CN118470043B - Automatic medical image segmentation method based on cognitive deep learning - Google Patents
Automatic medical image segmentation method based on cognitive deep learning Download PDFInfo
- Publication number
- CN118470043B CN118470043B CN202410619372.6A CN202410619372A CN118470043B CN 118470043 B CN118470043 B CN 118470043B CN 202410619372 A CN202410619372 A CN 202410619372A CN 118470043 B CN118470043 B CN 118470043B
- Authority
- CN
- China
- Prior art keywords
- medical image
- output
- medical images
- fusion
- segmentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an automatic medical image segmentation method based on cognitive deep learning, and relates to the technical field of medical image segmentation. The method comprises the steps of obtaining medical images of different types, preprocessing the images, inputting the medical images into a trained multi-branch convolutional neural network to obtain output characteristics of the medical images of the corresponding types of each branch, recognizing the output characteristics through channel attention to obtain output characteristic fusion weights, merging the output characteristics according to the output characteristics and the output characteristic fusion weights to obtain fusion segmentation medical images, and carrying out post-processing on the fusion segmentation medical images to complete automatic segmentation of the medical images. The multi-branch convolutional neural network is used for processing different types of medical images, processing paths and parameters are optimized for each image type, segmentation accuracy and adaptability are improved, and through a channel attention mechanism, the characteristics output by each branch network can be effectively distributed with weights, so that dynamic fusion of the characteristics is realized, and information loss and interference are reduced.
Description
Technical Field
The invention relates to the technical field of medical image segmentation, in particular to an automatic medical image segmentation method based on cognitive deep learning.
Background
In modern medical diagnosis, medical image technology plays a crucial role, is commonly used for diagnosis, disease monitoring and treatment planning, automatic segmentation of medical images is an important research direction in the field of medical image processing, and aims to automatically identify and segment a lesion area or other interested anatomical structures from complex medical images.
Although deep learning has made remarkable progress in medical image segmentation, the prior art still has some limitations that most existing models are trained and optimized for specific types of medical images, such as CT, MRI or X-ray images, and the processing strategy of a single image type limits the universality and adaptability of the models, so that the models are difficult to process multi-mode medical image data, and the prior art often lacks an effective feature fusion mechanism when processing multi-mode data, and features between different modes are directly fused to possibly cause information loss or interference, so that the final segmentation accuracy is affected.
Disclosure of Invention
Based on the above-mentioned drawbacks of the prior art, an object of the present invention is to provide an automatic segmentation method for medical images based on cognitive deep learning, so as to solve the above-mentioned technical problems.
In order to achieve the above purpose, the invention provides the following technical scheme that the automatic medical image segmentation method based on cognitive deep learning comprises the following steps:
acquiring different types of medical images, and preprocessing the different types of medical images;
Inputting the preprocessed medical images of different types into a trained multi-branch convolutional neural network to obtain output characteristics of medical images of corresponding types of each branch, wherein each branch processes one type of medical image;
Cognizing the output characteristics through channel attention to obtain output characteristic fusion weights;
carrying out output feature combination according to the output features and the output feature fusion weights to obtain fusion segmentation medical images;
and carrying out post-processing on the fusion segmentation medical image to complete automatic segmentation of the medical image.
The invention is further arranged that the preprocessing of the different types of medical images comprises:
normalizing the medical images of different types, and carrying out contrast enhancement on the normalized medical images;
aligning the preprocessed medical images through elastic transformation to enable the positions of key features of different types of medical images to be consistent;
Resampling the medical images of different types after the elastic transformation to ensure that the spatial resolution and the size of the medical images of different types are consistent.
The invention is further arranged for training logic of the trained multi-branch convolutional neural network, comprising:
acquiring a historical data set, wherein the historical data set comprises a historical medical image data set and a historical segmentation medical image data set;
and performing branch training according to the historical data set to obtain the output characteristics of the medical image, calculating loss according to a loss function, repeating training until the loss is smaller than a loss threshold value, and completing training.
The invention is further arranged such that the loss function is: Where SL is the loss function, n is the total number of pixels, q j is the true value of pixel j, For the predicted value of pixel j, w1 is a weight coefficient, and the value logic of q t is:
The invention is further configured to obtain an output feature fusion weight by cognizing the output feature through channel attention, including:
The output characteristic is marked as an output characteristic set F, wherein F epsilon R C*H*W, R is a real number set symbol, elements representing a characteristic diagram belong to real numbers, C is the number of channels, H is the height of the output characteristic, and W is the width of the output characteristic;
Performing global average pooling on the output features in the output feature set F of each channel to obtain a channel descriptor;
And obtaining the fusion weight of the output characteristics according to the channel descriptor.
