Quality evaluation method, system and device of spine image, medium and program product

Abstract

The invention provides a spine image quality evaluation method, system and device, a medium and a program product. The quality evaluation method comprises the steps of obtaining a to-be-detected spine image; segmenting the to-be-detected spine image to obtain a plurality of to-be-detected spine images; determining whether the to-be-detected vertebral body image is a lateral vertebral body image or not by adopting a preset generative large model; obtaining a target number corresponding to the lateral vertebral body images in a plurality of to-be-detected vertebral body images corresponding to the to-be-detected vertebral body image; and based on the target number, evaluating the quality of the to-be-detected spine image. According to the method, the to-be-detected spine image is segmented to obtain the to-be-detected spine image, and whether the to-be-detected spine image is the lateral spine image is determined through the generative large model, so that the quality evaluation of the to-be-detected spine image is realized, the efficiency of classifying the to-be-detected spine image is improved, the cost is reduced, and the efficiency of classifying the to-be-detected spine image is improved. And the accuracy and reliability of spinal image quality evaluation are improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of spinal image processing, and in particular, to a method, system, device, medium, and program product for evaluating the quality of spinal images. Background Art

[0002] X-ray images are common data in medical practice, used for doctors to diagnose patients, and also for archiving and storing after X-ray machine shooting, and inputting into downstream systems for subsequent analysis. The standardization (normal shooting operation), integrity (shooting the complete part), and correctness (shooting the part concerned by diagnosis) of image data are crucial for doctors' diagnosis and further processing and analysis of downstream information systems.

[0003] In the diagnosis of spinal diseases, spinal lateral X-ray films are often used. It is very important to distinguish qualified lateral films because only qualified lateral films can provide accurate and comprehensive spinal image information for doctors to make correct diagnosis and treatment decisions; unqualified lateral films may have problems such as blurred images, missing parts, incorrect angles, etc., which may lead to misdiagnosis or missed diagnosis. At the same time, qualified lateral films are also a necessary condition for downstream information systems to perform data processing and analysis. Only by ensuring the image quality can the reliability and accuracy of subsequent data analysis be guaranteed.

[0004] Currently, using a pre-trained machine learning model to classify X-ray films can achieve a certain classification effect. However, due to the lack of in-depth analysis and understanding of the image content and the lack of relevant medical training data, directly applying it to the classification and recognition of medical images has a high error rate and cannot meet clinical needs. In addition, medical-related machine learning tasks often require the participation of professional medical staff for annotation, which is time-consuming and laborious. Especially in the medical field, the cost of professional annotators is high. Summary of the Invention

[0005] The technical problem to be solved by the present disclosure is to overcome the defects of low classification efficiency, low accuracy, and high cost in classifying spinal X-ray films using a machine learning model in the prior art, and to provide a method, system, device, medium, and program product for evaluating the quality of spinal images.

[0006] The present disclosure solves the above technical problem through the following technical solutions:

[0007] The present disclosure provides a method for evaluating the quality of spinal images, and the quality evaluation method includes:

[0008] Obtain a spinal image to be detected;

[0009] Perform segmentation processing on the spinal image to be detected to obtain a plurality of vertebral images to be detected;

[0010] Use a pre-set generative large model to determine whether the vertebral image to be detected is a lateral vertebral image;

[0011] Obtain the target quantity corresponding to the lateral vertebral image among a number of vertebral images to be detected corresponding to the spinal image to be detected;

[0012] Evaluate the quality of the spinal image to be detected based on the target quantity.

[0013] Preferably, the pre-set generative large model includes a multi-modal generative large model and a text generative large model;

[0014] The step of using the pre-set generative large model to determine whether the vertebral image to be detected is a lateral vertebral image includes:

[0015] Input the vertebral image to be detected into the multi-modal generative large model to obtain a first text description content;

[0016] Input the first text description content into the text generative large model to determine whether the vertebral image to be detected is the lateral vertebral image.

[0017] Preferably, the step of inputting the first text description content into the text generative large model to determine whether the vertebral image to be detected is the lateral vertebral image includes:

[0018] Input the first text description content into the text generative large model to determine whether the vertebral image to be detected is a reference vertebral image;

[0019] In response to the vertebral image to be detected being the reference vertebral image, use the feature information of the reference vertebral image as a prompt word for the text generative large model, so that the text generative large model determines whether the vertebral image to be detected is the lateral vertebral image.

[0020] Preferably, the step of performing segmentation processing on the spinal image to be detected to obtain a number of vertebral images to be detected includes:

[0021] Input the spinal image to be detected into a first pre-set target detection model to obtain a spinal region image;

[0022] Perform segmentation processing on the spinal region image to obtain a number of the vertebral images to be detected.

[0023] Preferably, the step of performing segmentation processing on the spinal region image to obtain a number of the vertebral images to be detected includes:

[0024] Perform segmentation processing on the spinal region image to obtain a number of segmentation images to be detected;

[0025] Obtain a first similarity between the first contour information of the to-be-detected segmented image and the second contour information of the lateral vertebral body image;

[0026] Based on the first similarity, determine whether the to-be-detected segmented image is the to-be-detected vertebral body image.

[0027] Preferably, the step of segmenting the spinal region image to obtain a plurality of to-be-detected segmented images includes:

[0028] Use a preset image segmentation model to segment the spinal region image to obtain a plurality of the to-be-detected segmented images;

[0029] And / or

[0030] The step of determining whether the to-be-detected segmented image is the to-be-detected vertebral body image based on the first similarity includes:

[0031] In response to the first similarity being less than a first preset similarity, determine that the to-be-detected segmented image is the to-be-detected vertebral body image;

[0032] In response to the first similarity being not less than the first preset similarity, determine that the to-be-detected segmented image is not the to-be-detected vertebral body image.

[0033] Preferably, the step of evaluating the quality of the to-be-detected spinal image based on the target quantity includes:

[0034] In response to the target quantity being greater than a preset quantity, determine that the quality of the to-be-detected spinal image is qualified;

[0035] In response to the target quantity being not greater than the preset quantity, determine that the quality of the to-be-detected spinal image is unqualified.

[0036] Preferably, before the step of segmenting the to-be-detected spinal image to obtain a plurality of to-be-detected vertebral body images, further includes:

[0037] Input the to-be-detected spinal image into a second preset target detection model to determine whether there is a foreign object in the to-be-detected spinal image;

[0038] In response to there being no foreign object in the to-be-detected spinal image, execute the step of segmenting the to-be-detected spinal image to obtain a plurality of to-be-detected vertebral body images;

[0039] In response to the presence of the foreign object in the spine image to be detected, based on the second preset object detection model, an image of the foreign object area is obtained; the image of the foreign object area is segmented to obtain a plurality of images of foreign objects to be detected; a second similarity between the third contour information of the image of the foreign object to be detected and the fourth contour information of the lateral foreign object image is obtained; in response to the second similarity being not less than a second preset similarity, perform the step of segmenting the spine image to be detected to obtain a plurality of vertebral body images to be detected.

[0040] Preferably, the quality evaluation method further includes:

[0041] In response to the second similarity being less than the second preset similarity, input the image of the foreign object to be detected into a multi-modal generative large model to obtain second text description content;

[0042] Input the second text description content into a text generative large model to determine whether the image of the foreign object to be detected is a reference foreign object image;

[0043] In response to the image of the foreign object to be detected not being the reference foreign object image, perform the step of segmenting the spine image to be detected to obtain a plurality of vertebral body images to be detected;

[0044] In response to the image of the foreign object to be detected being the reference foreign object image, evaluate the quality of the spine image to be detected using the multi-modal generative large model.

