CN115035019A - Neck lymph node analysis device based on convolutional neural network - Google Patents
Neck lymph node analysis device based on convolutional neural network Download PDFInfo
- Publication number
- CN115035019A CN115035019A CN202210293141.1A CN202210293141A CN115035019A CN 115035019 A CN115035019 A CN 115035019A CN 202210293141 A CN202210293141 A CN 202210293141A CN 115035019 A CN115035019 A CN 115035019A
- Authority
- CN
- China
- Prior art keywords
- layer
- lymph node
- neural network
- convolutional neural
- image
- 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
- 210000001165 lymph node Anatomy 0.000 title claims abstract description 70
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 38
- 238000004458 analytical method Methods 0.000 title claims abstract description 25
- 210000002751 lymph Anatomy 0.000 claims abstract description 17
- 238000002604 ultrasonography Methods 0.000 claims abstract description 6
- 230000008034 disappearance Effects 0.000 claims abstract description 5
- 238000000605 extraction Methods 0.000 claims abstract description 5
- 238000010606 normalization Methods 0.000 claims description 25
- 230000004913 activation Effects 0.000 claims description 17
- 239000000654 additive Substances 0.000 claims description 13
- 230000000996 additive effect Effects 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 5
- 238000007634 remodeling Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 230000007797 corrosion Effects 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 3
- 230000009849 deactivation Effects 0.000 claims description 3
- 230000010339 dilation Effects 0.000 claims description 2
- 230000004807 localization Effects 0.000 claims 1
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 230000001926 lymphatic effect Effects 0.000 description 14
- 238000012549 training Methods 0.000 description 10
- 238000000034 method Methods 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 5
- 210000004027 cell Anatomy 0.000 description 4
- 230000000877 morphologic effect Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000002708 enhancing effect Effects 0.000 description 3
- 230000001788 irregular Effects 0.000 description 3
- 230000003902 lesion Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 241000894006 Bacteria Species 0.000 description 1
- 241000700605 Viruses Species 0.000 description 1
- 230000017531 blood circulation Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 210000000987 immune system Anatomy 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 230000003211 malignant effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000001394 metastastic effect Effects 0.000 description 1
- 206010061289 metastatic neoplasm Diseases 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/10132—Ultrasound image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- 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
-
- 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
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a neck lymph node analysis device based on a convolution neural network, which comprises: an image acquisition module: for acquiring an ultrasound image with a cervical lymph node; a lymph node image extraction module: the neck lymph node image acquisition device is used for intercepting the neck lymph node in the ultrasonic image to obtain an interested neck lymph node image; a lymph node analysis module: and the system is used for inputting the neck lymph node image of interest into a convolutional neural network XDNetV2-C-NLH and determining the lymph gate structure of the neck lymph node, wherein the lymph gate structure comprises normal, eccentric or disappearance. The invention can effectively detect the lymph gate structure of the cervical lymph node.
Description
Technical Field
The invention relates to the technical field of auxiliary medical diagnosis, in particular to a neck lymph node analysis device based on a convolutional neural network.
Background
Lymph nodes are a part of the human immune system and help the body to fight the invasion of foreign bacteria, viruses, etc. The cervical lymph nodes can be affected by various etiological or metastatic lesions, and the changes in the structure of the lymphatic gates can be caused by the lesions of the cervical lymph nodes. Therefore, whether the lymphatic structure is normal or not is an important basis for judging whether the lesion is benign or malignant and for etiological diagnosis.
Ultrasound scanning is the most common clinical diagnostic means for cervical lymph nodes because of its low cost (compared to CT, MRI, etc.) and small damage to the body. The existing diagnosis methods are that doctors operate ultrasonic equipment and simultaneously carry out manual identification and judgment on neck lymph node ultrasonic images, and have the following defects: the diagnosis speed is slow, and the efficiency is low; the energy and physical strength of doctors are greatly consumed; and has high requirements on the experience level of individual doctors.
