Search

CN-122023313-A - Image segmentation method, system, equipment and medium based on multimode image data

CN122023313ACN 122023313 ACN122023313 ACN 122023313ACN-122023313-A

Abstract

The invention provides an image segmentation method, an image segmentation system, image segmentation equipment and an image segmentation medium based on multimode image data. The image segmentation method comprises the steps of performing skeleton line calculation on a voxelized structure based on multimode image data to obtain a discrete skeleton line, performing smoothing processing on the discrete skeleton line according to a first loss function of restraining smooth movement amplitude of skeleton points, a second loss function of controlling distance difference between different points and a third loss function of controlling center line smoothness to obtain a smooth skeleton line, extracting cross sections at each center point of the smooth skeleton line, randomly establishing a 2D coordinate system on the cross sections, jointly extracting the cross sections at each center point of the smooth skeleton line, performing rotation adjustment on the randomly generated 2D coordinate system by using a fourth loss function to obtain an adjusted skeleton line, performing straightening processing on the adjusted skeleton line to obtain a straightened image, and performing an image segmentation method on the straightened image.

Inventors

  • CHEN DUANDUAN

Assignees

  • 北京理工大学

Dates

Publication Date
20260512
Application Date
20260126

Claims (10)

  1. 1. An image segmentation method based on multimode image data, comprising: obtaining metadata in the target image; Normalizing the target image according to the metadata to obtain a normalization result, wherein coordinate normalization is performed on voxels of the target image according to the metadata, and the mode of the target image is determined according to the metadata, so that gray scale normalization is performed on the target image according to the determined mode; Determining segmentation instructions for the normalized results of the target image that are available for execution by a computing device based on a user's speech input; executing a pre-segmentation model on the normalization result of the target image according to the segmentation instruction to obtain a pre-segmentation result; Obtaining a complete segmentation result for the target image based on the pre-segmentation result using a first computing network and obtaining a straightened vessel model based on the complete segmentation result using a second computing network to obtain a simplified segmentation result, and Morphological parameters and hemodynamic parameters are extracted based on the simplified segmentation result.
  2. 2. The method of claim 1, wherein the coordinate normalization of voxels of the target image according to the metadata comprises: Determining whether the target image is a Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) image, and When the target image is a Computed Tomography (CT) image or a Magnetic Resonance Imaging (MRI) image, determining a row direction, a column direction and a slice direction of the target image according to the metadata, and performing spatial alignment according to the determined row direction, column direction and slice direction, wherein a slice direction unit vector can be obtained by cross multiplying a row vector and a column vector.
  3. 3. The method of claim 1, wherein the determining a modality of the target image from the metadata to gray scale normalize the target image from the determined modality comprises: When the target image is determined to be a CT image or an MRI image according to the metadata, gray scale normalization is performed according to the following formula: I_scaled=(I-I_min)/(I_max-I_min ) wherein I_ "min", I_ "max" is the lower and upper bounds of intensity set for each modality. I is the original voxel gray value. I_ "scaled" is the gray value mapped to the 0-1 interval after linear scaling.
  4. 4. The method of claim 1, further comprising: correlating the normalized result with information indicated by the corresponding ultrasound signal, and Correcting the hemodynamic parameter using information indicated by the associated ultrasound signal, wherein The information indicated by the ultrasonic signal at least comprises the position of an ultrasonic sampling point and the speed information of the corresponding sampling point; said correlating the normalized result with information indicated by the corresponding ultrasound signal comprises: and correlating the position of the ultrasonic sampling point with the central line of the blood vessel in the normalization result.
  5. 5. The method of any of claims 1-4, wherein using the second computing network to obtain a straightened vessel model based on the complete segmentation result to obtain a simplified segmentation result comprises: performing skeleton line calculation of the blood vessel on the voxel structure generated by the first calculation network segmentation to form a discrete skeleton line of the step size of the blood vessel; Performing skeleton smoothing on discrete skeleton lines of the blood vessel based on the loss function; extracting a cross section at the center point of each smoothed skeleton line of the blood vessel, and randomly establishing a 2D coordinate system on the cross section; combining all cross sections and rotating a randomly generated 2D coordinate system over the cross sections; And carrying out straightening mapping on the blood vessels according to the combined and rotated cross sections to generate straightened images, and carrying out segmentation and inverse transformation processing on the straightened images to obtain simplified segmentation results.
  6. 6. The method of claim 5, wherein the rotating the randomly generated 2D coordinate system over the cross-section comprises: The randomly generated 2D coordinate system over the cross section is rotated using the following loss function: Wherein the method comprises the steps of Represents the loss value based on the cross section, Is a set of parameterized discrete cross-sectional point sequences, The rotation parameter is indicated as such, Representing an offset or step size parameter for selecting a spacing point in the sequence of points, calculating a second order difference at different step sizes, Representing each discrete cross-section point sequence I.e. the number of points in the sequence. A sequence of points representing a single discrete cross section, belonging to One element in the collection.
  7. 7. The method of any one of claims 1-4, wherein The simplified segmentation results appear as a three-dimensional point cloud, Extracting hemodynamic parameters based on the simplified segmentation result includes: Inputting an internal point cloud of the reduced segmented result three-dimensional point cloud to a first channel of a dynamic graph rolling network (DGCNN), and inputting a surface point cloud of the reduced segmented result three-dimensional point cloud to a second channel of DGCNN, wherein the first channel is independent of the second channel; The first channel and the second channel respectively perform feature coding on the input point clouds, wherein the surface point clouds in the second channel are subjected to feature aggregation to form global feature tensors; And carrying out feature fusion on feature coding results of the first channel and the second channel, wherein the global features are spliced with point-by-point features of an internal point cloud through a broadcasting mechanism to form a comprehensive feature representation comprising local geometric details and global structure perception.
  8. 8. An image segmentation system based on multi-mode image data, comprising: the data management module is configured to be connected with the image database to obtain metadata in the target image; a normalization processing module configured to perform normalization processing on the target image according to the metadata to obtain a normalization result, wherein coordinate normalization processing is performed on voxels of the target image according to the metadata, and a modality of the target image is determined according to the metadata, so that gray scale normalization processing is performed on the target image according to the determined modality; a large language model module configured to determine segmentation instructions for the normalized results for the target image based on a user's speech input that are available for execution by a computing device; The image pre-segmentation module is configured to execute a pre-segmentation model on the normalization result of the target image according to the determined segmentation instruction so as to obtain a pre-segmentation result; An image segmentation module configured to obtain a complete segmentation result for the target image based on the pre-segmentation result obtained by the image pre-segmentation module 740 using a first computing network, and to obtain a straightened vessel model based on the complete segmentation result using a second computing network to obtain a simplified segmentation result; a parameter extraction module configured to extract morphological parameters and hemodynamic parameters based on the simplified segmentation result.
  9. 9. An electronic device comprising a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor in communication with the storage medium via the bus when the electronic device is in operation, the processor executing the machine-readable instructions to perform the steps of any one of claims 1 to 7 relating to a method of image segmentation based on multimodal image data.
  10. 10. A computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, performs the steps of a method of image segmentation based on multimode image data as claimed in any one of claims 1 to 7.

