CN117765330A - MRI image-based data labeling method and system - Google Patents
MRI image-based data labeling method and system Download PDFInfo
- Publication number
- CN117765330A CN117765330A CN202311810565.1A CN202311810565A CN117765330A CN 117765330 A CN117765330 A CN 117765330A CN 202311810565 A CN202311810565 A CN 202311810565A CN 117765330 A CN117765330 A CN 117765330A
- Authority
- CN
- China
- Prior art keywords
- image
- mri
- annotation
- region
- original
- 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
- 238000002372 labelling Methods 0.000 title claims description 61
- 238000012549 training Methods 0.000 claims abstract description 40
- 238000009499 grossing Methods 0.000 claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000002595 magnetic resonance imaging Methods 0.000 claims description 232
- 230000001629 suppression Effects 0.000 claims description 35
- 238000001514 detection method Methods 0.000 claims description 23
- 238000004422 calculation algorithm Methods 0.000 claims description 19
- 238000011282 treatment Methods 0.000 claims description 13
- 230000004044 response Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 3
- 230000005764 inhibitory process Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 10
- 230000008569 process Effects 0.000 description 10
- 238000004891 communication Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000004590 computer program Methods 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 6
- 238000010191 image analysis Methods 0.000 description 6
- 238000007726 management method Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000003902 lesion Effects 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000003708 edge detection Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010028980 Neoplasm Diseases 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000005316 response function Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention relates to the field of data annotation, and provides a data annotation method and a system based on an MRI image, wherein the method comprises the following steps: the method comprises the steps of obtaining an MRI denoising image and an MRI smooth image through image denoising and smoothing processing, dividing the MRI smooth image into a target area, extracting parameters of the area, calculating edge intensity of the target area, detecting corner points in the residual area of the MRI smooth image based on the edge intensity, extracting corner point inhibition values and information factors, identifying texture features of the residual area, constructing an image annotation model of an original MRI image according to the texture features, training the model by using a training set to obtain a trained image annotation model, carrying out annotation prediction on the original MRI image by using the trained model to obtain a prediction annotation image, and removing overlapping parts to obtain a final image annotation result. The invention can improve the accuracy of the MRI image data annotation.
Description
Technical Field
The invention relates to the field of data annotation, in particular to a data annotation method and system based on an MRI image.
Background
MRI images are a medical imaging technology, images are obtained by utilizing resonance phenomena of atomic nuclei under the action of strong magnetic fields and radio frequency pulses, the MRI images are used for observing structures and functions of internal tissues and organs of a human body, the MRI images have important clinical value for diagnosing diseases, the MRI images can also display high-contrast and high-definition anatomical information, and meanwhile, functional information and observation of dynamic processes can also be provided.
At present, the data labeling method in the MRI image can be realized by using an automatic labeling method, which mainly utilizes a computer vision technology and a machine learning algorithm to automatically identify and label the MRI image, and a large amount of training data with labels is needed to train a model to learn the characteristics of specific lesions or structures, so that the corresponding region in the automatic labeling image is realized, but when the complex MRI image is processed, certain errors exist due to the difference between different types of lesions and structures, so that the data labeling method and system based on the MRI image are needed to improve the accuracy of the data labeling of the MRI image.
Disclosure of Invention
The invention provides a data labeling method and a system based on an MRI image, which mainly aim to improve the accuracy of the MRI image data labeling.
In order to achieve the above object, the data labeling method based on MRI image provided by the present invention includes:
acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoising image, and performing smoothing treatment on the MRI denoising image to obtain an MRI smooth image;
dividing a target region in the MRI smooth image, extracting region parameters corresponding to the target region, calculating image gradients corresponding to the target region based on the region parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target region based on the gradient factors;
Performing corner detection on the residual region of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the residual region based on the information factor;
based on the texture characteristics, an image annotation model corresponding to the original MRI image is constructed, an image training set corresponding to the MRI smooth image is divided by using the image annotation model, and model training is carried out on the image annotation model by using the training set, so that a trained image annotation model is obtained;
and carrying out annotation prediction on the original MRI image by using the trained model to obtain a predicted annotation image, and carrying out overlapping removal on the predicted annotation image to obtain an image annotation result corresponding to the original MRI image.
Optionally, the acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoised image, includes:
performing MRI scanning on a patient by using preset medical equipment to obtain an original MRI image;
performing noise analysis on the original MRI image to obtain noise data;
identifying a noise direction corresponding to the noise data;
And carrying out image denoising on the original MRI image based on the noise direction to obtain an MRI denoising image.
Optionally, the performing smoothing processing on the MRI denoising image to obtain an MRI smoothed image includes:
identifying an image requirement corresponding to the MRI denoising image;
setting smoothing parameters corresponding to the MRI denoising image based on the image demand;
and on the basis of the smoothing parameters, starting a preset smoothing filter to carry out smoothing treatment on the MRI denoising image, so as to obtain an MRI smooth image.
