tag thresholds; determining the total tag score based on the scores of the tags and the corresponding weights. Specifically, different data types include the attributes and quantities of things. The tags are classified according to the data type, the scores of the tags are determined through the classified tags and the corresponding preset tag thresholds, and the total tag score is determined according to the scores of each tag and the corresponding weights.

[0049] More specifically, in an exemplary embodiment of the present application, the tags of different data types include qualitative tags and quantitative tags. Determining the scores of the tags based on the classified tags and the corresponding preset tag thresholds includes: when the tag is a qualitative tag, if the qualitative tag is equal to the preset qualitative tag value, then determining the score of the tag as the preset qualitative score; if the qualitative tag is not equal to the preset qualitative tag value, then determining the score of the tag as zero; when the tag is a quantitative tag, if the quantitative tag satisfies the preset quantitative tag value range, then determining the score of the tag as the preset quantitative score; if the quantitative tag does not satisfy the preset quantitative tag value range, then determining the score of the tag as zero; among them, the qualitative tag is the descriptive feature of the DR medical image, and the quantitative tag is the quantitative value of the DR medical image.

[0050] More specifically, in an exemplary embodiment of the present application, determining the total tag score based on the scores of the tags and the corresponding weights includes: assigning weights to the preset qualitative score and the preset quantitative score; performing a weighted calculation on the preset qualitative score and the preset quantitative score to obtain the total tag score.

[0051] It should be emphasized that tags of different data types include qualitative tags and quantitative tags. Qualitative tags describe the attributes or characteristics of DR medical images; quantitative tags involve quantified values (quantities or metrics) of DR medical images, which are represented by numbers and can be subjected to mathematical and statistical analysis. By classifying tags by data type, quantitative tags and qualitative tags are obtained. For example, the image type describes the characteristics of the image itself, and the image type is a qualitative tag. The tube current is related to the current magnitude during detection, and the tube current is a quantitative tag. Qualitative tags include image type, detection site, and examination position, etc.; quantitative tags include tube current, tube voltage, exposure time, exposure dose, SID (Source-to-Image Distance), and FOV (Field of View), etc.

[0052] Please refer to Figure 2 , Figure 2 which is a schematic diagram showing the determination of the total tag score shown in an exemplary embodiment of the present invention.

[0053] As Figure 2 shown, when the tag is a qualitative tag, the qualitative tag is the image type (Image Type). The tags "DX", "CT", "MRI", or other types of the corresponding field are extracted through the identifier (0008, 0008). If the image type is equal to the corresponding preset qualitative tag value in the DR medical image, the score of the image type is determined as X1. If the image type is not equal to the corresponding preset qualitative tag value in the DR medical image or the image type is empty, the score of the image type is determined as zero, and the preset weight corresponding to the score of the image type is α1; the qualitative tag is the examination site (Body Part Examined), and the tags "Chest" or "Elbow", etc. of the corresponding field are extracted through the identifier (0018, 0015). If the examination site is equal to the corresponding preset qualitative tag value in the DR medical image, the score of the examination site is determined as X2. If the examination site is not equal to the corresponding preset qualitative tag value in the DR medical image or the examination site is empty, the score of the examination site is determined as zero, and the preset weight corresponding to the score of the examination site is α2; the qualitative tag is the examination position (View Position), and the tags "AP" (anteroposterior) or "LAT" (lateral), etc. of the corresponding field are extracted through the identifier (0018, 5101). If the examination position is equal to the corresponding preset qualitative tag value in the DR medical image, the score of the examination position is determined as X3. If the examination position is not equal to the corresponding preset qualitative tag value in the DR medical image or the examination position is empty, the score of the examination position is determined as zero, and the preset weight corresponding to the score of the examination position is α3.

