CN114511502A - Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium - Google Patents
Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium Download PDFInfo
- Publication number
- CN114511502A CN114511502A CN202111657235.4A CN202111657235A CN114511502A CN 114511502 A CN114511502 A CN 114511502A CN 202111657235 A CN202111657235 A CN 202111657235A CN 114511502 A CN114511502 A CN 114511502A
- Authority
- CN
- China
- Prior art keywords
- module
- polyp
- image
- convolution
- output
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06T2207/10068—Endoscopic 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30028—Colon; Small intestine
-
- 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
- G06T2207/30092—Stomach; Gastric
-
- 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
- G06T2207/30096—Tumor; Lesion
Abstract
The invention discloses a gastrointestinal tract endoscope image polyp detection system based on artificial intelligence, a terminal and a storage medium. The image sampling module is included: the system is used for collecting complete gastrointestinal endoscope pictures of a focus area, and dividing the collected pictures into a training set and a testing set; an image preprocessing module: preprocessing pictures in a training set output by an image sampling module; gastrointestinal endoscope image polyp detection neural network module: extracting high-dimensional features from an image input into a network model through a coding network, completing feature optimization and denoising through a polyp focus feature enhancement module, and finally restoring semantic information through a decoding network to output the position and the edge of a polyp; a polyp detection module: and inputting the gastrointestinal endoscope picture to be detected into the trained model, and outputting a detection result. The invention can intelligently identify and detect the focus of polyp and the like in the endoscope image by building a polyp detection model, and can achieve higher accuracy by continuously training the neural network model.
Description
Technical Field
The invention belongs to the technical field of gastrointestinal endoscope image polyp detection, and particularly relates to a gastrointestinal endoscope image polyp detection system based on artificial intelligence, a terminal and a storage medium.
Background
Deep convolutional neural networks are successfully applied in the visual field. The deep convolutional neural network model provides the most advanced performance for many computer vision tasks such as image classification, image segmentation, image detection and the like due to the strong representation capability, fast reasoning and filter sharing characteristics of the deep convolutional neural network model. Recent studies have shown that deep convolutional neural networks can automatically learn hierarchies directly from data, thereby obtaining more and more complex features. With the development of the technology, the target detection method based on deep learning has achieved great success in medical image analysis. Through the learning of the input image samples, the network model continuously adjusts the parameter matrix of each layer, thereby identifying and detecting the real data.
Polyps are protrusions of the colon surface. Polyps appear in different shapes, from flat to pedicular. Early diagnosis of colorectal cancer can be achieved if a polyp of the large intestine (which may be a precursor of cancer) is detected and resected before it develops into a malignant tumor. The gold standard screening method for polyp detection and resection is optical colonoscopy. Colonoscopy is a successful preventative surgery. Colonoscopy is an operator-dependent procedure in which human factors such as fatigue and lack of concentration can lead to missed polyps during prolonged and back-to-back surgery. Computer-assisted polyp detection can assist colonoscopies in performing polyp detection tasks and can increase attention during surgery. A reliable polyp detection system must handle the variability of polyps in shape, size, color, and texture. Such a system may potentially save lives.
Disclosure of Invention
In order to solve the problems in the background art, the invention provides a gastrointestinal tract endoscope image polyp detection system based on artificial intelligence, a terminal and a storage medium, wherein the characteristics of an image are extracted from a gastrointestinal tract endoscope image through a coding network integrated by multi-scale context information, and effective information is provided for the identification and positioning of polyps; high-latitude features which are extracted from a coding network in a refined mode through a polyp focus feature enhancement module are subjected to feature transformation to further emphasize features related to a lesion region, so that the accuracy of identification and detection is increased; and summarizing effective characteristics extracted at different levels through a multi-level characteristic fusion decoding network to complete characteristic fusion, and recovering the regions of focuses such as polyps and the like in the original picture from the characteristics so as to complete the identification and the positioning of the polyps.