The invention is further configured to perform global averaging pooling on the output features in the output feature set F of each channel, obtain a channel descriptor, and calculate logic of the channel descriptor is as follows: Where S c is the channel descriptor of the c-th channel, and F (c, l, m) is the element value of the output feature at the height l, width m on the c-th channel.
The invention further provides that the output feature fusion weight is obtained according to the channel descriptor, the calculation logic is Z c＝σ(w2*δ(w3*Sc +b1) +b2, wherein Z c is the feature fusion weight, sigma is a sigmoid activation function, delta is a ReLU activation function, w2 and w3 are weight parameters, and b1 and b2 are bias parameters.
The invention is further configured to perform output feature merging according to the output feature and the output feature fusion weight to obtain a fusion segmentation medical image, including:
Acquiring the output characteristics and the corresponding characteristic fusion weights of each channel;
calculating a weighted output feature according to the output feature and the feature fusion weight;
And splicing the weighted output characteristics to obtain a fused medical image, and dividing the fused medical image to obtain the fused divided medical image.
The invention further provides that the post-processing of the fused segmented medical image comprises:
Smoothing the segmentation boundary and filling the small hole of the fusion segmentation medical image through morphological transformation;
and blurring processing is carried out on the fusion segmentation medical image after morphological transformation through a Gaussian filter, so that image noise is reduced.
The invention provides an automatic segmentation method based on cognitive deep learning medical images, which comprises the steps of preprocessing different types of medical images by acquiring the different types of medical images, inputting the preprocessed different types of medical images into a trained multi-branch convolutional neural network to obtain output characteristics of medical images of corresponding types of each branch, wherein each branch processes one type of medical image, cognizing the output characteristics through channel attention to obtain output characteristic fusion weights, merging the output characteristics according to the output characteristics and the output characteristic fusion weights to obtain fusion segmentation medical images, and performing post-processing on the fusion segmentation medical images to complete automatic segmentation of the medical images, wherein the generated beneficial effects comprise:
1. The method can optimize the processing path and parameters for each image type by using the multi-branch convolutional neural network to process different types of medical images, and the customized processing mode can more effectively extract key features in various images, thereby improving the segmentation accuracy and adaptability.
2. Dynamic feature fusion, namely, through a channel attention mechanism, the method can effectively distribute weights of the features output by each branch network, so that dynamic fusion of the features is realized, the dynamic fusion strategy not only considers the characteristics of each type of image, but also dynamically adjusts the fusion strategy according to specific contents, and the final fusion image can reduce information loss and interference while retaining important information.
3. Optimizing the use of computing resources through an efficient network architecture and post-processing technology, including global averaging pooling and feature fusion technology.
4. The morphological transformation and Gaussian filtering processing in the post-processing step are helpful for smoothing the segmentation boundary and filling the small holes, so that the visual effect of segmentation is improved, and the consistency and reliability of the segmentation result are also improved.
The foregoing description is only an overview of the present application, and is intended to be implemented in accordance with the teachings of the present application in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present application more readily apparent.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. In the drawings:
fig. 1 is a flowchart illustrating an automatic segmentation method based on a cognitive deep learning medical image according to an exemplary embodiment of the present invention.
Detailed Description
Further advantages and effects of the present invention will become readily apparent to those skilled in the art from the disclosure herein, by referring to the accompanying drawings and the preferred embodiments. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be understood that the preferred embodiments are presented by way of illustration only and not by way of limitation.
It should be noted that the illustrations provided in the following embodiments merely illustrate the basic concept of the present invention by way of illustration, and only the components related to the present invention are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
In the following description, numerous details are set forth in order to provide a more thorough explanation of embodiments of the present invention, it will be apparent, however, to one skilled in the art that embodiments of the present invention may be practiced without these specific details, in other embodiments, well-known structures and devices are shown in block diagram form, rather than in detail, in order to avoid obscuring the embodiments of the present invention.
