Zero-shot driven abdominal MRI multi-organ segmentation system
The zero-shot driven abdominal MRI multi-organ segmentation system utilizes zero-shot learning technology and deep learning models to achieve accurate segmentation of abdominal MRI images without labeled data. This solves the problems of resource waste and insufficient applicability in traditional methods, and improves segmentation accuracy and system applicability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ANHUI UNIV
- Filing Date
- 2026-01-20
- Publication Date
- 2026-06-19
AI Technical Summary
Traditional abdominal MRI image segmentation methods rely on manual annotation and large amounts of training data, resulting in a waste of time and resources. Furthermore, they are difficult to apply to unseen datasets, limiting their widespread use.
A zero-shot driven abdominal MRI multi-organ segmentation system is adopted, including an input device, a preprocessing device, a model training device, and a segmentation device. It utilizes zero-shot learning technology to extract image features and perform accurate segmentation under unlabeled data conditions. Through image preprocessing and deep learning model construction, combined with an adaptive multi-scale fusion module, the segmentation accuracy is improved.
Efficient and accurate multi-organ segmentation of abdominal MRI images was achieved without labeled data, improving the system's applicability and segmentation accuracy, reducing data labeling costs, and meeting the needs of high-quality image applications.
Smart Images
Figure CN122244430A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of abdominal MRI technology, and more specifically, to a zero-sample driven abdominal MRI multi-organ segmentation system. Background Technology
[0002] In the field of medical imaging diagnosis, abdominal magnetic resonance imaging (MRI) is an important non-invasive examination method. MRI can clearly observe the structures of multiple organs in the abdominal cavity, such as the liver, spleen, pancreas, and kidneys, which is of great significance for the early diagnosis and treatment of diseases.
[0003] Traditional abdominal MRI image segmentation methods mostly rely on manual annotation and large amounts of training data. This not only consumes a lot of time and resources but is also difficult to apply to previously unseen datasets. Furthermore, the difficulty and cost of data annotation may limit their widespread application. To address this issue, zero-shot learning techniques have emerged in recent years. These techniques can quickly adapt to new, unseen datasets using existing knowledge without additional labeled data, thereby improving the model's generalization ability and application scope. Summary of the Invention
[0004] To overcome the shortcomings of existing technologies, this invention provides a zero-sample driven abdominal MRI multi-organ segmentation system, which has the advantage of reducing the cost of data annotation.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a zero-sample driven abdominal MRI multi-organ segmentation system, comprising an input device, a preprocessing device, a model training device, and a segmentation device, wherein the output end of the input device is connected to the input end of the preprocessing device, and the output end of the preprocessing device is connected to the input end of the segmentation device.
[0006] Input device for receiving abdominal MRI images;
[0007] A preprocessing unit is used to preprocess MRI images to eliminate noise and improve image quality;
[0008] The model training device acquires a segmentation model through zero-shot learning;
[0009] The segmentation device uses a trained segmentation model to segment multiple organs in abdominal MRI images.
[0010] As a preferred embodiment of the present invention, the input device is connected to the nuclear magnetic resonance detection device via a cable, and the input device can also be connected to a database via a network.
[0011] As a preferred embodiment of the present invention, the preprocessing device includes an image input module, multiple image processing modules, and an image output module;
[0012] The image input module is used to receive abdominal MRI images transmitted by the input device and add time information to the abdominal MRI images;
[0013] The image processing module uses digital processing methods to suppress noise in abdominal MRI images while preserving detailed features.
[0014] The image input module outputs the processed abdominal MRI image to the segmentation device based on the time information.
[0015] As a preferred embodiment of the present invention, the preprocessing device further includes an image sharpening module, which is located between the image input module and the image processing module, and can sharpen abdominal MRI images.
[0016] As a preferred embodiment of the present invention, the model training device includes a data input module, a deep learning module, and a model building module;
[0017] The data input module is used to input a small amount of existing labeled data and a large number of unlabeled sample images, and then send the samples into the deep learning module for learning.
[0018] The deep learning module uses a deep learning-based neural network model to learn from a small number of labeled data sample images, automatically learns organ features in the sample images, and uses a multi-layer convolutional neural network structure to identify organ boundaries in unlabeled sample images, compares them with the aforementioned organ features, and then extracts relevant organ features.