Disclosure of Invention
The invention aims to provide a neck lymph node analysis device based on a convolutional neural network, which can effectively detect a lymph gate structure of a neck lymph node.
The technical scheme adopted by the invention for solving the technical problems is as follows: provided is a neck lymph node analysis apparatus based on a convolutional neural network, including:
an image acquisition module: for acquiring an ultrasound image with a cervical lymph node;
a lymph node image extraction module: the neck lymph node image acquisition device is used for intercepting the neck lymph node in the ultrasonic image to obtain an interested neck lymph node image;
lymph node analysis module: and the system is used for inputting the neck lymph node image of interest into a convolutional neural network XDNetV2-C-NLH and determining the lymph gate structure of the neck lymph node, wherein the lymph gate structure comprises normal, eccentric or disappearance.
The convolutional neural network XDNetV2-C-NLH comprises 1 first characteristic unit, 4 second characteristic units, 2 third characteristic units and 1 fourth characteristic unit, wherein the first characteristic unit is sequentially connected with the 2 second characteristic units, the 2 fifth characteristic units and the 1 fourth characteristic unit, and the other 2 second characteristic units are connected with the third characteristic unit to form the 2 fifth characteristic units.
The first feature cell includes a depth separable convolutional layer, a batch normalization layer, an active layer, a convolutional layer, an active layer, a layer normalization layer, an attention module, and an additive layer, which are sequentially connected, and an output of the 2 nd active layer is connected with the additive layer.
The second characteristic unit comprises a convolution layer, an active layer, a layer normalization layer, an attention module, an addition layer, a layer normalization layer, an attention module and an addition layer which are sequentially connected, wherein the output of the active layer is connected with the 1 st addition layer, and the output of the 1 st addition layer is connected with the 2 nd addition layer.
The third feature cell includes an active layer, a batch normalization layer, a depth-separable convolution layer, and an additive layer, which are connected in sequence.
The fourth characteristic unit comprises a batch normalization layer, an activation layer, a global mean pooling layer, an inactivation layer, a full-link layer and an activation layer which are sequentially connected.
The attention module comprises a relative position coding layer and 3 depth separable convolution layers, wherein the 2 nd depth separable layer and the 3 rd depth separable layer are subjected to point multiplication to obtain a first point multiplication result, the 3 rd depth separable layer and the relative position coding layer are subjected to point multiplication to obtain a second point multiplication result, the first point multiplication result and the second point multiplication result are added and then are connected with the attention activation layer, and the output of the activation layer and the 1 st depth separable layer are subjected to point multiplication and then are connected with the remodeling layer.
The convolutional neural network XDNetV2-C-NLH adopts weighted classification cross entropy as a loss function, and the formula of the weighted classification cross entropy is as follows:
wherein L represents the prediction loss of the convolutional neural network XDNetV2-C-NLH, N represents the sample size, K represents the classification number value of the convolutional neural network XDNetV2-C-NLH output layer, and w j Represents a weight of the jth class; y is ij Representing that the ith sample corresponds to the actual value of the jth class, and taking the value of 1 or 0;
the ith sample representing the convolutional neural network XDNetV2-C-NLH corresponds to the predicted value of the jth class.
Still include the orientation module: when the lymphatic gate structure is predicted to be normal or eccentric, the lymphatic gate structure is localized by connected domain analysis, open-close operation, swelling corrosion or threshold comparison.
Further comprising a verification module: the system is used for constructing a rectangular coordinate system according to the glottic structure of the cervical lymph node in the cervical lymph node image of interest, calculating the distance of the positioned glottic structure from the central point of the rectangular coordinate system, and determining whether the glottic structure is normal or eccentric according to the distance of the rectangular coordinate system from the central point.
