CN111429452A - Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network - Google Patents
Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network Download PDFInfo
- Publication number
- CN111429452A CN111429452A CN202010294218.8A CN202010294218A CN111429452A CN 111429452 A CN111429452 A CN 111429452A CN 202010294218 A CN202010294218 A CN 202010294218A CN 111429452 A CN111429452 A CN 111429452A
- Authority
- CN
- China
- Prior art keywords
- bladder
- image
- layer
- unet
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10132—Ultrasound image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a bladder ultrasonic image segmentation method and a device based on a UNet convolutional neural network, wherein the method comprises the steps of acquiring a bladder ultrasonic image of a human body by utilizing ultrasonic scanning equipment; constructing a training set and a testing set of bladder ultrasonic images, wherein the training set and the testing set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing; constructing a UNet convolutional neural network model, wherein the UNet convolutional neural network model comprises a down-sampling layer and an up-sampling layer; training the constructed UNet network model by using the training set image data of the bladder ultrasonic image to generate a network model, and testing the model effect by using the test set image data of the bladder ultrasonic image; and segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by using the trained UNet network model. The invention can eliminate the interference of the noise on the segmentation of the bladder ultrasonic image and improve the accuracy of the bladder volume calculation.
Description
Technical Field
The invention relates to a bladder ultrasonic image segmentation method and a bladder ultrasonic image segmentation device, in particular to a bladder ultrasonic image segmentation method and a bladder ultrasonic image segmentation device based on a UNet convolutional neural network.
Background
Image segmentation plays an important role in imaging diagnosis, and automatic segmentation can help doctors determine the size and volume of organs or lesions and can quantitatively evaluate the effect before and after treatment. In addition, the identification of organs and lesions is a daily task for imaging physicians, and if the organs and lesions are manually segmented, heavy workload is brought to the physicians.
The bladder is a muscular saccular organ for storing urine in a human body, and the volume of the bladder reflects the amount of urine stored in the human body and is also an important parameter for clinical application of urology. In the diagnosis and treatment of urinary system, the measurement results of residual urine volume of human bladder and full urine volume of bladder are important reference for doctors to diagnose cases.
Methods for non-invasive measurement of bladder volume using ultrasound devices have been widely used clinically, and some intelligent measurement methods are continuously available in the market, but most are semi-automatic measurement methods. In addition, the bladder image difference between different individuals is large, and the image signal-to-noise ratio difference is also large, so that the quality of an ultrasonic image is influenced to a great extent, the difficulty of edge segmentation of the bladder ultrasonic image is increased, the detected bladder boundary is not accurate enough, and the measurement and calculation of the bladder volume are directly influenced. Therefore, a new ultrasonic bladder image segmentation method is needed, which reduces the interference of noise to the image, improves the bladder image segmentation precision, further improves the bladder volume calculation accuracy, and avoids the occurrence of wrong diagnosis by doctors.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the bladder ultrasonic image segmentation caused by noise interference is difficult, and the segmentation boundary is inaccurate.
In order to solve the technical problems, the invention adopts the technical scheme that: a bladder ultrasonic image segmentation method based on a UNet convolutional neural network comprises the following steps,
s10, acquiring a bladder ultrasonic image of a human body by using ultrasonic scanning equipment;
s20, constructing a training set and a test set of the bladder ultrasonic image, wherein the training set and the test set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing;
s30, constructing a UNet convolutional neural network model, wherein the UNet convolutional neural network model comprises a downsampling layer and an upsampling layer;
s40, training the constructed UNet network model by using the image data of the training set of the bladder ultrasonic image to generate a network model, and testing the model effect by using the image data of the testing set of the bladder ultrasonic image;
and S50, segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by using the trained UNet network model.
Further, step S10 specifically includes,
scanning ultrasonic image data of the part where the human bladder is located by using ultrasonic image scanning equipment, and storing an ultrasonic image;
an ultrasound image containing the bladder site is selected from the stored ultrasound images and the other ultrasound images not containing the bladder site are deleted.
Further, in step S20, the training set and the test set of the bladder ultrasound image are constructed by,
segmenting a bladder area in the ultrasonic image by using the acquired bladder ultrasonic image to obtain a segmentation standard image;
simultaneously carrying out amplification, translation, rotation and contrast enhancement on the original bladder ultrasonic image and the segmentation standard image obtained by segmentation to obtain a group of changed bladder ultrasonic images and segmentation standard images, thereby expanding a data set of the bladder ultrasonic images;
the data set of the bladder ultrasound image is divided into a training set and a test set of bladder ultrasound images on a 4:1 scale.
Further, in step S30, the UNet convolutional neural network model includes 5 downsampling layers and 5 upsampling layers, and feature maps output by the 5 downsampling layers are respectively merged with feature maps output by the 5 upsampling layers.