[0019] The model building module constructs a segmentation model by extracting relevant features of organs and combining them with the adaptive multi-scale fusion module.
[0020] As a preferred embodiment of the present invention, the small number of existing labeled data sample images can be sample images of similar medical images from the past.
[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0022] 1. This invention improves the system's applicability by incorporating an input device that can receive abdominal MRI images from various sources. Since MRI images often contain noise or uneven quality, the input module requires preprocessing. Through filtering and denoising methods, the image clarity and consistency are maintained before segmentation. Therefore, the preprocessing module eliminates interfering factors in the image, improving image quality. The segmentation device uses a segmentation model acquired through zero-shot learning from a model training device to accurately segment multiple organs in the MRI image. This process utilizes zero-shot learning technology to effectively extract image features and apply them to the segmentation task even without labeled data, thereby improving the accuracy and robustness of segmentation.
[0023] 2. This invention, by setting up an image input module, can distinguish different abdominal MRI images by adding time information to the images, thus preventing image confusion. By setting up multiple image processing modules, multiple abdominal MRI images can be processed simultaneously, effectively improving processing efficiency. The image sharpening module sharpens the images, solving the problem of insufficient image clarity, so that the processed images can retain more details and edge information, improving the visual effect of the images, thereby meeting the application scenarios with high requirements for image quality. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of the structure of the present invention;
[0025] Figure 2 This is a schematic diagram of the pretreatment device structure of the present invention;
[0026] Figure 3 This is a schematic diagram of the model training device of the present invention. Detailed Implementation
[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] like Figures 1 to 3 As shown, the present invention provides a zero-shot driven abdominal MRI multi-organ segmentation system, including an input device, a preprocessing device, a model training device, and a segmentation device. The output end of the input device is connected to the input end of the preprocessing device, and the output end of the preprocessing device is connected to the input end of the segmentation device.
[0029] Input device for receiving abdominal MRI images;
[0030] A preprocessing unit is used to preprocess MRI images to eliminate noise and improve image quality;
[0031] The model training device acquires a segmentation model through zero-shot learning;
[0032] The segmentation device uses a trained segmentation model to segment multiple organs in abdominal MRI images.
[0033] By configuring the input device, this module can receive abdominal MRI images from various sources, improving the system's applicability. Since MRI images often contain noise or uneven quality, the input module requires preprocessing. Through filtering and denoising methods, the image clarity and consistency are maintained before segmentation. Therefore, the preprocessing module eliminates interfering factors in the image, improving image quality. Subsequently, the segmentation module uses a segmentation model acquired through zero-shot learning by the training module to accurately segment multiple organs in the MRI image. This process, by utilizing zero-shot learning technology, can effectively extract image features and apply them to the segmentation task even without labeled data, thereby improving the accuracy and robustness of segmentation.
[0034] As a preferred embodiment of the present invention, the input device is connected to the nuclear magnetic resonance detection device via a cable, and the input device can also be connected to a database via a network.
[0035] By setting up a database, it is possible to receive abdominal MRI images from different locations, thereby increasing the diversity of data sources for the system and improving its applicability.
[0036] The preprocessing device includes an image input module, multiple image processing modules, and an image output module.
[0037] The image input module is used to receive abdominal MRI images transmitted by the input device and add time information to the abdominal MRI images;
[0038] The image processing module uses digital processing methods to suppress noise in abdominal MRI images while preserving detailed features.
[0039] The image input module outputs the processed abdominal MRI image to the segmentation device based on the time information.
[0040] By setting up an image input module, time information can be added to abdominal MRI images to distinguish different abdominal MRI images and prevent image confusion. By setting up multiple image processing modules, multiple abdominal MRI images can be processed simultaneously, effectively improving processing efficiency.
[0041] The preprocessing device further includes an image sharpening module, which is located between the image input module and the image processing module. The image sharpening module can sharpen abdominal MRI images.
[0042] The image sharpening module sharpens the image, solving the problem of insufficient image clarity. This allows the processed image to retain more details and edge information, improving the visual effect and thus meeting the needs of application scenarios with high image quality requirements.