The automatic segmentation method based on the cognitive deep learning medical image, as shown in fig. 1, comprises the following steps:
acquiring different types of medical images, and preprocessing the different types of medical images;
Inputting the preprocessed medical images of different types into a trained multi-branch convolutional neural network to obtain output characteristics of medical images of corresponding types of each branch, wherein each branch processes one type of medical image;
Cognizing the output characteristics through channel attention to obtain output characteristic fusion weights;
carrying out output feature combination according to the output features and the output feature fusion weights to obtain fusion segmentation medical images;
and carrying out post-processing on the fusion segmentation medical image to complete automatic segmentation of the medical image.
In particular, acquiring and preprocessing different types of medical images involves collecting different types of medical images, including X-ray (X-ray), one of the most common medical images for examination of bones and certain soft tissue structures, such as the lungs, computed Tomography (CT) scanning, which uses X-rays and computer technology to generate cross-sectional images that can provide more detailed tissue structure information for the diagnosis of various diseases, magnetic Resonance Imaging (MRI), which generates detailed internal body structure images using magnetic fields and harmless radio waves, and has a better display effect on soft tissues and organs, often for examination of brain, spine, joints, etc., ultrasound (Ultrasound) which generates images using high frequency sound waves, often for examination of fetuses, livers, hearts, blood vessels, etc., of pregnant women, positron Emission Tomography (PET) which provides information about tissue and organ functions by detecting the distribution of radiolabeled drugs in vivo, often for tumor diagnosis and evaluation of brain functions. A single photon emission tomography (SPECT) similar to PET scanning but using a single photon radioactive tracer for detecting cardiovascular and nervous system problems, a multi-branch convolutional neural network with multiple parallel convolutional branches for processing different types of medical images, a channel attention mechanism for learning the importance of different characteristic channels in the neural network, wherein the channel attention can help the network focus on the more important characteristic channels for improving performance, an output feature fusion weight obtained by the channel attention mechanism for fusing the output features of each branch, an output feature fusion for weighting and fusing the features according to the output features and the fusion weight to obtain final fused segmented medical images, and a post-processing for removing noise, filling holes, smoothing boundaries and the like applied on the fused segmented medical images to generate final medical image segmentation results.
The invention is further arranged that the preprocessing of the different types of medical images comprises:
The normalization is to scale the pixel value of the image to a fixed range so as to eliminate the brightness difference between different images, and the contrast enhancement is to enhance the details and contrast in the image by adjusting the dynamic range of the pixel value so as to improve the image quality;
The elastic transformation is an image transformation technology, can locally deform the image to realize the alignment of the image, and is generally used for aligning the images of different patients in the medical image, so that the key characteristics, including organ boundaries, are correspondingly consistent in space, thereby being beneficial to improving the generalization capability of a neural network;
the method comprises the steps of carrying out elastic transformation on medical images of different types, and carrying out resampling on the medical images of different types after the elastic transformation to enable the spatial resolution and the size of the medical images of different types to be consistent, wherein in particular, the resampling is to adjust the spatial resolution and the size of the images to enable the different images to have the same spatial scale, so that the neural network is ensured to be capable of processing the images of different resolutions and sizes, because the neural network generally requires the input to have the same size.
The invention is further arranged for training logic of the trained multi-branch convolutional neural network, comprising:
The method comprises the steps of obtaining a historical data set, wherein the historical data set comprises a historical medical image data set and a historical segmentation medical image data set, and specifically comprises a historical medical image data set and a historical segmentation medical image data set, the historical medical image data set comprises original medical image data, and the historical segmentation medical image data set comprises corresponding segmentation tag data, namely accurate boundaries of a region of interest in a medical image;
The method comprises the steps of obtaining a history data set, carrying out branch training according to the history data set to obtain output characteristics of medical images, calculating loss according to the loss function, repeating training until the loss is smaller than a loss threshold value, and completing training, specifically, using the history data set for training of a multi-branch convolutional neural network, wherein each branch is responsible for processing one type of medical image, inputting the medical image into the corresponding branch, carrying out network learning to extract the characteristics of the medical image and generate output characteristics matched with a segmentation label, evaluating the performance of the branch by using the loss function, wherein the loss function generally measures the difference between network output and the segmentation label, and in each training iteration, the network adjusts parameters according to the feedback of the loss function to reduce the difference between prediction and the label until the loss is reduced to meet a predefined loss threshold value.