Further, in the 5 upsampling layers and the 5 downsampling layers, each downsampling layer includes 2 convolution operation layers and 1 pooling layer, the size of a convolution kernel in the convolution operation layer of each downsampling layer is 3 × 3, the size of a convolution kernel in the pooling layer is 2 × 2, and the number of convolution kernels in the convolution operation layer of each downsampling operation layer is 32, 64, 128, 256 and 512;
each upsampling layer comprises 1 upsampling operation layer and 2 convolution operation layers, the size of a convolution kernel in the convolution operation layer of each upsampling layer is 3 x 3, the size of a convolution kernel in the upsampling operation layer is 2 x 2, and the number of the convolution kernels in each upsampling operation layer is 512, 256, 128, 64 and 32 respectively;
two convolution operation layers are arranged before the first up-sampling operation layer, and the number of convolution kernels of the two convolution operation layers is 1024.
Further, a Dropout layer is arranged before the pooling layer of the 5 th down-sampling layer and before the up-sampling operation layer of the first up-sampling layer in the UNet convolutional neural network model, and the Dropout discarding rate is set to be 0.5.
Further, all convolution operation layers are followed by an activation layer, and the activation function used by the activation layer is a Re L U function.
Further, in step S40, the UNet network model is trained by using the training set image data of the bladder ultrasound image, and the loss function used in the training is a cross entropy loss function.
Further, in step S40, the training of the constructed UNet network model with the training set image data of the bladder ultrasound image includes,
and training the network by using the image data in batches, adjusting the size of the training batch according to the change of the loss function in the training process, and setting the learning rate to be 0.0001 in the network training.
The invention also provides a bladder ultrasonic image segmentation device based on the UNet convolutional neural network, which comprises,
the image acquisition module is used for acquiring an ultrasonic image of the human bladder by utilizing ultrasonic scanning equipment;
the data set construction module is used for constructing a training set and a test set of the bladder ultrasonic image, wherein the training set and the test set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing;
the model building module is used for building a UNet convolutional neural network model, and the UNet convolutional neural network model comprises a down-sampling layer and an up-sampling layer;
the model training module is used for training the constructed UNet network model by utilizing the image data of the training set of the bladder ultrasonic image to generate a network model and testing the model effect by utilizing the image data of the testing set of the bladder ultrasonic image;
and the image segmentation module is used for segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by utilizing the trained UNet network model.
The invention has the beneficial effects that: according to the invention, after the obtained bladder ultrasonic image is labeled and data enhancement is carried out, a training set and a test set of the bladder ultrasonic image are constructed, the constructed UNet network model is trained by using the image data of the training set of the bladder ultrasonic image to generate the network model, and the model effect is tested by using the image data of the test set of the bladder ultrasonic image, so that the trained network model can eliminate the interference of noise on the segmentation of the bladder ultrasonic image, and the accuracy of bladder volume calculation is improved.
Drawings
The following detailed description of the invention refers to the accompanying drawings.
FIG. 1 is a flow chart of a bladder ultrasound image segmentation method based on a UNet convolutional neural network according to an embodiment of the present invention;
FIG. 2 is a block diagram of a bladder ultrasound image segmentation device based on a UNet convolutional neural network according to an embodiment of the present invention;
FIG. 3 is a diagram of a network model training process according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a UNet network model according to an embodiment of the present invention;
FIG. 5 is a schematic block diagram of a computer apparatus of an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
The first embodiment of the invention is as follows: as shown in fig. 1, a bladder ultrasonic image segmentation method based on UNet convolutional neural network comprises,
s10, acquiring a bladder ultrasonic image of a human body by using ultrasonic scanning equipment;
step S10 specifically includes scanning ultrasound image data of a location where the human bladder is located with an ultrasound image scanning device, and storing an ultrasound image; an ultrasound image containing the bladder site is selected from the stored ultrasound images and the other ultrasound images not containing the bladder site are deleted.
In this step, the bladder ultrasound image obtained by the ultrasound image scanning device includes three parts, a bladder region, an ultrasound sector region, and a background region.
S20, constructing a training set and a test set of the bladder ultrasonic image, wherein the training set and the test set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing;
further, in the step S20, the building process of the training set and the test set of the bladder ultrasound image is to segment the bladder area in the ultrasound image by using the obtained bladder ultrasound image to obtain a segmentation standard image; simultaneously carrying out amplification, translation, rotation and contrast enhancement on the original bladder ultrasonic image and the segmentation standard image obtained by segmentation to obtain a group of changed bladder ultrasonic images and segmentation standard images, thereby expanding a data set of the bladder ultrasonic images; the data set of the bladder ultrasound image is divided into a training set and a test set of bladder ultrasound images on a 4:1 scale.