Optionally, the segmenting the target region in the MRI smooth image, extracting a region parameter corresponding to the target region includes:
identifying a target factor corresponding to the MIR smooth image;
determining a target area corresponding to the MRI smooth image based on the target factor;
segmenting the target region in the MRI smooth image by using a preset segmentation algorithm;
and inquiring the region requirement corresponding to the target region, and extracting the region parameter corresponding to the region requirement.
Optionally, the calculating, based on the region parameter, an image gradient corresponding to the target region includes:
first, an average gray value corresponding to the target area is calculated using the following formula:
Wherein MG represents an average gray value corresponding to the target area, N represents the number of pixels in the target area, and I (x, y) represents a pixel gray value at coordinates (x, y).
Secondly, calculating the sum of squares of differences between each pixel of the target area and the adjacent pixels based on the average gray value:
SD＝∑[I(x,y)-I(x’,y’)] 2
wherein SD represents the sum of squares of differences between each pixel of the target region and its neighboring pixels, (x ', y') represents the neighboring pixel position of the (x, y) pixel, and I (x, y) and I (x ', y') represent the pixel gray values within the target region, respectively.
Finally, calculating the image gradient corresponding to the target area based on the sum of squares of the differences:
wherein GT represents the image gradient corresponding to the target region, and N represents the number of pixels in the target region.
Optionally, the calculating the edge strength of the target area based on the gradient factor includes:
where BQ represents the edge intensity of the target region, (i, j) represents the pixel coordinates in the image, (MB) represents the target region, gx (i, j) and Gy (i, j) represent the gradient values in the horizontal and vertical directions at the pixel point (i, j), respectively, and M represents the total number of pixel points in the target region.
Optionally, based on the edge intensity, performing corner detection on a remaining area of the MRI smooth image to obtain a corner suppression value, including:
Determining an edge image corresponding to the residual region based on the edge intensity;
performing corner monitoring on the edge image to obtain corner parameters;
calculating a corner response value corresponding to the residual region based on the corner parameter;
and carrying out non-maximum suppression on the angular point response value to obtain an angular point suppression value.
Optionally, the extracting the information factor corresponding to the corner suppression value, based on the information factor, identifies the texture feature corresponding to the remaining area, including:
determining a fixed neighborhood corresponding to the corner suppression value, and extracting an information factor corresponding to the fixed neighborhood;
judging the texture category corresponding to the information factor;
and identifying texture features corresponding to the residual area based on the texture category.
Optionally, the performing label prediction on the original MRI image by using the trained model to obtain a predicted label image includes:
acquiring an MRI image sample corresponding to the original MRI image;
equalizing the MRI image sample to obtain an equalized sample;
carrying out image prediction on the balanced sample by using the trained model to obtain a predicted image;
and carrying out image annotation on the predicted image to obtain a predicted and annotated image.
In order to solve the above problems, the present invention also provides a data labeling system based on MRI images, the system comprising:
the smoothing processing module is used for acquiring an original MRI image, carrying out image denoising on the original MRI image to obtain an MRI denoising image, and carrying out smoothing processing on the MRI denoising image to obtain an MRI smooth image;
the edge calculation module is used for dividing a target area in the MRI smooth image, extracting area parameters corresponding to the target area, calculating image gradients corresponding to the target area based on the area parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target area based on the gradient factors;
the texture recognition module is used for carrying out corner detection on the residual area of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and recognizing texture features corresponding to the residual area based on the information factor;
the model training module is used for constructing an image annotation model corresponding to the original MRI image based on the texture features, dividing an image training set corresponding to the MRI smooth image by using the image annotation model, and carrying out model training on the image annotation model by using the training set to obtain a trained image annotation model;
And the image labeling module is used for labeling and predicting the original MRI image by using the trained model to obtain a predicted labeling image, and overlapping and removing the predicted labeling image to obtain an image labeling result corresponding to the original MRI image.
The invention carries out image denoising on the original MRI image by acquiring the original MRI image, thereby obtaining the MRI denoising image, improving the image quality, improving the image analysis and diagnosis effect, reducing the misdiagnosis rate, optimizing the subsequent image processing process, reducing the influence of noise, enabling the image to be clearer and more definite in detail, thereby improving the accuracy of detection. Has important benefits for research and clinical practice in the medical field. Therefore, the data labeling method and the system based on the MRI image provided by the invention are used for improving the accuracy of the MRI image data labeling.
Drawings
FIG. 1 is a flow chart of a method for labeling MRI image-based data according to an embodiment of the present invention;
FIG. 2 is a schematic block diagram of a data labeling system based on MRI images according to an embodiment of the present invention;
fig. 3 is a schematic diagram of an internal structure of an electronic device according to an MRI image-based data labeling method according to an embodiment of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the application provides a data labeling method based on MRI images. The execution subject of the MRI image-based data labeling method includes, but is not limited to, at least one of a server, a terminal, and the like, which can be configured to execute the method provided by the embodiments of the present application. In other words, the MRI image-based data labeling method may be performed by software or hardware installed in a terminal device or a server device, and the software may be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.
