CN-117237378-B - Method for segmenting important organs in head and neck CT image of nasopharyngeal carcinoma patient

CN117237378BCN 117237378 BCN117237378 BCN 117237378BCN-117237378-B

Abstract

The invention discloses a processing method and a device for segmenting a head and neck CT image of a nasopharyngeal carcinoma patient, wherein the method comprises the steps of inputting the head and neck CT image of the nasopharyngeal carcinoma patient into a slice screening model, and screening to obtain CT image slices containing important organs; inputting the CT image slice containing the important organ into a slice segmentation model to obtain a segmentation result, and completing segmentation of the important organ in the head and neck CT image of the nasopharyngeal carcinoma patient. The invention adopts the slice screening model and the slice segmentation model built by the artificial intelligence technology to realize automatic segmentation of the important organs in the head and neck CT image of the nasopharyngeal carcinoma patient, solves the problem of time and labor consumption of traditional manual sketching of segmentation results, can assist doctors in designing diagnosis and treatment schemes before treatment, and can also reduce the damage of rays to key organs as much as possible in the radiation treatment process.

Inventors

ZHANG XIN
WU WENXUAN
XING XIAOFEN
XU XIANGMIN
CHEN BEI

Assignees

人工智能与数字经济广东省实验室(广州)
华南理工大学

Dates

Publication Date: 20260512
Application Date: 20230802

Claims (6)

1. A method for segmenting vital organs in a CT image of the head and neck of a patient with nasopharyngeal carcinoma, said method comprising: inputting a head and neck CT image of a nasopharyngeal carcinoma patient into a slice screening model, and screening to obtain CT image slices containing important organs; Inputting a CT image slice containing an important organ into a slice segmentation model to obtain a segmentation result, and completing segmentation of the important organ in the head and neck CT image of the nasopharyngeal carcinoma patient; the slice screening model is obtained by the following steps: Preprocessing a head and neck CT image of a patient with nasopharyngeal carcinoma to be segmented, wherein the preprocessing comprises background removal and data prescribing; designing a pseudo tag for marking whether a single slice contains an important organ or not by using a real segmentation result; training the pseudo tag and the preprocessed head and neck CT image of the nasopharyngeal carcinoma patient together to obtain a slice screening model; the slice segmentation model is obtained by the following steps: erasing partial area of the CT image after pretreatment, and matching the erased data with the complete CT image after pretreatment Building a U-Net full convolution segmentation network under a deep learning framework, using paired data, taking CT images of erased partial areas as input, taking original data as labels, enabling the U-Net network to obtain the image recovery capability, and completing the pre-training of the model; Changing the last layer of convolution layer of the pre-training model, setting the output channel number as the number of the corresponding segmented organ types, and further training on the pre-trained model to be converged through the real segmentation result of the manual labeling and the matched head and neck CT image of the nasopharyngeal carcinoma patient to obtain a slice segmentation model; the slice screening model consists of a shallow convolutional neural network, the activation function of an output layer is a sigmoid function, when the output result is more than or equal to 0.5, the current slice is considered to contain important organs, and when the output result is less than 0.5, the current slice is considered to not contain important organs; The pretreated CT image is erased by 100 x 100 square critical areas.
2. The method for segmenting vital organs in a head and neck CT image of a patient with nasopharyngeal carcinoma of claim 1, wherein said background removal comprises: Judging whether the row or column in which the sliding window is positioned is background information from the upper direction, the lower direction, the left direction and the right direction of the CT image by utilizing the sliding window; When the sliding window moves to a non-background area, the window stops moving, the passed area is considered as redundant background information, and the information of the passed area is removed.
3. The method for segmenting vital organs in a head and neck CT image of a patient with nasopharyngeal carcinoma of claim 2, wherein said data defining includes: Each voxel value Subtracting the mean of all voxels Divided by standard deviation Obtaining a specified voxel value : 。
4. The method for segmenting vital organs in a head and neck CT image of a patient suffering from nasopharyngeal carcinoma of claim 1, wherein said pseudo-labels are binary pseudo-labels, 0 representing that the slice does not contain vital organs, and 1 representing that the slice contains vital organs.
5. An apparatus for segmenting vital organs in a CT image of the head and neck of a patient suffering from nasopharyngeal carcinoma, comprising a memory, a processor and a computer program stored in said memory and executable on said processor, characterized in that said processor implements the steps of the method according to any one of claims 1 to 4 when said computer program is executed.
6. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method according to any one of claims 1 to 4.

