CN111461233A - Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network - Google Patents
Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network Download PDFInfo
- Publication number
- CN111461233A CN111461233A CN202010255476.5A CN202010255476A CN111461233A CN 111461233 A CN111461233 A CN 111461233A CN 202010255476 A CN202010255476 A CN 202010255476A CN 111461233 A CN111461233 A CN 111461233A
- Authority
- CN
- China
- Prior art keywords
- stm
- data set
- mdc
- densenet
- network
- 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
- 238000005481 NMR spectroscopy Methods 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims description 46
- 210000004556 brain Anatomy 0.000 claims description 36
- 238000012360 testing method Methods 0.000 claims description 34
- 238000007781 pre-processing Methods 0.000 claims description 28
- 238000012795 verification Methods 0.000 claims description 28
- 238000011176 pooling Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 210000001175 cerebrospinal fluid Anatomy 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 3
- 208000024827 Alzheimer disease Diseases 0.000 description 18
- 238000004574 scanning tunneling microscopy Methods 0.000 description 16
- 208000010877 cognitive disease Diseases 0.000 description 10
- 230000019771 cognition Effects 0.000 description 9
- 208000027061 mild cognitive impairment Diseases 0.000 description 9
- 238000010586 diagram Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 7
- 238000003745 diagnosis Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000002595 magnetic resonance imaging Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000003759 clinical diagnosis Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 208000028698 Cognitive impairment Diseases 0.000 description 1
- 206010039966 Senile dementia Diseases 0.000 description 1
- 210000004727 amygdala Anatomy 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000003920 cognitive function Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 210000001320 hippocampus Anatomy 0.000 description 1
- 230000002427 irreversible effect Effects 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 230000004770 neurodegeneration Effects 0.000 description 1
- 208000015122 neurodegenerative disease Diseases 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 210000004872 soft tissue Anatomy 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000004304 visual acuity Effects 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a nuclear magnetic resonance image automatic classification method and device based on an MDC L STM-L DenseNet network, and belongs to the field of medical image processing.
Description
Technical Field
The invention relates to the field of medical image processing, in particular to a nuclear magnetic resonance image automatic classification method and device based on an MDC L STM-L DenseNet network.
Background
Alzheimer's Disease (AD), also known as senile dementia, is an irreversible cerebral neurodegenerative disease. AD is high in the elderly, and patients often have symptoms such as memory deterioration and cognitive impairment until losing daily life capacity. Clinical diagnosis of alzheimer's disease is mainly based on medical images and clinical indicators. In medical image-based diagnosis, a doctor diagnoses by observing morphology of a brain-related region, particularly morphology information of a hippocampus, an amygdala and other regions; in clinical index-based diagnosis, physicians perform diagnosis by measuring biomarkers in cerebrospinal fluid, in combination with cognitive function scales. Generally, the diagnosis of alzheimer's disease can be viewed as a classification problem, i.e. to determine to which of normal cognition, mild cognitive impairment and alzheimer's disease a subject belongs.
Currently, Magnetic Resonance Imaging (MRI) is widely used in clinical diagnosis of alzheimer's disease, MRI is a 3D image composed of a series of 2D slice images, and has the characteristics of high resolution, high contrast, strong soft tissue resolving power, and the like, and provides a powerful aid for diagnosis of alzheimer's disease.
One solution to the above problem is to use a deep learning based approach, and it is not simple to diagnose alzheimer's disease using deep learning techniques: firstly, a common deep learning method is widely used for natural image processing, and medical images and natural images have great difference; secondly, MRI is a 3D image, and continuous change information between internal slices thereof, particularly continuous information of a lesion region, needs to be considered; finally, convolutional neural networks generally have a more complex structure and huge parameters, and the alzheimer data set is small in size and is prone to overfitting.
Disclosure of Invention
According to the problems in the prior art, the invention discloses a nuclear magnetic resonance image automatic classification method based on an MDC L STM-L DenseNet network, which comprises the following steps:
s1, acquiring nuclear magnetic resonance images of different types of subjects;
s2, preprocessing the nuclear magnetic resonance image to obtain a training data set, a verification data set and a test data set;
s3, constructing an MDC L STM-L Densenet network model by combining a cavity convolution and a Conv L STM model on the basis of a Densenet-121 network;
s4, inputting the training data set into an MDC L STM-L DenseNet network model for training, and adjusting the hyper-parameters of the MDC L STM-L DenseNet network model by using the verification data set to obtain the trained MDC L STM-L DenseNet network model;
and S5, inputting the test data set into the MDC L STM-L Densennet network model after training for testing, and obtaining the classification result of the subject corresponding to the nuclear magnetic resonance image and the classification accuracy of the MDC L STM-L Densennet network model.