Referring to fig. 1, a flowchart of a data labeling method based on MRI images according to an embodiment of the present invention is shown. In this embodiment, the data labeling method based on MRI image includes:
s1, acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoising image, and performing smoothing treatment on the MRI denoising image to obtain an MRI smooth image.
According to the invention, the original MRI image is obtained, and image denoising is carried out on the original MRI image, so that an MRI denoising image is obtained, the image quality can be improved, the image analysis and diagnosis effects are improved, the misdiagnosis rate is reduced, the subsequent image processing process is optimized, the influence of noise can be reduced, the image is clearer, the details are clearer, and the detection accuracy is improved.
Wherein the original MRI image refers to an MRI image which is acquired directly from a patient and is not subjected to any treatment; the MRI denoising image refers to an image obtained by applying an image processing algorithm to an original MRI image to reduce or remove unwanted image features such as noise, artifacts and the like in the image.
As an embodiment of the present invention, the acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoised image, includes: performing MRI scanning on a patient by using preset medical equipment to obtain an original MRI image; performing noise analysis on the original MRI image to obtain noise data; identifying a noise direction corresponding to the noise data; and carrying out image denoising on the original MRI image based on the noise direction to obtain an MRI denoising image.
The preset medical device refers to medical image devices predetermined to be used for performing MRI scanning on a patient, for example: a magnetic resonance imaging machine; the noise data refers to interference signals present in the original MRI image; the noise direction refers to determining the directionality of noise in an image according to the characteristics and distribution of noise data.
Further, the noise data may be obtained by a noise analysis tool implementation, such as: audacity, adobe audio and other tools; the noise direction may be obtained by a beamforming algorithm implementation, such as: delay-and-Sum, beamforming, etc.
The MRI denoising image is subjected to smoothing treatment to obtain an MRI smooth image, and the image with clearer contrast and accuracy can be obtained by enhancing the image quality, so that more accurate information can be obtained, and repeated scanning is avoided.
Wherein the MRI smooth image refers to an MRI image that reduces or eliminates noise in the image while preserving key image features.
As one embodiment of the present invention, the performing a smoothing process on the MRI denoising image to obtain an MRI smoothed image includes: identifying an image requirement corresponding to the MRI denoising image; setting smoothing parameters corresponding to the MRI denoising image based on the image demand; and on the basis of the smoothing parameters, starting a preset smoothing filter to carry out smoothing treatment on the MRI denoising image, so as to obtain an MRI smooth image.
The image requirement refers to an input image required when the image is processed or analyzed, and the input image can be an original image or a preprocessed image; the smoothing parameters refer to parameters used to control the smoothing filter to smooth the image, such as: window size, convolution kernel size, etc.; the preset smoothing filter refers to a filter designed according to a specific mathematical model and used for image smoothing, such as: an average filter, a gaussian filter, etc.
Further, the image requirements may be obtained by an image processing tool, such as: openCV, pillow, scikit-image, etc.; the smoothing parameters may be obtained by statistical analysis tools such as: matplotlib, numPy and SciPy, etc.
S2, segmenting a target region in the MRI smooth image, extracting region parameters corresponding to the target region, calculating image gradients corresponding to the target region based on the region parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target region based on the gradient factors.
According to the invention, the target region in the MRI smooth image is segmented, the region parameters corresponding to the target region are extracted, the accurate positioning and analysis of the target region can be facilitated, powerful support is provided for disease diagnosis, treatment and research, and automatic image analysis can be realized, so that the cost of manpower and material resources is saved, and the efficiency is improved.
Wherein the target region refers to a specific region of interest in the MRI-smoothed image, for example: tumor area, brain structure, etc.; the regional parameters refer to quantitative parameters obtained by analyzing the characteristics of the target region.
As one embodiment of the present invention, the segmenting the target region in the MRI smoothing image, extracting the region parameters corresponding to the target region, includes: identifying a target factor corresponding to the MIR smooth image; determining a target area corresponding to the MRI smooth image based on the target factor; segmenting the target region in the MRI smooth image by using a preset segmentation algorithm; inquiring the region requirement corresponding to the target region; and extracting the region parameters corresponding to the region requirements.
Wherein the target factor refers to a specific factor or feature used to identify an object of interest in the MRI smooth image; the preset segmentation algorithm is a specific algorithm for separating a target area from a background area in an MRI smooth image, and the common preset segmentation algorithm comprises: threshold segmentation, edge detection, region growing, graph theory-based methods, and the like; the area requirements refer to specific properties or parameters that need to be analyzed, measured or extracted in the target area.
Further, the target factor may be obtained through a machine learning model implementation, such as: support vector machines SVM, random forest RF, deep learning models, etc.; the region requirement may be obtained by a threshold segmentation algorithm, such as: otsu, adaptive threshold, etc.