In this step, the acquired bladder ultrasound image is labeled, and 3 categories in the bladder ultrasound image are labeled by using a via image standard labeling tool, wherein the categories are a background area, an effective bladder area and a sector area except for the bladder. And (4) after the bladder ultrasonic image is labeled by using a via standard labeling tool, generating and storing file data of the categories and the corresponding coordinate points. Reading the file data, detecting and labeling an image by using a pointPolygontest function for detecting the position of a coordinate point in an OpenCV computer vision library, wherein a bladder area is labeled with red, a sector area except the bladder is labeled with green, and a background area is labeled with black, so that an artificially segmented bladder ultrasonic labeling image is generated.
And performing data expansion by using the acquired bladder ultrasonic image and the manually segmented bladder ultrasonic annotation image, wherein the data expansion method is to perform operations of amplification, translation, rotation and contrast enhancement on the conventional image. The bladder ultrasound image magnification should be performed within a suitable range to prevent the image from being magnified too much, thereby losing the effective ultrasound image of the bladder area, such as 0.1-0.5 times magnification. The translation operation of the image should also pay attention to the translation size, so as to prevent the loss of the effective area. The rotation operations of the image are mainly mirror rotation along a vertical axis, rotation by 90 ° and 180 °. The gray scale range of the image is widened through various methods, so that the contrast enhancement operation is carried out on the bladder ultrasonic image. The bladder ultrasonic images are subjected to data expansion by using the methods, so that a bladder ultrasonic image data set with rich samples is established, the diversity of sample characteristics can be enriched by the expanded image data, and the universality and the stability of the algorithm are improved.
The data set is randomly divided into a training set and a testing set of bladder ultrasonic images according to the ratio of 4:1, and is used for training and testing subsequent convolution nerves.
S30, constructing a UNet convolutional neural network model, wherein the UNet convolutional neural network model comprises a downsampling layer and an upsampling layer;
further, in step S30, as shown in fig. 4, the UNet convolutional neural network model includes 5 downsampling layers and 5 upsampling layers, and feature maps output by the 5 downsampling layers are respectively merged with feature maps output by the 5 upsampling layers.
Further, in the 5 upsampling layers and the 5 downsampling layers, each downsampling layer includes 2 convolution operation layers and 1 pooling layer, the size of a convolution kernel in the convolution operation layer of each downsampling layer is 3 × 3, the size of a convolution kernel in the pooling layer is 2 × 2, and the number of convolution kernels in the convolution operation layer of each downsampling operation layer is 32, 64, 128, 256 and 512;
each upsampling layer comprises 1 upsampling operation layer and 2 convolution operation layers, the size of a convolution kernel in the convolution operation layer of each upsampling layer is 3 x 3, the size of a convolution kernel in the upsampling operation layer is 2 x 2, and the number of the convolution kernels in each upsampling operation layer is 512, 256, 128, 64 and 32 respectively;
two convolution operation layers are arranged before the first up-sampling operation layer, and the number of convolution kernels of the two convolution operation layers is 1024.
Further, a Dropout layer is arranged before the pooling layer of the 5 th down-sampling layer and before the up-sampling operation layer of the first up-sampling layer in the UNet convolutional neural network model, and the Dropout discarding rate is set to be 0.5.
Further, all convolution operation layers are followed by an activation layer, and the activation function used by the activation layer is a Re L U function.
In the step, the UNet network is a U-shaped network comprising a down-sampling layer and an up-sampling layer, wherein a structure formed by the down-sampling layer and the down-sampling layer is also called as an encoder-decoder structure, the characteristic information of the bladder ultrasound image is extracted through convolution operation and pooling operation in the down-sampling layer, and then the bladder ultrasound image size is reduced through the up-sampling layer of the network.
The method comprises the steps that a UNet network comprises 5 downsampling layers and 5 upsampling layers, a jump connection mode is used for performing feature fusion operation on feature maps output by the downsampling layers and the upsampling layers, each downsampling layer comprises 2 convolution operation layers and 1 pooling layer, the convolution operation layers use convolution kernel sizes of 3 x 3 and 1, the convolution operation layers perform zero padding on the convolved feature maps in a mode of padding or same for the upper and lower sides of the convolved feature maps to ensure that the sizes of the convolved feature maps are unchanged, an activation layer is added after each convolution operation layer, the activation layer uses a Re L U activation function, the pooling layer uses a maximum pooling function, the convolution kernel size of the pooling layer is 2 x 2, a Dropout layer is added before the pooling layer of the 5 th downsampling layer, the discarding rate of the Dropout layer is 0.5, the addition of the Dropout layer can effectively prevent the network model from being over-filtered, the robustness is enhanced, the upsampling layers mainly comprise the upsampling operation layers, the upsampling operation layers use the characteristic maps with the size of the upper samples, the upsampling layers respectively, the upsampling operation layers, the characteristic maps with the upper samples, the jump connection operation layers, the convolution operation layers contain the upper samples, the characteristic maps, the jump connection operation layers contain the convolution kernel sizes of 256, the convolution kernel sizes of the convolution operation layers are increased by 512, the convolution operation layers, the jump connection operation layers, the convolution operation layers are increased by 512, the jump connection operation layers, the convolution operation layers, the jump connection operation layers, the characteristic maps, the jump connection operation layers are increased by the jump connection operation layers, the characteristic maps of the jump connection operation layers, the convolution operation layers, the jump connection operation.