Advantageous effects
Due to the adoption of the technical scheme, compared with the prior art, the invention has the following advantages and positive effects: the convolutional neural network XDNetV2-C-NLH constructed by the method can quickly judge whether the cervical lymph node is normal, eccentric or disappeared, and has the characteristics of high analysis speed, high efficiency, high accuracy and stable performance; the development of the convolutional neural network XDNetV2-C-NLH network constructed by the method belongs to small sample training, the light network can effectively avoid the over-fitting problem of the small sample training, and the generalization capability is improved; the invention can assist the diagnosis of doctors, reduce the workload and effectively relieve the problem of medical resource shortage.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a schematic diagram of a convolutional neural network architecture of an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an attention module according to an embodiment of the present invention;
fig. 4 is a schematic diagram of verifying whether the lymphatic gate structure is normal or eccentric in an embodiment of the present invention.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention can be made by those skilled in the art after reading the teaching of the present invention, and these equivalents also fall within the scope of the claims appended to the present application.
The embodiment of the present invention relates to a neck lymph node analysis device based on a convolutional neural network, please refer to fig. 1, which includes:
an image acquisition module: for acquiring an ultrasound image with a cervical lymph node;
a lymph node image extraction module: the neck lymph node image acquisition device is used for intercepting the neck lymph node in the ultrasonic image to obtain an interested neck lymph node image;
a lymph node analysis module: and inputting the neck lymph node image of interest into a convolutional neural network XDNetV2-C-NLH, and determining the lymph gate structure of the neck lymph node, wherein the lymph gate structure is a blood circulation structure and comprises normal, eccentric or disappearance.
The present embodiment may further include:
a positioning module: locating a lymph gate structure by connected domain analysis, open and close operations, dilation corrosion or threshold comparison when the lymph gate structure is predicted to be normal or eccentric.
A verification module: and the system is used for constructing a rectangular coordinate system according to the glottic structure of the cervical lymph node in the cervical lymph node image of interest, calculating the distance of the positioned glottic structure from the central point of the rectangular coordinate system, and determining whether the glottic structure is normal or eccentric according to the distance of the rectangular coordinate system from the central point.
The present embodiment is described in detail below:
1. training data composition
Tens of thousands of desensitized cervical lymph node ultrasound images were obtained from multiple hospitals. A plurality of senior capital officers mark data and verify and confirm, and the marked contents are as follows: delineating contours, lymphatic portal regions, and lymphatic portal structure types; there are 3 classes of the lymphatic portal structural types: normal, eccentric, disappearance; the classification is obtained through the study of a plurality of senior capital officers, and the 3 types are considered to have typical clinical significance and meet the clinical diagnosis requirement.
2. Data Preprocessing (Data Preprocessing)
Obtaining irregular polygon outlines of the nodules and circumscribed rectangles of the irregular polygon outlines according to the drawing results of doctors; according to the rectangle cutting original image, cutting out a region of interest (ROI), and only keeping pixel values within the polygon, and setting all the other pixel values to be zero; the ROI was then resized to (160, 3) and normalized.
3. Training data enhancement (DataAugmentation)
And enhancing the training data and improving the robustness.
The data enhancement method used comprises: random rotation (random rotation), random lateral shift (random width shift), random vertical shift (random height shift), random brightness shift (random brightness shift), random cross-cut transform (random shear), random zoom (random zoom), random channel shift (random channel shift), random lateral flip (random width flip), and random vertical flip (random height flip).
4. Network architecture (network architecture)
In the lymph node analysis module, the present embodiment uses a convolutional neural network XDNetV2-C-NLH, which is a semi-attention convolutional neural network (semi-self-attention CNN), and is analyzed for the task target, i.e., the lymphatic gate structure of the cervical lymph nodes, as shown in detail in fig. 2.
Further, the partial structure used by the convolutional neural network XDNetV2-C-NLH includes:
A) a convolution layer (convolution layer);
B) a depth-wise separable convolution layer (depth-wise separable convolution layer);
C) batch normalization layer (batch normalization layer): is the normalization of a single neuron between a batch of training data;
D) layer normalization layer (layer normalization layer): is the normalization of a single training data to all neurons in a certain layer;
E) activation layer (activation layer, swish and softmax);
swish function formula:
softmax function formula:
wherein e is a natural constant; x is the number of i Representing the ith element of the input.