Optionally, as an embodiment of the present invention, the calculating, based on the region parameter, an image gradient corresponding to the target region includes:
first, an average gray value corresponding to the target area is calculated using the following formula:
wherein MG represents an average gray value corresponding to the target area, N represents the number of pixels in the target area, and I (x, y) represents a pixel gray value at coordinates (x, y).
Secondly, calculating the sum of squares of differences between each pixel of the target area and the adjacent pixels based on the average gray value:
SD＝∑[I(x,y)-I(x’,y’)] 2
wherein SD represents the sum of squares of differences between each pixel of the target region and its neighboring pixels, (x ', y') represents the neighboring pixel position of the (x, y) pixel, and I (x, y) and I (x ', y') represent the pixel gray values within the target region, respectively.
Finally, calculating the image gradient corresponding to the target area based on the sum of squares of the differences:
Wherein GT represents the image gradient corresponding to the target region, and N represents the number of pixels in the target region.
According to the method, the gradient factors corresponding to the image gradients are inquired, the edge intensity of the target area is calculated based on the gradient factors, the targets can be separated from the background, and different objects in the images can be identified and classified, so that the edges are more obvious, and the visual effect of the target area is enhanced.
Wherein, the image gradient refers to the direction and intensity of the fastest change of the pixel value in the image; the gradient factor refers to the mode length of the image gradient, also referred to as gradient magnitude or gradient magnitude; the edge intensity refers to the degree or intensity of edge saliency of an image.
Alternatively, the gradient factor may be obtained by a detection algorithm implementation, such as: prewitt operator, laplacian operator.
As an embodiment of the present invention, the calculating the edge intensity of the target area based on the gradient factor includes:
where BQ represents the edge intensity of the target region, (i, j) represents the pixel coordinates in the image, (MB) represents the target region, gx (i, j) and Gy (i, j) represent the gradient values in the horizontal and vertical directions at the pixel point (i, j), respectively, and M represents the total number of pixel points in the target region.
And S3, carrying out corner detection on the residual region of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the residual region based on the information factor.
According to the invention, based on the edge intensity, the corner detection is carried out on the residual area of the MRI smooth image to obtain the corner inhibition value, so that richer image characteristics can be provided, the accuracy of target positioning is enhanced, the registration and alignment of the images are assisted, the image segmentation is assisted and other applications are assisted, and the medical image processing effect and quality are improved.
The residual area refers to an image area which is not covered by smoothing after the smoothing operation is performed; the corner suppression value refers to the position and attribute information of the corner obtained by performing corner detection on the residual region.
As an embodiment of the present invention, the detecting the corner point of the remaining area of the MRI smooth image based on the edge intensity to obtain a corner point suppression value includes: determining an edge image corresponding to the residual region based on the edge intensity; performing corner monitoring on the edge image to obtain corner parameters; calculating a corner response value corresponding to the residual region based on the corner parameter; and carrying out non-maximum suppression on the angular point response value to obtain an angular point suppression value.
The edge image is a binary image obtained according to the edge intensity (such as an edge extracted by a Canny algorithm) of the image; the corner parameters refer to the characteristic parameters of each corner obtained by calculation when the corner detection is carried out on the edge image; the corner response value refers to an important value representing each corner in the image, which is calculated according to the corner parameters.
Further, the edge image may be obtained through an edge detection calculation implementation, such as: canny algorithm, sobel operator, laplacian operator, etc.: the corner parameters can be obtained through a corner detection algorithm, such as: harris corner detection algorithm, shi-Tomasi corner detection algorithm, FAST corner detection algorithm and the like: the corner response value may be obtained by a corner response function implementation, such as: functions cv2.corerharris (), cv2.goodfeaturestrack (), and the like in the OpenCV library.
According to the invention, by extracting the information factors corresponding to the corner suppression values and identifying the texture features corresponding to the residual regions based on the information factors, finer, accurate and robust texture feature extraction and classification results can be brought, so that the application effects and performances of the fields of image analysis, computer vision, pattern recognition and the like are improved.
Wherein, the information factor refers to a quantization index of the corner suppression value and surrounding pixels thereof by calculation and analysis; the texture features are used for describing and distinguishing different materials, textures and morphological features.
As one embodiment of the present invention, the extracting the information factor corresponding to the corner suppression value, and identifying the texture feature corresponding to the remaining area based on the information factor, includes: determining a fixed neighborhood corresponding to the corner suppression value; extracting information factors corresponding to the fixed neighborhood; judging the texture category corresponding to the information factor; and identifying texture features corresponding to the residual area based on the texture category.
Wherein, the fixed neighborhood refers to a window or neighborhood with fixed size defined for extracting information factors in the process of calculating corner suppression values; the texture class refers to a classification class to which the texture features are judged according to the information factors.
Further, the fixed neighborhood may be obtained through a Python processing library implementation, such as: PIL, scikit-image, etc.; the texture class may be obtained by an LBP algorithm implementation.
S4, constructing an image annotation model corresponding to the original MRI image based on the texture features, dividing an image training set corresponding to the MRI smooth image by using the image annotation model, and performing model training on the image annotation model by using the training set to obtain a trained image annotation model.
