CN-122023599-A - Image processing method, device and computer equipment

Abstract

The application relates to an image processing method, an image processing device and computer equipment. The method comprises the steps of obtaining image processing results and an excitation model, wherein the image processing results are obtained by correcting metal artifacts of a metal implant in a scanned image based on a correction model, segmenting each image processing result through the excitation model to obtain a target tissue segmentation image and a metal segmentation image, calculating tissue characteristic loss between the target tissue segmentation image and an original tissue image based on the excitation model, calculating metal characteristic loss between the metal segmentation image and the original metal image, exciting and optimizing an application process or a training process of the correction model according to the metal characteristic loss and the tissue characteristic loss, and using the excited and optimized correction model for image processing. By adopting the method, the accuracy of image processing can be improved.

Inventors

• YUAN HUISHU
• ZHANG YAN
• YE KAI

Assignees

• 北京大学第三医院（北京大学第三临床医学院）

Dates

Publication Date

20260512

Application Date

20260115

Claims (10)

1. 1. An image processing method, the method comprising: the method comprises the steps of obtaining image processing results and an excitation model, wherein the image processing results are obtained by correcting metal artifacts of a metal implant in a scanned image based on a correction model; Dividing each image processing result through the excitation model to obtain a target tissue division image and a metal division image; calculating tissue characteristic loss between the target tissue segmentation image and the original tissue image based on the excitation model, and calculating metal characteristic loss between the metal segmentation image and the original metal image; and performing excitation optimization on the application process or training process of the correction model according to the metal characteristic loss and the tissue characteristic loss, wherein the correction model after excitation optimization is used for image processing.
2. 2. The method of claim 1, wherein the obtaining each image processing result comprises: Acquiring each scanning image, each original tissue image and each original metal image; Acquiring a correction model based on each of the original metal images and each of the original tissue images; and inputting each scanning image into the correction model, and correcting the metal artifact of the metal implant in the scanning image through the correction model to obtain each image processing result.
3. 3. The method of claim 2, wherein the acquiring each scanned image, each original tissue image, and each original metal image comprises: obtaining each scanning image from a database, wherein the scanning image comprises a scanned part of a scanned object and a metal implant; Acquiring an original metal image corresponding to the metal implant for each scanned image; And acquiring the original tissue image of the scanned object and the scanned part which belong to the same scanned object and the scanned part from the database.
4. 4. The method of claim 2, wherein the obtaining a correction model based on each of the original metal images and each of the original tissue images comprises: splicing each original tissue image and an original metal image corresponding to the original tissue image to obtain a spliced image; performing forward projection processing and reverse projection processing on the spliced image to obtain a simulation image, and combining the spliced image and the simulation image to obtain a training sample set; And training a preset machine learning model according to each training sample group to obtain a correction model.
5. 5. The method according to claim 1, wherein the excitation model includes an image segmentation algorithm and an interpolation restoration algorithm, and the segmenting each image processing result by the excitation model to obtain a target tissue segmentation image and a metal segmentation image includes: Dividing each image processing result into a tissue segmentation image and a metal segmentation image according to the image segmentation algorithm; and updating the metal in the tissue segmentation image into the tissue of the scanned part based on the interpolation restoration algorithm to obtain a target tissue segmentation image.
6. 6. The method of claim 1, wherein the excitation model includes a feature extraction algorithm and a feature difference algorithm, and wherein the calculating tissue feature loss between the target tissue segmentation image and the original tissue image based on the excitation model comprises: Transforming the target tissue segmentation image into a tissue segmentation sectional image according to the excitation model, and transforming the original tissue image into an original tissue sectional image; Extracting a first tissue feature of the original tissue section image based on the feature extraction algorithm, and extracting a second tissue feature of the tissue segmentation section image; and calculating the tissue characteristic loss between the first tissue characteristic and the second tissue characteristic through the characteristic difference algorithm.
7. 7. The method of any one of claims 1 or 6, wherein the calculating a metal feature loss between the metal segmented image and the original metal image comprises: Extracting first metal features of the metal segmentation image according to a feature extraction algorithm, and extracting second metal features of the original metal image; a metal feature loss between the first metal feature and the second metal feature is calculated based on a feature difference algorithm.
8. 8. The method of claim 1, wherein said excitation optimization of the application process or training process of the correction model based on each of the metal feature loss and each of the tissue feature loss comprises: Weighting each tissue characteristic loss and the metal characteristic loss corresponding to the tissue characteristic loss to obtain total loss; In the training process or the application process of the correction model, exciting and optimizing the correction model according to the total loss and the back propagation algorithm until the correction model meets a preset optimization stopping condition; And determining the correction model as a target correction model, wherein the target correction model is used for processing metal artifacts in the scanned image.
9. 9. An image processing apparatus, characterized in that the apparatus comprises: the acquisition module is used for acquiring each image processing result and an excitation model, wherein the image processing result is obtained by correcting metal artifacts of a metal implant in a scanned image based on a correction model; The segmentation module is used for segmenting each image processing result through the excitation model to obtain a target tissue segmentation image and a metal segmentation image; A calculation module for calculating tissue feature loss between the target tissue segmentation image and the original tissue image based on the excitation model, and calculating metal feature loss between the metal segmentation image and the original metal image; the optimization module is used for performing excitation optimization on the application process or the training process of the correction model according to the metal characteristic loss and the tissue characteristic loss, and the correction model after excitation optimization is used for image processing.
10. 10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 8 when the computer program is executed.

Description

Image processing method, device and computer equipment Technical Field The present application relates to the field of large model technology, and in particular, to an image processing method, an image processing apparatus, a computer device, a computer readable storage medium, and a computer program product. Background In CT (Computed Tomography ) imaging examinations, metal implants absorb and scatter X-rays, resulting in significant metal artifacts in the captured image and thus in reduced image quality. Therefore, in order to improve the quality of an image, it is necessary to correct metal artifacts in the image by correcting the model. In the conventional technology, a computer device acquires each spliced image and each simulated image, and trains an initial correction model according to each spliced image and each simulated image until the initial correction model meets a preset first training stop condition, so as to obtain a correction model. The simulation image is obtained by constructing metal artifacts of metal implants in the spliced image based on experience of an operator and adding the metal artifacts into the spliced image, and the spliced image is obtained by splicing an original tissue image and an original metal image. However, in the conventional technology, the simulation image is obtained based on the manual experience process, and has a certain limitation, so the simulation image is inaccurate, and further, the correction model obtained based on the training of the simulation image is inaccurate. Therefore, the accuracy of the current image processing method is low. Disclosure of Invention Based on this, it is necessary to provide an image processing method, an apparatus, a computer device, a computer readable storage medium, and a computer program product in view of the above technical problems. In a first aspect, the present application provides an image processing method, including: the method comprises the steps of obtaining image processing results and an excitation model, wherein the image processing results are obtained by correcting metal artifacts of a metal implant in a scanned image based on a correction model; Dividing each image processing result through the excitation model to obtain a target tissue division image and a metal division image; calculating tissue characteristic loss between the target tissue segmentation image and the original tissue image based on the excitation model, and calculating metal characteristic loss between the metal segmentation image and the original metal image; and performing excitation optimization on the application process or training process of the correction model according to the metal characteristic loss and the tissue characteristic loss, wherein the correction model after excitation optimization is used for image processing. In one embodiment, the acquiring each image processing result includes: Acquiring each scanning image, each original tissue image and each original metal image; Acquiring a correction model based on each of the original metal images and each of the original tissue images; and inputting each scanning image into the correction model, and correcting the metal artifact of the metal implant in the scanning image through the correction model to obtain each image processing result. In one embodiment, the acquiring each scanned image, each original tissue image, and each original metal image includes: obtaining each scanning image from a database, wherein the scanning image comprises a scanned part of a scanned object and a metal implant; Acquiring an original metal image corresponding to the metal implant for each scanned image; And acquiring the original tissue image of the scanned object and the scanned part which belong to the same scanned object and the scanned part from the database. In one embodiment, the acquiring a correction model based on each of the original metal images and each of the original tissue images includes: splicing each original tissue image and an original metal image corresponding to the original tissue image to obtain a spliced image; performing forward projection processing and reverse projection processing on the spliced image to obtain a simulation image, and combining the spliced image and the simulation image to obtain a training sample set; And training a preset machine learning model according to each training sample group to obtain a correction model. In one embodiment, the excitation model includes an image segmentation algorithm and an interpolation restoration algorithm, and the segmenting each image processing result by the excitation model to obtain a target tissue segmentation image and a metal segmentation image includes: Dividing each image processing result into a tissue segmentation image and a metal segmentation image according to the image segmentation algorithm; and updating the metal in the tissue segmentation image into the tissue of the scanned part based on the int