F) Relative Position encoding (Relative Position encoding): common convolutional layers extract image features of pixel intensity (i.e., color) and edge contour; the attention module uses relative position coding to obtain the relative position relation between pixels, thereby extracting the characteristics of shape structure and the like;
G) adding (add);
H) dot product (dotproduct);
I) a remodeling layer (reshape layer);
J) a global averaging pooling layer (global averaging pooling layer);
K) deactivation layer (dropout layer): the method is used for improving generalization ability and preventing overfitting;
l) fully connected layer (full connected layer).
Further, the local structures are spliced to obtain 1 first feature unit, 4 second feature units, 2 third feature units and 1 fourth feature unit, wherein the first feature unit is sequentially connected with the 2 second feature units, the 2 fifth feature units and the 1 fourth feature unit, and the other 2 second feature units are connected with the third feature unit to form the 2 fifth feature units. The first characteristic unit and the second characteristic unit have similar functions, mainly use an attention module, have stronger image characteristic capacity compared with a simple convolution layer, and realize down-sampling through convolution with the step length of 2; the third feature unit belongs to a common convolution and jump structure and is used for enhancing the extraction of high-dimensional and macroscopic image features; the fourth feature unit is used for converting the 2D feature layer into 1D and outputting the result.
The first feature cell includes a depth separable convolution layer, a batch normalization layer, an active layer (swish), a convolution layer, an active layer (swish), a layer normalization layer, an attention module, and an additive layer, which are connected in this order, and an output of the 2 nd active layer (swish) is connected to the additive layer.
The second characteristic unit comprises a convolution layer, an active layer (swish), a layer normalization layer, an attention module, an addition layer, a layer normalization layer, an attention module and an addition layer which are connected in sequence, wherein the output of the active layer (swish) is connected with the 1 st addition layer, and the output of the 1 st addition layer is connected with the 2 nd addition layer.
The third feature cell includes an active layer, a batch normalization layer, a depth-separable convolution layer, an active layer (swish), a batch normalization layer, a depth-separable convolution layer, and an additive layer, which are connected in sequence.
The fourth characteristic unit comprises a batch normalization layer, an activation layer (swish), a global mean pooling layer, a deactivation layer, a full connection layer and an activation layer (softmax) which are sequentially connected.
The attention module comprises a relative position coding layer and 3 depth separable convolution layers, wherein the 2 nd depth separable layer and the 3 rd depth separable layer are subjected to point multiplication to obtain a first point multiplication result, the 3 rd depth separable layer and the relative position coding layer are subjected to point multiplication to obtain a second point multiplication result, the first point multiplication result and the second point multiplication result are added and then connected with an attention active layer (softmax), and the output of the active layer and the 1 st depth separable layer are subjected to point multiplication and then connected with a remodeling layer, which is detailed in fig. 3.
The embodiment constructs an attention module suitable for being applied to an ultrasonic image through a self-attention (self-attention) mechanism in a transducer; combining with full pre-activation skip connection in ResNet, realizing feature enhancement (feature enhancement) and giving random depth to the network; and the application of the depth separable convolution layer greatly reduces the number of parameters and the complexity of calculation.
5. Loss Function (Loss Function)
In order to solve the data imbalance, the loss function used in this embodiment is a weighted classified cross entropy (weighted classified cross entropy) formula as follows:
wherein L represents the prediction loss of the convolutional neural network XDNetV 2-C-NLH; n represents the sample size; k represents the number of output values (i.e., the number of neurons in the output layer of the neural network, i.e., the number of classes, which is 3 in this embodiment); w is a j Represents the weight of class j (i.e., normal, off-center, vanished weight); y is ij The actual value of the ith sample corresponding to the jth class is represented as 1 or 0, and the ith sample belongs to the jth class or does not belong to the jth class respectively;
the predicted value of the ith sample corresponding to the jth class is represented, and the value range is [0,1 ]]The real number of (2).
