CN109285128A - A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning - Google Patents
A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning Download PDFInfo
- Publication number
- CN109285128A CN109285128A CN201811037162.7A CN201811037162A CN109285128A CN 109285128 A CN109285128 A CN 109285128A CN 201811037162 A CN201811037162 A CN 201811037162A CN 109285128 A CN109285128 A CN 109285128A
- Authority
- CN
- China
- Prior art keywords
- cholelithiasis
- medical image
- image data
- deep learning
- training set
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4046—Scaling the whole image or part thereof using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/60—Rotation of a whole image or part thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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]
Abstract
The present invention provides a kind of based on the cholelithiasis CT medical image data Enhancement Method based on deep learning, it include: building cholelithiasis CT medical image data enhancing convolutional neural networks, the network is made of four part convolution units, input of the building cholelithiasis CT medical image data sets as neural network first, image is subjected to the operation such as enhancing marginal information and removal redundancy, and then image is cut according to segmentation threshold, many image blocks are formed, data set is expanded by the operation such as scaling, rotation and translation.Convolutional neural networks are constantly trained using data set, adaptive promotion network generates the convolutional neural networks model that can be used for the enhancing of cholelithiasis CT medical image data to functions such as the feature extraction of image, contrast stretching, histogram equalization and image reconstructions.This method can complete the function of cholelithiasis CT medical image data enhancing in real time, and obtain good visual effect and medical effect.
Description
Technical field
The present invention relates to field of image processing, specially a kind of cholelithiasis CT medical image data based on deep learning increases
Strong method.
Background technique
The process of image recognition includes the links such as pretreatment, feature extraction, characteristic matching, Similarity measures.In pretreatment
An important link be exactly image enhancement processing, its purpose is, enhancing the useful information in image improves the view of image
Effect is felt, for the application of given image.Image data enhancing is a critically important part, its treatment effect is direct
Influence subsequent image recognition processes.
Cholelithiasis according to the present invention is a kind of common disease of digestive system, and disease is various, and pathogenic factors is crisscross multiple
It is miscellaneous, have disease incidence high, the features such as molten row's stone is difficult.In addition, the type and form of cholelithiasis are varied, part cholelithiasis
Lesion form be also it is closely similar, this hinders the correct diagnosis and treatment of cholelithiasis significantly.In this case, the liver of some youths
Gallbladder section doctor needs the technical ability for grasping diagnosis cholelithiasis that prolonged study can be skilled, this gives the clinic of hepatology doctor
Diagnosis brings huge challenge.And for the patient, cholelithiasis cannot be according to itself illness in online access data, if not
It can obtain medical treatment in time, there is carcinogenic danger.
It is created in combination with cholelithiasis big data and knowledge etc. based on deep learning by artificial intelligence deep learning
Cholelithiasis CT medical image data Enhancement Method, and then more effective help is provided for the correct diagnosis and treatment of cholelithiasis.Cholelith disease
Class and form are varied, so that needing very more cholelithiasis medical image datas during model training.It is common
Way is to have initially set up cholelithiasis CT medical image training set, in the method using deep learning using model to a large amount of figure
As training set is trained, trained model is generated, new image authentication the set pair analysis model is then constructed and is verified, obtained most
Whole recognition result.But the type of cholelithiasis is varied with form so that its clinical diagnostic process is extremely difficult, to gallbladder
The clinical diagnosis of stone disease brings great challenge.Using image data enhancing method to cholelithiasis CT medical image at
Reason, can effectively alleviate this difficulty, and the diagnosis efficiency of doctor can be improved.It is not found also currently on the market for cholelith
The data of sick CT medical image enhance processing method.
Summary of the invention
The purpose of the present invention is to solve difficulties present in above-mentioned disease in the liver and gallbladder field, provide a kind of based on depth
The cholelithiasis CT medical image data Enhancement Method of habit can help the doctor of cholelithiasis clinic to carry out correct diagnosis and treatment, improve
The efficiency of diagnosis.
To achieve the above object, a kind of technical solution provided by the present invention are as follows: cholelithiasis CT doctor based on deep learning
Treat image data Enhancement Method, comprising the following steps:
1) cholelithiasis CT medical image is chosen, training set is constructed；
2) data image is split, expanding data training set；
3) cholelithiasis CT medical image data of the building based on deep learning enhances convolutional neural networks, and the network is by four
Bundling product unit composition；
4) convolutional neural networks model is trained using the training set of building, generates trained cholelithiasis CT medical treatment
Image data enhances convolutional neural networks model.
