CN113469961A - Neural network-based carpal tunnel image segmentation method and system - Google Patents
Neural network-based carpal tunnel image segmentation method and system Download PDFInfo
- Publication number
- CN113469961A CN113469961A CN202110704545.0A CN202110704545A CN113469961A CN 113469961 A CN113469961 A CN 113469961A CN 202110704545 A CN202110704545 A CN 202110704545A CN 113469961 A CN113469961 A CN 113469961A
- Authority
- CN
- China
- Prior art keywords
- layer
- carpal tunnel
- convolution
- block
- segmentation
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 29
- 238000003709 image segmentation Methods 0.000 title claims abstract description 29
- 230000011218 segmentation Effects 0.000 claims abstract description 68
- 238000012545 processing Methods 0.000 claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000000877 morphologic effect Effects 0.000 claims abstract description 11
- 230000003044 adaptive effect Effects 0.000 claims abstract description 8
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 7
- 238000011176 pooling Methods 0.000 claims description 47
- 238000005096 rolling process Methods 0.000 claims description 33
- 238000010606 normalization Methods 0.000 claims description 10
- 230000017105 transposition Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 238000004804 winding Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 description 17
- 230000006870 function Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 5
- 238000003745 diagnosis Methods 0.000 description 4
- 238000002595 magnetic resonance imaging Methods 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 3
- 208000003295 carpal tunnel syndrome Diseases 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 210000001617 median nerve Anatomy 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 208000024891 symptom Diseases 0.000 description 2
- 206010003694 Atrophy Diseases 0.000 description 1
- 206010016654 Fibrosis Diseases 0.000 description 1
- 208000010428 Muscle Weakness Diseases 0.000 description 1
- 206010028372 Muscular weakness Diseases 0.000 description 1
- 208000028389 Nerve injury Diseases 0.000 description 1
- 230000037444 atrophy Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004761 fibrosis Effects 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002427 irreversible effect Effects 0.000 description 1
- 208000028867 ischemia Diseases 0.000 description 1
- 201000005518 mononeuropathy Diseases 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 230000008764 nerve damage Effects 0.000 description 1
- 238000012148 non-surgical treatment Methods 0.000 description 1
- 231100000862 numbness Toxicity 0.000 description 1
- 235000016709 nutrition Nutrition 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
- 230000008719 thickening 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a neural network-based carpal tunnel image segmentation method, which comprises the following steps of: constructing and training a processing model by using a convolutional neural network, wherein the processing model at least comprises a classification module and a segmentation module; acquiring a carpal tunnel image to be segmented; preprocessing a carpal tunnel image to be segmented; using a classification module to perform morphological classification on the preprocessed image; and according to the form classification result, using an adaptive segmentation module to segment the carpal tunnel image to be segmented and outputting a segmentation result. The invention classifies and segments the images of the carpal tunnel to be segmented, thereby improving the accuracy of the segmentation of the carpal tunnel images by a processing model.
Description
Technical Field
The invention relates to a carpal tunnel image segmentation method and system, in particular to a carpal tunnel image segmentation method and system based on a neural network, and belongs to the technical field of medical image processing.
Background
Carpal tunnel syndrome, caused by the compression of the median nerve as it passes through the carpal tunnel, is manifested by pain, numbness, muscle weakness in the innervated area, is the most common clinically compressive focal mononeuropathy, and the incidence increases year by year. Early non-surgical treatment can bring certain curative effect, and more patients can obtain remarkable improvement of symptoms from surgical treatment. However, due to the non-specificity of the symptoms, misdiagnosis and missed diagnosis occur frequently, and the condition is continuously progressed due to the delay of diagnosis, and finally irreversible nerve injury and muscle nutritional loss atrophy are developed, so that defects on the appearance and the function of hands of patients are brought.