The technical scheme is as follows:
the utility model provides a gastrointestinal tract endoscope image polyp detecting system based on artificial intelligence, includes:
an image sampling module: the method is used for collecting gastrointestinal endoscope pictures with complete polyp focus areas, good picture brightness and intensity, less or no interference factors, and dividing the collected pictures into a training set and a testing set;
an image preprocessing module: preprocessing pictures in a training set output by an image sampling module;
gastrointestinal endoscope image polyp detection neural network module: the polyp focus characteristic enhancement module comprises a coding network, a polyp focus characteristic enhancement module and a decoding network; the coding network mainly comprises four context information aggregation modules which are connected in sequence; the decoding network mainly comprises four convolutional layers which are sequentially connected, wherein an up-sampling operation is carried out after each convolutional layer, and the output characteristics after up-sampling are fused with the output characteristics of a context information aggregation module in the coding network through residual connection and then input into the next convolutional layer;
extracting high-dimensional features from an image input into a gastrointestinal endoscope image polyp detection neural network module through a coding network, completing feature optimization and denoising through a polyp focus feature enhancement module, and finally restoring semantic information through a decoding network and outputting the position and the edge of polyp;
a neural network training module: the image output by the image preprocessing module is input into a gastrointestinal endoscope image polyp detection neural network module for autonomous training, and network model parameters with the optimal accuracy are stored to finish training; inputting the pictures in the test set output by the image sampling module into the trained neural network model, testing the generalization effect of the network, and verifying the identification and detection accuracy of the neural network model;
a polyp detection module: and loading a model file output by the neural network training module to obtain a trained gastrointestinal endoscope image polyp detection neural network model, inputting a gastrointestinal endoscope image to be detected into the gastrointestinal endoscope image polyp detection neural network model, and outputting the position and the edge of the polyp in the gastrointestinal endoscope image to finish detection.
The gastrointestinal endoscope image polyp detection neural network module comprises a multi-scale context information integrated coding network, a polyp focus characteristic enhancement module and a multi-level characteristic fusion decoding network. The coding network integrated by the multi-scale context information is composed of a convolution layer, a pooling layer and an activation function which are continuous, and the polyp detection and positioning method is used for extracting and accumulating characteristic information through a continuously enlarged receptive field. The polyp focus feature enhancement module applies an attention algorithm and consists of two attention modules, wherein one attention module extracts a space attention operator from the dimension of a space, and the other attention module extracts a channel attention operator from the dimension of a channel; the two operators guide the network model to be concentrated in the region of the polyp focus, so that redundant information is filtered, and the detection and the identification are more accurate. The decoding network with the multi-level feature fusion is composed of a plurality of continuous convolution layers and an upper sampling layer, the positions of polyps in an original image are continuously recovered from complex feature information of high latitude by learning and adjusting parameters, and semantic information of a feature map is continuously supplemented through jumping connection, so that the precision of positioning detection is improved.
The coding network consists of four context information aggregation modules, each context information aggregation module comprises a plurality of convolution layers connected in parallel, and the convolution kernels of the convolution layers are different in size, so that the reception fields of different sizes are brought, and the context information aggregation modules can capture the characteristics of target objects of different sizes. The characteristics of the picture are compressed and integrated to high latitude after passing through four context information aggregation modules, which is the basis for identifying and detecting polyp focuses by the model.
The polyp focus characteristic enhancement module uses continuous full-connection layers, convolution layers, Softmax layers, transposition, dot multiplication and other operations to further perform characteristic transformation on the output of the coding network, and enables the network model to focus attention on a region related to the focus through automatically learned parameters, so that favorable characteristic information is further enhanced, and the accuracy of identification and detection is increased.
The decoding network of the multi-level feature fusion uses the cross connection to fuse the feature information from different scales, the shallow feature contains more detail information, the deep feature contains more semantic information, and the multi-scale feature is fused, thereby completing the identification and the positioning of polyps.
In the image preprocessing module, the preprocessing process specifically includes: the method comprises the steps of marking polyp focus areas contained in all pictures, then normalizing the resolution of all the pictures, carrying out format conversion, then carrying out data augmentation, and finally carrying out rotation and mirror image conversion on the data.
Secondly, a terminal comprising a memory and a processor is provided
The memory for storing a computer program;
the processor is used for realizing the functions of the gastrointestinal tract endoscope image polyp detection system based on artificial intelligence when executing the computer program.
