CN112233058A - Method for detecting lymph nodes in head and neck CT image - Google Patents
Method for detecting lymph nodes in head and neck CT image Download PDFInfo
- Publication number
- CN112233058A CN112233058A CN201910636816.6A CN201910636816A CN112233058A CN 112233058 A CN112233058 A CN 112233058A CN 201910636816 A CN201910636816 A CN 201910636816A CN 112233058 A CN112233058 A CN 112233058A
- Authority
- CN
- China
- Prior art keywords
- lymph node
- image
- segmentation
- neck
- head
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Abstract
The invention discloses a lymph node detection method in a head and neck CT image, which adopts a method of firstly segmenting lymph node partitions and then identifying segmented lymph nodes from the partitions, fully utilizes the prior knowledge of the head and neck lymph node partitions, and reduces the interference of muscle tissues outside the lymph node partitions and other tissues with similar lymph node density on the lymph node partitions; based on the condition that the lymph node morphology is relatively smooth and the change between CT image layers is not obvious, a three-dimensional convolution kernel is adopted in lymph node identification and segmentation, and the segmentation accuracy is improved by means of complementary information between image layers; the interference of muscle tissues on lymph node identification and segmentation can be avoided, and therefore the accuracy of lymph node segmentation is effectively improved.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a lymph node detection method in a head and neck CT image.
Background
The imaging by the CT technique is widely used in the field of clinical medicine because it can well display organs made of soft tissue and show lesions on the background of anatomical images. Accurate segmentation of target regions from CT images is a key step in computer-aided diagnosis and surgical planning, and inspired by deep learning and convolutional neural networks to succeed in computer image understanding and analysis, researchers have applied them to the processing of medical images such as CT/MR, and have obtained good results in the segmentation of various tissues and organs. For example, Dou Q et al, using a 3D convolutional neural network, achieved the identification and segmentation of lung nodules in LDCT images (Dou Q, Chen H, Jin Y, et al, automated Pulmonary node Detection via 3D Convnets with Online Sample Filtering and Hybrid-Loss resolution leading [ C ]// International Conference on Medical Image Computing & Computer-assisted interaction in spring, Cham,2017.), HuP et al, also achieved the segmentation of liver in CT images using a 3D convolutional neural network (Hu P, Wu F, Pen J, et al, automated 3D volumetric segmentation based segmentation and segmentation of liver in CT images (bag J, Sample J, study J. automated 3D volumetric segmentation of liver segmentation and segmentation of prostate tissue in prostate J, 8676, et al, achieved by using an N-K convolutional network, (Sau Q et al, tissue K2016, K76, et al, segmentation of lesion regions in Nasopharyngeal carcinoma CT Images using deep deconvolution Neural networks (Kuo M, Xinyuan C, Ye Z, et al. deep deconvolution Neural networks for Target Segmentation of Nasorgenomic Cancer in planar Computed Tomography [ J ]. Frontiers in organic in science, 2017,7: 315-), and so on.
Exploration and identification of cervical lymph nodes is extremely important and challenging. Healthy lymph nodes can help the lymph flow and filter foreign substances and initiate necessary immune responses, however, these physiological activities, which are important to the human body, form the physiological basis for the lymph nodes to be susceptible to various diseases. Because the pathological types of the lymph node lesions are various, the distribution is wide and various, the morbidity is high, and the head and neck anatomical structures are fine and complex, the characteristic performance of various image inspection methods at each disease stage of various common pathological types of the neck lymph nodes needs to be known by radiologists engaged in differential diagnosis of the neck lymph lesions by keeping the head and neck anatomical structures, the distribution conditions of the neck lymph nodes and partition standards in mind, mastering lymph reflux passages. Therefore, even a specialist radiologist working on the cranial nerves, considers it difficult to accurately diagnose lesions in the cervical lymph nodes. Although the deep learning has been primarily applied to medical image segmentation and has been successful in various tissues, due to the characteristics of various forms, complex distribution and difficult differentiation from other tissues, no report on the head and neck CT lymph node identification segmentation by using the deep learning technology is available at present.