Clinical assessment is currently considered the gold standard for clinical and research, but it depends on the clinical experience and expertise of the physician, and therefore more objective imaging results are emerging. The Magnetic Resonance Imaging (MRI) technology can accurately measure the diameter of the carpal tunnel and the nerve, so as to distinguish whether the median nerve has swelling and thickening when entering the carpal tunnel and whether the median nerve enters the carpal tunnel and is pressed to be flat, and meanwhile, the ischemia and fibrosis of the nerve can show different degrees of signal changes in the MRI. However, it is time-consuming and labor-consuming to segment the carpal tunnel image manually, and different doctors may obtain different results when performing image segmentation, and even if the same doctor segments the same hand MRI image at different time periods, there are slight differences.
In recent years, a neural network has been developed rapidly in an image processing direction, and a large number of segmentation models have been proposed, such as network structures of FCN, UNet, SegNet, deep lab, and the like. Medical image processing is also increasingly applied to neural networks. Therefore, research and development of a neural network-based carpal tunnel image segmentation method and system are of great significance to the field.
Disclosure of Invention
The invention aims to provide a neural network-based carpal tunnel image segmentation method, which can obtain objective and stable image segmentation results, assist doctors in making better judgment on disease conditions and solve the problems in the background art.
The invention further aims to provide a carpal tunnel image segmentation system based on the neural network.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a carpal tunnel image segmentation method based on a neural network comprises the following steps:
constructing and training a processing model by using a convolutional neural network, wherein the processing model at least comprises a classification module and a segmentation module;
acquiring a carpal tunnel image to be segmented;
preprocessing a carpal tunnel image to be segmented;
using a classification module to perform morphological classification on the preprocessed image;
and according to the form classification result, using an adaptive segmentation module to segment the carpal tunnel image to be segmented and outputting a segmentation result.
Preferably, the processing model comprises a plurality of segmentation modules, and is used for segmenting the carpal tunnel images to be segmented in different forms according to the form classification result and outputting the segmentation result.
Preferably, the classification module and the segmentation module each include at least a convolution block, and the convolution block includes a first convolution layer, a first batch normalization layer, a first active layer, a second convolution layer, a second batch normalization layer, and a second active layer, which are sequentially connected and encapsulated.
Preferably, the parameters of the volume block are set as:
a first winding layer: inputting a channel number in _ ch and an output channel number out _ ch, wherein the size of the adopted convolution kernel is 3, padding is 1 and stride is 1;
a first batch normalization layer and a second batch normalization layer: the number of channels is out _ ch;
first and second active layers: the activation function is a ReLU function, and the infill is True;
a second convolution layer: the number of input channels out _ ch and the number of output channels out _ ch are 3, padding is 1, and stride is 1.
Preferably, the classification module includes a first volume block, a first pooling layer, a second volume block, a second pooling layer, a first full-connection layer and a second full-connection layer, which are sequentially connected and encapsulated.
Preferably, the parameters of the classification module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 16;
first and second pooling layers: adopting maximum pooling, wherein the size of a pooling nucleus is 8;
a first rolling block: the number of input channels is 16, and the number of output channels is 32;
first fully-connected layer: input feature number 960, output feature number 64;
second full connection layer: the feature number 64 is input, and the feature number 1 is output.
Preferably, the dividing module includes a first convolution block, a first pooling layer, a second convolution block, a second pooling layer, a third convolution block, a third pooling layer, a fourth convolution block, a fourth pooling layer, a fifth convolution block, a first transposing convolution layer, a sixth convolution block, a second transposing convolution layer, a seventh convolution block, a third transposing convolution layer, an eighth convolution block, a fourth transposing convolution layer, a ninth convolution block, and a first convolution layer, which are connected and encapsulated in sequence.
