CN109919098B - Target object identification method and device - Google Patents
Target object identification method and device Download PDFInfo
- Publication number
- CN109919098B CN109919098B CN201910176768.7A CN201910176768A CN109919098B CN 109919098 B CN109919098 B CN 109919098B CN 201910176768 A CN201910176768 A CN 201910176768A CN 109919098 B CN109919098 B CN 109919098B
- Authority
- CN
- China
- Prior art keywords
- level set
- value
- target object
- image
- obtaining
- 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
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a target object identification method and a target object identification device. Wherein, the method comprises the following steps: acquiring a first image containing a target object; and carrying out image segmentation on the first image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function. The invention solves the technical problems of low identification accuracy and poor repeatability of the target object in the prior art.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a target object recognition method and device.
Background
In image recognition in the medical field, Anterior segment optical coherence tomography (AS-OCT) is used to assist in diagnosing many ophthalmic diseases, such AS corneal diseases, cataract, glaucoma, etc., and is a non-invasive, non-invasive imaging method. The crystalline lens, which is the main refractive structure of the eyeball, causes visual impairment when clouding (density increase) of the crystalline lens occurs, resulting in cataract. The density of crystalline lens is an important index for measuring the severity of cataract and other diseases, and the crystalline lens is a multi-layer structure and can be roughly divided into the following parts from outside to inside: the corneal layer, the cortex layer, and the nucleus. At present, most of lens structure segmentation based on AS-OCT images is carried out manually, and the defects of poor repeatability, high labor cost and the like are overcome. Blurring between the boundaries of lens structures in AS-OCT images leads to difficulties in fully automated segmentation, especially at the boundaries of the nucleus (nucleous) and cortex (cortix). Moreover, the boundaries can become more blurred in AS-OCT images taken in the eyes of cataract patients.
Aiming at the problems of low identification accuracy and poor repeatability of a target object in the prior art, an effective solution is not provided at present.
Disclosure of Invention
The embodiment of the invention provides a target object identification method and device, which at least solve the technical problems of low target object identification accuracy and poor repeatability in the prior art.
According to an aspect of the embodiments of the present invention, there is provided a target object identification method, including: acquiring a first image containing a target object; and carrying out image segmentation on the first image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function.
Further, the U-shaped full convolution neural network model includes: an encoding model and a decoding model, the encoding model comprising: a plurality of first network blocks, the first network blocks comprising: at least two first convolution layers connected in sequence, wherein the first convolution layers adopt a modified linear unit activation function and a pooling operation, and the last first convolution layer in a first network block is connected with the first convolution layer in the next first network block; the decoding model includes: a plurality of second network blocks and an output layer, the number of the first network blocks and the number of the second network blocks being the same, the second network blocks comprising: the cascade layer is connected with the corresponding first network block and the last second convolution layer in the last second network block in a copying and merging jump connection mode, the output layer is connected with the side output layer of the last second in the last second network block, and the output layer adopts a level set loss function.
Further, the pooling operation includes one of: a maximum pooling operation and a mean pooling operation.
Further, the level set loss function is determined by: determining that the result input to the output layer is a first level set and determining that the result output by the output layer is a second level set; acquiring the shape of the first level set and the shape of the second level set; a level set loss function is derived based on the shape of the first level set and the shape of the second level set.
Further, obtaining the shape of the first level set comprises: acquiring a first function value of a first level set and a second function value of a truth level set of a first image; based on the first function value and the second function value, a shape of the first level set is obtained.
Further, deriving a shape of the first level set based on the first function value and the second function value, comprising: obtaining the difference between the second function value and the first function value to obtain a difference value; obtaining an absolute value of the difference value to obtain an absolute value; obtaining the square of the absolute value to obtain a square value; the integral of the squared value is obtained, resulting in the shape of the first level set.
Further, after obtaining the second function value of the truth level set, the method further includes: processing the second function value based on the level set conversion function to obtain a processed second function value; normalizing the processed second function value to obtain a normalized function value; based on the first function value and the normalized function value, a shape of the first level set is obtained.