[0045] Preferably, the step of evaluating the quality of the spine image to be detected using the multi-modal generative large model includes:

[0046] Input the image of the foreign object to be detected into the multi-modal generative large model to determine whether the image of the foreign object to be detected is a lateral foreign object image;

[0047] In response to the image of the foreign object to be detected being the lateral foreign object image, determine that the quality of the spine image to be detected is qualified.

[0048] The present disclosure also provides a quality evaluation system for spine images, the quality evaluation system includes:

[0049] A spine image acquisition module, configured to acquire a spine image to be detected;

[0050] A vertebral body image acquisition module, configured to segment the spine image to be detected to obtain a plurality of vertebral body images to be detected;

[0051] A lateral vertebral body determination module, configured to use a preset generative large model to determine whether the vertebral body image to be detected is a lateral vertebral body image;

[0052] A target quantity acquisition module, configured to acquire the target quantity corresponding to the lateral vertebral body image among a plurality of to-be-detected vertebral body images corresponding to the to-be-detected spinal image;

[0053] A quality evaluation module, configured to evaluate the quality of the to-be-detected spinal image based on the target quantity.

[0054] Preferably, the preset generative large model includes a multi-modal generative large model and a text generative large model;

[0055] The lateral vertebral body determination module includes:

[0056] A first content acquisition unit, configured to input the to-be-detected vertebral body image into the multi-modal generative large model to obtain a first text description content;

[0057] A lateral vertebral body determination unit, configured to input the first text description content into the text generative large model to determine whether the to-be-detected vertebral body image is the lateral vertebral body image.

[0058] Preferably, the lateral vertebral body determination unit includes:

[0059] A reference vertebral body determination subunit, configured to input the first text description content into the text generative large model to determine whether the to-be-detected vertebral body image is a reference vertebral body image;

[0060] A lateral vertebral body determination subunit, configured to, in response to the to-be-detected vertebral body image being the reference vertebral body image, use the feature information of the reference vertebral body image as a prompt word of the text generative large model, so that the text generative large model determines whether the to-be-detected vertebral body image is the lateral vertebral body image.

[0061] Preferably, the vertebral body image acquisition module includes:

[0062] A region image acquisition unit, configured to input the to-be-detected spinal image into a first preset target detection model to obtain a spinal region image;

[0063] A vertebral body image acquisition unit, configured to perform segmentation processing on the spinal region image to obtain a plurality of the to-be-detected vertebral body images.

[0064] Preferably, the vertebral body image acquisition unit includes:

[0065] A segmented image acquisition subunit, configured to perform segmentation processing on the spinal region image to obtain a plurality of to-be-detected segmented images;

[0066] A first similarity obtaining subunit, configured to obtain a first similarity between first contour information of the to-be-detected segmented image and second contour information of the lateral vertebral body image;

[0067] A vertebral body image determining subunit, configured to determine whether the to-be-detected segmented image is the to-be-detected vertebral body image based on the first similarity.

[0068] Preferably, the segmented image obtaining subunit is further configured to perform segmentation processing on the spinal region image by using a preset image segmentation model to obtain a plurality of the to-be-detected segmented images;

[0069] And / or

[0070] The vertebral body image determining subunit is further configured to determine that the to-be-detected segmented image is the to-be-detected vertebral body image in response to the first similarity being less than a first preset similarity; and determine that the to-be-detected segmented image is not the to-be-detected vertebral body image in response to the first similarity being not less than the first preset similarity.

[0071] Preferably, the quality evaluation module includes:

[0072] A first response unit, configured to determine that the quality of the to-be-detected spinal image is qualified in response to the target quantity being greater than a preset quantity;

[0073] A second response unit, configured to determine that the quality of the to-be-detected spinal image is unqualified in response to the target quantity being not greater than the preset quantity.

[0074] Preferably, the quality evaluation system further includes:

[0075] A foreign object determining module, configured to input the to-be-detected spinal image into a second preset target detection model to determine whether there is a foreign object in the to-be-detected spinal image;

[0076] A first response module, configured to, in response to there being no such foreign object in the to-be-detected spinal image, call the vertebral body image obtaining module to perform the step of segmenting the to-be-detected spinal image to obtain a plurality of to-be-detected vertebral body images;

[0077] A second response module, configured to, in response to there being such a foreign object in the to-be-detected spinal image, obtain a foreign object region image based on the second preset target detection model; perform segmentation processing on the foreign object region image to obtain a plurality of to-be-detected foreign object images; obtain a second similarity between third contour information of the to-be-detected foreign object images and fourth contour information of a lateral foreign object image; and in response to the second similarity being not less than a second preset similarity, call the vertebral body image obtaining module to perform the step of segmenting the to-be-detected spinal image to obtain a plurality of to-be-detected vertebral body images.

[0078] Preferably, the quality evaluation system further includes:

[0079] A second content acquisition module, configured to input the image of the foreign object to be detected into a multi-modal generative large model in response to the second similarity being less than the second preset similarity, so as to obtain second text description content;

[0080] A reference foreign object determination module, configured to input the second text description content into a text generative large model to determine whether the image of the foreign object to be detected is a reference foreign object image;

[0081] A third response module, configured to call the vertebral body image acquisition module to execute the step of segmenting the spine image to be detected to obtain a plurality of vertebral body images to be detected in response to the image of the foreign object to be detected not being the reference foreign object image;

[0082] A fourth response module, configured to evaluate the quality of the spine image to be detected by using the multi-modal generative large model in response to the image of the foreign object to be detected being the reference foreign object image.

[0083] Preferably, the fourth response module includes:

[0084] A lateral foreign object determination unit, configured to input the image of the foreign object to be detected into the multi-modal generative large model to determine whether the image of the foreign object to be detected is a lateral foreign object image;

[0085] An image qualification determination unit, configured to determine that the quality of the spine image to be detected is qualified in response to the image of the foreign object to be detected being the lateral foreign object image.

[0086] The present disclosure further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and configured to run on the processor, where the processor implements the above-mentioned method for evaluating the quality of a spine image when executing the computer program.

[0087] The present disclosure further provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the above-mentioned method for evaluating the quality of a spine image when being executed by a processor.

[0088] The present disclosure further provides a computer program product, including a computer program, and the computer program implements the method for evaluating the quality of a spine image as described above when being executed by a processor.

[0089] On the basis of conforming to the common knowledge in the art, the above preferred conditions can be combined arbitrarily to obtain various preferred examples of the present disclosure.

[0090] The positive and progressive effects of the present disclosure are:

[0091] The present disclosure performs segmentation processing on the spine image to be detected to obtain the vertebral body image to be detected, and determines whether the vertebral body image to be detected is a lateral vertebral body image through a generative large model, thereby realizing the quality evaluation of the spine image to be detected, improving the efficiency of classifying the spine image to be detected, reducing the cost, and enhancing the accuracy and reliability of the quality evaluation of the spine image. Description of the Drawings

[0092] Figure 1 It is a flowchart of the method for quality evaluation of spine images in Embodiment 1 of the present disclosure;

[0093] Figure 2 It is a flowchart of step S13 in the method for quality evaluation of spine images in Embodiment 2 of the present disclosure;

[0094] Figure 3 It is the first example diagram of using a multi-modal generative large model to perform text description on the vertebral body image to be detected in Embodiment 2 of the present disclosure;

[0095] Figure 4 It is the second example diagram of using a multi-modal generative large model to perform text description on the vertebral body image to be detected in Embodiment 2 of the present disclosure;

[0096] Figure 5 It is a flowchart of step S132 in the method for quality evaluation of spine images in Embodiment 2 of the present disclosure;

[0097] Figure 6 It is a flowchart of step S12 in the method for quality evaluation of spine images in Embodiment 2 of the present disclosure;

[0098] Figure 7 It is an example diagram of the spine region image obtained after target detection of the lateral spine X-ray film in Embodiment 2 of the present disclosure.