According to the invention, based on the texture characteristics, the image annotation model corresponding to the original MRI image is constructed, and the image annotation model is utilized to divide the image training set corresponding to the MRI smooth image, so that the image annotation and classification can be automatically carried out, the processing efficiency is improved, different types of images can be better analyzed and processed, and the corresponding research and treatment scheme can be developed in a targeted manner.
The image annotation model can automatically assign corresponding annotation or classification to the image according to the extracted texture characteristics and other information; the image training set refers to a set of known labels or classes of image samples used to train and optimize an image annotation model.
Alternatively, the image annotation model may be provided by a model building tool, such as: tensorFlow, pyTorch, etc.; the image training set may be obtained by a partitioning tool implementation, such as: labelImg, rectLabel, etc.
According to the method, the training set is utilized to carry out model training on the image annotation model, so that a trained image annotation model is obtained, the annotation accuracy, the automatic annotation flow and the generalization capability and adaptability of the model can be improved, and meanwhile, a foundation is provided for optimizing and improving the model, so that the model is more suitable for specific tasks and application scenes.
The trained image annotation model is a model with higher accuracy and generalization capability after training and optimizing by a training set, and optionally, the trained image annotation model can be obtained by training the model of the image annotation model by using the training set.
And S5, performing annotation prediction on the original MRI image by using the trained model to obtain a predicted annotation image, and performing overlapping removal on the predicted annotation image to obtain an image annotation result corresponding to the original MRI image.
According to the invention, the original MRI image is subjected to annotation prediction by using the trained model, so that a predicted annotation image is obtained, the structure, lesion or characteristic in the image can be intuitively displayed, a doctor is helped to quickly know the condition of a patient, and corresponding diagnosis and treatment decisions are made.
The prediction labeling image is an image generated by labeling and predicting an original MRI image through the trained model.
As one embodiment of the present invention, the performing label prediction on the original MRI image by using the trained model to obtain a predicted label image includes: acquiring an MRI image sample corresponding to the original MRI image; equalizing the MRI image sample to obtain an equalized sample; carrying out image prediction on the balanced sample by using the trained model to obtain a predicted image; and carrying out image annotation on the predicted image to obtain a predicted and annotated image.
Wherein the MRI image sample refers to an image sample obtained from a medical database or other source, and contains information about tissue type, density, shape, etc.; the balanced sample is a sample obtained after the MRI image sample is subjected to the equalization processing.
Further, the MRI image sample may be obtained by an MRI scanner implementation; the equalized samples may be obtained by an equalization algorithm implementation, such as: histogram equalization, adaptive histogram equalization, etc.
According to the invention, the image annotation result corresponding to the original MRI image is obtained by overlapping and removing the prediction annotation image, so that the image quality can be improved, the readability can be improved, the image analysis and processing effects can be promoted, the lesion detection and diagnosis accuracy can be improved, and the method has important benefits for research and clinical practice in the medical field.
The image labeling result refers to a result of labeling and annotating an original MRI image, and optionally, the image labeling result may be obtained through a neural network model, for example: U-Net, resNet, VGG, etc.
The invention carries out image denoising on the original MRI image by acquiring the original MRI image, thereby obtaining the MRI denoising image, improving the image quality, improving the image analysis and diagnosis effect, reducing the misdiagnosis rate, optimizing the subsequent image processing process, reducing the influence of noise, enabling the image to be clearer and more definite in detail, thereby improving the accuracy of detection. Has important benefits for research and clinical practice in the medical field. Therefore, the data labeling method and the system based on the MRI image provided by the invention are used for improving the accuracy of the MRI image data labeling. Fig. 2 is a functional block diagram of a method and a system for labeling MRI image-based data according to an embodiment of the present invention.
The MRI image-based data annotation system 200 of the present invention may be installed in an electronic device. Depending on the functionality implemented, the MRI image-based data labeling system 200 may include a smoothing module 201, an edge calculation module 202, a texture recognition module 203, a model training module 204, and an image labeling module 205. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.
In the present embodiment, the functions concerning the respective modules/units are as follows:
the smoothing module 201 is configured to obtain an original MRI image, perform image denoising on the original MRI image to obtain an MRI denoised image, and perform smoothing on the MRI denoised image to obtain an MRI smoothed image;
the edge calculation module 202 is configured to segment a target region in the MRI smooth image, extract a region parameter corresponding to the target region, calculate an image gradient corresponding to the target region based on the region parameter, query a gradient factor corresponding to the image gradient, and calculate an edge strength of the target region based on the gradient factor;
The texture recognition module 203 is configured to perform corner detection on a remaining region of the MRI smooth image based on the edge intensity, obtain a corner suppression value, extract an information factor corresponding to the corner suppression value, and recognize texture features corresponding to the remaining region based on the information factor;
the model training module 204 is configured to construct an image labeling model corresponding to the original MRI image based on the texture feature, divide an image training set corresponding to the MRI smooth image by using the image labeling model, and perform model training on the image labeling model by using the training set to obtain a trained image labeling model;
the image labeling module 205 is configured to label and predict the original MRI image by using the trained model to obtain a predicted label image, and overlap and remove the predicted label image to obtain an image labeling result corresponding to the original MRI image.