Preferably, the parameters of the segmentation module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 32;
a first, second, third, and fourth pooling layer: adopting maximum pooling, wherein the size of a pooling nucleus is 2;
a second rolling block: input channel number 32, output channel number 64;
a third rolling block: input lane number 64, output lane number 128;
a fourth rolling block: input lane number 128, output lane number 256;
a fifth rolling block: input lane number 256, output lane number 512;
a first transfer buildup layer: the number of input channels is 512, the number of output channels is 256, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
a sixth rolling block: splicing the output results of the fourth convolution block and the first transfer convolution layer as input, inputting the channel number 512 and outputting the channel number 256;
the second transpose convolution layer: 256 input channels and 128 output channels, wherein the size of the adopted convolution kernel is 2, padding is 0 and stride is 2;
a seventh rolling block: the output result of the third convolution block and the second transposition convolution layer is spliced as input, the number of input channels is 256, and the number of output channels is 128;
the third transpose convolution layer: the number of input channels is 128, the number of output channels is 64, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
eighth volume block: the output result of the second convolution block and the third transposed convolution layer is spliced as input, the number of input channels is 128, and the number of output channels is 64;
the fourth transpose convolution layer: the number of input channels is 64, the number of output channels is 32, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
ninth volume block: splicing the output results of the first convolution block and the fourth transposed convolution layer as input, and inputting 64 channels and 32 channels;
a first winding layer: the number of input channels 32, the number of output channels 1, the size of the adopted convolution kernel is 1, padding is 0, stride is 1, and the segmentation result is given at the pixel level.
Preferably, the training process model comprises the steps of:
acquiring a training set, wherein the training set comprises images and labels corresponding to the images;
inputting the image into a processing model to obtain a segmentation result, and calculating a loss value and a gradient corresponding to each module parameter of the processing model according to the label;
updating parameters of each module of the processing model according to the gradient;
and repeating the steps until the training iteration threshold of the processing model is reached.
Preferably, the training set of the classification module comprises a carpal tunnel image to be segmented and a corresponding first label, and the first label is a morphological category of the carpal tunnel image to be segmented; the training set of the segmentation module comprises various forms of carpal tunnel images to be segmented and corresponding second labels, and the second labels are carpal tunnel segmentation results marked by professional doctors; the loss value calculation of the classification module adopts a two-classification cross entropy loss function, and the loss value calculation of the segmentation module adopts a Dice loss function.
Preferably, the classification module and the segmentation module are trained by adopting an Adam optimizer, and the learning rate is set to be 0.001.
A neural network-based carpal tunnel image segmentation system, comprising:
the input module is used for acquiring a carpal tunnel image to be segmented;
the preprocessing module is used for preprocessing the carpal tunnel image to be segmented;
the classification module is used for carrying out morphological classification on the preprocessed image;
and the segmentation module is used for segmenting the carpal tunnel image to be segmented according to the form classification result and outputting the segmentation result.
The invention has the beneficial effects that:
(1) according to the method, the carpal tunnel images to be segmented are classified firstly and then segmented, so that the accuracy of the segmentation of the carpal tunnel images by the processing model is improved;
(2) the invention applies the image processing technology of deep learning to the medical image processing based on the processing model constructed and trained by the convolutional neural network, and realizes end-to-end carpal tunnel image segmentation by utilizing the high accuracy of the segmentation result, so that doctors without professional experience can finish the diagnosis of the carpal tunnel syndrome.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a schematic diagram of the system of the present invention;
FIG. 3 is a schematic view of a carpal tunnel image to be segmented in accordance with the present invention;
FIG. 4 is a schematic view of a pre-processed carpal tunnel image to be segmented in accordance with the present invention;
FIG. 5 is a schematic diagram of the operation of the classification module and segmentation module of the present invention;
FIG. 6 is a schematic diagram of the classification module of the present invention;
FIG. 7 is a schematic diagram of the partitioning module operation of the present invention.
In the figure: 1. the device comprises an input module, 2, a preprocessing module, 3, a classification module, 4 and a segmentation module.
Detailed Description
The technical solution of the present invention is further specifically described below by way of specific examples in conjunction with the accompanying drawings. It is to be understood that the practice of the invention is not limited to the following examples, and that any variations and/or modifications may be made thereto without departing from the scope of the invention.
In the present invention, all parts and percentages are by weight, unless otherwise specified, and the equipment and materials used are commercially available or commonly used in the art. The methods in the following examples are conventional in the art unless otherwise specified. The components or devices in the following examples are, unless otherwise specified, standard parts or parts known to those skilled in the art, the structure and principle of which are known to those skilled in the art through technical manuals or through routine experimentation.
