Method for segmenting important organs in a head and neck ct image of a nasopharyngeal carcinoma patient

Abstract

The application discloses a processing method and device for segmenting a head and neck CT image of a nasopharyngeal carcinoma patient, the method comprising inputting the head and neck CT image of the nasopharyngeal carcinoma patient to a slice screening model to screen a CT image slice containing important organs; inputting the CT image slice containing the important organs to a slice segmentation model to obtain a segmentation result, and completing segmentation of the important organs in the head and neck CT image of the nasopharyngeal carcinoma patient. The slice screening model and the slice segmentation model built by using artificial intelligence technology are used to realize automatic segmentation of the important organs in the head and neck CT image of the nasopharyngeal carcinoma patient, thereby solving the problem of time and labor consumption of traditional manual delineation of a segmentation result, and not only can assist doctors in designing a diagnosis and treatment scheme before treatment, but also can reduce damage of rays to key organs in the process of radiotherapy.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and medical image segmentation, and in particular to a method and apparatus for processing head and neck CT images of nasopharyngeal carcinoma patients. Background Technology

[0002] Medical imaging plays a crucial role in cancer treatment, with different modalities providing varying information about cancerous areas. For example, computed tomography (CT) and magnetic resonance imaging (MRI) display anatomical information of tissues and organs, while positron emission tomography (PET) provides functional information. CT images, characterized by high spatial resolution and fast imaging speed, can clearly display the location, size, and structure of tumors in tumor diagnosis.

[0003] Nasopharyngeal carcinoma (NPC) is a malignant tumor that occurs in the nasopharynx or upper pharynx. During radiotherapy for NPC, damage to tissues and organs near the cancerous site is common. Therefore, marking important organs in the head and neck region of NPC patients can effectively assist doctors in protecting them. However, the key organs for NPC radiotherapy still rely on manual delineation on CT images. The accuracy of this delineation depends heavily on the physician's experience and skill, and is highly subjective. Furthermore, manual delineation is slow, labor-intensive, and places a heavy workload on doctors. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and apparatus for processing head and neck CT images of nasopharyngeal carcinoma patients to segment important organs more accurately, thus solving the problem of time-consuming and labor-intensive traditional manual segmentation.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, the method comprising:

[0007] Head and neck CT images of nasopharyngeal carcinoma patients are input into the slice selection model to obtain CT image slices containing important organs.

[0008] The CT image slices containing important organs are input into the slice segmentation model to obtain the segmentation results, thus completing the segmentation of important organs in the head and neck CT images of nasopharyngeal carcinoma patients.

[0009] Furthermore, the slice selection model is obtained in the following way:

[0010] Preprocessing of head and neck CT images of nasopharyngeal carcinoma patients to be segmented, including background removal and data specification;

[0011] We designed pseudo-labels based on real segmentation results to indicate whether a single slice contains important organs;

[0012] The pseudo-labels were trained together with preprocessed head and neck CT images of nasopharyngeal carcinoma patients to obtain a slice selection model.

[0013] Furthermore, the background removal includes:

[0014] Use a sliding window to determine whether the row or column in a CT image is background information from four directions: top, bottom, left, and right.

[0015] When the sliding window moves to a non-background area, the window stops moving, considering the area it has passed through as redundant background information, and removes the information from the area it has passed through.

[0016] Furthermore, the data specification includes:

[0017] For each voxel value v i Subtract the mean of all voxels (avg) A Divide by the standard deviation δ A The standardized voxel value v is obtained. i ':

[0018]

[0019] Furthermore, the pseudo-label is a binary pseudo-label, where 0 represents that the slice does not contain important organs, and 1 represents that the slice contains important organs.

[0020] Furthermore, the slice screening model consists of a shallow convolutional neural network, and the activation function of its output layer is the sigmoid function. When the output result is greater than or equal to 0.5, the current slice is considered to contain an important organ; when the output result is less than 0.5, the current slice is considered not to contain an important organ.

[0021] Furthermore, the slice segmentation model is obtained in the following way:

[0022] Partial areas of the preprocessed CT image were erased, and the erased data was paired with the preprocessed complete CT image.

[0023] A U-Net fully convolutional segmentation network was built within a deep learning framework. Paired data was used, with CT images of partially erased regions as input and the original data as labels, enabling the U-Net network to acquire image restoration capabilities and complete the model pre-training.

[0024] The last convolutional layer of the pre-trained model was modified to have the number of output channels equal to the number of organ types to be segmented. The model was then further trained until convergence using manually labeled real segmentation results and paired head and neck CT images of nasopharyngeal carcinoma patients to obtain the slice segmentation model.

[0025] Furthermore, the preprocessed CT image is erased by removing a 100*100 square key area.