In detail, each module in the MRI image-based data labeling system 200 in the embodiment of the present invention adopts the same technical means as the MRI image-based data labeling method in the drawings, and can produce the same technical effects, which are not described herein.
Fig. 3 is a schematic structural diagram of an electronic device for implementing the MRI image-based data labeling method according to the present invention.
The electronic device 1 may comprise a processor 30, a memory 31, a communication bus 32 and a communication interface 33, and may further comprise a computer program stored in the memory 31 and executable on the processor 30, such as an engineering safety supervisor based on artificial intelligence.
The processor 30 may be formed by an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or may be formed by a plurality of integrated circuits packaged with the same function or different functions, including one or more central processing units (Central Processing unit, CPU), a microprocessor, a digital processing chip, a graphics processor, a combination of various control chips, and so on. The processor 30 is a Control Unit (Control Unit) of the electronic device 1, connects various components of the entire electronic device using various interfaces and lines, executes or executes programs or modules (e.g., an artificial intelligence-based engineering safety supervision program, etc.) stored in the memory 31, and invokes data stored in the memory 31 to perform various functions of the electronic device and process the data.
The memory 31 includes at least one type of readable storage medium including flash memory, a removable hard disk, a multimedia card, a card memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The memory 31 may in some embodiments be an internal storage unit of the electronic device, such as a mobile hard disk of the electronic device. The memory 31 may also be an external storage device of the electronic device in other embodiments, for example, a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the electronic device. Further, the memory 31 may also include both an internal storage unit and an external storage device of the electronic device. The memory 31 may be used not only for storing application software installed in an electronic device and various types of data, such as codes of a database-configured connection program, but also for temporarily storing data that has been output or is to be output.
The communication bus 32 may be a peripheral component interconnect standard (peripheral component interconnect, PCI) bus, or an extended industry standard architecture (extended industry standard architecture, EISA) bus, among others. The bus may be classified as an address bus, a data bus, a control bus, etc. The bus is arranged to enable a connection communication between the memory 31 and at least one processor 30 or the like.
The communication interface 33 is used for communication between the electronic device 1 and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and/or a wireless interface (e.g., WI-FI interface, bluetooth interface, etc.), typically used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a Display (Display), an input unit such as a Keyboard (Keyboard), or alternatively a standard wired interface, a wireless interface. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like. The display may also be referred to as a display screen or display unit, as appropriate, for displaying information processed in the electronic device and for displaying a visual user interface.
Fig. 3 shows only an electronic device with components, it being understood by a person skilled in the art that the structure shown in fig. 3 does not constitute a limitation of the electronic device 1, and may comprise fewer or more components than shown, or may combine certain components, or may be arranged in different components.
For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for supplying power to each component, and preferably, the power source may be logically connected to the at least one processor 30 through a power management device, so that functions of charge management, discharge management, power consumption management, and the like are implemented through the power management device. The power supply may also include one or more of any of a direct current or alternating current power supply, recharging device, power failure detection circuit, power converter or inverter, power status indicator, etc. The electronic device 1 may further include various sensors, bluetooth modules, wi-Fi modules, etc., which will not be described herein.
It should be understood that the examples are for illustrative purposes only.
The database-configured connection program stored in the memory 31 in the electronic device 1 is a combination of a plurality of computer programs, which, when run in the processor 30, can implement:
acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoising image, and performing smoothing treatment on the MRI denoising image to obtain an MRI smooth image;
dividing a target region in the MRI smooth image, extracting region parameters corresponding to the target region, calculating image gradients corresponding to the target region based on the region parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target region based on the gradient factors;
Performing corner detection on the residual region of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the residual region based on the information factor;
based on the texture characteristics, an image annotation model corresponding to the original MRI image is constructed, an image training set corresponding to the MRI smooth image is divided by using the image annotation model, and model training is carried out on the image annotation model by using the training set, so that a trained image annotation model is obtained;
and carrying out annotation prediction on the original MRI image by using the trained model to obtain a predicted annotation image, and carrying out overlapping removal on the predicted annotation image to obtain an image annotation result corresponding to the original MRI image.
In particular, the specific implementation method of the processor 30 on the computer program may refer to the description of the relevant steps in the corresponding embodiment of fig. 1, which is not repeated herein.
Further, the integrated modules/units of the electronic device 1 may be stored in a non-volatile computer readable storage medium if implemented in the form of software functional units and sold or used as a stand alone product. The storage medium may be volatile or nonvolatile. For example, the computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM).