Example 1:
fig. 1 shows a neural network-based carpal tunnel image segmentation method, which comprises the following steps:
constructing and training a processing model by using a convolutional neural network, wherein the processing model at least comprises a classification module and a segmentation module;
acquiring a carpal tunnel image to be segmented;
preprocessing a carpal tunnel image to be segmented;
using a classification module to perform morphological classification on the preprocessed image;
and according to the form classification result, using an adaptive segmentation module to segment the carpal tunnel image to be segmented and outputting a segmentation result.
With the method, a carpal tunnel image segmentation system based on a neural network is shown in fig. 2, and comprises:
the input module 1 is used for acquiring a carpal tunnel image to be segmented;
the preprocessing module 2 is used for preprocessing a carpal tunnel image to be segmented;
the classification module 3 is used for carrying out morphological classification on the preprocessed image;
and the segmentation module 4 is used for segmenting the carpal tunnel image to be segmented according to the form classification result and outputting the segmentation result.
By the technical scheme, the carpal tunnel image to be segmented is classified firstly and then segmented, so that the accuracy of segmenting the carpal tunnel image by the processing model is improved.
Example 2:
fig. 1 shows a neural network-based carpal tunnel image segmentation method, which comprises the following steps:
constructing and training a processing model by using a convolutional neural network, wherein the processing model at least comprises a classification module and a segmentation module;
acquiring a carpal tunnel image to be segmented;
preprocessing a carpal tunnel image to be segmented;
using a classification module to perform morphological classification on the preprocessed image;
and according to the form classification result, using an adaptive segmentation module to segment the carpal tunnel image to be segmented and outputting a segmentation result.
The processing model comprises a plurality of segmentation modules, and is used for segmenting the carpal tunnel images to be segmented in different forms according to form classification results and outputting segmentation results. The embodiment selects two segmentation modules: the first segmentation module and the second segmentation module are used for respectively processing and segmenting the carpal tunnel images to be segmented in two forms.
The obtained images of the carpal tunnel to be segmented are shown in fig. 3, and all the images are gray images.
The method for preprocessing the carpal tunnel image to be segmented comprises the following steps: changing the input image to a uniform size 512x 512; keeping the central point of the image unchanged, and cutting the image into a uniform size of 320x 384; adaptive histogram equalization is used on the image. In order to reduce the number of training rounds of the neural network, the example adopts a batch input method to input data into the neural network, so that the sizes of pictures need to be unified. In addition, in order to reduce the calculation amount and memory occupation of the neural network, the original image is subjected to center cropping in the present example, and only the portion of the center containing useful information is intercepted, so that the unified size is 320x 384. To enhance the contrast of the grey scale map, an adaptive histogram equalization technique is used in this example. The picture is modified to a uniform size of 512x512 using the pile library in a Python3 environment and center clipping is performed. And then performing adaptive histogram equalization by using a sketch library. The pre-processed image is shown in fig. 4.
The working of the classification module and the segmentation module is schematically shown in fig. 5. In fig. 5, MRI is an input carpal tunnel image to be segmented, classification is a classification module, Category1 is a first form, Category2 is a second form, Seg1 is a first segmentation module, Seg2 is a second segmentation module, Predict is a segmentation result, and Output is an Output segmentation result.
As shown in fig. 6, the sorting module includes a first stacking block, a first pooling layer, a second stacking block, a second pooling layer, a first full-connection layer, and a second full-connection layer, which are sequentially connected from front to back and encapsulated. In fig. 6, Conv Block is a volume Block, Pooling is a Pooling layer, Flatten is a Flatten layer (for flattening input and unifying multidimensional input), full connectivity is a full Connection layer, and Category is an output morphological classification result.
The parameters of the classification module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 16;
first and second pooling layers: adopting maximum pooling, wherein the size of a pooling nucleus is 8;
a first rolling block: the number of input channels is 16, and the number of output channels is 32;
first fully-connected layer: input feature number 960, output feature number 64;
second full connection layer: the feature number 64 is input, and the feature number 1 is output.