The invention is further arranged such that the loss function is: Where SL is the loss function, n is the total number of pixels, q j is the true value of pixel j, For the predicted value of pixel j, w1 is a weight coefficient, and the value logic of q t is:
The invention is further configured to obtain an output feature fusion weight by cognizing the output feature through channel attention, including:
The output characteristic is marked as an output characteristic set F, wherein F epsilon R C*H*W, R is a real number set symbol, elements representing a characteristic diagram belong to real numbers, C is the number of channels, H is the height of the output characteristic, and W is the width of the output characteristic;
For the output characteristics of each channel, global average pooling operation is carried out, which is the average pooling operation carried out on the whole characteristic diagram, all elements of each channel are averaged to obtain one characteristic descriptor of the channel, and for C channels, C channel descriptors are obtained;
And specifically, the channel descriptor is used as input, and the output feature fusion weight is obtained through a preset calculation logic.
The invention is further configured to perform global averaging pooling on the output features in the output feature set F of each channel, obtain a channel descriptor, and calculate logic of the channel descriptor is as follows: Where S c is the channel descriptor of the c-th channel, and F (c, l, m) is the element value of the output feature at the height l, width m on the c-th channel.
The invention further provides that the output feature fusion weights are obtained according to the channel descriptors, the calculation logic is Z c＝σ(w2*δ(w3*Sc +b1) +b2, wherein Z c is the feature fusion weight, sigma is a sigmoid activation function, delta is a ReLU activation function, w2 and w3 are weight parameters, b1 and b2 are bias parameters, the values of the weight parameters w2 and w3 and the bias parameters b1 and b2 are obtained by learning through a back propagation algorithm in the training process, so that the network can better learn the relation between the features according to input data, the weight parameters w2 and w3 and the bias parameters b1 and b2 are random, then iterative optimization is carried out in the training process through the back propagation algorithm, reLU is a modified linear unit activation function, the sigmoid activation function is used for ensuring that the output is non-negative, the sigmoid activation function maps w3 to the space between (0, 1), the calculation logic can calculate the output feature fusion weights of each channel according to the channel descriptors, the final fusion performance of the output feature fusion can be improved, and the fusion performance of the output feature fusion can be more important, and the fusion performance of the network can be improved
The invention is further configured to perform output feature merging according to the output feature and the output feature fusion weight to obtain a fusion segmentation medical image, including:
And specifically, for each channel k, acquiring a corresponding output characteristic F k and a characteristic fusion weight Z c.
Calculating a weighted output characteristic according to the output characteristic and the characteristic fusion weight, specifically, for each channel k, multiplying the output characteristic F k by the characteristic fusion weight Z c to obtain the weighted output characteristic The weighted output characteristics take the importance of each channel into consideration, and the characteristics of the corresponding channels are weighted according to the weights;
the weighted output characteristics are spliced to obtain a fused medical image, and the fused segmented medical image is obtained by segmentation, and specifically, all the weighted output characteristics are processed The fusion feature map F s is obtained by stitching along the channel dimension, and further, stitching refers to that a plurality of tensors are connected along a certain dimension to form a larger tensor, in the embodiment of the present invention, stitching is performed on the channel dimension, that is, the feature maps of a plurality of channels are connected according to the channel direction, specifically, in the medical image segmentation task, a multi-branch convolutional neural network is used as a model, and in the process of feature extraction, the multi-branch convolutional neural network generates the feature maps of a plurality of channels, and each channel represents different feature representations, so when merging different features, each piece of learned information is reserved to make full use of features of different types and dimensions, and the feature maps of each channel need to be stitched together according to the channel dimension to form a more comprehensive feature representation, so that the model can more comprehensively understand the input image.
The invention further provides that the post-processing of the fused segmented medical image comprises:
The method comprises the steps of carrying out a morphological transformation on the fusion segmentation medical image to smooth the segmentation boundary and fill the small holes, wherein the morphological transformation is an operation based on the shape of the image and is commonly used for smoothing the boundary and filling the holes;
The Gaussian filter is a common image blurring processing method, can effectively reduce noise in an image, and in the embodiment of the invention, the Gaussian filter can be applied to carry out blurring processing on a segmentation result after morphological transformation so as to further reduce noise in the image and enable a segmentation boundary to be smoother and natural.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more sets of available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium may be a solid state disk.
It should be understood that the term "and/or" is merely an association relationship describing the associated object, and means that three relationships may exist, for example, a and/or B, and may mean that a exists alone, while a and B exist alone, and B exists alone, wherein a and B may be singular or plural. In addition, the character "/" herein generally indicates that the associated object is an "or" relationship, but may also indicate an "and/or" relationship, and may be understood by referring to the context.