5) new cholelithiasis CT medical image test set is constructed, trained model is tested.
2. further, pair in a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning
Data image is split, and specifically, the CT medical image of cholelithiasis lesion is had to every, the segmentation threshold set carries out
Segmentation, forms fixed-size image block, and constitute initial data training set.
3. it is further, the data training set is expanded, what is included the following steps is any one or more:
1) the step of image in the training set being zoomed in or out；
2) spin step is carried out to the image in the training set；
3) translation step is carried out to the image in the training set；
4) the step of radiation transformation being carried out to the image in the training set.
4. a kind of further, cholelithiasis CT medical image data Enhancement Method based on deep learning, neural network
Structure convolution unit includes:
1) first part's convolution unit is made of convolutional layer, average pond layer and Normalization layers of Batch, wherein
The size of convolution is 8*8, and Normalization layers of main function of Batch are that data are normalized with operation, guarantees net
The capability of fitting of network；
2) second part convolution unit is made of two identical branching networks, wherein each branch by convolutional layer, most
Great Chiization layer and LN layers of composition, convolution kernel are reduced, and 3*3 is become；
3) Part III convolution unit is made of convolutional layer, maximum pond layer, and convolution kernel size is constant, is 3*3；
4) Part IV convolution unit is made of one layer of convolutional layer, exports the image after final enhancing.
5. further, the instruction of the cholelithiasis CT medical image data enhancing convolutional neural networks model based on deep learning
Practice the stage, using back-propagation algorithm, it is therefore an objective to tuning constantly be carried out to the parameter of model, improve the capability of fitting of model, together
The regularization algorithms such as Shi Caiyong Dropout, Bagging are trained model, it is therefore an objective to which the capability of fitting of Optimized model prevents
Model over-fitting.
6. further, the cholelithiasis CT medical image data Enhancement Method based on deep learning, described in step 5)
It is identical for constructing new test set and the building training set described in step 1).
The present invention have the following advantages that with the utility model has the advantages that
1, image divide using the segmentation threshold of setting and obtained corresponding image block, using efficient side
Method expands data set, improves the validity of training set, and the capability of fitting of model then can be improved.
2, based on regularization algorithms training convolutional neural networks moulds such as back-propagation algorithm and Dropout, Bagging
Type can adaptively improve the learning efficiency of network, improve the reconstruction ability of network image.
3, the cholelithiasis CT medical image data enhancing based on deep learning of building being made of four part convolution units
Convolutional neural networks complete feature extraction, contrast stretching, histogram equalization and the image weight of cholelithiasis CT medical image
It the functions such as builds, good visual effect and medical effect can be obtained.
Detailed description of the invention
Fig. 1 is the framework that the cholelithiasis CT medical image data of the invention based on deep learning enhances convolutional neural networks
Schematic diagram.
Specific embodiment
The present invention is further explained in the light of specific embodiments
As shown in Figure 1, the cholelithiasis CT medical image data Enhancement Method described in the present embodiment based on deep learning,
Concrete condition:
1) 90 cholelithiasis CT medical images are chosen as the cholelithiasis CT medical image data enhancing based on deep learning
The initial training collection of convolutional neural networks, next improves training set, specific steps are as follows:
1-1) using the segmentation threshold set, every CT medical image with cholelithiasis lesion is split, often
Four pieces of image blocks of different sizes will be divided by opening image.
1-2) above-mentioned data set is expanded using any one or more of following steps at random:
The step of image in the training set is zoomed in or out；
Spin step is carried out to the image in the training set；
Translation step is carried out to the image in the training set；
The step of radiation transformation is carried out to the image in the training set.
2) cholelithiasis CT medical image data of the building based on deep learning enhances convolutional neural networks, by four part convolution
Unit composition, is respectively completed feature extraction, contrast stretching, histogram equalization and the image weight of cholelithiasis CT medical image
The functions such as build.Wherein first part's convolution unit is made of convolutional layer, average pond layer and Normalization layers of Batch,
Second part convolution unit is made of two identical branching networks, wherein each branch by convolutional layer, maximum pond layer with
And LN layers of composition, Part III convolution unit is by Pooling layers of convolutional layer, maximum pond layer and Global average structure
At Part IV convolution unit is made of one layer of convolutional layer, exports the image after final enhancing.Specifically, original gallbladder
Stone disease CT medical image size is 64*64, and the image size exported after the processing of convolutional neural networks is 36*
36.Specifically, the convolution unit of first part has the convolutional layer of 64 8*8 convolution kernels by one, and an operation core is the flat of 4*4
Equal pond layer and Normalization layers of a Batch composition；The convolution unit of second part is by two identical branched networks
Network is constituted, and wherein each branch has the convolutional layer of 64 6*6 convolution kernels, the maximum pond that an operation core is 3*3 by one
Layer and a LN layers of composition；The convolution unit of Part III has the convolutional layer of 32 3*3 convolution kernels by one, and an operation core is
The maximum pond layer of 3*3 is constituted；The convolution unit of Part IV is made of one layer of convolutional layer.
3) the cholelithiasis CT medical image data enhancing convolutional neural networks model based on deep learning is trained, is adopted
With regularization algorithms such as back-propagation algorithm and Dropout, Bagging, completely trained using 9000 iteration as one,
Tuning constantly is carried out to the parameter of model, reduces error, the capability of fitting of Optimized model.
4) new cholelithiasis CT medical image test set is constructed, the trained cholelithiasis CT based on deep learning is input to
Medical image data enhances in convolutional neural networks model, the image after output data enhancing.
Embodiment described above is only the preferred embodiments of the invention, and but not intended to limit the scope of the present invention, therefore
Change made by all principles according to the present invention, should all be included within the scope of protection of the present invention.
Claims (6)
1. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning, which comprises the following steps:
1) cholelithiasis CT medical image is chosen, training set is constructed；
2) data image is split, expanding data training set；
3) cholelithiasis CT medical image data of the building based on deep learning enhances convolutional neural networks, which is rolled up by four parts
Product unit composition；
4) convolutional neural networks model is trained using the training set of building, generates trained cholelithiasis CT medical image
Data enhance convolutional neural networks model.
5) new cholelithiasis CT medical image test set is constructed, trained model is tested.
2. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning according to claim 1, special
Sign is, described to be split to data image, specifically, the CT medical image of cholelithiasis lesion is had to every, has been set
Segmentation threshold be split, form fixed-size image block, and constitute initial data training set.
3. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning according to claim 1, special
Sign is that the expanding data training set, what is included the following steps is any one or more:
1) the step of image in the training set being zoomed in or out；
2) spin step is carried out to the image in the training set；
3) translation step is carried out to the image in the training set；
4) the step of radiation transformation being carried out to the image in the training set.
4. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning according to claim 1, special
Sign is that four part convolution unit includes:
1) first part's convolution unit is made of convolutional layer, average pond layer and Normalization layers of Batch, wherein convolution
Size be 5*5, Normalization layers of main function of Batch are that data are normalized with operation, guarantee network
Capability of fitting；
2) second part convolution unit is made of two identical branching networks, and wherein each branch is by convolutional layer, maximum pond
Change layer and LN layers of composition, convolution kernel are reduced, becomes 3*3；
3) Part III convolution unit is made of convolutional layer, maximum pond layer, and convolution kernel size is constant, is 3*3；
4) Part IV convolution unit is made of one layer of convolutional layer, exports the image after final enhancing.
5. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning according to claim 1, special
Sign is: the training stage of the cholelithiasis CT medical image data enhancing convolutional neural networks model, being calculated using backpropagation
Method, it is therefore an objective to tuning constantly be carried out to the parameter of model, improve the capability of fitting of model, while using Dropout, Bagging
Equal regularization algorithms are trained model, it is therefore an objective to which the capability of fitting of Optimized model prevents model over-fitting.
6. a kind of cholelithiasis CT medical image data Enhancement Method based on deep learning according to claim 1, special
Sign is that the building described in step 5) new test set and the building training set described in step 1) are identical.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811037162.7A CN109285128A (en) | 2018-09-06 | 2018-09-06 | A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811037162.7A CN109285128A (en) | 2018-09-06 | 2018-09-06 | A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109285128A true CN109285128A (en) | 2019-01-29 |
Family
ID=65183670
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811037162.7A Pending CN109285128A (en) | 2018-09-06 | 2018-09-06 | A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109285128A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109544585A (en) * | 2018-12-19 | 2019-03-29 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image data Enhancement Method based on lightweight convolutional neural networks |
| CN109902734A (en) * | 2019-02-22 | 2019-06-18 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image method for quickly identifying based on YOLO convolutional neural networks |
| CN110458246A (en) * | 2019-08-27 | 2019-11-15 | 上海理工大学 | A kind of Biliary Calculi classification method based on deep learning |
| CN110648288A (en) * | 2019-08-27 | 2020-01-03 | 上海理工大学 | Choledochoscope image enhancement method based on residual transposition deconvolution neural network |
| CN112489029A (en) * | 2020-12-10 | 2021-03-12 | 深圳先进技术研究院 | Medical image segmentation method and device based on convolutional neural network |
| CN112508804A (en) * | 2020-11-19 | 2021-03-16 | 中国石油大学(华东) | Ocean vortex image automatic enhancement system based on image noise adding and denoising, computer equipment and storage medium |
-
2018
- 2018-09-06 CN CN201811037162.7A patent/CN109285128A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109544585A (en) * | 2018-12-19 | 2019-03-29 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image data Enhancement Method based on lightweight convolutional neural networks |
| CN109902734A (en) * | 2019-02-22 | 2019-06-18 | 中国石油大学(华东) | A kind of cholelithiasis CT medical image method for quickly identifying based on YOLO convolutional neural networks |
| CN110458246A (en) * | 2019-08-27 | 2019-11-15 | 上海理工大学 | A kind of Biliary Calculi classification method based on deep learning |
| CN110648288A (en) * | 2019-08-27 | 2020-01-03 | 上海理工大学 | Choledochoscope image enhancement method based on residual transposition deconvolution neural network |
| CN110648288B (en) * | 2019-08-27 | 2023-02-03 | 上海理工大学 | Choledochoscope image enhancement method based on residual transposition deconvolution neural network |
| CN112508804A (en) * | 2020-11-19 | 2021-03-16 | 中国石油大学(华东) | Ocean vortex image automatic enhancement system based on image noise adding and denoising, computer equipment and storage medium |
| CN112489029A (en) * | 2020-12-10 | 2021-03-12 | 深圳先进技术研究院 | Medical image segmentation method and device based on convolutional neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109285128A (en) | A kind of cholelithiasis CT medical image data Enhancement Method based on deep learning | |
| CN109544585A (en) | A kind of cholelithiasis CT medical image data Enhancement Method based on lightweight convolutional neural networks | |
| Lal et al. | NucleiSegNet: Robust deep learning architecture for the nuclei segmentation of liver cancer histopathology images | |
| Oliveira et al. | Retinal vessel segmentation based on fully convolutional neural networks | |
| Kumar et al. | Breast cancer classification of image using convolutional neural network | |
| WO2020168820A1 (en) | Yolo convolutional neural network-based cholelithiasis ct medical image data enhancement method | |
| CN107154043A (en) | A kind of Lung neoplasm false positive sample suppressing method based on 3DCNN | |
| Nanda et al. | Computer-aided segmentation of liver lesions in CT scans using cascaded convolutional neural networks and genetically optimised classifier | |
| Avci et al. | An expert diagnosis system for classification of human parasite eggs based on multi-class SVM | |
| El-Shafai et al. | Efficient Deep-Learning-Based Autoencoder Denoising Approach for Medical Image Diagnosis. | |
| Otálora et al. | Training deep convolutional neural networks with active learning for exudate classification in eye fundus images | |
| CN109858429A (en) | A kind of identification of eye fundus image lesion degree and visualization system based on convolutional neural networks | |
| Chung et al. | Liver segmentation in abdominal CT images via auto-context neural network and self-supervised contour attention | |
| Chougrad et al. | Convolutional neural networks for breast cancer screening: transfer learning with exponential decay | |
| CN109948575B (en) | Eyeball area segmentation method in ultrasonic image | |
| CN107909588A (en) | Partition system under MRI cortex based on three-dimensional full convolutional neural networks | |
| Priyadharshini et al. | A novel hybrid Extreme Learning Machine and Teaching–Learning-Based Optimization algorithm for skin cancer detection | |
| Jiang et al. | Automatic screening of diabetic retinopathy images with convolution neural network based on caffe framework | |
| Mai et al. | Few-shot transfer learning for hereditary retinal diseases recognition | |
| Li et al. | Double attention U-Net for brain tumor MR image segmentation | |
| CN106709869A (en) | Dimensionally reduction method based on deep Pearson embedment | |
| Fu et al. | Automatic grading of Diabetic macular edema based on end-to-end network | |
| Chen et al. | Multi-threshold image segmentation based on an improved differential evolution: case study of thyroid papillary carcinoma | |
| CN111340760B (en) | Knee joint positioning method based on multitask two-stage convolution neural network | |
| CN116977323A (en) | Vitiligo lesion area segmentation method and device based on image editing and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190129 |