After a preprocessed image is input into the classification module, the classification module processes the preprocessed image to obtain the type of the image form. The specific process is as follows:
inputting a grayscale image of 1x320x 384;
the output shape is 16x320x384 after the first volume block action;
the output shape is 16x40x48 after the first pooling layer;
the output shape is 32x40x48 after the action of the first volume block;
the output shape is 32x5x6 by the action of the first pooling layer;
after the data is converted into one dimension, the obtained shape is 960;
the output shape is 64 under the action of the first full connection layer;
and under the action of the second full-connection layer, the output shape is 2, namely the type of the input image.
As shown in fig. 7, each partitioning module includes a first convolution block, a first pooling layer, a second convolution block, a second pooling layer, a third convolution block, a third pooling layer, a fourth convolution block, a fourth pooling layer, a fifth convolution block, a first transposing convolution layer, a sixth convolution block, a second transposing convolution layer, a seventh convolution block, a third transposing convolution layer, an eighth convolution block, a fourth transposing convolution layer, a ninth convolution block, and a first convolution layer, which are sequentially connected from front to back and encapsulated. In FIG. 7, Conv Block is a convolutional Block, Max Pooling is a Pooling layer (Max Pooling), Transposed Conv is a Transposed convolutional layer, and Skip connection is Skip connection.
The parameters of the segmentation module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 32;
a first, second, third, and fourth pooling layer: adopting maximum pooling, wherein the size of a pooling nucleus is 2;
a second rolling block: input channel number 32, output channel number 64;
a third rolling block: input lane number 64, output lane number 128;
a fourth rolling block: input lane number 128, output lane number 256;
a fifth rolling block: input lane number 256, output lane number 512;
a first transfer buildup layer: the number of input channels is 512, the number of output channels is 256, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
a sixth rolling block: splicing the output results of the fourth convolution block and the first transfer convolution layer as input, inputting the channel number 512 and outputting the channel number 256;
the second transpose convolution layer: 256 input channels and 128 output channels, wherein the size of the adopted convolution kernel is 2, padding is 0 and stride is 2;
a seventh rolling block: the output result of the third convolution block and the second transposition convolution layer is spliced as input, the number of input channels is 256, and the number of output channels is 128;
the third transpose convolution layer: the number of input channels is 128, the number of output channels is 64, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
eighth volume block: the output result of the second convolution block and the third transposed convolution layer is spliced as input, the number of input channels is 128, and the number of output channels is 64;
the fourth transpose convolution layer: the number of input channels is 64, the number of output channels is 32, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
ninth volume block: splicing the output results of the first convolution block and the fourth transposed convolution layer as input, and inputting 64 channels and 32 channels;
a first winding layer: the number of input channels 32, the number of output channels 1, the size of the adopted convolution kernel is 1, padding is 0, stride is 1, and the segmentation result is given at the pixel level.
And calling the corresponding segmentation module to segment the carpal tunnel image to be segmented according to the form classification result of the classification module and outputting the segmentation result. The specific process is as follows:
inputting a grayscale image of 1x320x 384;
the output shape is 32x320x384 under the action of the first volume block, and the output shape is 32x160x192 under the action of the first pooling layer;
the output shape is 64x160x192 through the second convolution block, and the output shape is 64x80x96 through the second pooling layer;
the output shape is 128x80x96 through the third convolution block, and the output shape is 128x40x48 through the third pooling layer;
the output shape is 256x40x48 under the action of a fourth convolution block, and the output shape is 256x20x24 under the action of a fourth pooling layer;
the output shape is 512x20x24 through the function of a fifth convolution block, and the output shape is 256x40x48 through the function of a first transposition convolution layer;
splicing the output of the first transpose convolution layer and the output of the fourth convolution block, wherein the output shape is 256x40x48 through the action of a sixth convolution block, and the output shape is 128x80x96 through the action of the second transpose convolution layer;
splicing the output of the second transpose convolutional layer and the output of the third convolutional block, wherein the output shape is 128x80x96 through the action of a seventh convolutional block, and the output shape is 64x160x192 through the action of the third transpose convolutional layer;
the output of the third transposed convolutional layer and the output of the second convolutional block are spliced, the output shape is 64x160x192 through the action of an eighth convolutional block, and the output shape is 32x320x384 through the action of a fourth transposed convolutional layer;
and splicing the output of the fourth transposed convolution layer and the output of the first convolution block, wherein the output shape is 32x320x384 through the action of the ninth convolution block, and the output shape is 1x320x384 through the action of the first convolution layer, namely the hand carpal tunnel image segmentation result.