In the present application, "at least one" means one or more, and "a plurality" means two or more. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (a, b, or c) of a, b, c, a-b, a-c, b-c, or a-b-c may be represented, wherein a, b, c may be single or plural.
It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. 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 application.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided by the present application, it should be understood that the disclosed system may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. The storage medium includes a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (6)
1. The automatic segmentation method for the medical image based on the cognitive deep learning is characterized by comprising the following steps of:
acquiring different types of medical images, and preprocessing the different types of medical images;
Inputting the preprocessed medical images of different types into a trained multi-branch convolutional neural network to obtain output characteristics of medical images of corresponding types of each branch, wherein each branch processes one type of medical image;
Cognizing the output features through channel attention to obtain an output feature fusion weight, wherein the output features are marked as an output feature set F, Wherein R is a real number set symbol, elements representing the feature map belong to real numbers, C is the number of channels, H is the height of an output feature, W is the width of the output feature, global average pooling is performed on the output features in the output feature set F of each channel to obtain a channel descriptor, and the calculation logic of the channel descriptor is as follows: Wherein, the method comprises the steps of, wherein, For the channel descriptor of the c-th channel,Height on the c-th channel for output featuresAcquiring output feature fusion weights according to the channel descriptors, wherein the calculation logic of the feature fusion weights is as follows: Wherein, the method comprises the steps of, wherein, For the feature fusion weights to be used,For the sigmoid activation function,For the function to be activated by the ReLU,AndAs a parameter of the weight-bearing element,AndIs a bias parameter;
carrying out output feature combination according to the output features and the output feature fusion weights to obtain fusion segmentation medical images;
and carrying out post-processing on the fusion segmentation medical image to complete automatic segmentation of the medical image.
2. The automatic segmentation method for medical images based on cognitive deep learning according to claim 1, wherein the preprocessing of the different types of medical images comprises:
normalizing the medical images of different types, and carrying out contrast enhancement on the normalized medical images;
aligning the preprocessed medical images through elastic transformation to enable the positions of key features of different types of medical images to be consistent;
Resampling the medical images of different types after the elastic transformation to ensure that the spatial resolution and the size of the medical images of different types are consistent.
3. The automatic segmentation method based on cognitive deep learning medical images according to claim 1, wherein the training logic of the trained multi-branch convolutional neural network comprises:
acquiring a historical data set, wherein the historical data set comprises a historical medical image data set and a historical segmentation medical image data set;
and performing branch training according to the historical data set to obtain the output characteristics of the medical image, calculating loss according to a loss function, repeating training until the loss is smaller than a loss threshold value, and completing training.
4. The automatic segmentation method based on the cognitive deep learning medical image according to claim 3, wherein the loss function is: Where SL is the loss function, n is the total number of pixels, Is the true value of the pixel j,As the predicted value of the pixel j,As the weight coefficient of the light-emitting diode,The value logic of (a) is as follows:。
5. The automatic segmentation method based on cognitive deep learning medical image according to claim 1, wherein the step of merging output features according to the output features and the output feature fusion weights to obtain a fusion segmented medical image comprises the steps of:
Acquiring the output characteristics and the corresponding characteristic fusion weights of each channel;
calculating a weighted output feature according to the output feature and the feature fusion weight;
And splicing the weighted output characteristics to obtain a fused medical image, and dividing the fused medical image to obtain the fused divided medical image.