Description

Image segmentation method, system, equipment and medium based on multimode image data Technical Field The present invention relates to the field of image processing, and more particularly, to an image segmentation method, system, device, and medium based on multimode image data. Background Common to the core pathological mechanisms of cardiovascular and cerebrovascular diseases are vascular wall injury, hemodynamic abnormalities and thrombosis. For example, in cerebrovascular disease, cerebral aneurysms are locally dilated due to degeneration of the elastic layer of the vessel wall, and rupture can cause subarachnoid hemorrhage. For example, in cardiovascular diseases, coronary heart disease is a thrombosis caused by rupture of coronary plaque, and aortic diseases occur as a result of rupture of elastic fibers in the wall of the blood vessel, as a result of dissection (rupture of intima to form true and false double lumens) or aneurysm (local vessel wall expansion). Taking aortic dissection as an example, the pressing of a false cavity into a true cavity during lesion development can cause organ ischemia, the fracture mortality rate is as high as 90 percent, and the false cavity can share a pathological chain of 'abnormal vascular structure, dysfunction and fatal complications' with other cardiovascular and cerebrovascular diseases. Image reconstruction (i.e., three-dimensional reconstruction based on medical images) techniques are the core support for the overall process of "diagnosis-study" of cardiovascular and cerebrovascular diseases. In the aspect of diagnosis, through image reconstruction such as CT angiography (Computed Tomography Angiography, CTA), magnetic resonance angiography (Magnetic Resonance Angiography, MRA) and the like, a two-dimensional image can be converted into a three-dimensional vascular model, focus (such as a cerebral aneurysm position and an aortic dissection breach) can be accurately positioned, and missed diagnosis caused by the limitation of a two-dimensional image view angle is avoided. For example, intracranial arterial CTA reconstruction can clearly show the extent of arterial stenosis in the brain, and aortic CTA reconstruction can quantify the true lumen volume. In terms of treatment, the reconstruction model provides an anatomical basis for surgical planning, for example, cerebral aneurysm occlusion needs to be determined based on reconstruction, and thoracic aortic endoluminal repair (TEVAR) surgery needs to rely on reconstruction to select stent diameters and anchoring regions. In the follow-up aspect, the curative effect can be quantified through the comparison of the pre-operation reconstruction model and the post-operation reconstruction model. In addition, in the scientific research aspect, the standardized image reconstruction can unify vessel morphological parameters of a large number of patients, and a unified data base is provided for intelligent algorithm training. However, if the accurate image reconstruction is not available, the diagnosis and treatment of cardiovascular and cerebrovascular diseases are degenerated into "empirical driving". For example, patients with steep aortic arch may have insufficient stent anchoring due to lack of mastering the three-dimensional morphology, increasing the risk of postoperative endoleak. In addition, the data formats are not uniform in scientific research, so that multi-center research is difficult to develop. The current image reconstruction technology is difficult to meet the requirements of large-scale multimode data diagnosis and treatment. The current image reconstruction system is mostly limited to a single mode (for example, only CTA images are stored), mechanical test data (for example, mooney-Rivlin constitutive coefficients) of tissue samples, post-operation follow-up DSA images and hemodynamic calculation results (for example, wall shear stress) cannot be associated, a large amount of patient data are stored in a scattered mode, cross-system integration is needed during calling, efficiency is low, and algorithm training of a thousand-person-level queue is difficult to support. Therefore, a new image segmentation technique based on multimode image data is needed to solve the above technical problems. Disclosure of Invention In view of the above problems, the present invention provides an image segmentation method, system, device and medium based on multimode image data. According to the image segmentation method, system, equipment and medium based on the multimode image data, the image data with different modes of a large number of patients can be effectively managed, so that the interested areas of related images can be segmented rapidly, and required morphological and mechanical result parameters can be obtained. According to one aspect of the disclosure, an image segmentation method based on multimode image data is provided, comprising obtaining metadata in a target image, normalizing the target image accor