[0054] As Figure 2As shown, when the label is a quantitative label, the quantitative label is the peak kilovoltage. The peak kilovoltage is extracted through the identifiers (0018, 0060). If the peak kilovoltage meets the pre-set quantitative label value range (for example, when the examination part is the elbow, the pre-set quantitative label value range of the peak kilovoltage for the elbow examination part is 55 - 65 Kv), then the score of the peak kilovoltage is determined as X4. If the peak kilovoltage does not meet the pre-set quantitative label value range, then the score of the peak kilovoltage is determined as zero, and the corresponding preset weight for the score of the peak kilovoltage is α4; the quantitative label is the x-ray tube current. The x-ray tube current is extracted through the identifiers (0018, 1151). If the x-ray tube current meets the pre-set quantitative label value range (for example, when the examination part is the elbow, the pre-set quantitative label value range of the x-ray tube current for the elbow examination part is 3 - 5 mA), then the score of the x-ray tube current is X5. If the x-ray tube current does not meet the pre-set quantitative label value range, then the score of the x-ray tube current is determined as zero, and the corresponding preset weight for the score of the x-ray tube current is α5; the quantitative label is the exposure time. The exposure time is extracted through the identifiers (0018, 1150). If the exposure time meets the pre-set quantitative label value range (for example, when the examination part is the elbow, the pre-set quantitative label value range of the exposure time for the elbow examination part is 5 - 20 ms), then the score of the exposure time is determined as X6. If the exposure time does not meet the pre-set quantitative label value range, then the score of the exposure time is determined as zero, and the corresponding preset weight for the score of the exposure time is α6; the quantitative label is the image and fluoroscopy area dose product. The image and fluoroscopy area dose product is extracted through the identifiers (0018, 115E). If the image and fluoroscopy area dose product meets the pre-set quantitative label value range (for example, when the examination part is the elbow, the pre-set quantitative label value range of the image and fluoroscopy area dose product for the elbow examination part is less than 2 dGy·cm 2) If the exposure dose meets the requirements, the score for the exposure dose is determined as X7; if the exposure dose does not meet the pre-set quantitative label value range, the score for the exposure dose is determined as zero, and the preset weight corresponding to the score of the exposure dose is α7. The quantitative label is the source-to-image distance (SID). The source-to-image distance is extracted through the identifier (0018, 1110). If the source-to-image distance meets the pre-set quantitative label value range (for example, when the examination part is Elbow, the pre-set quantitative label value range of the source-to-image distance for the examination part is 100 - 120 cm), the score for the source-to-image distance is determined as X8; if the source-to-image distance does not meet the pre-set quantitative label value range, the score for the source-to-image distance is determined as zero, and the preset weight corresponding to the score of the source-to-image distance is α8. The quantitative label is the field of view (FOV). The field of view is extracted through the identifier (0018, 1149). If the field of view meets the pre-set quantitative label value range (for example, when the examination part is Elbow, the pre-set quantitative label value range of the field of view for the examination part is 15 cm × 15 cm - 25 cm × 25 cm), the score for the field of view is determined as X9; if the field of view does not meet the pre-set quantitative label value range, the score for the field of view is determined as zero, and the preset weight corresponding to the score of the field of view is α9.

[0055] Specifically, as Figure 2 shown, the steps to determine the total label score based on the label score and the corresponding weight are as follows: According to the specific situation of the label, configure the corresponding weights for the pre-set qualitative score and the pre-set quantitative score, and perform weighted calculation on the pre-set qualitative score and the pre-set quantitative score to obtain the total label score. The determination expression of the total label score is as shown in expression (1):

[0056] L = X1α1 + X2α2 + … + X9α9 (1)

[0057] In expression (1), L is the total label score, α1 - α9 are the scores of the label, and X1 - X9 are the corresponding weights of the label scores.

[0058] Step S130: Determine the image score based on the key points in the DR medical image.

[0059] It should be noted that the scoring of the DR medical image uses artificial intelligence technology to perform quality assessment on the DR medical image, including but not limited to key point detection, image segmentation, and classification processing.

[0060] Please refer to Figure 3 , Figure 3 which is a schematic diagram showing the determination of the image score shown in an exemplary embodiment of the present invention.

[0061] Specifically, in an exemplary embodiment of the present application, as Figure 3As shown, after obtaining the DR medical image, preprocess the DR medical image: perform format conversion on the DR medical image based on a preset image format to obtain a preset format image; crop the size of the preset format image to obtain a standard image; normalize the pixels of the standard image to obtain a normalized image. For example, the preset image format is JPEG format. Perform format conversion on the DR medical image according to the preset image format to obtain a JPEG format preset format image in the RGB color space and 24-bit depth. Adjust the size of the preset format image to 640x640. Normalize the pixels of the standard image to [0, 1] to obtain a normalized image.

[0062] Specifically, in an exemplary embodiment of the present application, as Figure 3 shown, determining an image score based on key points in the DR medical image includes: providing an image quality assessment model, where the image quality assessment model is used to represent the correspondence between key points in the image and the score; inputting the DR medical image into the image quality assessment model to score the key points of the DR medical image based on the correspondence to obtain an image score. Specifically, provide an image quality assessment model. The image quality assessment model includes the correspondence between key points in the image and the corresponding scores. Input the DR medical image into the image quality assessment model and apply the constructed correspondence to accurately score the key points in the DR medical image to obtain an image score. It should be emphasized that the key point scoring of the image quality assessment model includes the scoring of the target detection area and the scoring of the target point. The scoring of the target detection area includes the scoring of the position and range of the target detection area. The scoring of the target point includes the position of the target point, the distance of the target point, and the distance and direction of multiple target points. Assign corresponding weights to the scoring of the target detection area and the scoring of the target point, and perform weighted calculation on the scoring of the target detection area and the scoring of the target point. The image quality assessment model outputs an image score.

[0063] Specifically, in an exemplary embodiment of the present application, the steps for establishing the image quality assessment model include: obtaining sample DR medical images; performing key point localization and recognition on the sample DR medical images to obtain the key points of the sample DR medical images; inputting the key points of the sample DR medical images into a neural network model for feature extraction and score prediction to obtain a sample correspondence; constructing a loss function based on the sample correspondence and the actual correspondence; updating the parameters of the neural network model according to the loss value of the loss function until the loss value is within a preset loss threshold range to obtain the image quality assessment model. Specifically, multiple sample DR medical images are obtained, key points of the DR medical images are localized to obtain a target detection region, key points are recognized in the target detection region to obtain the key points of the sample DR medical images, the key points of the sample DR medical images are input into the neural network model for feature extraction and the corresponding scores are output, a correspondence between the key points and scores in the image is established based on the neural network module to obtain a sample correspondence; a loss function is established through the sample correspondence and the actual correspondence; the parameters of the neural network model are updated based on the loss value of the loss function to improve the accuracy of the image scores output by the image quality assessment model until the loss value is within a preset loss threshold range to obtain the image quality assessment model. Particularly, the neural network model can be the YOLOv8n model, which analyzes the relationship between the key points and scores of the sample DR medical images. The function of the YOLOv8n model is not limited to key point detection, but also covers multiple tasks such as image segmentation and classification, and has the ability of continuous learning and self-optimization, thus ensuring a high degree of accuracy of the evaluation results.

[0064] More specifically, in an exemplary embodiment of the present application, the loss value includes the loss of the target detection region and the loss of the target point. The determination method of the target detection region loss is shown in Expression (2):

[0065] Loss box =λ coord [(x pred -x true ) 2 +(y pred -y true ) 2 +(w pred -w true )2 +(h pred -h true ) 2 (2)

[0066] In Expression (2), Loss box is the loss of the target detection region, x pred represents the predicted X coordinate of the center of the target detection region, x trueRepresents the actual X coordinate of the center of the target detection area, y pred Represents the predicted y coordinate of the center of the target detection area, y true Represents the actual y coordinate of the center of the target detection area, w pred Represents the predicted width of the target detection area, w true Represents the actual width of the target detection area, h pred Represents the predicted height of the target detection area, h true Represents the actual height of the target detection area, λ coord Is a hyperparameter.

[0067] The loss of the target point is determined as shown in Expression (3):

[0068]

[0069] In Expression (3), Loss keypoints Is the loss of the target point, n represents the number of target points in the sample DR medical image, x pred，i , y pred，i Represents the predicted coordinates of the i-th target point; x true，i , y true，i Represents the true coordinates of the i-th target point.

[0070] Step S140, determine the quality assessment result of the DR medical image according to the total label score and the image score.

[0071] Specifically, weights are configured for the total label score and the image score, and weighted calculations are performed on the total label score and the image score to obtain the total score of the DR medical image. The total score is matched with the preset quality assessment standard to obtain the quality assessment result of the DR medical image. Since different DR medical images and their corresponding labels have different importance levels in different examination parts, the weights need to be set specifically according to the actual situation.

[0072] The determination method of the quality assessment result of the DR medical image is shown in Expression (4):

[0073] S = w L L + w A A (4)

[0074] In Expression (4), S is the total score of the DR medical image, L is the total label score, w L Is the weight corresponding to the total label score, A is the image score, w A Is the weight corresponding to the image score.

[0075] It should be emphasized that the determination steps of the total label score and the image score are carried out simultaneously, without a specific order.

[0076] The present application provides a technical solution for quality assessment of imaging DR examination results. DR medical images and corresponding labels are obtained. The labels include multiple data types. The total label score is determined according to the scores of labels of different data types. The image score is determined based on key points in the DR medical images. The quality assessment result of the DR medical images is determined according to the total label score and the image score. The present application conducts quality assessment on DR medical images from multiple aspects of DR medical images and corresponding labels, increases the reference materials for quality assessment of DR medical images, can automatically conduct quality assessment on the quality of DR medical images, improves the accuracy, stability and efficiency of quality assessment of DR medical images, reduces the differences and subjectivity of manual evaluation, is beneficial to enhancing the homogenized management ability of imaging examination processes and details, ensures the effective implementation of mutual recognition of examination results, and reduces the medical risks during patients' visits.

[0077] Please refer to Figure 4 , Figure 4 which is an evaluation device for imaging DR examination results shown in an exemplary embodiment of the present invention.

[0078] As Figure 4 shown, the exemplary evaluation device for imaging DR examination results includes: a collection module 410, a label evaluation module 420, an image evaluation module 430, and a comprehensive evaluation module 440.

[0079] The collection module 410 is used to obtain DR medical images and corresponding labels, and the labels include multiple data types;

[0080] The label evaluation module 420 is used to determine the total label score based on the scores of labels of different data types;

[0081] The image evaluation module 430 is used to determine the image score based on key points in the DR medical images;

[0082] The comprehensive scoring module 440 is used to determine the quality assessment result of the DR medical images according to the total label score and the image score.

[0083] It should be noted that the evaluation device for imaging DR examination results provided in the above embodiment and the evaluation method for imaging DR examination results provided in the above embodiment belong to the same concept. The specific manners in which each module and unit perform operations have been described in detail in the method embodiment and will not be elaborated here. In practical applications, the evaluation device for imaging DR examination results provided in the above embodiment can, as needed, allocate the above functions to different functional modules, that is, divide the internal structure of the device into different functional modules to complete all or part of the above-described functions, and no limitation will be made here either.

[0084] Please refer to Figure 5 ,Figure 5 It is a comprehensive evaluation system designed for the evaluation device based on the imaging DR examination results shown in an exemplary embodiment of the present invention.

[0085] By expanding the design of the evaluation device for imaging DR examination results, such as Figure 5 shown, relevant staff can view the quality evaluation results in the comprehensive scoring module 440 of the evaluation device for imaging DR examination results through a user-friendly interface. After viewing the results, provide feedback and rectification opinions on the evaluation results. Before the next evaluation, upload the DR medical image and the corresponding label to the acquisition module 410, and adjust the parameters in the evaluation device for imaging DR examination results according to the feedback results and modification opinions; this system ensures that the quality evaluation results of the evaluation device for imaging DR examination results are both accurate and efficient.

[0086] An embodiment of the present invention also provides an electronic device, including: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, enabling the electronic device to implement the evaluation method for imaging DR examination results provided in each of the above embodiments.

[0087] Please refer to Figure 6 , which shows a schematic structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present invention. It should be noted that Figure 6 the shown computer system 600 of the electronic device is only an example and should not bring any limitations to the functions and usage scope of the embodiments of the present invention.

[0088] As Figure 6 shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage section 608 into the random access memory (RAM) 603, such as executing the method in the above embodiments. In the RAM 603, various programs and data required for system operation are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other through a bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0089] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read therefrom can be installed into the storage section 608 as needed.

[0090] Specifically, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 609, and / or installed from the removable medium 611. When the computer program is executed by a central processing unit (CPU) 601, various functions defined in the system of the present invention are executed.

[0091] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the method for evaluating the imaging DR examination results as described above. The computer-readable storage medium may be included in the electronic device described in the above embodiment, or may exist separately without being assembled into the electronic device.

[0092] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0093] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. Among them, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0094] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those of ordinary skill in the art without departing from the spirit and technical idea disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. An evaluation method for the results of imaging DR examination, characterized in that Including: Obtain DR medical images and corresponding labels, where the labels include multiple data types; Determine the total label score based on the scores of labels of different data types; Determine the image score based on the key points in the DR medical image; Determine the quality assessment result of the DR medical image according to the total label score and the image score.