Wherein, the structure of all the above-mentioned volume blocks all includes: the first rolling layer, the first batch normalization layer, the first activation layer, the second rolling layer, the second batch normalization layer and the second activation layer are connected and packaged in sequence. The parameters of the volume block are set as:
a first winding layer: inputting a channel number in _ ch and an output channel number out _ ch, wherein the size of the adopted convolution kernel is 3, padding is 1 and stride is 1;
a first batch normalization layer and a second batch normalization layer: the number of channels is out _ ch;
first and second active layers: the activation function is a ReLU function, and the infill is True;
a second convolution layer: the number of input channels out _ ch and the number of output channels out _ ch are 3, padding is 1, and stride is 1.
Wherein, training the treatment model includes the following steps:
acquiring a training set, wherein the training set comprises images and labels corresponding to the images;
inputting the image into a processing model to obtain a segmentation result, and calculating a loss value and a gradient corresponding to each module parameter of the processing model according to the label;
updating parameters of each module of the processing model according to the gradient;
and repeating the steps until the training iteration threshold of the processing model is reached.
The training set of the classification module comprises a carpal tunnel image to be segmented and a corresponding first label, and the first label is the morphology category of the carpal tunnel image to be segmented. The loss value calculation of the classification module adopts a two-classification cross entropy loss function, and the expression is as follows:
where y is the true category of the image,
is the class predicted by the classification module.
The optimizer adopts an Adam optimizer, and the learning rate is set to be 0.001; the number of batch input data is 12, 30 epochs are trained, and after 20 epochs, the learning rate is reduced to one tenth of the original rate.
The training set of the first segmentation module and the training set of the second segmentation module comprise carpal tunnel images to be segmented in two forms and corresponding second labels, and the second labels are carpal tunnel segmentation results marked by professional doctors. The loss value calculation of the segmentation module adopts a Dice loss function, and the expression is as follows:
wherein X is a carpal tunnel segmentation label, Y is a neural network prediction result, | X | and | Y | represent the number of elements in X and Y, respectively, | X | N.Y | represents the number in the intersection of X and Y, and epsilon is a sufficiently small number for ensuring that the denominator is not 0.
The optimizer adopts an Adam optimizer, and the learning rate is set to be 0.001; the number of batch input data is 12, 120 epochs are trained, and after 90 epochs, the learning rate is reduced to one tenth of the original rate.
According to the technical scheme, the processing model constructed and trained on the basis of the convolutional neural network is used for applying the image processing technology of deep learning to medical image processing, and end-to-end carpal tunnel image segmentation is realized by utilizing the high accuracy of segmentation results, so that doctors without professional experience can finish diagnosis of carpal tunnel syndrome.
The above-described embodiments are only preferred embodiments of the present invention, and are not intended to limit the present invention in any way, and other variations and modifications may be made without departing from the spirit of the invention as set forth in the claims.
Claims (10)
1. A carpal tunnel image segmentation method based on a neural network is characterized in that: the method comprises the following steps:
constructing and training a processing model by using a convolutional neural network, wherein the processing model at least comprises a classification module and a segmentation module;
acquiring a carpal tunnel image to be segmented;
preprocessing a carpal tunnel image to be segmented;
using a classification module to perform morphological classification on the preprocessed image;
and according to the form classification result, using an adaptive segmentation module to segment the carpal tunnel image to be segmented and outputting a segmentation result.
2. The neural network-based carpal tunnel image segmentation method as set forth in claim 1, wherein: the processing model comprises a plurality of segmentation modules, and is used for segmenting the carpal tunnel images to be segmented in different forms according to form classification results and outputting segmentation results.