The down-sampling layer in the network is used for extracting local and global characteristic information of the bladder ultrasonic images through convolution operation and pooling operation, outputting characteristic maps of a plurality of bladder ultrasonic images and extracting an interested region in the bladder ultrasonic images by combining the characteristic information. The 5 down-sampling operations adopt a maximum pooling function, the maximum pooling operation extracts the global features of the image, the dimension reduction is carried out on the data, and the size of the output feature map is half of that of the input feature map. The function of the up-sampling in the network is to reduce the size of the bladder ultrasonic image, and the feature maps output by the down-sampling layer and the up-sampling layer are fused, so that the diversity of image features can be increased, the segmentation precision of the network model can be improved, the diversity of the features can be increased, and the segmentation precision of the network model can be improved.
The activation function adopted by the activation layer in the network is a Re L U activation function, and a Re L U activation layer is connected behind each convolution operation layer in the network, and the Re L U activation layer is used for increasing the nonlinear relation between each layer in the convolution neural network, preventing the occurrence of a network overfitting phenomenon, improving the convergence rate of the network and preventing the problem that the network cannot be trained due to gradient disappearance in back propagation.
The jump connection in the network is to splice and fuse the feature graph output by each down-sampling layer after convolution operation, activation operation and pooling operation and the output feature graph of the up-sampling layer after up-sampling operation, so that the feature information of the image is enriched, the segmentation precision of the network is improved, and the universality of the model is improved. In this example, 5 times of the merging operation of the up-sampling layer and down-sampling layer output feature maps is performed.
S40, as shown in fig. 3, training the constructed UNet network model by using the training set image data of the bladder ultrasonic image to generate a network model, and testing the model effect by using the test set image data of the bladder ultrasonic image;
further, in step S40, the UNet network model is trained by using the training set image data of the bladder ultrasound image, and the loss function used in the training is a cross entropy loss function.
Training the constructed UNet network model by using the image data of the training set of the bladder ultrasonic image comprises the steps of training the network by using the image data in batches, adjusting the size of a training batch according to the change of a loss function in the training process, and setting the learning rate to be 0.0001 in network training.
In this embodiment, a bladder ultrasound image data set is used to perform model training and testing of UNet convolutional neural network, the sizes of the bladder ultrasound image data sets input by the network are normalized to 256 × 256, 1200 bladder ultrasound image data are used as training samples, the training sample set includes an original bladder ultrasound image and an artificial standard bladder segmentation image, and 300 bladder ultrasound images are used as a test sample set.
In this embodiment, the loss function adopted by the UNet convolutional neural network is a cross entropy loss function, the cross entropy is used to evaluate the difference between the predicted probability distribution obtained by training the current network and the actual distribution, and the size of the cross entropy loss function reflects the distance between the actual output probability and the expected output, that is, the smaller the value of the cross entropy, the closer the two probability distributions are, which means the more accurate the prediction is.
The training of the convolutional neural network by utilizing the bladder ultrasonic image comprises two stages, wherein the first stage is a forward propagation stage, the input bladder ultrasonic image data is subjected to convolution operation and pooling operation, characteristic vectors are extracted, and an output result is finally obtained, namely the process of propagation from a low level to a high level. And the second stage is a back propagation stage, when the output result of the current direction propagation is not consistent with the true value, the error between the output predicted value and the true value is back propagated from the high level to the low level, so that the weight and the bias in the network are updated, wherein the calculation function of the error is the cross entropy loss function. Through repeated iterative training of forward propagation and backward propagation, a network model with errors meeting expectations is obtained, so that the network training is stopped and the generated model is stored, the iterative training of 1200 bladder ultrasound image training data is performed for 1 cycle, 60 cycles are performed in total, and a network segmentation model is obtained. Testing the trained network model by using the test set bladder ultrasonic image data, inputting an original bladder ultrasonic image, outputting a bladder segmentation result, counting whether the test result meets the expectation, if so, saving the model, performing the segmentation task of the bladder ultrasonic image by using the network model, if not, adjusting the training strategy of the neural network, and then continuing to train the network by using the training set data until outputting a convolution neural network model meeting the expectation.