6. The automatic segmentation method based on the cognitive deep learning medical image according to claim 1, wherein the post-processing of the fusion segmented medical image comprises:
Smoothing the segmentation boundary and filling the small hole of the fusion segmentation medical image through morphological transformation;
and blurring processing is carried out on the fusion segmentation medical image after morphological transformation through a Gaussian filter, so that image noise is reduced.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410619372.6A CN118470043B (en) | 2024-05-19 | 2024-05-19 | Automatic medical image segmentation method based on cognitive deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410619372.6A CN118470043B (en) | 2024-05-19 | 2024-05-19 | Automatic medical image segmentation method based on cognitive deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118470043A CN118470043A (en) | 2024-08-09 |
| CN118470043B true CN118470043B (en) | 2025-03-14 |
Family
ID=92166316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410619372.6A Active CN118470043B (en) | 2024-05-19 | 2024-05-19 | Automatic medical image segmentation method based on cognitive deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118470043B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119359835B (en) * | 2024-09-29 | 2025-04-01 | 广东省第二人民医院（广东省卫生应急医院） | Imaging method and system of double tracer models |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117078692A (en) * | 2023-10-13 | 2023-11-17 | 山东未来网络研究院(紫金山实验室工业互联网创新应用基地) | Medical ultrasonic image segmentation method and system based on self-adaptive feature fusion |
| CN117474933A (en) * | 2023-10-20 | 2024-01-30 | 江苏科技大学 | A medical image segmentation method based on dilated convolution attention mechanism |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109872328B (en) * | 2019-01-25 | 2021-05-07 | 腾讯科技（深圳）有限公司 | Brain image segmentation method, device and storage medium |
| CN115115662B (en) * | 2022-07-01 | 2025-04-08 | 重庆邮电大学 | Multi-mode medical image target segmentation method based on knowledge and data fusion driving |
| US11776240B1 (en) * | 2023-01-27 | 2023-10-03 | Fudan University | Squeeze-enhanced axial transformer, its layer and methods thereof |
| CN117893479A (en) * | 2023-12-21 | 2024-04-16 | 桂林理工大学 | Medical image segmentation method and system integrating multiple attention mechanisms |
| CN117746145A (en) * | 2023-12-22 | 2024-03-22 | 桂林电子科技大学 | Fusion model classification method based on convolutional neural network and attention mechanism |
-
2024
- 2024-05-19 CN CN202410619372.6A patent/CN118470043B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117078692A (en) * | 2023-10-13 | 2023-11-17 | 山东未来网络研究院(紫金山实验室工业互联网创新应用基地) | Medical ultrasonic image segmentation method and system based on self-adaptive feature fusion |
| CN117474933A (en) * | 2023-10-20 | 2024-01-30 | 江苏科技大学 | A medical image segmentation method based on dilated convolution attention mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118470043A (en) | 2024-08-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yousef et al. | A holistic overview of deep learning approach in medical imaging | |
| Sorin et al. | Creating artificial images for radiology applications using generative adversarial networks (GANs)–a systematic review | |
| Mazurowski et al. | Deep learning in radiology: An overview of the concepts and a survey of the state of the art with focus on MRI | |
| US11127137B2 (en) | Malignancy assessment for tumors | |
| RU2571523C2 (en) | Probabilistic refinement of model-based segmentation | |
| CN112885453A (en) | Method and system for identifying pathological changes in subsequent medical images | |
| EP2401719B1 (en) | Methods for segmenting images and detecting specific structures | |
| WO2012072129A1 (en) | Longitudinal monitoring of pathology | |
| Bandhyopadhyay et al. | Segmentation of brain MRI image–a review | |
| Gu et al. | Segmentation of coronary arteries images using global feature embedded network with active contour loss | |
| US20240303973A1 (en) | Actor-critic approach for generating synthetic images | |
| CN118470043B (en) | Automatic medical image segmentation method based on cognitive deep learning | |
| Zhai et al. | Automatic quantitative analysis of pulmonary vascular morphology in CT images | |
| Dong et al. | Identifying carotid plaque composition in MRI with convolutional neural networks | |
| US20220044454A1 (en) | Deep reinforcement learning for computer assisted reading and analysis | |
| Bhushan | Liver cancer detection using hybrid approach-based convolutional neural network (HABCNN) | |
| Faisal et al. | Computer assisted diagnostic system in tumor radiography | |
| Rasool et al. | Multimodal brain tumor segmentation using 3D-U-Net | |
| CN118864858A (en) | Spine CT image data segmentation method and system based on machine learning | |
| Nasim et al. | Review on multimodality of different medical image fusion techniques | |
| Sharma et al. | Importance of deep learning models to perform segmentation on medical imaging modalities | |
| Pandey et al. | A framework for mathematical methods in medical image processing | |
| Braik et al. | Exploration of genetic algorithms and particle swarm optimization in improving the quality of medical images | |
| Sreelekshmi et al. | A Review on Multimodal Medical Image Fusion. | |
| Ayshwarya et al. | MRI Image Analysis for Brain Tumor Detection Using Convolutional Neural Network |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20250205 Address after: 215000 No. 158-116 Huashan Road, high tech Zone, Suzhou, Jiangsu Applicant after: Yinnuo high tech (Suzhou) Co.,Ltd. Country or region after: China Address before: 518000 Huafeng Building 1806-MM30, No. 6006 Shennan Avenue, Zijing Community, Lianhua Street, Futian District, Shenzhen, Guangdong Province Applicant before: Shenzhen Jiukang Kexin Technology Partnership (L.P.) Country or region before: China |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |