Artificial intelligence-based visual automatic marking system and method for eye burst

Abstract

The invention relates to an artificial intelligence-based visual automatic marking system and method for eye burst, and belongs to the technical field of iconography evaluation. The system comprises a data module, a display module, an AI automatic labeling module, a manual labeling module, a data quantitative analysis module and an image three-dimensional reconstruction module. The method can shorten the time of organ sketching, improve the working efficiency of doctors, support the remote sketching mode of Internet medical treatment, effectively shorten the treatment cycle of patients, and is significantly improved compared with the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of imaging evaluation, and particularly relates to an artificial intelligence-based visualization automatic annotation system and method for exophthalmos. Background Art

[0002] Thyroid-associated ophthalmopathy (TAO), also known as Graves ophthalmopathy (GO), is the most common extra-thyroid manifestation of Graves disease (GD), with an incidence rate accounting for 25% - 40% of GD. It can also be seen in 2% of patients with chronic lymphocytic thyroiditis, a small number of patients with hypothyroidism, and the euthyroid population. The prevalence of TAO is estimated to be 90 to 305 cases per 100,000 people. The main clinical manifestations include unilateral or bilateral eyelid retraction, exophthalmos, diplopia, restrictive strabismus, exposure keratopathy, and optic neuropathy, which seriously affect the quality of life of patients.

[0003] The natural course of TAO is divided into an active inflammatory stage and an inflammatory stable stage, and the severity of the disease usually progresses during the active inflammatory stage. The pathological changes in active TAO are mainly inflammatory reactions, including extraocular muscle swelling, tissue edema, inflammatory cell infiltration, hyaluronic acid deposition, etc., while the pathological changes in non-active TAO are mainly fat infiltration and tissue fibrosis. The condition of mild TAO patients will improve with the development of the natural course, while moderate to severe cases and those endangering vision need to receive intervention treatment for TAO. The clinical outcomes and prognosis of TAO vary greatly. Therefore, the management of TAO must be based on a correct assessment of the disease state. The assessment methods recommended by the guidelines for the disease state include the disease activity score for assessing activity and the EUGOGO method and NOSPECS classification for assessing severity, which have limitations due to relying on the subjective observation of the examiner.

[0004] MRI can effectively prevent various damages caused by ionizing radiation and has high soft tissue resolution. The advantages of its multi-sequence and multi-parameter can not only well evaluate the morphological changes of the orbit but also evaluate the changes of the pathological components of each structure through quantitative parameters. However, in the existing technology, mainly by manually and continuously delineating the region of interest (ROI) layer by layer on the MRI image, calculating parameters such as the signal intensity (SI) of tissues such as the medial rectus muscle (MR), lateral rectus muscle (LR), superior rectus muscle (SR), inferior rectus muscle (IR), lacrimal gland (LG), and orbital fat (OF) on both sides, the manual time cost is high and the acquisition difficulty is high. In view of this, the present invention is specifically proposed to solve the above problems. Summary of the Invention

[0005] The object of the present invention is to provide a visualization system for automatic annotation of exophthalmos based on artificial intelligence, with low cost and low acquisition difficulty.

[0006] To achieve the above object, the present invention provides the following technical solutions: A visualization automatic annotation system for exophthalmos based on artificial intelligence, including a data module, a display module, an AI automatic annotation module, a manual annotation module, a data quantitative analysis module, and an image three-dimensional reconstruction module; The data module includes a data import unit, a data management unit, and a data export unit; The AI automatic annotation module calls an AI algorithm to perform image segmentation on medical images and automatically annotate organs according to the medical image data in the data module; The manual annotation module includes a semi-automatic segmentation algorithm and a manual drawing tool for correcting the annotation of the AI automatic annotation module; The data quantitative analysis module measures the length, angle, area, volume, and exophthalmos parameters of organs or regions through measuring tools; The image three-dimensional reconstruction module three-dimensionally reconstructs the segmentation data of organs into three-dimensional surface data and saves the three-dimensional surface data into the data management unit.

[0007] Further, the organs include the superior rectus muscle, inferior rectus muscle, medial rectus muscle, lateral rectus muscle, optic nerve, retrobulbar fat, and lacrimal gland.

[0008] Further, the semi-automatic segmentation algorithm includes a threshold algorithm, a region growing algorithm, and a morphological algorithm.

[0009] The present invention also provides a visualization automatic annotation method for exophthalmos based on artificial intelligence, which is applied to the above system and includes the following steps: S1: Data import and preprocessing Import the MRI images of the patient into the data module and automatically analyze the image parameters; S2: AI automatic annotation Call a pre-trained model through the AI automatic annotation module to segment the MRI images and automatically annotate key organs, and the target regions include the superior rectus muscle, inferior rectus muscle, optic nerve, retrobulbar fat, and lacrimal gland; Step 3: Manual correction and interaction Medical staff view the AI annotation results through the system interface and use the semi-automatic segmentation algorithm and manual drawing tool to correct the annotation if errors are found; Step 4: Three-dimensional reconstruction and visualization Convert the annotated image file into a three-dimensional model through the image three-dimensional reconstruction module, generate the three-dimensional surface data of the organs, and display the spatial relationship of the orbital structure in a three-dimensional view through the display module; Step 5: Quantitative Analysis Calculate the volume, surface area, and eye convexity of each key organ through the data quantitative analysis module, generate a structured report, and mark abnormal parameters for doctors' reference; Step 6: Data Export and Remote Collaboration Export the 3D model, quantitative analysis results, and marked imaging data for remote consultation or 3D printing. The patient data is transmitted to the partner hospital through an encrypted link for secondary confirmation by experts.