The present invention also provides a storage medium storing a computer program which, when executed by a processor of an electronic device, can implement:
acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoising image, and performing smoothing treatment on the MRI denoising image to obtain an MRI smooth image;
dividing a target region in the MRI smooth image, extracting region parameters corresponding to the target region, calculating image gradients corresponding to the target region based on the region parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target region based on the gradient factors;
performing corner detection on the residual region of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the residual region based on the information factor;
based on the texture characteristics, an image annotation model corresponding to the original MRI image is constructed, an image training set corresponding to the MRI smooth image is divided by using the image annotation model, and model training is carried out on the image annotation model by using the training set, so that a trained image annotation model is obtained;
And carrying out annotation prediction on the original MRI image by using the trained model to obtain a predicted annotation image, and carrying out overlapping removal on the predicted annotation image to obtain an image annotation result corresponding to the original MRI image.
In the several embodiments provided in the present invention, it should be understood that the disclosed apparatus, device and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical function division, and there may be other manners of division when actually implemented.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in the embodiments of the present invention 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 integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
It should be noted that in this document, relational terms such as "first" and "second" and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for labeling data based on MRI images, the method comprising:
acquiring an original MRI image, performing image denoising on the original MRI image to obtain an MRI denoising image, and performing smoothing treatment on the MRI denoising image to obtain an MRI smooth image;
dividing a target region in the MRI smooth image, extracting region parameters corresponding to the target region, calculating image gradients corresponding to the target region based on the region parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target region based on the gradient factors;
performing corner detection on the residual region of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the residual region based on the information factor;
Based on the texture characteristics, an image annotation model corresponding to the original MRI image is constructed, an image training set corresponding to the MRI smooth image is divided by using the image annotation model, and model training is carried out on the image annotation model by using the training set, so that a trained image annotation model is obtained;
and carrying out annotation prediction on the original MRI image by using the trained model to obtain a predicted annotation image, and carrying out overlapping removal on the predicted annotation image to obtain an image annotation result corresponding to the original MRI image.
2. The MRI image-based data labeling method of claim 1, wherein said obtaining an original MRI image, image denoising said original MRI image, obtaining an MRI denoised image, comprises:
performing MRI scanning on a patient by using preset medical equipment to obtain an original MRI image;
performing noise analysis on the original MRI image to obtain noise data;
identifying a noise direction corresponding to the noise data;
and carrying out image denoising on the original MRI image based on the noise direction to obtain an MRI denoising image.
3. The MRI image-based data labeling method of claim 1, wherein said smoothing said MRI denoised image to obtain an MRI smoothed image comprises:
Identifying an image requirement corresponding to the MRI denoising image;
setting smoothing parameters corresponding to the MRI denoising image based on the image demand;
and on the basis of the smoothing parameters, starting a preset smoothing filter to carry out smoothing treatment on the MRI denoising image, so as to obtain an MRI smooth image.
4. The MRI image-based data labeling method as set forth in claim 1, wherein said segmenting the target region in the MRI smooth image, extracting the region parameters corresponding to the target region, comprises:
identifying a target factor corresponding to the MIR smooth image;
determining a target area corresponding to the MRI smooth image based on the target factor;
segmenting the target region in the MRI smooth image by using a preset segmentation algorithm;
and inquiring the region requirement corresponding to the target region, and extracting the region parameter corresponding to the region requirement.
5. The MRI image-based data labeling method according to claim 1, wherein said calculating an image gradient corresponding to said target region based on said region parameter comprises:
first, an average gray value corresponding to the target area is calculated using the following formula:
wherein MG represents an average gray value corresponding to the target area, N represents the number of pixels in the target area, and I (x, y) represents a pixel gray value at coordinates (x, y).
Secondly, calculating the sum of squares of differences between each pixel of the target area and the adjacent pixels based on the average gray value:
SD＝∑[I(x,y)-I(x’,y’)] 2
wherein SD represents the sum of squares of differences between each pixel of the target region and its neighboring pixels, (x ', y') represents the neighboring pixel position of the (x, y) pixel, and I (x, y) and I (x ', y') represent the pixel gray values within the target region, respectively.
Finally, calculating the image gradient corresponding to the target area based on the sum of squares of the differences:
wherein GT represents the image gradient corresponding to the target region, and N represents the number of pixels in the target region.
6. The MRI image-based data labeling method of claim 1, wherein said calculating edge intensity of said target region based on said gradient factor comprises:
where BQ represents the edge intensity of the target region, (i, j) represents the pixel coordinates in the image, (MB) represents the target region, gx (i, j) and Gy (i, j) represent the gradient values in the horizontal and vertical directions at the pixel point (i, j), respectively, and M represents the total number of pixel points in the target region.
7. The MRI image-based data labeling method of claim 1, wherein the performing corner detection on the remaining region of the MRI smooth image based on the edge intensity to obtain a corner suppression value comprises:
Determining an edge image corresponding to the residual region based on the edge intensity;
performing corner monitoring on the edge image to obtain corner parameters;
calculating a corner response value corresponding to the residual region based on the corner parameter;
and carrying out non-maximum suppression on the angular point response value to obtain an angular point suppression value.