Further, obtaining a shape of the second level set includes: acquiring a plurality of probability values input to an output layer and a true value corresponding to each probability value; obtaining a product of each probability value based on each probability value and the corresponding true value; and acquiring the sum of the products of the probability values to obtain the shape of the second level set.
Further, obtaining a product of each probability value based on each probability value and the corresponding true value, including: obtaining the logarithm of each probability value to obtain the logarithm value of each probability value; and obtaining the product of the logarithm value of each probability value and the corresponding true value to obtain the product of each probability value.
Further, deriving a level set loss function based on the shape of the first level set and the shape of the second level set, comprising: obtaining a product of the shape of the first level set and the first parameter to obtain a first product; obtaining a product of the shape of the second level set and the second parameter to obtain a second product; obtaining the product of the length of the boundary of the target object and the third parameter to obtain a third product; acquiring a product of an integral value of the area where the target object is located and a fourth parameter to obtain a fourth product, wherein the integral value is used for representing the inside and the outside of the area where the target object is located; and acquiring the sum of the first product, the second product, the third product and the fourth product to obtain a level set loss function.
Further, before the image segmentation is performed on the first image by using the U-shaped full convolution neural network model to obtain the boundary of the target object, the method further includes: processing the first image by using an edge detection algorithm to obtain a second image of the region where the target object is located; and carrying out image segmentation on the second image by utilizing the U-shaped full convolution neural network model to obtain the boundary of the target object.
Further, processing the first image by using an edge detection algorithm to obtain a second image of the region where the target object is located, including: and processing the first image by utilizing a multi-level edge detection algorithm to obtain a second image.
Further, acquiring a first image containing a target object, comprises: and scanning the target object by utilizing a front-segment optical coherence tomography technology to obtain a first image.
Further, the target object is the lens nucleus.
According to another aspect of the embodiments of the present invention, there is also provided an apparatus for identifying a target object, including: an acquisition module for acquiring an image containing a target object; and the image segmentation module is used for carrying out image segmentation on the image by utilizing the U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function.
According to another aspect of the embodiments of the present invention, there is also provided a storage medium including a stored program, wherein when the program runs, a device on which the storage medium is located is controlled to execute the above-mentioned target object identification method.
According to another aspect of the embodiments of the present invention, there is also provided a processor, configured to execute a program, where the program executes the method for identifying a target object.
In the embodiment of the present invention, after the first image including the target object is obtained, the first image may be subjected to image segmentation by using a U-shaped full convolution neural network model using a level set loss function, so as to obtain a boundary of the target object, that is, obtain a final segmentation result. It is easy to notice that the full-automatic lens structure segmentation based on deep learning is realized by combining the U-shaped full-convolution neural network model and the level set algorithm to perform the lens structure segmentation, so that the technical effect of effectively improving the accuracy and repeatability of the lens structure segmentation is achieved, and the technical problems of low identification accuracy and poor repeatability of a target object in the prior art are solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
fig. 1 is a flowchart of a target object identification method according to an embodiment of the present invention;
FIG. 2a is a schematic illustration of an alternative AS-OCT image according to an embodiment of the invention;
FIG. 2b is a schematic diagram of an area where an optional target object is located according to an embodiment of the present invention;
FIG. 2c is a diagram illustrating the results of an alternative segmentation according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an alternative network architecture for a U-net network according to an embodiment of the present invention;
FIG. 4a is a schematic diagram of an alternative truth level set in accordance with an embodiment of the present invention;
FIG. 4b is a schematic diagram of an alternative truth label in accordance with an embodiment of the present invention;
FIG. 4c is a graphical representation of an alternative normalized result according to an embodiment of the present invention; and
fig. 5 is a schematic diagram of a target object recognition apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example 1
In accordance with an embodiment of the present invention, there is provided an embodiment of a method for identifying a target object, it should be noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.
Fig. 1 is a flowchart of a target object identification method according to an embodiment of the present invention, as shown in fig. 1, the method includes the following steps:
step S102, a first image including a target object is acquired.