3. The neural network-based carpal tunnel image segmentation method as set forth in claim 1, wherein: the classification module and the segmentation module at least comprise a rolling block, and the rolling block comprises a first rolling layer, a first batch normalization layer, a first activation layer, a second rolling layer, a second batch normalization layer and a second activation layer which are sequentially connected and packaged.
4. The neural network-based carpal tunnel image segmentation method as set forth in claim 1, wherein: the classification module comprises a first volume block, a first pooling layer, a second volume block, a second pooling layer, a first full-connection layer and a second full-connection layer which are sequentially connected and packaged.
5. The neural network-based carpal tunnel image segmentation method as set forth in claim 4, wherein: the parameters of the classification module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 16;
first and second pooling layers: adopting maximum pooling, wherein the size of a pooling nucleus is 8;
a first rolling block: the number of input channels is 16, and the number of output channels is 32;
first fully-connected layer: input feature number 960, output feature number 64;
second full connection layer: the feature number 64 is input, and the feature number 1 is output.
6. The neural network-based carpal tunnel image segmentation method as set forth in claim 1, wherein: the segmentation module comprises a first convolution block, a first pooling layer, a second convolution block, a second pooling layer, a third convolution block, a third pooling layer, a fourth convolution block, a fourth pooling layer, a fifth convolution block, a first transposition convolution layer, a sixth convolution block, a second transposition convolution layer, a seventh convolution block, a third transposition convolution layer, an eighth convolution block, a fourth transposition convolution layer, a ninth convolution block and a first convolution layer which are sequentially connected and packaged.
7. The neural network-based carpal tunnel image segmentation method as set forth in claim 6, wherein: the parameters of the segmentation module are set as follows:
a first rolling block: the number of input channels is 1, and the number of output channels is 32;
a first, second, third, and fourth pooling layer: adopting maximum pooling, wherein the size of a pooling nucleus is 2;
a second rolling block: input channel number 32, output channel number 64;
a third rolling block: input lane number 64, output lane number 128;
a fourth rolling block: input lane number 128, output lane number 256;
a fifth rolling block: input lane number 256, output lane number 512;
a first transfer buildup layer: the number of input channels is 512, the number of output channels is 256, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
a sixth rolling block: splicing the output results of the fourth convolution block and the first transfer convolution layer as input, inputting the channel number 512 and outputting the channel number 256;
the second transpose convolution layer: 256 input channels and 128 output channels, wherein the size of the adopted convolution kernel is 2, padding is 0 and stride is 2;
a seventh rolling block: the output result of the third convolution block and the second transposition convolution layer is spliced as input, the number of input channels is 256, and the number of output channels is 128;
the third transpose convolution layer: the number of input channels is 128, the number of output channels is 64, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
eighth volume block: the output result of the second convolution block and the third transposed convolution layer is spliced as input, the number of input channels is 128, and the number of output channels is 64;
the fourth transpose convolution layer: the number of input channels is 64, the number of output channels is 32, the size of the adopted convolution kernel is 2, padding is 0, and stride is 2;
ninth volume block: splicing the output results of the first convolution block and the fourth transposed convolution layer as input, and inputting 64 channels and 32 channels;
a first winding layer: the number of input channels 32, the number of output channels 1, the size of the adopted convolution kernel is 1, padding is 0, stride is 1, and the segmentation result is given at the pixel level.
8. The neural network-based carpal tunnel image segmentation method as set forth in claim 1, wherein: the training process model comprises the following steps:
acquiring a training set, wherein the training set comprises images and labels corresponding to the images;
inputting the image into a processing model to obtain a segmentation result, and calculating a loss value and a gradient corresponding to each module parameter of the processing model according to the label;
updating parameters of each module of the processing model according to the gradient;
and repeating the steps until the training iteration threshold of the processing model is reached.
9. The neural network-based carpal tunnel image segmentation method as set forth in claim 8, wherein: the training set of the classification module comprises a carpal tunnel image to be segmented and a corresponding first label, wherein the first label is the morphological category of the carpal tunnel image to be segmented; the training set of the segmentation module comprises various forms of carpal tunnel images to be segmented and corresponding second labels, and the second labels are carpal tunnel segmentation results marked by professional doctors; the loss value calculation of the classification module adopts a two-classification cross entropy loss function, and the loss value calculation of the segmentation module adopts a Dice loss function.