Further, the process of preprocessing the nuclear magnetic resonance image is as follows:
s2-1, performing head correction, registration and segmentation on the nuclear magnetic resonance image to obtain three images of grey brain matter, white brain matter and cerebrospinal fluid;
s2-2, carrying out space standardization and Gaussian smoothing operation on the grey brain matter image to obtain a processed grey brain matter image;
s2-3, dividing the processed gray brain matter image to obtain a training data set, a verification data set and a test data set;
and S2-4, respectively performing data expansion on the training data set, the verification data set and the test data set to obtain the training data set, the verification data set and the test data set after the data expansion.
Further, the constructed MDC L STM-L DenseNet network model has the following process:
s3-1, on the basis of a DenseNet-121 network, expanding the DenseNet-121 network to a 3D-DenseNet-121 network, reducing the number of dense blocks of the 3D-DenseNet-121 network, reducing the number of convolutional layers in the remaining dense blocks, connecting the output of the third dense block to a global pooling layer, and connecting the output of the global pooling layer to an output layer to obtain a 3D light dense convolutional network;
s3-2, adding a hole convolution and a convolution duration memory network between adjacent dense blocks of the 3D light dense convolution network to form an MDC L STM-L DenseNet network model.
Further: the system comprises an image acquisition unit, an image preprocessing unit, a network building unit and an image classification unit;
the image acquisition unit acquires brain nuclear magnetic resonance images of different types of subjects;
the image preprocessing unit preprocesses the brain nuclear magnetic resonance images of the subjects of different classes acquired by the image acquisition unit to obtain a training data set, a verification data set and a test data set;
the network building unit inputs the training data set obtained by the image preprocessing unit into a built MDC L STM-L Densennet network model for training, and the verification data set obtained by the image preprocessing unit is used for adjusting the hyper-parameters of the MDC L STM-L Densennet network model to obtain the trained MDC L STM-L Densennet network model;
and the image classification unit inputs the test data set obtained by the preprocessing unit into the MDC L STM-L DenseNet network model obtained by the network building unit for testing to obtain the classification result of the subject corresponding to the nuclear magnetic resonance image and the classification accuracy of the MDC L STM-L DenseNet network model.
By adopting the technical scheme, the method and the device for automatically classifying the nuclear magnetic resonance images based on the MDC L STM-L DenseNet network provided by the invention integrate the feature extraction and classification uniformly to form an end-to-end deep neural network for training, expand the DenseNet-121 network to the 3D-DenseNet-121 network, add the void volume and the memory network when the convolution is long and short between adjacent dense blocks of the obtained 3D light dense convolutional network to form an MDC L STM-L DenseNet network model, not only enhance the extraction of multi-scale spatial features, but also enhance the learning of continuous change information between slices, and fuse the two information to enhance the learning capacity of the network on a focus area.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be 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 some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a process flow diagram;
FIG. 2 is a diagram of a MDC L STM-L DenseNet model;
FIG. 3 is a block diagram of the dense block in the MDC L STM-L DenseNet model;
fig. 4 is a structure diagram of MDC L STM in the MDC L STM-L DenseNet network model.
Detailed Description
In order to make the technical solutions and advantages of the present invention clearer, the following describes the technical solutions in the embodiments of the present invention clearly and completely with reference to the drawings in the embodiments of the present invention:
FIG. 1 is a flow chart of a method, and a nuclear magnetic resonance image automatic classification method based on an MDC L STM-L DenseNet network comprises the following steps:
s1, acquiring brain nuclear magnetic resonance images of three subjects of normal cognition, mild cognitive impairment and Alzheimer disease;
s2, preprocessing the nuclear magnetic resonance image to obtain a training data set, a verification data set and a test data set;
s3, constructing an MDC L STM-L Densenet network model by combining a cavity convolution and a Conv L STM model on the basis of a Densenet-121 network;
s4, inputting the training data set into an MDC L STM-L DenseNet network model for training, and adjusting the hyper-parameters of the MDC L STM-L DenseNet network model by using the verification data set to obtain the trained MDC L STM-L DenseNet network model;
and S5, inputting the test data set into the MDC L STM-L Densennet network model after training for testing, and obtaining the classification results of the nuclear magnetic resonance images of three types of subjects with normal cognition, mild cognitive impairment and Alzheimer' S disease and the classification accuracy of the MDC L STM-L Densennet network model.
Further, the process of preprocessing the magnetic resonance image is as follows:
s2-1, performing head correction, registration and segmentation on the NIFTI-format nuclear magnetic resonance image to obtain three images of grey brain matter, white brain matter and cerebrospinal fluid;
s2-2, carrying out space standardization and Gaussian smoothing operation on the grey brain matter image to obtain a processed grey brain matter image;
s2-3, dividing the processed gray brain matter image to obtain a training data set, a verification data set and a test data set;
and S2-4, respectively performing data expansion on the training data set, the verification data set and the test data set to obtain the training data set, the verification data set and the test data set after the data expansion.
Further, the process of the constructed MDC L STM-L densnet network model is as follows:
s3-1, on the basis of a DenseNet-121 network, expanding the DenseNet-121 network to a 3D-DenseNet-121 network, reducing the number of dense blocks of the 3D-DenseNet-121 network, reducing the number of convolutional layers in the remaining dense blocks, connecting the output of the third dense block to a global pooling layer, and connecting the output of the global pooling layer to an output layer to obtain a 3D light dense convolutional network;
s3-2, adding a hole convolution and a convolution duration memory network between adjacent dense blocks of the 3D light dense convolution network to form an MDC L STM-L DenseNet network model.
An automatic nuclear magnetic resonance image classification device based on an MDC L STM-L DenseNet network comprises an image acquisition unit, an image preprocessing unit, a network construction unit and an image classification unit;
the image acquisition unit acquires brain nuclear magnetic resonance images of different types of subjects;
the image preprocessing unit preprocesses the brain nuclear magnetic resonance images of the subjects of different classes acquired by the image acquisition unit to obtain a training data set, a verification data set and a test data set;
the network building unit inputs the training data set obtained by the image preprocessing unit into a built MDC L STM-L Densennet network model for training, and the verification data set obtained by the image preprocessing unit is used for adjusting the hyper-parameters of the MDC L STM-L Densennet network model to obtain the trained MDC L STM-L Densennet network model;
and the image classification unit inputs the test data set obtained by the preprocessing unit into the MDC L STM-L DenseNet network model obtained by the network building unit for testing to obtain the classification results of the nuclear magnetic resonance images of three subjects with normal cognition, mild cognitive impairment and Alzheimer's disease and the classification accuracy of the MDC L STM-L DenseNet network model.
Embodiment 1, a nuclear magnetic resonance image automatic classification method based on MDC L STM-L densnet network, comprising the following steps,
s1, acquiring brain nuclear magnetic resonance images of three types of subjects with normal cognition, mild cognitive impairment and Alzheimer 'S disease, wherein a nuclear magnetic resonance image data set is selected from an ADNI data set, and 571 subjects with data are selected, wherein 192 subjects with Alzheimer' S disease, 171 subjects with mild cognitive impairment and 208 subjects with normal cognition are selected;
s2, preprocessing the nuclear magnetic resonance image to obtain a training data set, a verification data set and a test data set; the pretreatment process is as follows:
s2-1, preprocessing the nuclear magnetic resonance image obtained in the step S1 by using SPM12, and performing head correction, registration and segmentation on the nuclear magnetic resonance image in the NIFTI format to obtain three images of grey brain matter, white brain matter and cerebrospinal fluid;
s2-2, spatially normalizing and 3 × 3 gaussian smoothing the gray brain matter image with size 256 × 166 × 256, resulting in a gray brain matter image with size 121 × 145 × 121;
s2-3, resizing the grey brain matter image to obtain a grey brain matter image data set with the size of 112 × 112;
s2-4, dividing the gray brain matter image data set according to the ratio of 8:1:1 to obtain a training data set, a verification data set and a test data set;
and S2-5, respectively performing data expansion on the training data set, the verification data set and the test data set, wherein the three expanded data sets comprise 5710 images.
S3, constructing an MDC L STM-L DenseneNet model by combining a cavity convolution and a Conv L STM model on the basis of a DenseneET-121 net, wherein the process of the constructed MDC L STM-L DenseneNet model is as follows:
s3-1, based on the DenseNet-121 network, expanding the DenseNet-121 network to the 3D-DenseNet-121 network, removing any one dense block in the 3D-DenseNet-121 network, uniformly reducing the number of convolution layers of 1 x 1 and 3 x 3 in each dense block to 4 layers for the remaining three dense blocks, connecting the output of the third dense block to the global pooling layer, and connecting the output of the global pooling layer to the output layer to obtain the 3D light dense convolution network;
and S3-2, adding an MDC L STM module consisting of a hole volume and Conv L STM between the first dense block and the second dense block and between the second dense block and the third dense block of the 3D light dense convolutional network, and keeping other structures of the network unchanged to form an MDC L STM-L DenseNet network model.
Table 1 shows the overall structure of the MDC L STM-L DenseNet model, as follows
Fig. 2 is a diagram of an MDC L STM-L densnet model, the first layer of the MDC L STM-L densnet model is a convolutional layer with 32 convolutional kernels, the convolutional kernels are 7 × 7, the output of the first layer is connected to the second layer, the second layer is a pooling layer, the pooling kernels are 2 × 2, the output of the second layer is connected to a first dense block, the output of the first dense block is connected to a first MDC L STM, the output of the first MDC L STM is connected to a second dense block, the output of the second dense block is connected to a second MDC L STM, the output of the second MDC L STM is connected to a third dense block, the output of the third dense block is connected to the penultimate layer, the penultimate layer is a global pooling layer, the size of the pooling kernels is 7 × 7, the output of the second MDC L STM is connected to a third dense block, the output of the second dense block is connected to a softnet layer, and the global output of the pool is an activation function of the softnet layer.
FIG. 3 is a diagram of dense blocks in an MDC L STM-L DenseNet model, the MDC L STM-L DenseNet model has the same internal structure of three dense blocks, each dense block is composed of 4 sets of convolution operations, each set is formed by stacking Batch Normalization (BN), linear correction Unit (Rectified L initial Unit, Re L U), convolution kernel size 1 x 1 convolution layer and 3 x 3 convolution layer, wherein BN layer and Re L U are used to prevent gradient diffusion and maintain network nonlinearity, 1 x 1 convolution layer is also called bottleneck layer, which can reduce the number of output feature maps, achieve the goal of dimensionality reduction and computational reduction, and 3 x 3 nuclear magnetic resonance layer is a current convolution layer for slice extraction and continuous change between slices, and 3 x 3 convolution layer is a network that is used to effectively enhance the ability of each dense block, and not only to enhance the ability of each previous convolution layer, but also to enhance the ability of each previous convolution layer by using the dense layer 3, thus not only enhancing the ability of the previous convolution layer, but also enhancing the ability of each previous convolution layer.
FIG. 4 is a diagram of an MDC L STM structure in an MDC L STM-L Densennet network model, wherein the internal structures of two MDC L STMs in the MDC L STM-L Densennet network model are the same, each MDC L STM module is composed of a multi-scale hole Convolution, a Convolution layer with Convolution kernel size of 1, a Conv L STM and an average pooling layer, wherein the multi-scale hole Convolution comprises 3 hole Convolution layers (scaled Convolution) with different expansion rates, the three hole Convolution layers perform feature extraction on the output of the previous dense block in a parallel mode, the purpose of expanding the network receptive field is achieved by setting the expansion rates with different sizes, then the image is subjected to feature extraction under the receptive fields with different sizes to obtain a multi-scale feature map, the outputs of the three hole Convolution layers are connected with the output of the previous dense block, the three hole Convolution layers are input into the Convolution layer with kernel size of 1, the feature extraction is further performed on the image under the receptive field with different sizes, the three hole Convolution layers are connected with the output of the previous dense block, the Conv Convolution layer, the output of the Conv is connected to an average Convolution layer output of a subsequent Convolution layer, the Conv 26, and the subsequent Convolution operation is performed on the subsequent Convolution layer, and the subsequent Convolution operation is performed on the subsequent Convolution operation of the subsequent Convolution layer, the subsequent information is changed by the subsequent Convolution operation, wherein the subsequent Conv 26, the subsequent Convolution operation, and the subsequent Convolution.
The MDC L STM-L DenseNet model training process comprises the following steps that a TensorFlow deep learning platform is utilized, an optimization function is Adam, a basic learning rate is set to be 0.0005, an attenuation rate is 1e-4, a training period is 500epoch, one epoch means traversing all samples in a training data set once, a growth rate k is set to be 16, and a GPU is adopted for accelerated training.
Formal training, namely, the integral network model integrates feature extraction and classification, the training process is end-to-end, 3D grey brain matter images and labels in a training data set are used as input, the 3D grey brain matter images with the size of 112 × 112 are input into the MDC L STM-L DenseNet network model, a series of convolution operation is performed, classification results of three subjects with normal cognition, mild cognitive impairment and Alzheimer's disease are obtained through a Softmax activation function, the classification results and the labels are substituted into a cross entropy loss function, the loss value of the training is calculated, back propagation is performed according to the loss value, the weight parameters in the MDC L STM-L DenseNet network model are updated, each time an epoch training is completed, the verification data set is used for verifying the STM L-L DenseNet network model, the hyper parameters of the model are adjusted according to the verification results until the classification performance of the model reaches the best.
And finally, taking the 3D grey brain matter image and the label in the test data set as input, sending the input into a trained MDC L STM-L DensenNet network model to obtain the classification results of the nuclear magnetic resonance images of three types of subjects with normal cognition, mild cognitive impairment and Alzheimer's disease, and calculating evaluation indexes for measuring model performance according to the classification results and the label, wherein the evaluation indexes comprise accuracy, sensitivity, specificity and AUC (AUC) values.
An automatic nuclear magnetic resonance image classification device based on an MDC L STM-L DenseNet model comprises an image acquisition unit, an image preprocessing unit, a network building unit and an image classification unit;
the image acquisition unit acquires brain nuclear magnetic resonance images of different types of subjects;
the image preprocessing unit preprocesses the brain nuclear magnetic resonance images of the subjects of different classes acquired by the image acquisition unit to obtain a training data set, a verification data set and a test data set;
the network building unit inputs the training data set obtained by the image preprocessing unit into a built MDC L STM-L Densennet network model for training, and the verification data set obtained by the image preprocessing unit is used for adjusting the hyper-parameters of the MDC L STM-L Densennet network model to obtain the trained MDC L STM-L Densennet network model;
and the image classification unit inputs the test data set obtained by the preprocessing unit into the MDC L STM-L DenseNet network model obtained by the network building unit for testing to obtain the classification results of the nuclear magnetic resonance images of three subjects with normal cognition, mild cognitive impairment and Alzheimer's disease and the classification accuracy of the MDC L STM-L DenseNet network model.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (4)
1. An automatic nuclear magnetic resonance image classification method based on an MDC L STM-L DenseNet is characterized by comprising the following steps:
s1, acquiring nuclear magnetic resonance images of different types of subjects;
s2, preprocessing the nuclear magnetic resonance image to obtain a training data set, a verification data set and a test data set;
s3, constructing an MDC L STM-L Densenet network model by combining a cavity convolution and a Conv L STM model on the basis of a Densenet-121 network;
s4, inputting the training data set into an MDC L STM-L DenseNet network model for training, and adjusting the hyper-parameters of the MDC L STM-L DenseNet network model by using the verification data set to obtain the trained MDC L STM-L DenseNet network model;
and S5, inputting the test data set into the MDC L STM-L DenseNet network model after training for testing to obtain the classification result of the nuclear magnetic resonance image and the classification accuracy of the MDC L STM-L DenseNet network model.
2. The method for automatically classifying the nuclear magnetic resonance images based on the MDC L STM-L DenseNet network is further characterized in that the process of preprocessing the nuclear magnetic resonance images is as follows:
s2-1, performing head correction, registration and segmentation on the nuclear magnetic resonance image to obtain three images of grey brain matter, white brain matter and cerebrospinal fluid;
s2-2, carrying out space standardization and Gaussian smoothing operation on the grey brain matter image to obtain a processed grey brain matter image;
s2-3, dividing the processed gray brain matter image to obtain a training data set, a verification data set and a test data set;
and S2-4, respectively performing data expansion on the training data set, the verification data set and the test data set to obtain the training data set, the verification data set and the test data set after the data expansion.
3. The method for automatically classifying nuclear magnetic resonance images based on the MDC L STM-L DenseNet network is further characterized in that the constructed MDC L STM-L DenseNet network model is processed as follows:
s3-1, on the basis of a DenseNet-121 network, expanding the DenseNet-121 network to a 3D-DenseNet-121 network, reducing the number of dense blocks of the 3D-DenseNet-121 network, reducing the number of convolutional layers in the remaining dense blocks, connecting the output of the third dense block to a global pooling layer, and connecting the output of the global pooling layer to an output layer to obtain a 3D light dense convolutional network;
s3-2, adding a hole convolution and a convolution duration memory network between adjacent dense blocks of the 3D light dense convolution network to form an MDC L STM-L DenseNet network model.
4. An automatic nuclear magnetic resonance image classification device based on an MDC L STM-L DenseNet network is characterized by comprising an image acquisition unit, an image preprocessing unit, a network construction unit and an image classification unit;
the image acquisition unit acquires brain nuclear magnetic resonance images of different types of subjects;
the image preprocessing unit preprocesses the brain nuclear magnetic resonance images of the subjects of different classes acquired by the image acquisition unit to obtain a training data set, a verification data set and a test data set;
the network building unit inputs the training data set obtained by the image preprocessing unit into a built MDC L STM-L Densennet network model for training, and the verification data set obtained by the image preprocessing unit is used for adjusting the hyper-parameters of the MDC L STM-L Densennet network model to obtain the trained MDC L STM-L Densennet network model;
and the image classification unit inputs the test data set obtained by the preprocessing unit into the MDC L STM-L DenseNet network model obtained by the network building unit for testing to obtain the classification result of the nuclear magnetic resonance image and the classification accuracy of the MDC L STM-L DenseNet network model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010255476.5A CN111461233A (en) | 2020-04-02 | 2020-04-02 | Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010255476.5A CN111461233A (en) | 2020-04-02 | 2020-04-02 | Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111461233A true CN111461233A (en) | 2020-07-28 |
Family
ID=71682332
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010255476.5A Pending CN111461233A (en) | 2020-04-02 | 2020-04-02 | Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111461233A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950443A (en) * | 2020-08-10 | 2020-11-17 | 北京师范大学珠海分校 | Dense crowd counting method of multi-scale convolutional neural network |
CN112036454A (en) * | 2020-08-17 | 2020-12-04 | 上海电力大学 | Image classification method based on multi-core dense connection network |
CN112950624A (en) * | 2021-03-30 | 2021-06-11 | 太原理工大学 | Rectal cancer T stage automatic diagnosis method and equipment based on deep convolutional neural network |
CN113077456A (en) * | 2021-04-20 | 2021-07-06 | 北京大学 | Training method and device for constructing network model based on functional magnetic resonance imaging, computer equipment and storage medium |
CN113223014A (en) * | 2021-05-08 | 2021-08-06 | 中国科学院自动化研究所 | Brain image analysis system, method and equipment based on data enhancement |
CN113989551A (en) * | 2021-10-25 | 2022-01-28 | 哈尔滨理工大学 | Alzheimer disease classification method based on improved ResNet network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107145727A (en) * | 2017-04-26 | 2017-09-08 | 中国人民解放军总医院 | The medical image processing devices and method of a kind of utilization convolutional neural networks |
US20180310870A1 (en) * | 2017-05-01 | 2018-11-01 | The Charles Stark Draper Laboratory, Inc. | Deep learning architecture for cognitive examination subscore trajectory prediction in alzheimer's disease |
CN109389587A (en) * | 2018-09-26 | 2019-02-26 | 上海联影智能医疗科技有限公司 | A kind of medical image analysis system, device and storage medium |
-
2020
- 2020-04-02 CN CN202010255476.5A patent/CN111461233A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107145727A (en) * | 2017-04-26 | 2017-09-08 | 中国人民解放军总医院 | The medical image processing devices and method of a kind of utilization convolutional neural networks |
US20180310870A1 (en) * | 2017-05-01 | 2018-11-01 | The Charles Stark Draper Laboratory, Inc. | Deep learning architecture for cognitive examination subscore trajectory prediction in alzheimer's disease |
CN109389587A (en) * | 2018-09-26 | 2019-02-26 | 上海联影智能医疗科技有限公司 | A kind of medical image analysis system, device and storage medium |
Non-Patent Citations (5)
Title |
---|
LIN WANG 等: ""A Multiplication Optimization Level Set Algorithm for Image segmentation"", 《IEEE》 * |
刘振丙 等: ""基于多尺度残差神经网络的阿尔茨海默病诊断分类"", 《山东大学学报(工学版)》 * |
庞树茂: ""基于局部线性表达和深度学习的医学图像分析"", 《中国博士学位论文全文数据库 医药卫生科技辑》 * |
江宗康 等: ""MRI脑肿瘤图像分割的深度学习方法综述"", 《中国图象图形学报》 * |
贾龙飞: ""基于CNN和GWAS的阿尔茨海默症早期诊断研究"", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950443A (en) * | 2020-08-10 | 2020-11-17 | 北京师范大学珠海分校 | Dense crowd counting method of multi-scale convolutional neural network |
CN111950443B (en) * | 2020-08-10 | 2023-12-29 | 北京师范大学珠海分校 | Dense crowd counting method of multi-scale convolutional neural network |
CN112036454A (en) * | 2020-08-17 | 2020-12-04 | 上海电力大学 | Image classification method based on multi-core dense connection network |
CN112036454B (en) * | 2020-08-17 | 2022-12-16 | 上海电力大学 | Image classification method based on multi-core dense connection network |
CN112950624A (en) * | 2021-03-30 | 2021-06-11 | 太原理工大学 | Rectal cancer T stage automatic diagnosis method and equipment based on deep convolutional neural network |
CN113077456A (en) * | 2021-04-20 | 2021-07-06 | 北京大学 | Training method and device for constructing network model based on functional magnetic resonance imaging, computer equipment and storage medium |
CN113223014A (en) * | 2021-05-08 | 2021-08-06 | 中国科学院自动化研究所 | Brain image analysis system, method and equipment based on data enhancement |
CN113989551A (en) * | 2021-10-25 | 2022-01-28 | 哈尔滨理工大学 | Alzheimer disease classification method based on improved ResNet network |
CN113989551B (en) * | 2021-10-25 | 2024-05-03 | 哈尔滨理工大学 | Alzheimer's disease classification method based on improved ResNet network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111461233A (en) | Automatic nuclear magnetic resonance image classification method and device based on MDC L STM-L DenseNet network | |
Li et al. | Alzheimer's disease classification based on combination of multi-model convolutional networks | |
CN110236543B (en) | Alzheimer disease multi-classification diagnosis system based on deep learning | |
CN111461232A (en) | Nuclear magnetic resonance image classification method based on multi-strategy batch type active learning | |
CN110853111B (en) | Medical image processing system, model training method and training device | |
CN112348785B (en) | Epileptic focus positioning method and system | |
CN110969614B (en) | Brain age prediction method and system based on three-dimensional convolutional neural network | |
CN112465905A (en) | Characteristic brain region positioning method of magnetic resonance imaging data based on deep learning | |
CN111415361B (en) | Method and device for estimating brain age of fetus and detecting abnormality based on deep learning | |
CN114332462A (en) | MRI segmentation method for integrating attention mechanism into cerebral lesion | |
CN115375711A (en) | Image segmentation method of global context attention network based on multi-scale fusion | |
CN111932522B (en) | Alzheimer's disease classifier based on brain imaging big data deep learning | |
CN111523617B (en) | Epilepsy detection system based on white matter fusion characteristic diagram and residual error attention network | |
CN115147600A (en) | GBM multi-mode MR image segmentation method based on classifier weight converter | |
CN115272295A (en) | Dynamic brain function network analysis method and system based on time domain-space domain combined state | |
Asyhar et al. | Implementation LSTM Algorithm for Cervical Cancer using Colposcopy Data | |
Baskar et al. | An Accurate Prediction and Diagnosis of Alzheimer’s Disease using Deep Learning | |
CN113344991A (en) | Unsupervised medical image registration method and system based on deep learning | |
CN116452593B (en) | Method, device and system for constructing AI evaluation model of vascular cognitive disorder | |
CN116823613A (en) | Multi-mode MR image super-resolution method based on gradient enhanced attention | |
CN111652843A (en) | Decision support method and system based on nuclear magnetic resonance structure image weighting | |
KR102604217B1 (en) | A method and apparatus for image segmentation using residual convolution based deep learning network | |
CN116128890A (en) | Pathological cell image segmentation method and system based on self-adaptive fusion module and cross-stage AU-Net network | |
CN113160256B (en) | MR image placenta segmentation method for multitasking countermeasure model | |
CN111932523B (en) | Gender classifier based on brain imaging big data deep learning |
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 | ||
AD01 | Patent right deemed abandoned | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20240315 |