Although there are many target region segmentation convolutional neural network architectures in the existing CT image based on deep learning, the basic ideas are consistent, and the target region segmentation convolutional neural network architectures all comprise the steps of data input, convolutional calculation feature extraction, ReLU excitation nonlinear processing, pooled data compression, full-connection classification and the like, and finally end-to-end direct output is realized, namely the CT image is input, the segmentation target is output, and the whole process is completed in one step.
For the targets of lung nodules and livers which are obviously different from the density and the form of background tissues and organs in the CT image, the effect of performing one-step in-place segmentation by adopting the convolutional neural network is better. However, the density of lymph nodes is similar to that of muscles, the CT value is between 20 and 50Hu, in addition, the shape and the size of the lymph nodes at different positions are greatly different, which brings difficulty for segmentation, and the effect of direct segmentation by adopting a convolutional neural network in one step is not good. Therefore, how to distinguish lymph nodes from muscle tissue is significant for the correct segmentation of lymph nodes in CT images. About 300 lymph nodes out of about 800 lymph nodes are located in the neck, and in anatomy, the lymph nodes are clustered and distributed in groups and chains in the neck at specific anatomical positions. Currently, VII divisions commonly defined by the 2009 American Thyroid Association (ATA) surgical group, the American Association of Endocrinologists (AAES), the american academy of otorhinolaryngology-head and neck surgery (AAOHNS), and the american Association of Head and Neck (AHNS) are widely used clinically for cervical lymph node division. Briefly, region i (Level i) includes the lymph nodes under the chin and under the jaw, region ii (Level ii) is the upper group of the lymph nodes in the internal jugular vein, region iii (Level iii) is the middle group of the lymph nodes in the internal jugular vein, region iv (Level iv) is the lower group of the lymph nodes in the internal jugular vein, region v (Level v) includes the upper group of the lymph nodes in the posterior triangular region of the neck and the supraclavicular region, region vi (Level vi) is also called the central lymph node, and includes the anterior roar lymph node, the anterior tracheal lymph node, the paratracheal lymph node, the perithyroid lymph node and the retropharyngeal lymph node, and region VII (Level VII) is the upper longitudinal compartment from the upper edge of the sternum to the upper edge of the aortic arch. Examples of the regions in the CT image are shown in fig. 1: it can be seen that the distribution positions of the head and neck lymph nodes are regular, and the distribution areas of the lymph nodes can be used as a priori knowledge to guide the identification and segmentation of the neck lymph nodes.
Disclosure of Invention
In view of the above, the present invention provides a method for lymph node detection in head and neck CT images to solve the deficiencies of the prior art.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a method for detecting lymph nodes in head and neck CT images is provided, which comprises the following steps:
s1, lymph node partition identification and segmentation in CT images: identifying and segmenting each lymph node partition in the head and neck CT image by adopting a convolutional neural network;
s1.1, data acquisition: taking the CT image of the head and neck of the patient as input data;
s1.2, feature extraction: calculating and extracting high-dimensional characteristic vectors by using a full convolution neural network comprising a plurality of convolution layers through convolution kernels with specific sizes, and then obtaining a result through pooling layer compression, wherein the architecture and parameters in a network model are obtained by training a large amount of data;
s1.3, image segmentation: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; simultaneously introducing a jump structure, and performing deconvolution on feature layers of different levels simultaneously and then fusing;
s2, identifying and segmenting the lymph nodes in the subareas: adopting a convolutional neural network to identify and partition a single lymph node in a lymph node partition of the head and neck CT image;
s2.1, data acquisition: using the lymph node partition information divided in step S1 as input data;
s2.2, feature extraction: using a full convolution neural network, including a plurality of convolution layers, extracting high-dimensional characteristic vectors through three-dimensional convolution kernels, wherein the architecture and parameters in the network model are obtained by training a large amount of data;
s2.3, image segmentation: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; and simultaneously introducing a jump structure, and performing deconvolution and fusion on the feature layers of different levels at the same time.
The method for detecting the lymph nodes in the head and neck CT image comprises the following steps of U-Net, R-CNN, FCN and DenseNet.