[0026] In a second aspect, the present invention provides an apparatus for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0027] Thirdly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the methods described above.

[0028] Compared with the prior art, the advantages of this invention are as follows:

[0029] This invention utilizes artificial intelligence technology to build slice selection and segmentation models for automatic segmentation of vital organs in head and neck CT images of nasopharyngeal carcinoma patients, solving the problem of time-consuming and labor-intensive traditional manual segmentation. It not only assists doctors in designing treatment plans before treatment but also minimizes radiation damage to critical organs during radiotherapy. Attached Figure Description

[0030] Figure 1 This is a flowchart of a method for segmenting head and neck CT images of nasopharyngeal carcinoma patients, provided in Embodiment 1 of the present invention.

[0031] Figure 2 Flowchart for training the slice selection model;

[0032] Figure 3 Training flowchart for slice segmentation model training;

[0033] Figure 4 This is a schematic diagram of CT image data preprocessing;

[0034] Figure 5 This is a schematic diagram of the overall framework of the method of the present invention;

[0035] Figure 6 The results of partial parotid and submandibular gland segmentation obtained using the method of this invention;

[0036] Figure 7 This is a three-dimensional visualization of the segmentation results obtained using the method of this invention;

[0037] Figure 8 This is a schematic diagram of the processing device for segmenting head and neck CT images of nasopharyngeal carcinoma patients provided in Embodiment 1 of the present invention. Detailed Implementation

[0038] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] Example 1:

[0040] See Figure 1 The diagram shown is a flowchart of a method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, as provided in this embodiment, including the following steps:

[0041] 102. Input the head and neck CT images of nasopharyngeal carcinoma patients into the slice selection model to select CT image slices containing important organs; that is, in this step, the slice selection model selects CT image slices containing important organs and removes redundant information.

[0042] 102. Input CT image slices containing important organs into the slice segmentation model to obtain segmentation results, thus completing the segmentation of important organs in the head and neck CT images of nasopharyngeal carcinoma patients. In this step, the slice segmentation model is used to segment key organs in the selected CT slices, which can accurately segment the important organs.

[0043] Therefore, this method, which uses artificial intelligence to build slice selection and segmentation models, automatically segments important organs in head and neck CT images of nasopharyngeal carcinoma patients, solving the problem of time-consuming and labor-intensive traditional manual segmentation. It can not only assist doctors in designing treatment plans before treatment but also minimize radiation damage to critical organs during radiotherapy.

[0044] In one specific embodiment, such as Figure 2 As shown, the above slice filtering model is obtained in the following way:

[0045] 201. Preprocess the head and neck CT images of nasopharyngeal carcinoma patients to be segmented, such as... Figure 4As shown, this includes background removal and data specification to enable the network to better fit the data. Since CT images contain a large amount of redundant background information, background removal is performed using sliding windows in four directions: top, bottom, left, and right. The sliding windows move towards the center and stop when they touch a non-background area; the area traversed by the sliding window represents the background information to be removed. The voxel values ​​of CT images in a computer are hu values, which reflect the different photon absorption capabilities of different tissues. In a computer, the voxel values ​​of CT images have a wide distribution range, generally between -1024 and 2000. To enable the segmentation network to better fit the data, each voxel value is... i Subtract the mean of all voxels (avg) A Divide by the standard deviation δ A The standardized voxel value v is obtained. i ':

[0046]

[0047] 202. Design pseudo-labels to indicate whether a single slice contains important organs using actual segmentation results. Slices containing important organs are labeled with a "1" and slices not containing important organs are labeled with a "0" by manual annotation.

[0048] 203. The pseudo-labels were trained together with the preprocessed head and neck CT images of nasopharyngeal carcinoma patients to obtain the slice selection model.

[0049] The slice selection model consists of a shallow convolutional neural network with five convolutional layers and one fully connected layer. The activation function of its output layer is the sigmoid function. When the output result is greater than or equal to 0.5, the current slice is considered to contain an important organ; when the output result is less than 0.5, the current slice is considered not to contain an important organ.

[0050] In one specific embodiment, such as Figure 3 As shown, the slice segmentation model is obtained in the following way:

[0051] 301. Erasure a portion of the preprocessed CT image data and pair the erased data with the preprocessed complete CT image.

[0052] 302. A U-Net fully convolutional segmentation network is built under the deep learning framework. Paired data is used, with CT images of partially erased regions as input and the original data as labels, so that the U-Net network can obtain the ability to restore images and complete the pre-training of the model.

[0053] 303. Modify the last convolutional layer of the pre-trained model and set its output channel number to the number of the corresponding organ types to be segmented; further train the pre-trained model until convergence using manually labeled real segmentation results and paired head and neck CT images of nasopharyngeal carcinoma patients to obtain the slice segmentation model.