[0010] The beneficial effects of the present invention are as follows: 1. High efficiency: AI significantly shortens the annotation time and supports rapid clinical decision-making.

[0011] 2. Accuracy: Combine AI with manual correction to balance automation and flexibility.

[0012] 3. Visualization value: The 3D model intuitively shows the spatial relationship of the lesions and assists in personalized treatment.

[0013] 4. Remote collaboration: Data export is compatible with Internet medical care and promotes hierarchical diagnosis and treatment.

[0014] In summary, the present invention can shorten the time for organ delineation, improve the working efficiency of doctors, support the remote delineation method of Internet medical care, effectively shorten the patient treatment cycle, and has a significant improvement compared with the prior art.

[0015] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly and implement it in accordance with the content of the specification, the following takes the preferred embodiments of the present invention and combines with the accompanying drawings to describe in detail as follows. Description of the Drawings

[0016] Figure 1 It is a flowchart of the visualization automatic annotation method for exophthalmos shown in an embodiment of the present invention. Detailed Embodiment

[0017] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0019] The exophthalmos visualization automatic annotation system based on artificial intelligence shown in a preferred embodiment of the present application includes a data module, a display module, an AI automatic annotation module, a manual annotation module, a data quantitative analysis module, and an image three-dimensional reconstruction module.

[0020] Among them, the data module includes three units, including: 1. Data import unit: Reads the medical image data of the patient for display and subsequent functions. 2. Data management unit: Saves and manages relevant data such as medical image data, organ segmentation data, three-dimensional surface data, and measurement data imported externally or generated by the platform. 3. Data export unit: Exports the data managed by the data management module separately or as an engineering file as a whole to the user-specified location.

[0021] Among them, the display module can display the two-dimensional views such as medical images and organ segmentation data stored in the data management unit, as well as the three-dimensional views such as three-dimensional surface data, and the relevant data of data quantitative analysis to the corresponding view interfaces.

[0022] Among them, the AI automatic annotation module can call the AI algorithm to perform image segmentation on the medical image according to the image data in the data module, and automatically annotate organs including but not limited to the superior rectus muscle, inferior rectus muscle, medial rectus muscle, lateral rectus muscle, optic nerve, retrobulbar fat, lacrimal gland, etc.

[0023] The manual annotation module includes a series of semi-automatic segmentation algorithms and manual drawing tools. Among them, the semi-automatic segmentation algorithms include but are not limited to: 1. Threshold algorithm: Selects N thresholds according to the image and divides the image into multiple regions. 2. Region growing algorithm: Selects M seed points in the image, and starts from the set of seed points, and merges adjacent pixels with similar attributes such as intensity, gray level, texture color, etc. to this region. 3. Morphological algorithm: Performs image morphological operations such as erosion, dilation, opening operation, and closing operation on the segmentation data. The manual drawing tools include a brush, contour drawing, magic wand, etc. When the AI automatic annotation and semi-automatic algorithm annotation cannot meet the user's needs, the manual drawing tools can be used to modify the organs annotated by the AI automatic annotation and semi-automatic algorithms or manually draw the specified organs or regions.

[0024] The data quantitative analysis module includes a series of automatic measurement tools and manual measurement tools. Among them, the automatic measurement tools can automatically measure parameters including but not limited to: 1. Volume and surface area measurement: Calculates the volume of a closed surface using the divergence theorem and calculates the surface area of a closed surface using the discrete integral method.

[0025] 2. Measurement of eye protrusion parameters: (1) Distance between the left and right eye protrusions: Respectively find the farthest protruding points A and B of the left and right eyeballs, and calculate the shortest distance between points A and B as the eye protrusion distance; (2) Connecting line of the outer bony orbital margin vertices of the left and right eyes: Respectively find the lowest edge points C and D of the lateral fat of the left and right eyes in the cross-sectional data where the eye protrusion points are located, calculate the shortest distance between points C and D as the distance between the left and right eye axes, and then calculate the shortest distance between the line segment AB formed by the farthest protruding points of the left and right eyeballs and the line segment CD formed by the lowest edge points of the lateral fat of the left and right eyes.

[0026] Manual measurement tools include length measurement tools, angle measurement tools, area measurement tools, gray value measurement tools, etc. When the automatic measurement tools cannot meet the user's needs, parameters such as length, angle, area, and volume of the specified organ or region can be measured in the two-dimensional view and three-dimensional view according to the user's needs.