[0043] The model training device includes a data input module, a deep learning module, and a model building module.
[0044] The data input module is used to input a small amount of existing labeled data and a large number of unlabeled sample images, and then send the samples into the deep learning module for learning.
[0045] The deep learning module uses a deep learning-based neural network model to learn from a small number of labeled data sample images, automatically learns organ features in the sample images, and uses a multi-layer convolutional neural network structure to identify organ boundaries in unlabeled sample images, compares them with the aforementioned organ features, and then extracts relevant organ features.
[0046] The model building module constructs a segmentation model by extracting relevant features of organs and combining them with the adaptive multi-scale fusion module.
[0047] By setting up an adaptive multi-scale fusion module, which can combine information from different scales, the segmentation accuracy of the segmentation model is enhanced, thereby promoting the accurate identification of the edges of different organs.
[0048] Among them, the small number of existing labeled data sample images can be sample images of similar medical images from the past.
[0049] By using sample images of similar medical images from the past, the time required for manual annotation of sample data can be effectively reduced, and the detection efficiency can be improved.
[0050] Working principle and usage process of this invention:
[0051] By configuring the input device, this module can receive abdominal MRI images from various sources, improving the system's applicability. Since MRI images often contain noise or uneven quality, the input module requires preprocessing. Through filtering and denoising methods, the image clarity and consistency are maintained before segmentation. Therefore, the preprocessing module eliminates interfering factors in the image, improving image quality. Subsequently, the segmentation module uses a segmentation model acquired through zero-shot learning by the training module to accurately segment multiple organs in the MRI image. This process, by utilizing zero-shot learning technology, can effectively extract image features and apply them to the segmentation task even without labeled data, thereby improving the accuracy and robustness of segmentation.
[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A zero-shot-driven abdominal MRI multi-organ segmentation system, characterized in that: It includes an input device, a preprocessing device, a model training device, and a segmentation device. The output end of the input device is connected to the input end of the preprocessing device, and the output end of the preprocessing device is connected to the input end of the segmentation device. Input device for receiving abdominal MRI images; A preprocessing unit is used to preprocess MRI images to eliminate noise and improve image quality; The model training device acquires a segmentation model through zero-shot learning; The segmentation device uses a trained segmentation model to segment multiple organs in abdominal MRI images.
2. The zero-shot driven abdominal MRI multi-organ segmentation system according to claim 1, characterized in that: The input device is connected to the nuclear magnetic resonance imaging device via a cable, and the input device can also be connected to a database via a network.
3. The zero-shot driven abdominal MRI multi-organ segmentation system according to claim 1, characterized in that: The preprocessing device includes an image input module, multiple image processing modules, and an image output module; The image input module is used to receive abdominal MRI images transmitted by the input device and add time information to the abdominal MRI images; The image processing module uses digital processing methods to suppress noise in abdominal MRI images while preserving detailed features. The image input module outputs the processed abdominal MRI image to the segmentation device based on the time information.
4. The zero-shot driven abdominal MRI multi-organ segmentation system according to claim 3, characterized in that: The preprocessing device also includes an image sharpening module, which is located between the image input module and the image processing module. The image sharpening module can sharpen abdominal MRI images.
5. The zero-shot driven abdominal MRI multi-organ segmentation system according to claim 1, characterized in that: The model training device includes a data input module, a deep learning module, and a model building module; The data input module is used to input a small amount of existing labeled data and a large number of unlabeled sample images, and then send the samples into the deep learning module for learning. The deep learning module uses a deep learning-based neural network model to learn from a small number of labeled data sample images, automatically learns organ features in the sample images, and uses a multi-layer convolutional neural network structure to identify organ boundaries in unlabeled sample images, compares them with the aforementioned organ features, and then extracts relevant organ features. The model building module constructs a segmentation model by extracting relevant features of organs and combining them with the adaptive multi-scale fusion module.
6. The zero-shot driven abdominal MRI multi-organ segmentation system according to claim 5, characterized in that: The limited number of existing labeled data sample images can be sample images of similar medical images from the past.