The training process of the neural network is an iterative process, and weight initialization has a great influence on the final training result of the convolutional neural network. The too large or too small initial value of the parameter will have bad influence on the convergence result of the network training, the too large initial value of the weight will cause gradient explosion, making the network difficult to converge, the too small initial value of the weight will cause the gradient to disappear, resulting in slow convergence speed of the network, the initialization method adopted in this example is He normal distribution initialization He _ normal, which extracts samples from the normal distribution with 0 as the center and stddev as the standard deviation sqrt (2/fan _ in), where fan _ in is the number of input units in the weight tensor.
The learning rate is an important hyper-parameter in the convolutional neural network, the learning rate determines whether the target function can be converged to a local minimum value and the rate of convergence of the target function to the minimum value, and an appropriate learning rate is selected for network model training, so that the target function can be converged to the local minimum value in an appropriate time. The learning rate in the embodiment of the present application is set to 0.0001.
And S50, segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by using the trained UNet network model.
In the step, a network model is generated through training of a large number of bladder ultrasonic images, the bladder ultrasonic images acquired by ultrasonic equipment are directly segmented by the network model, the segmented bladder ultrasonic images are output, early-stage complex image preprocessing work is avoided, the acquired bladder ultrasonic images can be directly used as the input of the model, and compared with a traditional image processing segmentation method, the UNet convolutional neural network-based bladder ultrasonic image segmentation method reduces the influence of noise on image segmentation and greatly improves the bladder segmentation precision in the bladder ultrasonic images.
As shown in fig. 2, the second embodiment of the present invention is: a bladder ultrasonic image segmentation device based on a UNet convolutional neural network comprises,
an image acquisition module 10, configured to acquire an ultrasound image of a bladder of a human body by using an ultrasound scanning device;
a data set constructing module 20, configured to construct a training set and a test set of bladder ultrasound images, where the training set and the test set include all bladder ultrasound image data subjected to labeling and data enhancement processing;
a model construction module 30, configured to construct a UNet convolutional neural network model, where the UNet convolutional neural network model includes a downsampling layer and an upsampling layer;
the model training module 40 is used for training the constructed UNet network model by using the training set image data of the bladder ultrasonic image to generate a network model, and testing the model effect by using the test set image data of the bladder ultrasonic image;
and the image segmentation module 50 is used for segmenting the actual bladder ultrasonic image acquired by the ultrasonic device by using the trained UNet network model.
Further, the image acquisition module 10 specifically includes,
the image scanning unit is used for scanning ultrasonic image data of the part where the human bladder is located by utilizing ultrasonic image scanning equipment and storing an ultrasonic image;
and the image screening unit is used for selecting the ultrasonic images containing the bladder part from the stored ultrasonic images and deleting other ultrasonic images not containing the bladder part.
Further, the data set building block 20 includes,
the image segmentation unit is used for segmenting the bladder in the ultrasound by using the acquired bladder ultrasound image to obtain a segmentation standard image;
the data set expansion unit is used for simultaneously carrying out amplification, translation, rotation and contrast enhancement on the original bladder ultrasonic image and the segmentation standard image obtained by segmentation to obtain a group of changed bladder ultrasonic images and segmentation standard images so as to expand the data set of the bladder ultrasonic images;
and the data set dividing unit is used for dividing the data set of the bladder ultrasonic image into a training set and a testing set of the bladder ultrasonic image according to the ratio of 4: 1.
It should be noted that, as can be clearly understood by those skilled in the art, for the concrete implementation process of the bladder ultrasound image segmentation apparatus based on the UNet convolutional neural network and each unit, reference may be made to the corresponding description in the foregoing method embodiment, and for convenience and brevity of description, no further description is provided herein.
The apparatus for segmenting an ultrasound image of a bladder based on a UNet convolutional neural network as described above may be implemented in the form of a computer program which can be run on a computer device as shown in fig. 5.
Referring to fig. 5, fig. 5 is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device 500 may be a terminal or a server, where the terminal may be an electronic device with a communication function, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, and a wearable device. The server may be an independent server or a server cluster composed of a plurality of servers.
Referring to fig. 5, the computer device 500 includes a processor 502, memory, and a network interface 505 connected by a system bus 501, where the memory may include a non-volatile storage medium 503 and an internal memory 504.
The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 comprises program instructions that, when executed, cause the processor 502 to perform a method of bladder ultrasound image segmentation based on UNet convolutional neural network.
The processor 502 is used to provide computing and control capabilities to support the operation of the overall computer device 500.
The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503, and when the computer program 5032 is executed by the processor 502, the processor 502 can be enabled to perform a bladder ultrasound image segmentation method based on the UNet convolutional neural network.