[0099] Figure 8 It is an example diagram of the spine region image obtained after target detection of the non-lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0100] Figure 9 It is a flowchart of step S122 in the method for quality evaluation of spine images in Embodiment 2 of the present disclosure;

[0101] Figure 10 It is the first example diagram of the segmented image to be detected after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0102] Figure 11 It is the second example diagram of the segmented image to be detected after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0103] Figure 12It is the third example diagram of the to-be-detected segmented image after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0104] Figure 13 It is the fourth example diagram of the to-be-detected segmented image after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0105] Figure 14 It is the example diagram of the second contour information of the lateral vertebral body image in Embodiment 2 of the present disclosure;

[0106] Figure 15 It is the example diagram of the to-be-detected segmented image obtained after semantic segmentation of the lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0107] Figure 16 It is the example diagram of the to-be-detected segmented image obtained after semantic segmentation of the non-lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0108] Figure 17 It is the flowchart of the method for evaluating the quality of the spine image in Embodiment 2 of the present disclosure;

[0109] Figure 18 It is the example diagram of the foreign object area image obtained after target detection of the bone nail in the lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0110] Figure 19 It is the example diagram of the foreign object area image obtained after target detection of the bone nail in the non-lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0111] Figure 20 It is the example diagram of the to-be-detected foreign object image obtained after semantic segmentation of the bone nail part in the lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0112] Figure 21 It is the example diagram of the to-be-detected foreign object image obtained after semantic segmentation of the non-lateral spine X-ray film in Embodiment 2 of the present disclosure;

[0113] Figure 22 It is the first example diagram of the to-be-detected foreign object image after black-and-white transformation processing in Embodiment 2 of the present disclosure.

[0114] Figure 23 It is the second example diagram of the to-be-detected foreign object image after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0115] Figure 24 It is the third example diagram of the to-be-detected foreign object image after black-and-white transformation processing in Embodiment 2 of the present disclosure;

[0116] Figure 25An exemplary diagram for using a multi-modal generative large model to generate a text description of a lateral foreign object image to be detected in Embodiment 2 of the present disclosure;

[0117] Figure 26 An exemplary diagram for using a multi-modal generative large model to generate a text description of a non-lateral foreign object image to be detected in Embodiment 2 of the present disclosure;

[0118] Figure 27 A flowchart of step S18 in the method for evaluating the quality of a spinal image according to Embodiment 2 of the present disclosure;

[0119] Figure 28 An exemplary diagram for the method for evaluating the quality of a spinal image according to Embodiment 2 of the present disclosure;

[0120] Figure 29 A schematic diagram of the modules of the system for evaluating the quality of a spinal image according to Embodiment 3 of the present disclosure;

[0121] Figure 30 A schematic diagram of the modules of the system for evaluating the quality of a spinal image according to Embodiment 4 of the present disclosure;

[0122] Figure 31 A schematic diagram of the structure of an electronic device according to Embodiment 5 of the present disclosure. Detailed implementation manners

[0123] The present disclosure will be further described below by way of embodiments, but the present disclosure is not limited to the scope of the described embodiments.

[0124] In the embodiments of the present disclosure, prefix words such as "first" and "second" are only used to distinguish different described objects, and do not limit the position, order, priority, quantity, content, etc. of the described objects. The use of ordinal numbers and other prefix words for distinguishing described objects in the embodiments of the present disclosure does not limit the described objects. The description of the described objects refers to the description in the claims or the context of the embodiments, and should not constitute unnecessary limitations due to the use of such prefix words. In addition, in the description of this embodiment, unless otherwise specified, "a plurality of" means two or more.

[0125] In the embodiments of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information and other processes all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0126] Embodiment 1

[0127] This embodiment provides a method for evaluating the quality of a spinal image, as Figure 1 shown, the quality evaluation method includes:

[0128] S11. Obtain a spinal image to be detected;

[0129] S12. Segment the spine image to be detected to obtain a number of vertebral body images to be detected;

[0130] S13. Use a preset generative large model to determine whether the vertebral body image to be detected is a lateral vertebral body image;

[0131] S14. Obtain the target quantity corresponding to the lateral vertebral body images among the number of vertebral body images to be detected corresponding to the spine image to be detected;

[0132] S15. Evaluate the quality of the spine image to be detected based on the target quantity.

[0133] Specifically, the spine is composed of multiple vertebrae, that is, vertebral bodies. The vertebrae are divided into thoracic vertebrae and lumbar vertebrae. These vertebrae are often important parts for diagnosis and analysis in spine imaging.

[0134] The spine image to be detected can be a spine X-ray film, that is, a spine X-ray image. The X-ray image is an image formed by using X-rays to penetrate human tissues and imaging onto a photosensitive material or a digital sensor. X-ray images are often used to examine internal structures of the human body, such as bones, lungs, etc.

[0135] The generative large model refers to a generative model with a large number of parameters. Usually based on deep learning, it can learn the complex distribution of data and generate new data similar to or related to the original data. For example, the generative large model can generate new images, describe images, or generate corresponding images according to text based on the learned image features.

[0136] In this embodiment, the spine image to be detected is segmented to obtain the vertebral body image to be detected. Whether the vertebral body image to be detected is a lateral vertebral body image is determined by the generative large model, thereby realizing the quality evaluation of the spine image to be detected, improving the efficiency of classifying the spine image to be detected, reducing the cost, and enhancing the accuracy and reliability of the spine image quality evaluation.

[0137] Embodiment 2

[0138] This embodiment provides a method for evaluating the quality of a spine image, which is a further improvement of Embodiment 1.

[0139] In an implementable solution, the preset generative large model includes a multimodal generative large model and a text generative large model;

[0140] As Figure 2 shown, step S13 includes:

[0141] S131. Input the vertebral body image to be detected into the multimodal generative large model to obtain the first text description content;

[0142] S132. Input the first text description content into the text generation large model to determine whether the vertebral image to be detected is a lateral vertebral image.

[0143] Specifically, the multimodal generation large model is a generation large model that can simultaneously process and generate multiple modalities of data (such as images, text, speech, etc.). For example, the multimodal generation large model can generate corresponding text descriptions based on the input images, or generate corresponding images based on text descriptions, realizing the conversion and association between different modalities.

[0144] Utilize the language generation ability of the multimodal generation large model to describe the tile for auxiliary judgment.

[0145] Since the accuracy of general image segmentation methods in medical images is limited, errors or misidentifications may occur. To further verify and identify the content of the tile, a multimodal generation large model is introduced. These models are trained with a large amount of general graphic and text data and a certain amount of medical graphic and text data, and have certain medical knowledge, and can describe the image in text, thereby assisting in judging whether the tile is a spinal vertebra.

[0146] Use multimodal generation large models, such as GPT-4V (a multimodal model), GLM-4V (a multimodal vision large model), LlaVA (a large multimodal model), to describe the vertebral image to be detected. When using the multimodal generation large model, the guiding words (prompt words) do not preset the parts or content, and let it use its own integrated knowledge to judge.