Thirdly, a computer readable storage medium is provided
The storage medium has stored thereon a computer program that, when executed by a processor, implements the functionality of the artificial intelligence based gastrointestinal endoscope image polyp detection system described above.
The invention has the beneficial effects that:
1. compared with the prior art, the method has the advantages that the polyp in the endoscope image is detected by building the gastrointestinal tract endoscope image polyp detection neural network, and meanwhile, the higher accuracy is achieved by continuously training the constructed convolutional neural network model, so that the judgment accuracy of the network model is not lower than the judgment level of an experienced doctor.
2. The invention adopts the attention mechanism and concentrates on the region of the polyp focus, thereby filtering redundant information and further improving the accuracy of identification and detection.
3. The gastrointestinal endoscope image polyp detection neural network total parameter number is in a reasonable range, the calculation time is rapid, and the targets such as polyps in the image can be rapidly identified.
Drawings
FIG. 1 is a schematic structural diagram of a gastrointestinal endoscope image polyp detection neural network module according to the present invention;
FIG. 2 is a diagram of a context information aggregation module according to the present invention;
fig. 3 is a schematic diagram of a polyp lesion feature enhancement module according to the present invention.
FIG. 4 is a schematic flow chart of the present invention.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
Fig. 1 shows a gastrointestinal endoscope image polyp detection neural network module according to the present invention. The gastrointestinal endoscope image polyp detection neural network module consists of a coding network integrated by multi-scale context information, a polyp focus characteristic enhancement module and a decoding network fused with multi-level characteristics. The coding network consists of four context information aggregation modules; the polyp focus characteristic enhancement module applies an attention algorithm and consists of two attention modules, wherein one attention module extracts a space attention operator from the dimension of a space, and the other attention module extracts a channel attention operator from the dimension of a channel. The decoding network is composed of four continuous convolution layers, each convolution layer is then subjected to up-sampling operation, the output characteristics after up-sampling are fused with the characteristics from the coding layer through residual connection, then the next convolution layer is input, and the detection result is output after four times. Each convolution layer performs convolution calculation on input, and the input of the model is an endoscope picture. The input image firstly passes through a coding network, high-dimensional feature vectors are extracted, then passes through a polyp focus feature enhancement module, feature optimization and denoising are completed, finally semantic information is restored through a decoding network, and the position and the edge of polyp are detected.
Existing network architectures are limited to a single reception field, and each layer can only see a fixed size region. This simple structure cannot capture polyp lesions of varying sizes and abundance. Therefore, we use the context information aggregation module shown in fig. 2 to adopt parallel connected convolution kernels with different sizes to form reception field branches with different sizes, and capture polyp feature information with different sizes. Used as the judgment basis for detecting the focus by the network model. Specifically, we complete the coding by concatenating 3 × 3, 5 × 5, and 7 × 7 convolution kernels, and finally adding all the results, aggregating the features through one maximum pooling layer, and reducing the spatial size of the features. Where a 1 x 1 convolution kernel is used to adjust the feature channel.
Many conventional neural networks are difficult to identify due to the complex characteristics of the focal region. In addition, noise characteristics are introduced due to uncertain factors such as contrast and illumination in the image, and detection is interfered. We design a polyp lesion feature enhancement module as shown in fig. 3, and apply the attention mechanism to focus the network on the lesion region, enhance the features and filter the noise region. Specifically, after feature transformation is carried out on high-dimensional features output by the coding network through two full-connection layers, transposition is completed, the high-dimensional features are multiplied by the high-dimensional features, an attention matrix is obtained after a result passes through a softmax function, and the attention matrix guides the network to focus on a focus area. Finally, after multiplying the original input high latitude characteristic with the attention moment array, further characteristic transformation is completed through a convolution layer of 3 multiplied by 3, the enhanced optimization characteristic is obtained, and the influence of noise information is filtered. The calculation formula of the polyp focus characteristic enhancement module is as follows:
M=I*{I*[MLP2(MLP1(I))]T}
O=Conv3×3(M)
wherein I represents input high-altitude features, MLP1 and MLP2 are two-layer fully-connected layer transformation functions, M is a middle feature vector, and O represents output features of the module.
The decoding network with multi-level feature fusion combines the fine image features from the coding network and the high-latitude semantic features in the decoder by using residual connection, and extracts the features with stronger expression for recovering the position information of the focus in the original image.
The invention discloses a method for constructing a gastrointestinal tract endoscope image polyp detection module, which comprises the following steps: the method comprises the following steps: constructing a gastrointestinal endoscope image polyp detection neural network model, wherein the network model comprises a coding network, a polyp focus characteristic enhancement module and a decoding network; step two: collecting 800 gastrointestinal tract endoscope pictures, selecting pictures with better brightness and intensity, complete focus areas and few or no interference factors as samples, preprocessing 600 pictures, marking the focus areas contained in all the pictures during processing, normalizing the resolution of all the pictures, converting formats, augmenting data, and converting the data by rotation, mirror image and the like to finally obtain 1000 pictures as a training data set; step three: and sending the processed 1000 gastrointestinal tract endoscope pictures into a network model for computer autonomous training. In the training process, the learning rate and other hyper-parameters are adjusted, and the convolutional neural network model parameters with the optimal accuracy are stored to finish training; step four: and (4) sending the remaining 200 gastrointestinal endoscope pictures into the convolutional neural network model, testing the generalization effect of the network, and verifying the identification and detection accuracy of the model. The test result shows that the detection accuracy rate reaches more than 90 percent. The real-time performance of the algorithm with the highest accuracy reaches 35FPS, and the system can quickly identify objects such as polyps in the image. The number of model parameters is 35M, and the storage size requirement of a computer is met.
The detection result output by our CNN output layer is a binary image with the same size as the original image, the pixel value of the background region is 0, and the pixel value of the lesion region is 255. The position and edge information of the focus is indicated through the binary image, and the semantic segmentation effect is achieved.
FIG. 4 is a schematic flow chart of the present invention:
1) firstly, a high-quality endoscope image database is established through acquired gastrointestinal endoscope image data, and the database is opened to a doctor. And marking and sketching the focus area by a doctor expert. And after the labeling is finished, preprocessing the image and the label together, and finishing the data augmentation.
2) And (3) inputting the input data into the built coding network, outputting a detection result after passing through the focus characteristic enhancement module and the decoding network, calculating the accuracy and finishing the training. And then sending the trained model into a test database for testing. If the accuracy is not enough, returning to adjust the parameters and retraining. And if the accuracy reaches the requirement, adding the test data set into the training database to further improve the performance of the model.
3) And then testing the speed performance of the model, if the speed of the model is lower, compressing the model, returning to retrain until the speed requirement is met. And sending the model meeting the requirements of accuracy and speed into an online application to finish auxiliary diagnosis, and adding new data into an endoscope image database to perform a new iteration.
Claims (8)
1. An artificial intelligence based gastrointestinal endoscope image polyp detection system, comprising:
an image sampling module: the system is used for collecting complete gastrointestinal endoscope pictures of polyp focus areas and dividing the collected pictures into a training set and a testing set;
an image preprocessing module: preprocessing pictures in a training set output by an image sampling module;
gastrointestinal endoscope image polyp detection neural network module: the polyp focus characteristic enhancement module comprises a coding network, a polyp focus characteristic enhancement module and a decoding network; the coding network mainly comprises four context information aggregation modules which are connected in sequence; the decoding network mainly comprises four convolution layers which are connected in sequence, wherein each convolution layer is subjected to up-sampling operation, and the output characteristics after up-sampling are fused with the output characteristics of a context information aggregation module in the coding network through residual connection and then input into the next convolution layer;
extracting high-dimensional features from an image input into a gastrointestinal endoscope image polyp detection neural network module through a coding network, completing feature optimization and denoising through a polyp focus feature enhancement module, and finally restoring semantic information through a decoding network and outputting the position and the edge of polyp;
a neural network training module: the image output by the image preprocessing module is input into a gastrointestinal endoscope image polyp detection neural network module for autonomous training, and network model parameters with the optimal accuracy are stored to finish training;
a polyp detection module: and loading a model file output by the neural network training module to obtain a trained gastrointestinal endoscope image polyp detection neural network model, inputting a gastrointestinal endoscope picture to be detected into the trained gastrointestinal endoscope image polyp detection neural network model, and outputting the position and the edge of the polyp in the gastrointestinal endoscope picture to finish detection.
2. The system of claim 1, wherein each context information aggregation module of the gi tract endoscopic image polyp detection neural network module is mainly composed of five convolutional layers, an activation layer and a max-pooling layer, and the second convolutional layer, the third convolutional layer and the fourth convolutional layer are connected in parallel; the input of the context information aggregation module is input into the second convolution layer, the third convolution layer and the fourth convolution layer respectively after passing through the first convolution layer, and the output of the second convolution layer, the third convolution layer and the fourth convolution layer is added and then input into the maximum pooling layer through the fifth convolution layer and the ReLU activation layer in sequence.
3. The system of claim 2, wherein convolution kernels of the second convolution layer, the third convolution layer and the fourth convolution layer are 3 x 3, 5 x 5 and 7 x 7 respectively, and convolution kernels of the first convolution layer and the fifth convolution layer are 1 x 1 respectively.
4. The system of claim 1, wherein the neural network module for gastrointestinal endoscope image polyp detection comprises two attention modules: a channel attention operator and a spatial attention operator;
the high-dimensional features output by the coding network are transposed after feature transformation of two fully-connected layers in a channel attention operator, then multiplied by the originally input high-dimensional features, the multiplied result obtains an attention matrix through a softmax function of a space attention operator, the attention matrix is multiplied by the originally input high-dimensional features, further feature transformation is completed through a 3 x 3 convolution layer, and the enhanced optimized features are obtained and serve as output of a polyp focus feature enhancement module.
5. The system as claimed in claim 1, wherein in the neural network module for detecting polyps in gastrointestinal tract endoscope images, the decoding network comprises four convolution modules connected in sequence, each convolution module comprises a convolution layer and an up-sampling layer, and the convolution layer is followed by the up-sampling operation;
the output of the first convolution module is fused with the output of the fourth context information aggregation module through residual connection and then input into the second convolution module; the output of the second convolution module is fused with the output of the third context information aggregation module through residual connection and then input into the third convolution module; the output of the third convolution module is fused with the output of the second context information aggregation module through residual connection and then input into the fourth convolution module; and the output of the fourth convolution module is fused with the output of the first context information aggregation module through residual connection to be used as the output of the decoding network.
6. The system for gastrointestinal endoscope image polyp detection based on artificial intelligence of claim 1, wherein in the image preprocessing module, the preprocessing process is specifically as follows: the method comprises the steps of marking polyp focus areas contained in all pictures, then normalizing the resolution of all the pictures, carrying out format conversion, then carrying out data augmentation, and finally carrying out rotation and mirror image conversion on the data.
7. A terminal comprising a memory and a processor;
the memory for storing a computer program;
the processor, when executing the computer program, implements the functionality of the artificial intelligence based gastrointestinal endoscope image polyp detection system according to any one of claims 1 to 6.
8. A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, implements the functionality of an artificial intelligence based gastrointestinal endoscope image polyp detection system as recited in any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111657235.4A CN114511502A (en) | 2021-12-30 | 2021-12-30 | Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111657235.4A CN114511502A (en) | 2021-12-30 | 2021-12-30 | Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114511502A true CN114511502A (en) | 2022-05-17 |
Family
ID=81547970
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111657235.4A Pending CN114511502A (en) | 2021-12-30 | 2021-12-30 | Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114511502A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115359066A (en) * | 2022-10-24 | 2022-11-18 | 岱川医疗(深圳)有限责任公司 | Focus detection method and device for endoscope, electronic device and storage medium |
CN116030303A (en) * | 2023-01-09 | 2023-04-28 | 深圳市大数据研究院 | Video colorectal lesion typing method based on semi-supervised twin network |
CN116223661A (en) * | 2023-01-04 | 2023-06-06 | 江苏福多美生物科技有限公司 | Method for measuring content of allicin in garlic wastewater |
CN117830751A (en) * | 2024-03-06 | 2024-04-05 | 苏州凌影云诺医疗科技有限公司 | Intestinal polyp LST morphology identification method and device based on DenseNet |
CN117830751B (en) * | 2024-03-06 | 2024-05-07 | 苏州凌影云诺医疗科技有限公司 | DenseNet-based intestinal polyp LST morphology identification method and DenseNet-based intestinal polyp LST morphology identification device |
-
2021
- 2021-12-30 CN CN202111657235.4A patent/CN114511502A/en active Pending
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115359066A (en) * | 2022-10-24 | 2022-11-18 | 岱川医疗(深圳)有限责任公司 | Focus detection method and device for endoscope, electronic device and storage medium |
CN116223661A (en) * | 2023-01-04 | 2023-06-06 | 江苏福多美生物科技有限公司 | Method for measuring content of allicin in garlic wastewater |
CN116223661B (en) * | 2023-01-04 | 2023-12-15 | 江苏福多美生物科技有限公司 | Method for measuring content of allicin in garlic wastewater |
CN116030303A (en) * | 2023-01-09 | 2023-04-28 | 深圳市大数据研究院 | Video colorectal lesion typing method based on semi-supervised twin network |
CN116030303B (en) * | 2023-01-09 | 2024-01-30 | 深圳市大数据研究院 | Video colorectal lesion typing method based on semi-supervised twin network |
CN117830751A (en) * | 2024-03-06 | 2024-04-05 | 苏州凌影云诺医疗科技有限公司 | Intestinal polyp LST morphology identification method and device based on DenseNet |
CN117830751B (en) * | 2024-03-06 | 2024-05-07 | 苏州凌影云诺医疗科技有限公司 | DenseNet-based intestinal polyp LST morphology identification method and DenseNet-based intestinal polyp LST morphology identification device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111145170B (en) | Medical image segmentation method based on deep learning | |
Li et al. | Accurate retinal vessel segmentation in color fundus images via fully attention-based networks | |
CN111401384B (en) | Transformer equipment defect image matching method | |
CN111950649B (en) | Attention mechanism and capsule network-based low-illumination image classification method | |
CN114511502A (en) | Gastrointestinal endoscope image polyp detection system based on artificial intelligence, terminal and storage medium | |
CN111915573A (en) | Digestive endoscopy focus tracking method based on time sequence feature learning | |
Fayaz et al. | Underwater object detection: architectures and algorithms–a comprehensive review | |
CN111179275B (en) | Medical ultrasonic image segmentation method | |
Souaidi et al. | A new automated polyp detection network MP-FSSD in WCE and colonoscopy images based fusion single shot multibox detector and transfer learning | |
CN116309648A (en) | Medical image segmentation model construction method based on multi-attention fusion | |
CN111161271A (en) | Ultrasonic image segmentation method | |
CN114821682B (en) | Multi-sample mixed palm vein identification method based on deep learning algorithm | |
CN115294075A (en) | OCTA image retinal vessel segmentation method based on attention mechanism | |
Kareem et al. | Skin lesions classification using deep learning techniques | |
CN113344933B (en) | Glandular cell segmentation method based on multi-level feature fusion network | |
CN113538363A (en) | Lung medical image segmentation method and device based on improved U-Net | |
Yang et al. | Lesion classification of wireless capsule endoscopy images | |
Chou et al. | Representation learning on large and small data | |
CN116935044A (en) | Endoscopic polyp segmentation method with multi-scale guidance and multi-level supervision | |
Zhang et al. | Artifact detection in endoscopic video with deep convolutional neural networks | |
CN115358961A (en) | Multi-focus image fusion method based on deep learning | |
CN110555342B (en) | Image identification method and device and image equipment | |
MAHMOOD | Smart controlling devices with convolutional neural network | |
CN114998210B (en) | Retinopathy of prematurity detecting system based on deep learning target detection | |
Abdou et al. | Deep Learning and Computer Vision Techniques for Microcirculation Analysis: A Review |
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 |