Description

Method for segmenting important organs in head and neck CT image of nasopharyngeal carcinoma patient Technical Field The invention relates to the field of computer vision and the field of medical image segmentation, in particular to a method and a device for processing a head and neck CT image of a nasopharyngeal carcinoma patient. Background Medical images play an important role in the treatment of cancer, and medical images of different modalities provide different information on cancerous regions. For example, computer scan imaging (Computed Tomography, CT) and structural nuclear magnetic resonance (Magnetic Resonance Imaging, MRI) show anatomical information of tissue organs, and positron emission computed tomography (Positron Emission Computed Tomography, PET) provides functional information of tissue organs. The CT image has the characteristics of high spatial resolution, high imaging speed and the like, and can clearly display the position, the size and the structure of the tumor in the diagnosis of the tumor. Nasopharyngeal carcinoma (Nasopharyngeal Carcinoma, NPC) is a malignancy that occurs in the nasopharyngeal cavity or upper pharyngeal cavity. When radiation therapy of nasopharyngeal carcinoma is performed, tissue and organs near a cancerous region are easily damaged. Therefore, the labeling of important organs of the head and neck of the nasopharyngeal carcinoma patient can effectively assist doctors in protecting the patient. The critical organs of nasopharyngeal carcinoma radiotherapy still depend on manual sketching on CT images, the sketching precision depends on experience and proficiency of doctors, great subjectivity exists, and in addition, the manual sketching speed is low, the labor cost is high, and the workload of doctors is great. Disclosure of Invention The invention aims to overcome the defects of the prior art, and provides a processing method and a processing device for segmenting a head and neck CT image of a nasopharyngeal carcinoma patient, so that important organs in the head and neck CT image can be segmented more accurately, and the problem that the traditional manual sketching of segmentation results is time-consuming and labor-consuming is solved. In order to achieve the above purpose, the technical scheme of the invention is as follows: in a first aspect, the present invention provides a method for segmenting vital organs in a head and neck CT image of a patient with nasopharyngeal carcinoma, said method comprising: inputting a head and neck CT image of a nasopharyngeal carcinoma patient into a slice screening model, and screening to obtain CT image slices containing important organs; inputting the CT image slice containing the important organ into a slice segmentation model to obtain a segmentation result, and completing segmentation of the important organ in the head and neck CT image of the nasopharyngeal carcinoma patient. Further, the slice screening model is obtained by: Preprocessing a head and neck CT image of a patient with nasopharyngeal carcinoma to be segmented, wherein the preprocessing comprises background removal and data prescribing; designing a pseudo tag for marking whether a single slice contains an important organ or not by using a real segmentation result; training the pseudo tag and the pretreated head and neck CT image of the nasopharyngeal carcinoma patient together to obtain a slice screening model. Further, the background removal includes: Judging whether the row or column in which the sliding window is positioned is background information from the upper direction, the lower direction, the left direction and the right direction of the CT image by utilizing the sliding window; When the sliding window moves to a non-background area, the window stops moving, the passed area is considered as redundant background information, and the information of the passed area is removed. Further, the data prescribing includes: Subtracting the average avg A of all voxels from each voxel value v i, and dividing by the standard deviation delta A to obtain a specified voxel value v i': further, the pseudo tag is a binary pseudo tag, 0 represents that the slice does not contain an important organ, and 1 represents that the slice contains an important organ. Further, the slice screening model is composed of a shallow convolutional neural network, the activation function of an output layer is a sigmoid function, when the output result is more than or equal to 0.5, the current slice is considered to contain important organs, and when the output result is less than 0.5, the current slice is considered to contain no important organs. Further, the slice segmentation model is obtained by: erasing partial area of the CT image after pretreatment, and matching the erased data with the complete CT image after pretreatment Building a U-Net full convolution segmentation network under a deep learning framework, using paired data, taking CT images of erased partial areas as inpu