The network interface 505 is used for network communication with other devices. Those skilled in the art will appreciate that the configuration shown in fig. 5 is a block diagram of only a portion of the configuration associated with the present application and does not constitute a limitation of the computer device 500 to which the present application may be applied, and that a particular computer device 500 may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
Wherein the processor 502 is configured to run a computer program 5032 stored in the memory to implement the UNet convolutional neural network-based bladder ultrasound image segmentation method as described above.
It should be understood that, in the embodiment of the present Application, the Processor 502 may be a Central Processing Unit (CPU), and the Processor 502 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
It will be understood by those skilled in the art that all or part of the flow of the method implementing the above embodiments may be implemented by a computer program instructing associated hardware. The computer program includes program instructions, and the computer program may be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the flow steps of the embodiments of the method described above.
Accordingly, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program comprises program instructions. The program instructions, when executed by the processor, cause the processor to perform the above UNet convolutional neural network-based bladder ultrasound image segmentation method.
The storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk, which can store various computer readable storage media.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative. For example, the division of each unit is only one logic function division, and there may be another division manner in actual implementation. For example, various elements or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented.
The steps in the method of the embodiment of the invention can be sequentially adjusted, combined and deleted according to actual needs. The units in the device of the embodiment of the invention can be merged, divided and deleted according to actual needs. In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a storage medium. Based on such understanding, the technical solution of the present invention essentially or partially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a terminal, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A bladder ultrasonic image segmentation method based on a UNet convolutional neural network is characterized by comprising the following steps: comprises the steps of (a) preparing a mixture of a plurality of raw materials,
s10, acquiring a bladder ultrasonic image of a human body by using ultrasonic scanning equipment;
s20, constructing a training set and a test set of the bladder ultrasonic image, wherein the training set and the test set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing;
s30, constructing a UNet convolutional neural network model, wherein the UNet convolutional neural network model comprises a downsampling layer and an upsampling layer;
s40, training the constructed UNet network model by using the training set image data of the bladder ultrasonic image to generate a network model, and testing the model effect by using the test set image data of the bladder ultrasonic image;
and S50, segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by using the trained UNet network model.
2. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 1, wherein: the step S10 specifically includes the steps of,
scanning ultrasonic image data of the part where the human bladder is located by using ultrasonic image scanning equipment, and storing an ultrasonic image;
an ultrasound image containing the bladder site is selected from the stored ultrasound images and the other ultrasound images not containing the bladder site are deleted.
3. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 1, wherein: in step S20, the process of constructing the training set and the test set of the bladder ultrasound image is,
segmenting a bladder area in the ultrasonic image by using the acquired bladder ultrasonic image to obtain a segmentation standard image;
simultaneously carrying out amplification, translation, rotation and contrast enhancement on the original bladder ultrasonic image and the segmentation standard image obtained by segmentation to obtain a group of changed bladder ultrasonic images and segmentation standard images, thereby expanding a data set of the bladder ultrasonic images;
the data set of the bladder ultrasound image is divided into a training set and a test set of bladder ultrasound images on a 4:1 scale.
4. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 1, wherein: in step S30, the UNet convolutional neural network model includes 5 downsampling layers and 5 upsampling layers, and feature maps output by the 5 downsampling layers are respectively merged with feature maps output by the 5 upsampling layers.
5. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 1, wherein: in the 5 upsampling layers and the 5 downsampling layers, each downsampling layer comprises 2 convolution operation layers and 1 pooling layer, the size of a convolution kernel in the convolution operation layer of each downsampling layer is 3 x 3, the size of a convolution kernel in the pooling layer is 2 x 2, and the number of convolution kernels in the convolution operation layer of each downsampling operation layer is 32, 64, 128, 256 and 512 respectively;
each upsampling layer comprises 1 upsampling operation layer and 2 convolution operation layers, the size of a convolution kernel in the convolution operation layer of each upsampling layer is 3 x 3, the size of a convolution kernel in the upsampling operation layer is 2 x 2, and the number of the convolution kernels in each upsampling operation layer is 512, 256, 128, 64 and 32 respectively;
two convolution operation layers are arranged before the first up-sampling operation layer, and the number of convolution kernels of the two convolution operation layers is 1024.
6. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 5, wherein: a Dropout layer is provided before the pooling layer of the 5 th downsampling layer and before the upsampling operation layer of the first upsampling layer in the UNet convolutional neural network model, and the Dropout drop rate is set to 0.5.
7. The method for segmenting the ultrasound image of the bladder based on the UNet convolutional neural network as claimed in claim 6, wherein all the convolutional operation layers are followed by an activation layer, and the activation function used by the activation layer is a Re L U function.
8. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 1, wherein: in step S40, the UNet network model is trained using the training set image data of the bladder ultrasound image, and the loss function used in the training is a cross entropy loss function.
9. The UNet convolutional neural network-based bladder ultrasound image segmentation method as claimed in claim 8, wherein: in step S40, training the constructed UNet network model using the training set image data of the bladder ultrasound image includes,
and training the network by using the image data in batches, adjusting the size of the training batch according to the change of the loss function in the training process, and setting the learning rate to be 0.0001 in the network training.
10. A bladder ultrasonic image segmentation device based on a UNet convolutional neural network is characterized in that: comprises the steps of (a) preparing a mixture of a plurality of raw materials,
the image acquisition module is used for acquiring an ultrasonic image of the human bladder by utilizing ultrasonic scanning equipment;
the data set construction module is used for constructing a training set and a test set of the bladder ultrasonic image, wherein the training set and the test set comprise all bladder ultrasonic image data subjected to labeling and data enhancement processing;
the model building module is used for building a UNet convolutional neural network model, and the UNet convolutional neural network model comprises a down-sampling layer and an up-sampling layer;
the model training module is used for training the constructed UNet network model by using the training set image data of the bladder ultrasonic image to generate a network model and testing the model effect by using the test set image data of the bladder ultrasonic image;
and the image segmentation module is used for segmenting the actual bladder ultrasonic image acquired by the ultrasonic equipment by utilizing the trained UNet network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010294218.8A CN111429452A (en) | 2020-04-15 | 2020-04-15 | Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010294218.8A CN111429452A (en) | 2020-04-15 | 2020-04-15 | Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111429452A true CN111429452A (en) | 2020-07-17 |
Family
ID=71558019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010294218.8A Pending CN111429452A (en) | 2020-04-15 | 2020-04-15 | Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111429452A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112132813A (en) * | 2020-09-24 | 2020-12-25 | 中国医学科学院生物医学工程研究所 | Skin ultrasonic image segmentation method based on improved UNet network model |
| CN112164022A (en) * | 2020-08-24 | 2021-01-01 | 同济大学 | Skin injury area identification method applied to venous blood collection robot |
| CN112435198A (en) * | 2020-12-03 | 2021-03-02 | 西安交通大学 | Welding seam radiographic inspection negative image enhancement method, storage medium and equipment |
| CN112767355A (en) * | 2021-01-20 | 2021-05-07 | 北京小白世纪网络科技有限公司 | Method and device for constructing thyroid nodule Tirads grading automatic identification model |
| CN112773508A (en) * | 2021-02-04 | 2021-05-11 | 清华大学 | Robot operation positioning method and device |
| CN113298786A (en) * | 2021-05-26 | 2021-08-24 | 北京长木谷医疗科技有限公司 | Image recognition and model training method, and true mortar position recognition method and device |
| CN113409224A (en) * | 2021-07-09 | 2021-09-17 | 浙江大学 | Image target pertinence enhancing method, device, equipment and storage medium |
| CN113838083A (en) * | 2020-10-19 | 2021-12-24 | 南京航空航天大学 | Multi-scale enhanced UNet passive terahertz image fixed background subtraction method |
| CN114170128A (en) * | 2020-08-21 | 2022-03-11 | 张逸凌 | Bone segmentation method and system based on deep learning |
| CN114581463A (en) * | 2022-03-04 | 2022-06-03 | 中国科学院近代物理研究所 | Multi-phase 4D CT image segmentation method and system |
| CN115376205A (en) * | 2022-07-25 | 2022-11-22 | 武汉库柏特科技有限公司 | Ultrasonic scanning robot control method, device, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308695A (en) * | 2018-09-13 | 2019-02-05 | 镇江纳兰随思信息科技有限公司 | Based on the cancer cell identification method for improving U-net convolutional neural networks model |
| CN110136147A (en) * | 2019-05-21 | 2019-08-16 | 湖北工业大学 | A kind of method, apparatus and storage medium of the Medical Image Segmentation based on U-Net model |
| WO2019200753A1 (en) * | 2018-04-17 | 2019-10-24 | 平安科技（深圳）有限公司 | Lesion detection method, device, computer apparatus and storage medium |
-
2020
- 2020-04-15 CN CN202010294218.8A patent/CN111429452A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019200753A1 (en) * | 2018-04-17 | 2019-10-24 | 平安科技（深圳）有限公司 | Lesion detection method, device, computer apparatus and storage medium |
| CN109308695A (en) * | 2018-09-13 | 2019-02-05 | 镇江纳兰随思信息科技有限公司 | Based on the cancer cell identification method for improving U-net convolutional neural networks model |
| CN110136147A (en) * | 2019-05-21 | 2019-08-16 | 湖北工业大学 | A kind of method, apparatus and storage medium of the Medical Image Segmentation based on U-Net model |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114170128A (en) * | 2020-08-21 | 2022-03-11 | 张逸凌 | Bone segmentation method and system based on deep learning |
| CN112164022A (en) * | 2020-08-24 | 2021-01-01 | 同济大学 | Skin injury area identification method applied to venous blood collection robot |
| CN112132813B (en) * | 2020-09-24 | 2022-08-05 | 中国医学科学院生物医学工程研究所 | Skin ultrasonic image segmentation method based on improved UNet network model |
| CN112132813A (en) * | 2020-09-24 | 2020-12-25 | 中国医学科学院生物医学工程研究所 | Skin ultrasonic image segmentation method based on improved UNet network model |
| CN113838083A (en) * | 2020-10-19 | 2021-12-24 | 南京航空航天大学 | Multi-scale enhanced UNet passive terahertz image fixed background subtraction method |
| CN113838083B (en) * | 2020-10-19 | 2024-04-05 | 南京航空航天大学 | Multi-scale enhanced UNet passive terahertz image fixed background subtraction method |
| CN112435198A (en) * | 2020-12-03 | 2021-03-02 | 西安交通大学 | Welding seam radiographic inspection negative image enhancement method, storage medium and equipment |
| CN112767355A (en) * | 2021-01-20 | 2021-05-07 | 北京小白世纪网络科技有限公司 | Method and device for constructing thyroid nodule Tirads grading automatic identification model |
| CN112773508A (en) * | 2021-02-04 | 2021-05-11 | 清华大学 | Robot operation positioning method and device |
| CN113298786A (en) * | 2021-05-26 | 2021-08-24 | 北京长木谷医疗科技有限公司 | Image recognition and model training method, and true mortar position recognition method and device |
| CN113409224A (en) * | 2021-07-09 | 2021-09-17 | 浙江大学 | Image target pertinence enhancing method, device, equipment and storage medium |
| CN114581463A (en) * | 2022-03-04 | 2022-06-03 | 中国科学院近代物理研究所 | Multi-phase 4D CT image segmentation method and system |
| CN114581463B (en) * | 2022-03-04 | 2024-05-17 | 中国科学院近代物理研究所 | Multi-phase 4D CT image segmentation method and system |
| CN115376205A (en) * | 2022-07-25 | 2022-11-22 | 武汉库柏特科技有限公司 | Ultrasonic scanning robot control method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111429452A (en) | Bladder ultrasonic image segmentation method and device based on UNet convolutional neural network | |
| WO2022095895A1 (en) | Vascular stenosis analysis method and apparatus | |
| CN109410216B (en) | Ischemic stroke image region segmentation method and device | |
| CN111291825B (en) | Focus classification model training method, apparatus, computer device and storage medium | |
| CN111429451B (en) | Medical ultrasonic image segmentation method and device | |
| CN111951344B (en) | Magnetic resonance image reconstruction method based on cascade parallel convolution network | |
| CN108846829B (en) | Lesion site recognition device, computer device, and readable storage medium | |
| CN113298831B (en) | Image segmentation method and device, electronic equipment and storage medium | |
| Li et al. | Automated measurement network for accurate segmentation and parameter modification in fetal head ultrasound images | |
| CN109949318B (en) | Full convolution neural network epileptic focus segmentation method based on multi-modal image | |
| JP2023540910A (en) | Connected Machine Learning Model with Collaborative Training for Lesion Detection | |
| WO2023241031A1 (en) | Deep learning-based three-dimensional intelligent diagnosis method and system for osteoarthritis | |
| CN114581628B (en) | Cerebral cortex surface reconstruction method and readable storage medium | |
| CN113077479A (en) | Automatic segmentation method, system, terminal and medium for acute ischemic stroke focus | |
| CN110751187B (en) | Training method of abnormal area image generation network and related product | |
| CN110910335B (en) | Image processing method, image processing device and computer readable storage medium | |
| CN114758137B (en) | Ultrasonic image segmentation method and device and computer readable storage medium | |
| CN112465754B (en) | 3D medical image segmentation method and device based on layered perception fusion and storage medium | |
| CN113935943A (en) | Method, device, computer equipment and storage medium for intracranial aneurysm identification detection | |
| CN112750137A (en) | Liver tumor segmentation method and system based on deep learning | |
| CN113378984A (en) | Medical image classification method, system, terminal and storage medium | |
| CN113223015A (en) | Vascular wall image segmentation method, device, computer equipment and storage medium | |
| CN115439651A (en) | DSA (digital Signal amplification) cerebrovascular segmentation system and method based on space multi-scale attention network | |
| CN113724203B (en) | Model training method and device applied to target feature segmentation in OCT image | |
| CN113313728B (en) | Intracranial artery segmentation method and system |
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 |