6. Morphological analysis
In the positioning module, when the network classification prediction result is normal or eccentric, the lymphatic gates of the neck lymph nodes need to be positioned. By performing morphological analysis on the ROI, the lymphatic gate region is accurately located on the image using algorithms of conventional image processing such as connected component analysis, on-off operation, dilation-erosion, threshold comparison, and the like, see the white shaded portion in fig. 4.
7. Analysis of geometry
In the verification module, the rectangular coordinate system is established according to the outlined contour shape, and is established by finding the longest axis as a global long axis through the outlined points of the cervical lymph nodes, then finding the longest axis as a global short axis in the direction perpendicular to the global long axis, and finally taking the global long axis as the x axis and the global short axis as the y axis. As shown in fig. 4, the oval region is a cervical lymph node, the irregular figure in the oval is a lymph gate structure, the lymph gate region obtained by morphological analysis and positioning is calculated to be deviated from the central point, the threshold value comparison is performed to judge whether the lymph gate structure is normal or eccentric, and the network classification prediction result is verified.
8. Development and application flow
Preprocessing and enhancing the marked ultrasonic image, and inputting and training a neural network model; and adjusting a plurality of parameters and threshold values of morphological analysis and geometric analysis by a Kalman filtering method according to the lymphatic portal area marked by the doctor.
When the method is applied, the ultrasonic images and the nodule delineation data are input into an algorithm, and the algorithm outputs the structural type of the lymphatic gate and positions and visualizes the lymphatic gate area, which is shown in figure 4.
It can be easily found that the network (XDNetV2-C-NLH) of the invention mainly uses an attention module and a jumper as main components, is used for the self-adaptive network depth, largely uses depth separable convolution, reduces the number of calculation parameters, and belongs to a very light-weighted network structure. The lightweight network has two advantages in the present invention: 1: the network operation speed is high because the calculated amount is small; 2: the ultrasonic image of the cervical lymph node belongs to medical data, and not only is acquisition difficult, but also has the problems of ethics, privacy and the like. Therefore, the development of the XDNetV2-C-NLH network belongs to small sample training, the light-weight network can effectively avoid the over-fitting problem of the small sample training, and the generalization capability is improved. Therefore, compared with other network structures, the method has the advantages of high speed and high accuracy.
The foregoing descriptions of specific exemplary embodiments of the present invention have been presented for purposes of illustration and description. It is not intended to limit the invention to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain certain principles of the invention and its practical application to enable one skilled in the art to make and use various exemplary embodiments of the invention and various alternatives and modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims and their equivalents.
Claims (10)
1. A neck lymph node analysis apparatus based on a convolutional neural network, comprising:
an image acquisition module: for acquiring an ultrasound image with a cervical lymph node;
a lymph node image extraction module: the neck lymph node image acquisition device is used for intercepting the neck lymph node in the ultrasonic image to obtain an interested neck lymph node image;
a lymph node analysis module: and the system is used for inputting the interested cervical lymph node image into a convolutional neural network XDNetV2-C-NLH and determining the lymph gate structure of the cervical lymph node, wherein the lymph gate structure comprises normal, eccentric or disappearance.
2. The convolutional neural network-based cervical lymph node analyzing apparatus according to claim 1, wherein the convolutional neural network XDNetV2-C-NLH comprises 1 first feature unit, 4 second feature units, 2 third feature units and 1 fourth feature unit, wherein the first feature unit is sequentially connected with 2 second feature units, 2 fifth feature units and 1 fourth feature unit, and the other 2 second feature units are connected with the third feature unit to form 2 fifth feature units.
3. The convolutional neural network-based cervical lymph node analyzing apparatus as claimed in claim 2, wherein the first feature unit comprises a depth separable convolutional layer, a batch normalization layer, an activation layer, a convolutional layer, an activation layer, a layer normalization layer, an attention module and an additive layer which are connected in sequence, and an output of the 2 nd activation layer is connected to the additive layer.
4. The convolutional neural network-based cervical lymph node analyzing apparatus according to claim 2, wherein the second feature unit comprises a convolutional layer, an active layer, a layer normalization layer, an attention module, an additive layer, a layer normalization layer, an attention module, and an additive layer which are connected in sequence, and an output of the active layer is connected to the 1 st additive layer and an output of the 1 st additive layer is connected to the 2 nd additive layer.
5. The convolutional neural network-based cervical lymph node analyzing apparatus according to claim 2, wherein the third feature unit comprises an activation layer, a batch normalization layer, a depth separable convolution layer, and an addition layer, which are connected in sequence.
6. The convolutional neural network-based cervical lymph node analyzing apparatus as claimed in claim 2, wherein the fourth feature unit comprises a batch normalization layer, an activation layer, a global mean pooling layer, a deactivation layer, a full-link layer and an activation layer, which are connected in sequence.
7. The convolutional neural network-based cervical lymph node analysis apparatus as claimed in claims 3 to 6, wherein the attention module comprises a relative position coding layer and 3 depth separable convolutional layers, wherein the 2 nd depth separable layer and the 3 rd depth separable layer are dot-multiplied to obtain a first dot-multiplied result, the 3 rd depth separable layer and the relative position coding layer are dot-multiplied to obtain a second dot-multiplied result, the first dot-multiplied result and the second dot-multiplied result are added and then connected to the attention activation layer, and the output of the activation layer and the 1 st depth separable layer are dot-multiplied and then connected to the remodeling layer.
8. The convolutional neural network-based neck lymph node analyzing apparatus as claimed in claim 1, wherein the convolutional neural network XDNetV2-C-NLH adopts a weighted classification cross entropy as a loss function, and the formula of the weighted classification cross entropy is:
wherein L represents the prediction loss of the convolutional neural network XDNetV2-C-NLH, N represents the sample size, K represents the classification number value of the convolutional neural network XDNetV2-C-NLH output layer, and w j Represents a weight of the jth class; y is ij Representing that the ith sample corresponds to the actual value of the jth class, and taking the value of 1 or 0;
the ith sample representing the convolutional neural network XDNetV2-C-NLH corresponds to the predicted value of the jth class.
9. The convolutional neural network-based cervical lymph node analyzing apparatus as claimed in claim 1, further comprising a localization module: locating a lymph gate structure by connected domain analysis, open and close operations, dilation corrosion or threshold comparison when the lymph gate structure is predicted to be normal or eccentric.
10. The convolutional neural network-based cervical lymph node analyzing apparatus according to claim 9, further comprising a verification module: the system is used for constructing a rectangular coordinate system according to the glottic structure of the cervical lymph node in the cervical lymph node image of interest, calculating the distance of the positioned glottic structure from the central point of the rectangular coordinate system, and determining whether the glottic structure is normal or eccentric according to the distance of the rectangular coordinate system from the central point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210293141.1A CN115035019A (en) | 2022-03-24 | 2022-03-24 | Neck lymph node analysis device based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210293141.1A CN115035019A (en) | 2022-03-24 | 2022-03-24 | Neck lymph node analysis device based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115035019A true CN115035019A (en) | 2022-09-09 |
Family
ID=83119103
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210293141.1A Pending CN115035019A (en) | 2022-03-24 | 2022-03-24 | Neck lymph node analysis device based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115035019A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335261A (en) * | 2019-06-28 | 2019-10-15 | 山东科技大学 | It is a kind of based on when idle loop attention mechanism CT lymph node detection system |
| CN111369546A (en) * | 2020-03-10 | 2020-07-03 | 北京邮电大学 | Neck lymph node image classification and identification device and method |
| CN112201328A (en) * | 2020-10-09 | 2021-01-08 | 浙江德尚韵兴医疗科技有限公司 | Breast mass segmentation method based on cross attention mechanism |
| CN112529878A (en) * | 2020-12-15 | 2021-03-19 | 西安交通大学 | Multi-view semi-supervised lymph node classification method, system and equipment |
| KR102314270B1 (en) * | 2021-01-26 | 2021-10-19 | 주식회사 웨이센 | Ultrasonic bronchoscopy analysis method and apparatus |
-
2022
- 2022-03-24 CN CN202210293141.1A patent/CN115035019A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335261A (en) * | 2019-06-28 | 2019-10-15 | 山东科技大学 | It is a kind of based on when idle loop attention mechanism CT lymph node detection system |
| CN111369546A (en) * | 2020-03-10 | 2020-07-03 | 北京邮电大学 | Neck lymph node image classification and identification device and method |
| CN112201328A (en) * | 2020-10-09 | 2021-01-08 | 浙江德尚韵兴医疗科技有限公司 | Breast mass segmentation method based on cross attention mechanism |
| CN112529878A (en) * | 2020-12-15 | 2021-03-19 | 西安交通大学 | Multi-view semi-supervised lymph node classification method, system and equipment |
| KR102314270B1 (en) * | 2021-01-26 | 2021-10-19 | 주식회사 웨이센 | Ultrasonic bronchoscopy analysis method and apparatus |
Non-Patent Citations (2)
| Title |
|---|
| ZIYI LIU ET AL.: "Axillary lymph node metastasis prediction by contrast-enhanced computed tomography images for breast cancer patients based on deep learning", 《COMPUTERS IN BIOLOGY AND MEDICINE》, vol. 136, 10 August 2021 (2021-08-10), XP086773059, DOI: 10.1016/j.compbiomed.2021.104715 * |
| 晁晓菲: "基于深度学习的常见苹果叶片病害识别与病斑分割方法研究", 《中国博士学位论文全文数据库 农业科技辑》, no. 2022, 15 January 2022 (2022-01-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Brain tumor detection based on multimodal information fusion and convolutional neural network | |
| Liu et al. | A cascaded deep convolutional neural network for joint segmentation and genotype prediction of brainstem gliomas | |
| US20210118144A1 (en) | Image processing method, electronic device, and storage medium | |
| CN109166133A (en) | Soft tissue organs image partition method based on critical point detection and deep learning | |
| CN109389129A (en) | A kind of image processing method, electronic equipment and storage medium | |
| Al-Huda et al. | A hybrid deep learning pavement crack semantic segmentation | |
| CN103150723B (en) | The stomach CT image lymph node detection system of Shape-based interpolation and ellipse fitting and method | |
| Jiang et al. | Skin lesion segmentation based on multi-scale attention convolutional neural network | |
| Kaleem et al. | Segmentation of brain tumor tissue using marker controlled watershed transform method | |
| He et al. | Automated detection of boundary line in paddy field using MobileV2-UNet and RANSAC | |
| Liang et al. | Genetic programming for evolving figure-ground segmentors from multiple features | |
| CN112270667A (en) | TI-RADS-based integrated deep learning multi-tag identification method | |
| Fan et al. | Biomedical sensor image segmentation algorithm based on improved fully convolutional network | |
| Banerjee et al. | A CADe system for gliomas in brain MRI using convolutional neural networks | |
| Zuo et al. | A remote sensing image semantic segmentation method by combining deformable convolution with conditional random fields | |
| CN117274603A (en) | Liver focus image description method based on semantic segmentation network | |
| Lyu et al. | HRED-net: high-resolution encoder-decoder network for fine-grained image segmentation | |
| Hu et al. | A skin lesion segmentation method based on saliency and adaptive thresholding in wavelet domain | |
| Eun et al. | A study on the optimal artificial intelligence model for determination of urolithiasis | |
| Zhang et al. | Multi-scale aggregation networks with flexible receptive fields for melanoma segmentation | |
| CN115035019A (en) | Neck lymph node analysis device based on convolutional neural network | |
| Zhou et al. | BSMNet: Boundary-salience multi-branch network for intima-media identification in carotid ultrasound images | |
| Li et al. | Brown rice germ integrity identification based on deep learning network | |
| CN113408595B (en) | Pathological image processing method and device, electronic equipment and readable storage medium | |
| Wang et al. | Tongue semantic segmentation based on fully convolutional neural network |
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 |