[0147] As Figure 3 shown, it is the first example diagram of using the multimodal generation large model to perform text description on the vertebral image to be detected. Among them, the input is the vertebral image to be detected and the prompt word "Describe the details of this spinal vertebral X-ray film", and the output is as follows:

[0148] "This X-ray film shows a frontal (coronal) view of a spinal vertebra. The following are some observable details:

[0149] 1. Vertebral shape: The vertebra is rectangular, with relatively clear edges, showing a relatively normal anatomical structure. The outlines of the upper and lower edges are obvious, and the two sides and the central part of the vertebra can be seen.

[0150] 2. Vertebral bone density: In the image, the bone density of the vertebra seems uniform, without obvious manifestations of osteoporosis. The X-ray transparency of the entire vertebral region is uniform, without obvious density abnormal regions.

[0151] 3. Endplates of the vertebral body: The outlines of the upper and lower endplates can be seen, showing the structural integrity of the vertebral body. There is no obvious compression deformation or collapse in the endplate area.

[0152] 4. Intervertebral disc space: Since this is an X-ray of a single vertebra, detailed information about the intervertebral disc is not shown. However, it can be inferred that the intervertebral disc space is located in the upper and lower areas of the film.

[0153] 5. Trabecular structure: The trabecular structure inside the vertebral body can be seen with a certain texture, which is a typical feature of healthy bones and indicates that the bone structure is relatively intact.

[0154] 6. Pathological features: In this image, there are no obvious signs of fractures, tumors, or other pathological changes. "

[0155] like Figure 4 As shown in the figure, it is a second example of using a multimodal generative large model to describe the vertebral image to be detected. The input is the vertebral image to be detected and the prompt word "Describe the details of this spinal vertebral X-ray", and the output is as follows:

[0156] "This X-ray shows a side (sagittal) view of a spinal vertebra. The following details can be observed:

[0157] 1. Vertebral shape: The vertebral body is approximately rectangular in shape. From the side, the upper and lower edges and the front and back contours of the vertebral body are clearly visible. The height and width of the vertebral body appear to be uniform, with no obvious signs of compression or collapse.

[0158] 2. The front and rear edges of the vertebral body: The front and rear edges are clearly outlined, indicating that the vertebral body structure is intact. There is no sign of osteophytes (bone hyperplasia) on the front edge, and the rear edge does not show any abnormal protrusion or collapse.

[0159] 3. Intervertebral disc space: Since it is a side view, although the edges of the upper and lower vertebrae can be observed, this X-ray does not show the complete intervertebral disc space. If the intervertebral disc space is displayed, its thickness and structural state may be seen.

[0160] 4. Bone density: The bone density of the vertebral body appears uniform, showing no obvious signs of osteoporosis. The trabecular structure inside the vertebral body is clearly visible from the side, showing the typical characteristics of healthy bones.

[0161] 5. Pathological features: There are no obvious signs of vertebral lesions, such as fractures, tumors, or obvious deformation of the vertebral body. The anterior and posterior edges of the vertebral body are smooth, and no osteophytes or other abnormal bone are found.

[0162] In this solution, by introducing a multi-modal generative large model to generate a text description of the vertebral image to be detected, it is convenient to assist in judging whether the vertebral image to be detected is a lateral vertebral image, improving the accuracy and reliability of the judgment of the vertebral image to be detected.

[0163] In an implementable solution, as Figure 5 shown, step S132 includes:

[0164] S1321. Input the first text description content into the text generative large model to determine whether the vertebral image to be detected is a reference vertebral image;

[0165] S1322. In response to the vertebral image to be detected being a reference vertebral image, use the feature information of the reference vertebral image as a prompt word for the text generative large model to enable the text generative large model to determine whether the vertebral image to be detected is a lateral vertebral image.

[0166] When using the generative large model, in order to obtain the desired output result, some prompt words or guiding words are usually provided to the model to guide the generation process of the model. These prompt words can contain information such as the description of the task, the expected content or style, etc., to help the model more accurately understand the task requirements.

[0167] Through the understanding and reasoning ability of the text generative large model, further verify whether the tile is a vertebra. The text generative large model has powerful language understanding and reasoning abilities, and can perform logical analysis and judgment on the content described in the text. By inputting the first text description content and using a general text generative large model, judge whether the first text description content is related to the spinal vertebra. If the model judges it to be yes, keep the tile; otherwise, exclude it. In this way, non-spinal vertebra tiles can be further filtered, improving the overall recognition accuracy.

[0168] After determining that the vertebral image to be detected is a reference vertebral image, by introducing medical background knowledge, it is possible to more accurately judge whether the vertebral tile is a lateral film or a non-lateral film. First, sort out the different feature information of the spinal vertebra in the lateral film and the frontal film, such as the vertebral body width, symmetry, vertebral body height, the morphology of the anterior and posterior edges, pedicles, transverse processes and spinous processes, spinal canal contour, intervertebral disc space, intervertebral foramen, physiological curvature of the spine, lamina and articular processes, atlas and axis, etc. Then, use this feature information as a prompt word to guide the text generative large model to judge the first text description content. The model will infer whether the tile is a lateral film based on the description of the tile and the prompted features. This method combines the language understanding ability of the model and medical knowledge, improving the accuracy of the judgment. If the vertebral image to be detected is not a reference vertebral image, then the vertebral image to be detected is not a lateral vertebral image, and the quality evaluation of the vertebral image to be detected is unqualified.

[0169] In the process of using a pre - set generative large - model to determine whether the vertebral image to be detected is a lateral vertebral image, a step - by - step guiding method is adopted for the generative large - model. When dealing with complex tasks, if too much information is given to the generative large - model at one time, hallucination phenomena may occur, that is, inaccurate or irrelevant content is generated. By decomposing the task into multiple steps, with each step focusing on a specific subtask, the model can more accurately understand and process information at each stage. In this way, the judgment process of the model is more transparent, and it is also convenient to introduce artificial medical knowledge for guidance, improving the overall judgment accuracy.

[0170] In this solution, the feature information of the reference vertebral image is used as the prompt word of the text - generative large - model, combining the model's language understanding ability and medical knowledge, and improving the accuracy of the judgment of the vertebral image to be detected.

[0171] In an implementable solution, as Figure 6 shown, step S12 includes:

[0172] S121. Input the vertebral image to be detected into the first pre - set target detection model to obtain the vertebral region image;

[0173] S122. Perform segmentation processing on the vertebral region image to obtain a number of vertebral images to be detected.

[0174] Specifically, in image processing, the first pre - set target detection model obtains the vertebral region image by performing target detection on the vertebral image to be detected. Target detection refers to identifying and locating specific objects in an image. For example, in a vertebral X - ray film, the range of the vertebrae is detected, and this range can be a rectangular area.

[0175] The first pre - set target detection model is a general pre - trained target detection model, such as YOLO (a target detection algorithm), DINO (an unsupervised representation learning method), Grounding - DINO (an open - set detection model), etc. It performs target detection on the vertebral X - ray film, locates the position of the vertebrae in the X - ray film, and finds the approximate position of the vertebrae in the image to narrow the scope of attention and exclude other irrelevant image information.

[0176] As Figure 7 shown, it is an example diagram of the vertebral region image obtained after target detection of a lateral vertebral X - ray film.

[0177] As Figure 8 shown, it is an example diagram of the vertebral region image obtained after target detection of a non - lateral vertebral X - ray film.

[0178] In this solution, the target detection model is used to perform target detection on the spine image to be detected, obtaining the spine region image, which reduces the scope of attention of the image, and then segmentation processing is performed to ensure the accuracy and effectiveness of the segmentation process.