The technical scheme of the invention has the beneficial effects that:
the method of partitioning lymph node partitions firstly and then identifying and partitioning lymph nodes from the partitions is adopted, the priori knowledge of head and neck lymph node partitions is fully utilized, and the interference of muscle tissues outside the lymph node partitions and other tissues with similar lymph node density on lymph node partition is reduced; based on the condition that the lymph node morphology is relatively smooth and the change between CT image layers is not obvious, a three-dimensional convolution kernel is adopted in lymph node identification and segmentation, and the segmentation accuracy is improved by means of complementary information between image layers; the interference of muscle tissues on lymph node identification and segmentation can be avoided, and therefore the accuracy of lymph node segmentation is effectively improved.
Drawings
FIG. 1 is a diagram illustrating a prior art segmentation in a CT image;
FIG. 2 is a schematic flow chart of the method of the present invention.
Detailed Description
The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.
Referring to fig. 2, the present invention adopts a two-step method, that is, the first step is to outline the lymph node partition position in the head and neck CT image, and the second step is to perform lymph node segmentation in the lymph node partition, which includes the following steps:
the first step is as follows: lymph node partition identification and segmentation in CT images: and (3) adopting a convolution neural network including but not limited to U-Net, R-CNN, FCN, DenseNet and the like to identify and segment each lymph node partition in the head and neck CT image.
1.1 data acquisition step: taking the CT image of the head and neck of the patient as input data.
1.2 feature extraction step: the method comprises the steps of using a full convolution neural network, including a plurality of convolution layers, calculating and extracting high-dimensional characteristic vectors through convolution kernels with specific sizes, compressing results through a pooling layer to save calculated amount, and obtaining the framework and parameters in a network model through training a large amount of data, so that the result obtained through the model is reasonable, and certain scientific basis is provided.
1.3 image segmentation step: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; and simultaneously, a jump structure is introduced, and the feature layers of different levels are subjected to deconvolution simultaneously and then are fused so as to improve the segmentation accuracy.
The second step is that: identification and segmentation of intraregional lymph nodes: and (3) adopting a convolution neural network including but not limited to U-Net, R-CNN, FCN, DenseNet and the like to identify and segment single lymph nodes in the lymph node partition of the head and neck CT image.
2.1 data acquisition step: the lymph node partition information segmented in the first step is used as input data.
2.2 feature extraction: the full convolution neural network is used, the full convolution neural network comprises a plurality of convolution layers and extracts high-dimensional feature vectors through three-dimensional convolution kernels, and the architecture and parameters in the network model are obtained by training a large amount of data, so that the result obtained through the model is reasonable, and certain scientific basis is provided.
2.3 image segmentation step: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; and simultaneously, a jump structure is introduced, and the feature layers of different levels are subjected to deconvolution simultaneously and then are fused so as to improve the segmentation accuracy.
The invention adopts the method of segmenting the lymph node partition firstly and then identifying and segmenting the lymph node from the partition, fully utilizes the prior knowledge of the head and neck lymph node subsection, and reduces the interference of muscle tissues outside the lymph node partition and other tissues with similar lymph node density to the lymph node partition; based on the condition that the lymph node morphology is relatively smooth and the change between CT image layers is not obvious, a three-dimensional convolution kernel is adopted in lymph node identification and segmentation, and the segmentation accuracy is improved by means of complementary information between image layers; the interference of muscle tissues on lymph node identification and segmentation can be avoided, and therefore the accuracy of lymph node segmentation is effectively improved.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.
Claims (2)
1. A method for lymph node detection in head and neck CT images, comprising:
s1, lymph node partition identification and segmentation in CT images: identifying and segmenting each lymph node partition in the head and neck CT image by adopting a convolutional neural network;
s1.1, data acquisition: taking the CT image of the head and neck of the patient as input data;
s1.2, feature extraction: calculating and extracting high-dimensional characteristic vectors by using a full convolution neural network comprising a plurality of convolution layers through convolution kernels with specific sizes, and then obtaining a result through pooling layer compression, wherein the architecture and parameters in a network model are obtained by training a large amount of data;
s1.3, image segmentation: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; simultaneously introducing a jump structure, and performing deconvolution on feature layers of different levels simultaneously and then fusing;
s2, identifying and segmenting the lymph nodes in the subareas: adopting a convolutional neural network to identify and partition a single lymph node in a lymph node partition of the head and neck CT image;
s2.1, data acquisition: using the lymph node partition information divided in step S1 as input data;
s2.2, feature extraction: using a full convolution neural network, including a plurality of convolution layers, extracting high-dimensional characteristic vectors through three-dimensional convolution kernels, wherein the architecture and parameters in the network model are obtained by training a large amount of data;
s2.3, image segmentation: performing up-sampling on the feature map by using deconvolution to restore the feature map to the size of the original input image, thereby completing the prediction of each pixel point in the image; and simultaneously introducing a jump structure, and performing deconvolution and fusion on the feature layers of different levels at the same time.
2. The method of claim 1, wherein the convolutional neural network comprises U-Net, R-CNN, FCN and DenseNet.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636816.6A CN112233058A (en) | 2019-07-15 | 2019-07-15 | Method for detecting lymph nodes in head and neck CT image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910636816.6A CN112233058A (en) | 2019-07-15 | 2019-07-15 | Method for detecting lymph nodes in head and neck CT image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112233058A true CN112233058A (en) | 2021-01-15 |
Family
ID=74111177
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910636816.6A Pending CN112233058A (en) | 2019-07-15 | 2019-07-15 | Method for detecting lymph nodes in head and neck CT image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112233058A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113520437A (en) * | 2021-07-07 | 2021-10-22 | 上海中医药大学 | Cervical CT scanning image-based percutaneous positioning method for cricopharyngeal muscle |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106940816A (en) * | 2017-03-22 | 2017-07-11 | 杭州健培科技有限公司 | Connect the CT image Lung neoplasm detecting systems of convolutional neural networks entirely based on 3D |
| CN107103187A (en) * | 2017-04-10 | 2017-08-29 | 四川省肿瘤医院 | The method and system of Lung neoplasm detection classification and management based on deep learning |
| CN109035263A (en) * | 2018-08-14 | 2018-12-18 | 电子科技大学 | Brain tumor image automatic segmentation method based on convolutional neural networks |
| CN109009110A (en) * | 2018-06-26 | 2018-12-18 | 东北大学 | Axillary lymphatic metastasis forecasting system based on MRI image |
| CN109376756A (en) * | 2018-09-04 | 2019-02-22 | 青岛大学附属医院 | Upper abdomen metastatic lymph node section automatic recognition system, computer equipment, storage medium based on deep learning |
| CN109636807A (en) * | 2018-11-27 | 2019-04-16 | 宿州新材云计算服务有限公司 | A kind of grape disease blade split plot design of image segmentation and pixel recovery |
| CN109685811A (en) * | 2018-12-24 | 2019-04-26 | 北京大学第三医院 | PET/CT hypermetabolism lymph node dividing method based on dual path U-net convolutional neural networks |
| CN109934832A (en) * | 2019-03-25 | 2019-06-25 | 北京理工大学 | Liver neoplasm dividing method and device based on deep learning |
| WO2019134802A1 (en) * | 2018-01-03 | 2019-07-11 | Signify Holding B.V. | System and methods to share machine learning functionality between cloud and an iot network |
-
2019
- 2019-07-15 CN CN201910636816.6A patent/CN112233058A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106940816A (en) * | 2017-03-22 | 2017-07-11 | 杭州健培科技有限公司 | Connect the CT image Lung neoplasm detecting systems of convolutional neural networks entirely based on 3D |
| CN107103187A (en) * | 2017-04-10 | 2017-08-29 | 四川省肿瘤医院 | The method and system of Lung neoplasm detection classification and management based on deep learning |
| WO2019134802A1 (en) * | 2018-01-03 | 2019-07-11 | Signify Holding B.V. | System and methods to share machine learning functionality between cloud and an iot network |
| CN109009110A (en) * | 2018-06-26 | 2018-12-18 | 东北大学 | Axillary lymphatic metastasis forecasting system based on MRI image |
| CN109035263A (en) * | 2018-08-14 | 2018-12-18 | 电子科技大学 | Brain tumor image automatic segmentation method based on convolutional neural networks |
| CN109376756A (en) * | 2018-09-04 | 2019-02-22 | 青岛大学附属医院 | Upper abdomen metastatic lymph node section automatic recognition system, computer equipment, storage medium based on deep learning |
| CN109636807A (en) * | 2018-11-27 | 2019-04-16 | 宿州新材云计算服务有限公司 | A kind of grape disease blade split plot design of image segmentation and pixel recovery |
| CN109685811A (en) * | 2018-12-24 | 2019-04-26 | 北京大学第三医院 | PET/CT hypermetabolism lymph node dividing method based on dual path U-net convolutional neural networks |
| CN109934832A (en) * | 2019-03-25 | 2019-06-25 | 北京理工大学 | Liver neoplasm dividing method and device based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 秦品乐等: "基于级联全卷积神经网络的颈部淋巴结自动识别算法", 《计算机应用》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113520437A (en) * | 2021-07-07 | 2021-10-22 | 上海中医药大学 | Cervical CT scanning image-based percutaneous positioning method for cricopharyngeal muscle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hu et al. | Parallel deep learning algorithms with hybrid attention mechanism for image segmentation of lung tumors | |
| CN107230206B (en) | Multi-mode data-based 3D pulmonary nodule segmentation method for hyper-voxel sequence lung image | |
| CN106056595B (en) | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules | |
| Ji et al. | A level-set based approach for anterior teeth segmentation in cone beam computed tomography images | |
| WO2021115312A1 (en) | Method for automatically sketching contour line of normal organ in medical image | |
| CN109615636A (en) | Vascular tree building method, device in the lobe of the lung section segmentation of CT images | |
| CN105447872A (en) | Method for automatically identifying liver tumor type in ultrasonic image | |
| CN106133790A (en) | The method and apparatus generating one or more computed tomography image based on magnetic resonance image (MRI) with the help of separating in tissue types | |
| Xu et al. | Convolutional-neural-network-based approach for segmentation of apical four-chamber view from fetal echocardiography | |
| CN106651874B (en) | Space domain splitting method after brain tumor surgery based on multi-modal MRI data | |
| CN111340825A (en) | Method and system for generating mediastinal lymph node segmentation model | |
| CN109498046A (en) | The myocardial infarction quantitative evaluating method merged based on nucleic image with CT coronary angiography | |
| Gao et al. | Accurate lung segmentation for X-ray CT images | |
| Metlek et al. | ResUNet+: A new convolutional and attention block-based approach for brain tumor segmentation | |
| CN109215035B (en) | Brain MRI hippocampus three-dimensional segmentation method based on deep learning | |
| Hao et al. | Magnetic resonance image segmentation based on multi-scale convolutional neural network | |
| Ramos et al. | Fast and smart segmentation of paraspinal muscles in magnetic resonance imaging with CleverSeg | |
| CN112233058A (en) | Method for detecting lymph nodes in head and neck CT image | |
| CN113539476A (en) | Stomach endoscopic biopsy Raman image auxiliary diagnosis method and system based on artificial intelligence | |
| CN113570724B (en) | Grid sphere corner-protection parameterization method based on inverse sphere projection and application thereof | |
| CN110163847A (en) | Liver neoplasm dividing method and device based on CT/MR image | |
| CN103886580A (en) | Tumor image processing method | |
| CN104484874B (en) | Living animal lower limb vascular dividing method based on CT contrast imagings | |
| Yan et al. | Segmentation of pulmonary parenchyma from pulmonary CT based on ResU-Net++ model | |
| Lewis et al. | The role of anatomical context in soft‐tissue multi‐organ segmentation of cadaveric non‐contrast‐enhanced whole body CT |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210115 |
|
| RJ01 | Rejection of invention patent application after publication |