8. The MRI image-based data labeling method of claim 1, wherein the extracting an information factor corresponding to the corner suppression value, and identifying texture features corresponding to the remaining region based on the information factor, comprises:
determining a fixed neighborhood corresponding to the corner suppression value, and extracting an information factor corresponding to the fixed neighborhood;
judging the texture category corresponding to the information factor;
and identifying texture features corresponding to the residual area based on the texture category.
9. The MRI image-based data labeling method according to claim 1, wherein said labeling prediction of said original MRI image using a trained model to obtain a predicted labeling image comprises:
acquiring an MRI image sample corresponding to the original MRI image;
equalizing the MRI image sample to obtain an equalized sample;
Carrying out image prediction on the balanced sample by using the trained model to obtain a predicted image;
and carrying out image annotation on the predicted image to obtain a predicted and annotated image.
10. An MRI image-based data annotation system for performing an MRI image-based data annotation method according to any one of claims 1-9, said system comprising:
the smoothing processing module is used for acquiring an original MRI image, carrying out image denoising on the original MRI image to obtain an MRI denoising image, and carrying out smoothing processing on the MRI denoising image to obtain an MRI smooth image;
the edge calculation module is used for dividing a target area in the MRI smooth image, extracting area parameters corresponding to the target area, calculating image gradients corresponding to the target area based on the area parameters, inquiring gradient factors corresponding to the image gradients, and calculating edge strength of the target area based on the gradient factors;
the texture recognition module is used for carrying out corner detection on the residual area of the MRI smooth image based on the edge intensity to obtain a corner suppression value, extracting an information factor corresponding to the corner suppression value, and recognizing texture features corresponding to the residual area based on the information factor;
The model training module is used for constructing an image annotation model corresponding to the original MRI image based on the texture features, dividing an image training set corresponding to the MRI smooth image by using the image annotation model, and carrying out model training on the image annotation model by using the training set to obtain a trained image annotation model;
and the image labeling module is used for labeling and predicting the original MRI image by using the trained model to obtain a predicted labeling image, and overlapping and removing the predicted labeling image to obtain an image labeling result corresponding to the original MRI image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311810565.1A CN117765330A (en) | 2023-12-25 | 2023-12-25 | MRI image-based data labeling method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311810565.1A CN117765330A (en) | 2023-12-25 | 2023-12-25 | MRI image-based data labeling method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117765330A true CN117765330A (en) | 2024-03-26 |
Family
ID=90319808
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311810565.1A Pending CN117765330A (en) | 2023-12-25 | 2023-12-25 | MRI image-based data labeling method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117765330A (en) |
-
2023
- 2023-12-25 CN CN202311810565.1A patent/CN117765330A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110678903B (en) | System and method for analysis of ectopic ossification in 3D images | |
| CN108010021B (en) | Medical image processing system and method | |
| CN109522908B (en) | Image significance detection method based on region label fusion | |
| CN110334706B (en) | Image target identification method and device | |
| JP6660313B2 (en) | Detection of nuclear edges using image analysis | |
| Bergmeir et al. | Segmentation of cervical cell nuclei in high-resolution microscopic images: a new algorithm and a web-based software framework | |
| Poletti et al. | A review of thresholding strategies applied to human chromosome segmentation | |
| Abbas et al. | Computer‐aided pattern classification system for dermoscopy images | |
| EP3176751B1 (en) | Information processing device, information processing method, computer-readable recording medium, and inspection system | |
| EP3391284A1 (en) | Interpretation and quantification of emergency features on head computed tomography | |
| CN106296653A (en) | Brain CT image hemorrhagic areas dividing method based on semi-supervised learning and system | |
| US20090252429A1 (en) | System and method for displaying results of an image processing system that has multiple results to allow selection for subsequent image processing | |
| US8731278B2 (en) | System and method for sectioning a microscopy image for parallel processing | |
| WO2020253508A1 (en) | Abnormal cell detection method and apparatus, and computer readable storage medium | |
| US20090279778A1 (en) | Method, a system and a computer program for determining a threshold in an image comprising image values | |
| Son et al. | Morphological change tracking of dendritic spines based on structural features | |
| CN110490159B (en) | Method, device, equipment and storage medium for identifying cells in microscopic image | |
| CN112464803A (en) | Image comparison method and device | |
| CN114092450A (en) | Real-time image segmentation method, system and device based on gastroscopy video | |
| Feng et al. | Segmentation fusion based on neighboring information for MR brain images | |
| CN116433704A (en) | Cell nucleus segmentation method based on central point and related equipment | |
| CN108765399B (en) | Lesion site recognition device, computer device, and readable storage medium | |
| CN117274278B (en) | Retina image focus part segmentation method and system based on simulated receptive field | |
| Pham et al. | Extraction of fluorescent cell puncta by adaptive fuzzy segmentation | |
| CN110097071A (en) | The recognition methods in the breast lesion region based on spectral clustering in conjunction with K-means and device |
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 |