10. A carpal tunnel image segmentation system based on a neural network is characterized in that: the carpal tunnel image segmentation system based on the neural network comprises:
the input module is used for acquiring a carpal tunnel image to be segmented;
the preprocessing module is used for preprocessing the carpal tunnel image to be segmented;
the classification module is used for carrying out morphological classification on the preprocessed image;
and the segmentation module is used for segmenting the carpal tunnel image to be segmented according to the form classification result and outputting the segmentation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704545.0A CN113469961A (en) | 2021-06-24 | 2021-06-24 | Neural network-based carpal tunnel image segmentation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704545.0A CN113469961A (en) | 2021-06-24 | 2021-06-24 | Neural network-based carpal tunnel image segmentation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113469961A true CN113469961A (en) | 2021-10-01 |
Family
ID=77872792
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110704545.0A Pending CN113469961A (en) | 2021-06-24 | 2021-06-24 | Neural network-based carpal tunnel image segmentation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113469961A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114742917A (en) * | 2022-04-25 | 2022-07-12 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN117952962A (en) * | 2024-03-25 | 2024-04-30 | 南京科进实业有限公司 | Bone mineral density detection image processing method and system |
-
2021
- 2021-06-24 CN CN202110704545.0A patent/CN113469961A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114742917A (en) * | 2022-04-25 | 2022-07-12 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN114742917B (en) * | 2022-04-25 | 2024-04-26 | 桂林电子科技大学 | CT image segmentation method based on convolutional neural network |
| CN117952962A (en) * | 2024-03-25 | 2024-04-30 | 南京科进实业有限公司 | Bone mineral density detection image processing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111369563B (en) | Semantic segmentation method based on pyramid void convolutional network | |
| CN108898175B (en) | Computer-aided model construction method based on deep learning gastric cancer pathological section | |
| CN108268870B (en) | Multi-scale feature fusion ultrasonic image semantic segmentation method based on counterstudy | |
| WO2022077917A1 (en) | Instance segmentation model sample screening method and apparatus, computer device and medium | |
| CN111161273B (en) | Medical ultrasonic image segmentation method based on deep learning | |
| CN111242288B (en) | Multi-scale parallel deep neural network model construction method for lesion image segmentation | |
| CN109272048B (en) | Pattern recognition method based on deep convolutional neural network | |
| CN111369565B (en) | Digital pathological image segmentation and classification method based on graph convolution network | |
| CN110826596A (en) | Semantic segmentation method based on multi-scale deformable convolution | |
| CN110223304B (en) | Image segmentation method and device based on multipath aggregation and computer-readable storage medium | |
| CN112488234A (en) | End-to-end histopathology image classification method based on attention pooling | |
| CN114494195B (en) | Small sample attention mechanism parallel twin method for fundus image classification | |
| CN112150476A (en) | Coronary artery sequence vessel segmentation method based on space-time discriminant feature learning | |
| CN113469961A (en) | Neural network-based carpal tunnel image segmentation method and system | |
| CN108629772A (en) | Image processing method and device, computer equipment and computer storage media | |
| CN113344933B (en) | Glandular cell segmentation method based on multi-level feature fusion network | |
| CN111161271A (en) | Ultrasonic image segmentation method | |
| CN110930378A (en) | Emphysema image processing method and system based on low data demand | |
| CN108491863A (en) | Color image processing method based on Non-negative Matrix Factorization and convolutional neural networks | |
| CN114037699B (en) | Pathological image classification method, equipment, system and storage medium | |
| CN115100165A (en) | Colorectal cancer T staging method and system based on tumor region CT image | |
| CN114897782A (en) | Gastric cancer pathological section image segmentation prediction method based on generating type countermeasure network | |
| CN110910388A (en) | Cancer cell image segmentation method based on U-Net and density estimation | |
| CN114612709A (en) | Multi-scale target detection method guided by image pyramid characteristics | |
| CN115082840B (en) | Action video classification method and device based on data combination and channel correlation |
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 |