[0179] In an implementable solution, as Figure 9 shown, step S122 includes:

[0180] S1221. Perform segmentation processing on the spine region image to obtain a number of segmentation images to be detected;

[0181] S1222. Obtain the first similarity between the first contour information of the segmentation image to be detected and the second contour information of the lateral vertebral body image;

[0182] S1223. Based on the first similarity, determine whether the segmentation image to be detected is the vertebral body image to be detected.

[0183] Specifically, through shape matching, the true vertebral body region is screened out to exclude interference. Each tile corresponding to the segmentation image to be detected is extracted. Its contour is found through image processing and black-and-white transformation. As Figure 10 shown, it is the first example image of the segmentation image to be detected after black-and-white transformation processing. As Figure 11 shown, it is the second example image of the segmentation image to be detected after black-and-white transformation processing. As Figure 12 shown, it is the third example image of the segmentation image to be detected after black-and-white transformation processing. As Figure 13 shown, it is the fourth example image of the segmentation image to be detected after black-and-white transformation processing. As Figure 14 shown, it is the example image of the second contour information of the lateral vertebral body image. Obtain the prepared standard lateral vertebral body image, and calculate the first similarity between the segmentation image to be detected and the lateral vertebral body image, such as geometric values such as Hausdorff distance (a way to measure the distance between two sets) or Hu moment (a feature descriptor based on image moments). Determine whether the segmentation image to be detected is the vertebral body image to be detected according to the first similarity.

[0184] In this solution, determining whether the segmentation image to be detected is the vertebral body image to be detected according to the first similarity, and screening the segmentation image to be detected to obtain the vertebral body image to be detected, improves the accuracy and reliability of the vertebral body image to be detected.

[0185] In an implementable solution, step S1221 includes:

[0186] Use a preset image segmentation model to perform segmentation processing on the spine region image to obtain a number of segmentation images to be detected.

[0187] Specifically, crop out the spine part detected in the previous step of object detection, and use a general pre-trained image segmentation model, namely the preset image segmentation model, such as SAM (Spatial Attention Map), Medical-SAM (Medical Spatial Attention Map), or PaddleSeg (an end-to-end image segmentation toolkit), to segment the patches in different regions of the image. The preset image segmentation model performs segmentation processing on the spine region image through semantic segmentation. Semantic segmentation is a computer vision technique aimed at classifying each pixel in an image so that pixels of the same category are connected into regions spatially. For example, in a spine X-ray, semantic segmentation can segment out each vertebral body region.

[0188] This step extracts the part and shape of the vertebral body from the image, segments the vertebral body region, and obtains more detailed information, but it may also introduce noise and segment out some patches that are not vertebral bodies or are meaningless.

[0189] As Figure 15 shown, it is an example diagram of the segmentation image to be detected obtained after semantic segmentation of a lateral spine X-ray.

[0190] As Figure 16 shown, it is an example diagram of the segmentation image to be detected obtained after semantic segmentation of a non-lateral spine X-ray.

[0191] Both the first preset object detection model and the preset image segmentation model are general pre-trained models. A general pre-trained model refers to a model that has been pre-trained with a large amount of data. These models usually have wide applicability and can be used in multiple tasks. The pre-trained model is trained on a certain task or domain and learns general features and patterns. The core idea of these models is to utilize the training results of a large-scale dataset to reduce the training requirements for specific tasks, thereby improving efficiency and performance.

[0192] In this solution, the spine region image is segmented through the image segmentation model to obtain several segmentation images to be detected, improving the accuracy and reliability of the segmentation images to be detected.

[0193] In an implementable solution, step S1223 includes:

[0194] In response to the first similarity being less than the first preset similarity, determine that the segmentation image to be detected is the vertebral image to be detected;

[0195] In response to the first similarity being not less than the first preset similarity, determine that the segmentation image to be detected is not the vertebral image to be detected.

[0196] If the first similarity is less than a certain threshold, i.e., the first preset similarity, it can be considered that the contours of the two images are similar. The specific threshold can be determined through experiments in practice. For example, the threshold of the Hausdorff distance is generally between 20 and 50.

[0197] In this solution, according to the magnitude relationship between the first similarity and the first preset similarity, it is determined whether the image of the vertebra to be detected is the image of the vertebra to be detected, which improves the accuracy and reliability of the image of the vertebra to be detected.

[0198] In an implementable solution, step S15 includes:

[0199] In response to the target quantity being greater than the preset quantity, it is determined that the quality of the spinal image to be detected is qualified;

[0200] In response to the target quantity not being greater than the preset quantity, it is determined that the quality of the spinal image to be detected is unqualified.

[0201] Specifically, for each possible vertebral part obtained by segmenting the X-ray film, that is, for each image of the vertebra to be detected, it is judged whether it is a lateral vertebral image, and then the number of lateral vertebrae (or lateral vertebral bodies) in each X-ray film is counted. If the number of vertebral blocks judged to be lateral in an X-ray film exceeds a certain number, i.e., the preset quantity, it is considered a lateral film and the quality of the X-ray film is qualified; otherwise, it is not a lateral film and the quality of the X-ray film is unqualified. Generally, an ordinary spinal X-ray film includes 5 to 10 vertebral bodies. In practice, the number of detected lateral vertebral bodies is usually between 2 and 6. The specific threshold can be determined through experiments on the X-ray film dataset to be classified.

[0202] In this solution, by the magnitude relationship between the target quantity corresponding to the lateral vertebral image and the preset quantity, it is determined whether the quality of the spinal image to be detected is qualified, which improves the accuracy and reliability of the quality evaluation of the spinal image to be detected.

[0203] In an implementable solution, as Figure 17 shown, before step S12, it further includes:

[0204] S123. Input the spinal image to be detected into the second preset target detection model to determine whether there is a foreign object in the spinal image to be detected;

[0205] In response to there being no foreign object in the spinal image to be detected, execute step S12;

[0206] In response to there being a foreign object in the spinal image to be detected, execute step S124;

[0207] S124. Based on the second preset object detection model, obtain the foreign object area image; perform segmentation processing on the foreign object area image to obtain several foreign objects to be detected images; obtain the second similarity between the third contour information of the foreign objects to be detected images and the fourth contour information of the lateral foreign object images; in response to the second similarity being not less than the second preset similarity, execute step S12.

[0208] Specifically, a foreign object, that is, an implant, such as a bone nail, due to its geometric shape showing differently in lateral and non-lateral X-ray films, can be used to assist in judging the shooting direction of the X-ray film. In addition to bone nails, there can be other types of implants, as long as their geometric shapes have obvious differences in the anteroposterior and lateral X-ray films and can assist in the judgment.

[0209] Use a pre-trained general object detection model, that is, the second preset object detection model, to detect the position of the implant in the X-ray film. Since the implant has relatively clear geometric features, the general object detection model can also locate their positions. This step can provide important information for judging the direction of the X-ray film through the position and shape of the implant.

[0210] As Figure 18 shown, it is an example diagram of the foreign object area image obtained after target detection of the bone nail in the lateral spine X-ray film.

[0211] As Figure 19 shown, it is an example diagram of the foreign object area image obtained after target detection of the bone nail in the non-lateral spine X-ray film.

[0212] Segment the detected implant area, that is, the foreign object area image, which helps with subsequent shape matching. The method of performing segmentation processing on the foreign object area image to obtain several foreign objects to be detected images is similar to the method of performing segmentation processing on the spine area image to obtain several segmented images to be detected mentioned above, both using semantic segmentation, which will not be elaborated here.