[0054] like Figure 5 The diagram shown is a schematic representation of the overall framework of the method of the present invention. Figure 6 This is the partial parotid and submandibular gland segmentation result obtained using the method of this invention. Figure 7 This is a three-dimensional visualization of the segmentation result obtained using the method of this invention. Figure 6 and Figure 7 It can be seen that the method of the present invention can accurately segment important organs, solving the problem of time-consuming and labor-intensive traditional manual segmentation.

[0055] Example 2:

[0056] See Figure 8 As shown, the apparatus for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients provided in this embodiment includes a processor 81, a memory 82, and a computer program 83 stored in the memory 82 and executable on the processor 81, such as a program for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients. When the processor 81 executes the computer program 83, it implements the steps of Embodiment 1 described above, for example... Figure 1 The steps are shown.

[0057] For example, the computer program 83 may be divided into one or more modules / units, which are stored in the memory 82 and executed by the processor 81 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 83 in the apparatus for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients.

[0058] The device for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients can be a desktop computer, laptop, handheld computer, or cloud server, etc. The device may include, but is not limited to, a processor 81 and a memory 82. Those skilled in the art will understand that... Figure 8This is merely an example of a device for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, and does not constitute a limitation on the device for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the device for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients may also include input / output devices, network access devices, buses, etc.

[0059] The processor 81 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0060] The memory 82 can be an internal storage unit of the device for segmenting important organs in CT images of the head and neck of a patient with nasopharyngeal carcinoma, such as a hard disk or memory. The memory 82 can also be an external storage device of the device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 82 can include both internal storage units and external storage devices. The memory 82 is used to store the computer program and other programs and data required by the device for segmenting important organs in CT images of the head and neck of a patient with nasopharyngeal carcinoma. The memory 82 can also be used to temporarily store data that has been output or will be output.

[0061] Example 3:

[0062] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0063] The computer-readable medium shown can be any means that can contain, store, communicate, propagate, or transmit a program for use in or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, for example, by optically scanning the paper or other medium, then editing, interpreting, or otherwise processing it as necessary to obtain the program electronically, and then storing it in computer memory.

[0064] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the content of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, characterized in that, The method includes: Head and neck CT images of nasopharyngeal carcinoma patients are input into the slice selection model to obtain CT image slices containing important organs; CT image slices containing important organs are input into the slice segmentation model to obtain segmentation results, thus completing the segmentation of important organs in the head and neck CT images of nasopharyngeal carcinoma patients. The slice selection model is obtained in the following way: Preprocessing of head and neck CT images of nasopharyngeal carcinoma patients to be segmented, including background removal and data specification; We designed pseudo-labels based on real segmentation results to indicate whether a single slice contains important organs; The pseudo-labels were trained together with preprocessed head and neck CT images of nasopharyngeal carcinoma patients to obtain a slice selection model. The slice segmentation model is obtained in the following way: Partial areas of the preprocessed CT image were erased, and the erased data was paired with the preprocessed complete CT image. A U-Net fully convolutional segmentation network was built within a deep learning framework. Paired data was used, with CT images of partially erased regions as input and the original data as labels, enabling the U-Net network to acquire image restoration capabilities and complete the model pre-training. The last convolutional layer of the pre-trained model was modified, and its output channel number was set to the number of the corresponding organ types to be segmented. The model was further trained until convergence using manually labeled real segmentation results and paired head and neck CT images of nasopharyngeal carcinoma patients to obtain the slice segmentation model. The slice screening model consists of a shallow convolutional neural network, and the activation function of its output layer is the sigmoid function. When the output result is greater than or equal to 0.5, the current slice is considered to contain an important organ; when the output result is less than 0.5, the current slice is considered not to contain an important organ. The preprocessed CT image is erased by removing a 100*100 square key area.

2. The method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients as described in claim 1, characterized in that, The background removal includes: Use a sliding window to determine whether the row or column in a CT image is background information from four directions: top, bottom, left, and right. When the sliding window moves to a non-background area, the window stops moving, considering the area it has passed through as redundant background information, and removes the information from the area it has passed through.

3. The method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients as described in claim 2, characterized in that, The data specification includes: Each voxel value Subtract the mean of all voxels Divide by standard deviation The standardized voxel values ​​are obtained. : 。 4. The method for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients as described in claim 1, characterized in that, The pseudo-label is a binary pseudo-label, where 0 represents that the slice does not contain important organs, and 1 represents that the slice contains important organs.

5. An apparatus for segmenting important organs in head and neck CT images of nasopharyngeal carcinoma patients, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 4.

6. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 4.

Images