[0027] The image three-dimensional reconstruction module can three-dimensionally reconstruct the segmentation data of the organ into three-dimensional surface data and save the three-dimensional surface data to the data management unit.

[0028] The present invention will be further described below in combination with specific practical operations.

[0029] Step 1: Data import and preprocessing Import the DICOM format MRI images of the patient into the data module through the system interface, and automatically analyze the image parameters (such as slice thickness, resolution, sequence type).

[0030] The data management unit stores the original images and the subsequent generated annotation data, three-dimensional models, and quantitative analysis results.

[0031] Step 2: AI automatic annotation Call the pre-trained model through the AI automatic annotation module to segment the MRI images and automatically annotate the key organs. The target regions include the superior rectus muscle, inferior rectus muscle, optic nerve, retrobulbar fat, and lacrimal gland.

[0032] Time-consuming: It only takes 2 - 3 minutes to fully automatically process 120 layers of images (traditional manual work takes 3 hours).

[0033] Step 3: Manual correction and interaction Medical staff view the AI annotation results through the system interface, find that there are slight errors (due to MRI motion artifacts), and use the semi-automatic segmentation algorithm and manual drawing tools to correct the annotation.

[0034] Step 4: Three-dimensional reconstruction and visualization The annotated image files are converted into 3D models through the image 3D reconstruction module to generate 3D surface data of the organ, and the spatial relationship of the orbital structure is displayed in a 3D view through the display module.

[0035] Step 5: Quantitative Analysis The data quantitative analysis module calculates the volume, surface area, and convexity of each key organ, generates a structured report, and marks abnormal parameters for the doctor's reference.

[0036] ‌Step 6: Data Export and Remote Collaboration‌ The three-dimensional model, quantitative analysis results and annotated image data are exported through the data export unit for remote consultation or 3D printing. Patient data is transmitted to the cooperating hospital through an encrypted link for secondary confirmation by experts.

[0037] It has been verified that through the visualization system shown in the present invention, the total processing time is shortened from the traditional 3 hours to 10 minutes (2 minutes for AI labeling + 5 minutes for manual correction + 3 minutes for analysis); at the same time, the Dice coefficient of AI labeling reaches 0.93, and is increased to 0.98 after manual correction, which has significant advantages over the existing technology.

[0038] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0039] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. An automatic annotation system for exophthalmos visualization based on artificial intelligence, characterized in that, It includes a data module, a display module, an AI automatic annotation module, a manual annotation module, a data quantitative analysis module, and an image three-dimensional reconstruction module; The data module includes a data import unit, a data management unit, and a data export unit; The AI automatic annotation module calls an AI algorithm to segment the medical image and automatically annotate the organs according to the medical image data in the data module; The manual annotation module includes a semi-automatic segmentation algorithm and a manual drawing tool for correcting the annotations of the AI automatic annotation module; The data quantitative analysis module measures the length, angle, area, volume, and eye convexity parameters of organs or regions through measurement tools; The image three-dimensional reconstruction module three-dimensionally reconstructs the segmentation data of the organs into three-dimensional surface data and saves the three-dimensional surface data to the data management unit.

2. The automatic annotation system for exophthalmos visualization based on artificial intelligence according to claim 1, characterized in that, The organs include the superior rectus muscle, inferior rectus muscle, medial rectus muscle, lateral rectus muscle, optic nerve, retrobulbar fat, and lacrimal gland.

3. The automated visual annotation system for exophthalmos based on artificial intelligence according to claim 1, wherein The semi-automatic segmentation algorithm includes a threshold algorithm, a region growing algorithm, and a morphological algorithm.

4. A method for automatic visualization annotation of exophthalmos based on artificial intelligence, which is applied to the system for automatic visualization annotation of exophthalmos based on artificial intelligence according to any one of claims 1-3, characterized in that, It includes the following steps: S1: Data import and preprocessing Import the patient's MRI image into the data module and automatically analyze the image parameters; S2: AI automatic annotation Call a pre-trained model through the AI automatic annotation module to segment the MRI image and automatically annotate the key organs. The target regions include the superior rectus muscle, inferior rectus muscle, optic nerve, retrobulbar fat, and lacrimal gland; Step 3: Manual correction and interaction Medical staff view the AI annotation results through the system interface and use the semi-automatic segmentation algorithm and the manual drawing tool to correct the annotations if errors are found; Step 4: Three-dimensional reconstruction and visualization Convert the annotated image file into a three-dimensional model through the image three-dimensional reconstruction module, generate the three-dimensional surface data of the organs, and display the spatial relationship of the orbital structure in a three-dimensional view through the display module; Step 5: Quantitative analysis Calculate the volume, surface area, and eye convexity of each key organ through the data quantitative analysis module, generate a structured report, and mark abnormal parameters for doctors' reference; Step 6: Data export and remote collaboration Export the three-dimensional model, quantitative analysis results, and annotated image data for remote consultation or 3D printing. The patient data is transmitted to the partner hospital through an encrypted link for secondary confirmation by experts.

Images