[0213] As Figure 20 shown, it is an example diagram of the foreign object to be detected image obtained after semantic segmentation of the bone nail part in the lateral spine X-ray film.

[0214] As Figure 21 shown, it is an example diagram of the foreign object to be detected image obtained after semantic segmentation of the non-lateral spine X-ray film.

[0215] For the shape of the implant obtained by segmentation, that is, the image of the foreign object to be detected, contour extraction is performed and matched with the standard implant shape template, that is, the lateral foreign object image. By calculating the second similarity, the specific type and posture of the implant are judged. Since the shape of the implant in the lateral and anteroposterior X-ray films will be significantly different, shape matching can assist in judging the shooting direction of the X-ray film. The calculation and judgment process of the second similarity is similar to that of the first similarity, which will not be elaborated here.

[0216] As Figure 22 shown, it is the first example diagram of the foreign object image to be detected after black and white transformation processing.

[0217] As Figure 23 shown, it is the second example diagram of the foreign object image to be detected after black and white transformation processing.

[0218] As Figure 24 shown, it is the third example diagram of the foreign object image to be detected after black and white transformation processing.

[0219] In this solution, first judge whether there is a foreign object in the spine image to be detected, judge the second similarity between the foreign object image to be detected and the lateral foreign object image, and then determine whether to perform segmentation processing on the spine image to be detected to obtain a number of vertebral images to be detected, that is, first judge the foreign object situation in the spine image to be detected, and then perform segmentation processing on the image to obtain the vertebral images to be detected. The shape of the foreign object is more obvious in the lateral and non-lateral films. Judging by the foreign object first improves the efficiency of the spine image to be detected.

[0220] In an implementable solution, as Figure 17 shown, this quality evaluation method further includes:

[0221] S16. In response to the second similarity being less than the second preset similarity, input the foreign object image to be detected into a multi-modal generative large model to obtain a second text description content;

[0222] S17. Input the second text description content into a text generative large model to determine whether the foreign object image to be detected is a reference foreign object image;

[0223] In response to the foreign object image to be detected not being a reference foreign object image, execute step S12;

[0224] In response to the foreign object image to be detected being a reference foreign object image, execute step S18;

[0225] S18. Use a multi-modal generative large model to evaluate the quality of the spine image to be detected.

[0226] Specifically, a multi-modal generative large model is used to describe the extracted foreign object images to be detected. Since these models possess certain general medical knowledge, they can generate textual descriptions of the type and location of the foreign objects based on their shapes and characteristics, providing a basis for further judgment.

[0227] When using the multi-modal generative large model, a target detection box obtained through the second preset object detection model can be added to the original X-ray film, i.e., the spine image to be detected, to assist the large model in describing the target. At the same time, the prompting words of the large model do not use preset parts and content, allowing the large model to use the knowledge integrated within itself to make judgments.

[0228] Such as Figure 25 shown, it is an example diagram of using a multi-modal generative large model to generate a textual description of the lateral foreign object image to be detected. Among them, the input is the foreign object image to be detected and the prompting word "Describe in detail the part framed in this X-ray film, without other parts", and the output is as follows:

[0229] "In this X-ray film, the framed part is concentrated in the lower part of the spine, especially the surgical implant in the lumbar region. The following is a detailed description of the framed area:

[0230] 1. Framed steel screws:

[0231] · This area is labeled as "steel screw 0.65", indicating that this is a steel screw with a confidence level of 0.65.

[0232] · Inside the frame, two parallel pedicle screws can be seen. These screws are typical fixation instruments used in spinal surgery to stabilize the vertebral body.

[0233] The threaded part of the screw is very clear. It can be seen that the screw is implanted along the anterior-posterior direction of the vertebral body, and this method can provide a stable fixation effect. The screw presents a relatively long shape, and the thread design is used to tightly embed into the bone to ensure the firmness of the fixation.

[0234] 2. Connecting components: ……".

[0235] Such as Figure 26 shown, it is an example diagram of using a multi-modal generative large model to generate a textual description of the non-lateral foreign object image to be detected. Among them, the input is the foreign object image to be detected and the prompting word "Describe in detail the part framed in this X-ray film, without other parts", and the output is as follows:

[0236] "In this X-ray film, the two frames respectively mark the positions of the steel screws implanted during spinal surgery. The following is a detailed description of each framed part:

[0237] 1. The red box on the left (labeled "steel screw 0.60"):

[0238] · This box shows an implanted steel screw, with the screw part embedded in the spinal bone.

[0239] · The steel screw presents a helical thread structure, indicating that it is a pedicle screw used to fix the vertebral body.

[0240] · The screw head is relatively wide and clearly visible, showing the implantation angle and position. The screw head may be attached with a connecting rod for connection with other fixation devices.

[0241] · This steel screw is located on the left side of the spine and forms a certain angle with the central axis of the spine, usually used to stabilize the vertebral body and prevent abnormal movement of the spine.

[0242] 2. The green box on the right (labeled "steel screw 0.56"):

[0243] · This box also shows an implanted steel screw, with a shape and position similar to the left steel screw.

[0244] · The contour of the screw is clear, showing the thread structure and part of the screw head. Its shape and angle are opposite to those of the left steel screw...

[0245] In this solution, the second text description content is input into the text - generative large - model to determine whether the image of the foreign object to be detected is a reference foreign - object image. For example, to determine whether it is describing a bone screw rather than other things or abnormal blocks, improving the overall recognition accuracy.

[0246] In an implementable solution, as Figure 27 shown, step S18 includes:

[0247] S181. Input the image of the foreign object to be detected into the multi - modal generative large - model to determine whether the image of the foreign object to be detected is a lateral foreign - object image;

[0248] S182. In response to the image of the foreign object to be detected being a lateral foreign - object image, determine that the quality of the image of the spine to be detected is qualified;

[0249] S183. In response to the image of the foreign object to be detected not being a lateral foreign - object image, determine that the quality of the image of the spine to be detected is unqualified.

[0250] Specifically, due to the wide variety of implants, existing generative large models do not have sufficient medical device knowledge to describe their details. Therefore, by combining the object detection frames obtained through the second preset object detection model, the multi-modal generative large model is prompted to focus on specific parts, and combined with the image information of the entire X-ray film, it is comprehensively judged whether it is a lateral film. This step can improve the accuracy of judgment and make up for the lack of model knowledge.

[0251] In this solution, the multi-modal generative large model is used to determine whether the image of the foreign object to be detected is a lateral foreign object image, so as to determine whether the quality of the spine image to be detected is qualified, improving the accuracy and reliability of the quality evaluation of the spine image to be detected.