2. The evaluation method of the imaging DR examination result according to claim 1, characterized in that, The determining the total label score based on the scores of labels of different data types includes: Classify the labels according to the data types, so as to determine the scores of the labels based on the classified labels and the corresponding preset label thresholds; Determine the total label score based on the scores of the labels and the corresponding weights.

3. The evaluation method of the imaging DR examination result according to claim 2, characterized in that, The labels of different data types include qualitative labels and quantitative labels. The determining the scores of the labels based on the classified labels and the corresponding preset label thresholds includes: When the label is a qualitative label, if the qualitative label is equal to the preset qualitative label value, then determine that the score of the label is the preset qualitative score; if the qualitative label is not equal to the preset qualitative label value, then determine that the score of the label is zero; When the label is a quantitative label, if the quantitative label meets the preset quantitative label value range, then determine that the score of the label is the preset quantitative score; if the quantitative label does not meet the preset quantitative label value range, then determine that the score of the label is zero; Wherein, the qualitative label is a descriptive feature of the DR medical image, and the quantitative label is a quantitative value of the DR medical image.

4. The evaluation method for the imaging DR examination results according to claim 3, characterized in that, The determining the total label score based on the scores of the labels and the corresponding weights includes: Assign weights to the preset qualitative score and the preset quantitative score; Perform weighted calculation on the preset qualitative score and the preset quantitative score to obtain the total label score.

5. The evaluation method of the imaging DR examination result according to claim 1, wherein The determining the image score based on the key points in the DR medical image includes: Provide an image quality assessment model, which is used to represent the corresponding relationship between the key points in the image and the scores; Input the DR medical image into the image quality assessment model to score the key points of the DR medical image based on the corresponding relationship, and obtain the image score.

6. The evaluation method of the imaging DR examination result according to claim 5, characterized in that The steps for establishing the image quality assessment model include: Obtain sample DR medical images; Perform key point positioning and recognition on the sample DR medical images to obtain the key points of the sample DR medical images; Input the key points of the sample DR medical images into a neural network model for feature extraction and score prediction to obtain the sample corresponding relationship; Construct a loss function based on the sample corresponding relationship and the actual corresponding relationship; Update the parameters of the neural network model according to the loss value of the loss function until the loss value is within the preset loss threshold range to obtain the image quality assessment model.

7. The evaluation method of the imaging DR examination result according to claim 6, wherein, The loss value includes the loss of the target detection area and the loss of the target points, The determination method of the loss of the target detection area is as follows: Loss box = λ coord [(x pred - x true ) 2 +(y pred - y true ) 2 +(w pred - w true ) 2 +(h pred - h true ) 2 ​ In the above expression, Loss box is the loss of the target detection region, x pred represents the predicted X coordinate of the center of the target detection region, x true represents the actual X coordinate of the center of the target detection region, y pred represents the predicted y coordinate of the center of the target detection region, y true represents the actual y coordinate of the center of the target detection region, w pred represents the predicted width of the target detection region, w true represents the actual width of the target detection region, h pred represents the predicted height of the target detection region, h true represents the actual height of the target detection region, λ coord is a hyperparameter, that is, a preset parameter before starting training; The determination method of the loss of the target points is as follows: In the above expression, Loss keypoints is the loss of the target point, n represents the number of target points in the sample DR medical image, x pred，i , y pred，i represent the predicted coordinates of the i-th target point; x true，i , y true，i represent the true coordinates of the i-th target point.

8. An evaluation device for the results of imaging DR examination, characterized in that, Including: An acquisition module, which is used to obtain DR medical images and corresponding labels, where the labels include multiple data types; A label evaluation module, configured to determine the total label score based on the scores of labels of different data types; An image evaluation module, configured to determine an image score based on key points in the DR medical image; A comprehensive evaluation module, configured to determine a quality evaluation result of the DR medical image according to the total label score and the image score.

9. An electronic device, characterized in that, The electronic device includes: One or more processors; A storage device, configured to store one or more programs, which, when executed by the one or more processors, cause the electronic device to implement the evaluation method of the imaging DR examination result according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon, which, when executed by a processor of a computer, causes the computer to execute the evaluation method of the imaging DR examination result according to any one of claims 1 to 7.

Images