[0252] The working principle of the quality evaluation method for spine images in this embodiment is described below with specific examples:

[0253] Such as Figure 28As shown, start the quality evaluation of the spine image to be detected. First, perform the detection of the implant location. If the implant is not found, execute the steps of spine region target detection. If the implant is found, segment the implant image to generate the implant contour and obtain the foreign object image to be detected. Match the shape of the foreign object image to be detected with the lateral foreign object image to calculate the second similarity. If the second similarity does not meet the threshold condition, execute the steps of spine region target detection. If the second similarity meets the threshold condition, use the multi-modal generative large model to describe the foreign object image to be detected and obtain the second text description content. Use the text generative large model to determine whether the foreign object image to be detected is a reference foreign object image, such as whether it is a bone screw. If the foreign object image to be detected is not a reference foreign object image, execute the steps of spine region target detection. If the foreign object image to be detected is a reference foreign object image, comprehensively judge the anteroposterior and lateral positions of the foreign object image to be detected through the multi-modal generative large model. If the foreign object image to be detected is a lateral film, output that it is a lateral film and the quality of the spine image to be detected is qualified. After executing the steps of spine region target detection, if the spine region is not found, output that it is not a lateral film and the quality of the spine image to be detected is unqualified. If the spine region is found, obtain the spine region image. Perform semantic segmentation on the spine region image to obtain the vertebral body image to be detected. Match the shape of the vertebral body image to be detected with the lateral vertebral body image to calculate the first similarity. If the first similarity meets the threshold condition, use the multi-modal generative model to describe the vertebral body image to be detected and obtain the first text description content. Use the text generative large model to determine whether the vertebral body image to be detected is a reference vertebral body image. If the vertebral body image to be detected is a reference vertebral body image, use the text generative large model to determine whether the vertebral body image to be detected is a lateral vertebral body image. Count the number of targets corresponding to the lateral vertebral body images in the spine image to be detected and analyze whether the number of targets meets the threshold condition. If the number of targets meets the threshold condition, output that it is a lateral film and the quality of the spine image to be detected is qualified. If the number of targets does not meet the threshold condition, output that it is not a lateral film and the quality of the spine image to be detected is unqualified.

[0254] In this embodiment, the spine image to be detected is segmented to obtain the vertebral body image to be detected, and the generative large model is used to determine whether the vertebral body image to be detected is a lateral vertebral body image, thereby realizing the quality evaluation of the spine image to be detected, improving the efficiency of classifying the spine image to be detected, reducing the cost, enhancing the accuracy and reliability of the spine image quality evaluation, reducing the impact of abnormal X-ray films generated in spine-related medical practices on doctor diagnosis and downstream information processing, having medical interpretability, conforming to the cognition of medical practitioners, being automated, and not requiring cumbersome and costly medical sample annotation.

[0255] Example 3

[0256] This embodiment provides a quality evaluation system for spinal images, as follows Figure 29 shown, the quality evaluation system includes:

[0257] A spinal image acquisition module 11 for acquiring spinal images to be detected;

[0258] A vertebral body image acquisition module 12 for segmenting the spinal images to be detected to obtain a plurality of vertebral body images to be detected;

[0259] A lateral vertebral body determination module 13 for determining whether the vertebral body image to be detected is a lateral vertebral body image by using a preset generative large model;

[0260] A target quantity acquisition module 14 for acquiring the target quantity corresponding to the lateral vertebral body images among the plurality of vertebral body images to be detected corresponding to the spinal images to be detected;

[0261] A quality evaluation module 15 for evaluating the quality of the spinal images to be detected based on the target quantity.

[0262] In this embodiment, the spinal images to be detected are segmented to obtain vertebral body images to be detected, and whether the vertebral body images to be detected are lateral vertebral body images is determined by a generative large model, thereby realizing the quality evaluation of the spinal images to be detected, improving the efficiency of classifying the spinal images to be detected, reducing the cost, and enhancing the accuracy and reliability of the spinal image quality evaluation.

[0263] Embodiment 4

[0264] This embodiment provides a quality evaluation system for spinal images, which is a further improvement of Embodiment 3.

[0265] In an implementable solution, the preset generative large model includes a multimodal generative large model and a text generative large model;

[0266] As Figure 30 shown, the lateral vertebral body determination module 13 includes:

[0267] A first content acquisition unit 131 for inputting the vertebral body image to be detected into the multimodal generative large model to obtain a first text description content;

[0268] A lateral vertebral body determination unit 132 for inputting the first text description content into the text generative large model to determine whether the vertebral body image to be detected is a lateral vertebral body image.

[0269] In an implementable solution, the lateral vertebral body determination unit 132 includes:

[0270] A reference vertebral body determination subunit 1321 is configured to input the first text description content into a text generation large model to determine whether the vertebral body image to be detected is a reference vertebral body image;

[0271] A lateral vertebral body determination subunit 1322 is configured to, in response to the vertebral body image to be detected being a reference vertebral body image, use the feature information of the reference vertebral body image as a prompt word for the text generation large model, so that the text generation large model determines whether the vertebral body image to be detected is a lateral vertebral body image.

[0272] In an implementable solution, the vertebral body image acquisition module 12 includes:

[0273] A region image acquisition unit 121 is configured to input the vertebral column image to be detected into a first preset object detection model to obtain a vertebral column region image;

[0274] A vertebral body image acquisition unit 122 is configured to perform segmentation processing on the vertebral column region image to obtain a plurality of vertebral body images to be detected.

[0275] In an implementable solution, the vertebral body image acquisition unit 122 includes:

[0276] A segmented image acquisition subunit 1221 is configured to perform segmentation processing on the vertebral column region image to obtain a plurality of segmented images to be detected;

[0277] A first similarity acquisition subunit 1222 is configured to acquire a first similarity between the first contour information of the segmented image to be detected and the second contour information of the lateral vertebral body image;

[0278] A vertebral body image determination subunit 1223 is configured to determine whether the segmented image to be detected is a vertebral body image to be detected based on the first similarity.

[0279] In an implementable solution, the segmented image acquisition subunit 1221 is further configured to perform segmentation processing on the vertebral column region image by using a preset image segmentation model to obtain a plurality of segmented images to be detected.

[0280] In an implementable solution, the vertebral body image determination subunit 1223 is further configured to, in response to the first similarity being less than a first preset similarity, determine that the segmented image to be detected is a vertebral body image to be detected; in response to the first similarity being not less than the first preset similarity, determine that the segmented image to be detected is not a vertebral body image to be detected.

[0281] In an implementable solution, the quality evaluation module 15 includes:

[0282] A first response unit 151 is configured to, in response to the target quantity being greater than a preset quantity, determine that the quality of the vertebral column image to be detected is qualified;

[0283] A second response unit 152, configured to determine that the quality of the spine image to be detected is unqualified in response to the target quantity being not greater than the preset quantity.

[0284] In an implementable solution, the quality evaluation system further includes:

[0285] A foreign object determination module 16, configured to input the spine image to be detected into a second preset target detection model to determine whether there is a foreign object in the spine image to be detected;

[0286] A first response module 17, configured to, in response to the absence of a foreign object in the spine image to be detected, call the vertebral body image acquisition module to perform a segmentation process on the spine image to be detected to obtain a plurality of vertebral body images to be detected;

[0287] A second response module 18, configured to, in response to the presence of the foreign object in the spine image to be detected, obtain a foreign object region image based on the second preset target detection model; perform a segmentation process on the foreign object region image to obtain a plurality of foreign object images to be detected; obtain a second similarity between the third contour information of the foreign object image to be detected and the fourth contour information of the lateral foreign object image; and in response to the second similarity being not less than the second preset similarity, call the vertebral body image acquisition module to perform a segmentation process on the spine image to be detected to obtain a plurality of vertebral body images to be detected.

[0288] In an implementable solution, the quality evaluation system further includes:

[0289] A second content acquisition module 19, configured to, in response to the second similarity being less than the second preset similarity, input the foreign object image to be detected into a multi-modal generative large model to obtain second text description content;

[0290] A reference foreign object determination module 20, configured to input the second text description content into a text generative large model to determine whether the foreign object image to be detected is a reference foreign object image;

[0291] A third response module 21, configured to, in response to the foreign object image to be detected not being a reference foreign object image, call the vertebral body image acquisition module to perform a segmentation process on the spine image to be detected to obtain a plurality of vertebral body images to be detected;

[0292] A fourth response module 22, configured to, in response to the foreign object image to be detected being a reference foreign object image, evaluate the quality of the spine image to be detected using a multi-modal generative large model.

[0293] In an implementable solution, the fourth response module 22 includes:

[0294] A lateral foreign object determination unit 221, configured to input the foreign object image to be detected into a multi-modal generative large model to determine whether the foreign object image to be detected is a lateral foreign object image;

[0295] An image qualification determination unit 222 is configured to determine that the quality of the spine image to be detected is qualified in response to the foreign object image to be detected being a lateral foreign object image.

[0296] In this embodiment, the spine image to be detected is segmented to obtain the vertebral body image to be detected, and a generative large model is used to determine whether the vertebral body image to be detected is a lateral vertebral body image, thereby realizing the quality evaluation of the spine image to be detected, improving the efficiency of classifying the spine image to be detected, reducing the cost, and enhancing the accuracy and reliability of the spine image quality evaluation.

[0297] For the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present disclosure.

[0298] Embodiment 5

[0299] Figure 31 As shown in the structural schematic diagram of an electronic device according to an exemplary embodiment of the present disclosure, the electronic device includes a memory, a processor, and a computer program stored on the memory and configured to run on the processor. When the processor executes the computer program, it implements the method for evaluating the quality of the spine image described in any of the above embodiments. Figure 31 The electronic device 90 shown is merely an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0300] As Figure 31 shown, the electronic device 90 may be presented in the form of a general computing device, for example, it may be a server device. The components of the electronic device 90 may include, but are not limited to: at least one of the above processors 91, at least one of the above memories 92, and a bus 93 connecting different system components (including the memory 92 and the processor 91).

[0301] The bus 93 includes a data bus, an address bus, and a control bus.

[0302] The memory 92 may include volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922, and may further include a read-only memory (ROM) 923.

[0303] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) of program modules 924. Such program modules 924 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.

[0304] The processor 91 executes various functional applications and data processing by running computer programs stored in the memory 92, such as the method for evaluating the quality of spinal images provided in any of the above embodiments.

[0305] The electronic device 90 may also communicate with one or more external devices 94 (such as a keyboard, a pointing device, etc.). Such communication may be carried out through the input / output (I / O) interface 95. And, the electronic device 90 may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 96. As shown in the figure, the network adapter 96 communicates with other modules of the electronic device 90 through the bus 93. It should be understood that although not shown in the figure, other hardware and / or software modules may be used in combination with the electronic device 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (redundant array of independent disks) systems, tape drives, and data backup storage systems, etc.

[0306] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, such a division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above may be embodied in one unit / modules. Conversely, the features and functions of one unit / modules described above may be further divided and embodied by multiple units / modules.

[0307] Embodiment 6

[0308] The embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the method for evaluating the quality of spinal images provided in any of the above embodiments is implemented.

[0309] Among them, the more specific readable storage medium that can be adopted may include, but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0310] Embodiment 7

[0311] Embodiments of the present disclosure also provide a computer program product, including a computer program, which when executed by a processor implements the method for evaluating the quality of spinal images described in any one of the above.

[0312] Among them, the program code for executing the computer program product of the present disclosure can be written in any combination of one or more programming languages, and the program code can be executed entirely on the user device, partially on the user device, executed as an independent software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0313] Although the specific embodiments of the present disclosure have been described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Without departing from the principles and essence of the present disclosure, those skilled in the art can make various changes or modifications to these embodiments, but these changes and modifications all fall within the protection scope of the present disclosure.

Claims

1. A method for evaluating the quality of a spine image, characterized in that: The quality evaluation method comprises: Acquire a spine image to be detected; Segmenting the to-be-detected vertebral image to obtain a plurality of to-be-detected vertebral images; Using a preset generative large model to determine whether the vertebral image to be detected is a lateral vertebral image; Acquire the number of targets corresponding to the lateral vertebral images among the plurality of vertebral images to be detected corresponding to the vertebral image to be detected; Based on the target quantity, the quality of the spine image to be detected is evaluated.

2. The method for evaluating the quality of a spine image according to claim 1, wherein: The preset generative big model includes a multimodal generative big model and a text generative big model; The step of using a preset generative large model to determine whether the vertebral image to be detected is a lateral vertebral image comprises: Inputting the vertebral body image to be detected into the multimodal generative large model to obtain a first text description content; The first text description content is input into the text generation model to determine whether the vertebral image to be detected is the lateral vertebral image.

3. The method for evaluating the quality of a spine image according to claim 2, wherein: The step of inputting the first text description content into the text generation model to determine whether the vertebral image to be detected is the lateral vertebral image comprises: Inputting the first text description content into the text generation model to determine whether the vertebral image to be detected is a reference vertebral image; In response to the vertebral image to be detected being the reference vertebral image, feature information of the reference vertebral image is used as a prompt word of the text generation model, so that the text generation model determines whether the vertebral image to be detected is the lateral vertebral image.

4. The method for evaluating the quality of a spine image according to any one of claims 1 to 3, characterized in that: The step of evaluating the quality of the spine image to be detected based on the target quantity includes: In response to the target number being greater than a preset number, determining that the quality of the to-be-detected spine image is qualified; In response to the target number being not greater than the preset number, it is determined that the quality of the to-be-detected spine image is unqualified.

5. The method for evaluating the quality of a spine image according to any one of claims 1 to 3, characterized in that: Before the step of segmenting the to-be-detected vertebral image to obtain a plurality of to-be-detected vertebral images, the method further comprises: Inputting the spinal image to be detected into a second preset target detection model to determine whether there is a foreign body in the spinal image to be detected; In response to the foreign body not existing in the spinal image to be detected, performing the step of segmenting the spinal image to be detected to obtain a plurality of vertebral images to be detected; In response to the presence of the foreign body in the spinal image to be detected, a foreign body area image is obtained based on the second preset target detection model; the foreign body area image is segmented to obtain a number of foreign body images to be detected; a second similarity between the third contour information of the foreign body image to be detected and the fourth contour information of the lateral foreign body image is obtained; in response to the second similarity being not less than the second preset similarity, the step of segmenting the spinal image to be detected to obtain a number of vertebral images to be detected is performed.

6. The method for evaluating the quality of a spine image according to claim 5, wherein: The quality evaluation method further comprises: In response to the second similarity being less than the second preset similarity, inputting the foreign body image to be detected into a multimodal generative large model to obtain a second text description content; Inputting the second text description content into the text generation model to determine whether the foreign body image to be detected is a reference foreign body image; In response to the foreign body image to be detected not being the reference foreign body image, performing the step of segmenting the spinal image to be detected to obtain a plurality of vertebral images to be detected; In response to the foreign body image to be detected being the reference foreign body image, the multimodal generative large model is used to evaluate the quality of the spine image to be detected.

7. A quality evaluation system for spine images, characterized in that: The quality evaluation system comprises: A spine image acquisition module, used for acquiring a spine image to be detected; A vertebral image acquisition module, used for segmenting the spinal image to be detected to obtain a plurality of vertebral images to be detected; A lateral vertebral body determination module, used to determine whether the vertebral body image to be detected is a lateral vertebral body image by using a preset generative large model; A target quantity acquisition module, used for acquiring the target quantity corresponding to the lateral vertebral image among the plurality of vertebral images to be detected corresponding to the vertebral image to be detected; The quality evaluation module is used to evaluate the quality of the spine image to be detected based on the target quantity.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and used to run on the processor, characterized in that: When the processor executes the computer program, the spine image quality assessment method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the spine image quality assessment method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for evaluating the quality of a spine image according to any one of claims 1 to 6 is implemented.

Images