Alternatively, the target object may be a lens nucleus.
Specifically, the first image may be an AS-OCT fundus image, AS shown in fig. 2a, in which the image includes a lens structure, and the area where the white box is located is a lens area. Due to the blurring between the boundaries of the lens nucleus and the cortex, the accurate segmentation of the lens structure can be achieved by accurately segmenting the boundaries of the lens nucleus and the cortex, and therefore, the target object can be determined to be the lens nucleus, and the region where the target object is located is shown in fig. 2 b.
And S104, carrying out image segmentation on the first image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function.
Specifically, in the field of medical image segmentation, the image segmentation algorithm based on deep learning is developed rapidly at present, and particularly, a U-shaped full convolution neural network model, namely a U-net network, is developed greatly in the field of medical images such as lung nodule, tumor and blood vessel thickness segmentation. Thus, the lens nucleus region can be determined AS a segmentation of the initial contour of the lens structure in the AS-OCT image through the U-net network, the final segmentation result being shown in FIG. 2 c.
It should be noted that, by image segmentation of the AS-OCT fundus map through the U-net network, not only the nucleus region of the crystalline lens but also the cornea and cortex regions of the crystalline lens can be determined.
Because the boundary segmented by the U-net network is irregular, particularly the segmentation of the lens nucleus, and the level set method obtains better performance under the constraints of shape, noise, resolution, occlusion and the like, the level set loss function can be learned through the network, and the finally obtained zero level set represents the final segmentation result, namely, the zero level set identifies the boundary of the region where the target object is located.
By the above embodiment of the present invention, after the first image including the target object is obtained, the first image may be subjected to image segmentation by using the U-shaped full convolution neural network model using the level set loss function, so as to obtain the boundary of the target object, that is, obtain the final segmentation result. It is easy to notice that the full-automatic lens structure segmentation based on deep learning is realized by combining the U-shaped full-convolution neural network model and the level set algorithm to perform the lens structure segmentation, so that the technical effect of effectively improving the accuracy and repeatability of the lens structure segmentation is achieved, and the technical problems of low identification accuracy and poor repeatability of a target object in the prior art are solved.
Optionally, the U-shaped full convolution neural network model includes: an encoding model and a decoding model, the encoding model comprising: a plurality of first network blocks, the first network blocks comprising: at least two first convolution layers connected in sequence, wherein the first convolution layers adopt a modified linear unit activation function and a pooling operation, and the last first convolution layer in a first network block is connected with the first convolution layer in the next first network block; the decoding model includes: a plurality of second network blocks and an output layer, the number of the first network blocks and the number of the second network blocks being the same, the second network blocks comprising: the cascade layer is connected with the corresponding first network block and the last second convolution layer in the last second network block in a copying and merging jump connection mode, the output layer is connected with the side output layer of the last second in the last second network block, and the output layer adopts a level set loss function.
Optionally, the pooling operation described above may include one of: a maximum pooling operation and a mean pooling operation.
Specifically, because the size of the AS-OCT fundus map is larger, in order to improve the image segmentation accuracy and repeatability, the existing U-net network can be improved, and the improvement points mainly lie in that: the size of the input image is larger, e.g. the input size is 1024 x 1024, i.e. the size of the first image is 1024 x 1024; convolution layers use a smaller convolution kernel, which may be 3x3, for example; the overall U-net network has a large number of layers, for example, the overall network has 6 layers, that is, the number of the first network block and the second network block is 6.
In an embodiment of the invention, each convolutional layer may use an activation function (Relu) and max pooling operations.
The Network structure of the U-net Network is shown in fig. 3, the Network structure includes a coding model (shown on the left side in fig. 3) and a decoding model (shown on the right side in fig. 3), VGG19(Visual Geometry Group Network) can be used as a coding part of the Network, and includes six Network blocks (i.e. the first Network block described above), each Network block includes two to three convolutional layers, each convolutional layer uses an activation function Relu (modified linear unit) and a maximum pooling of 2 × 2 with a step size of 2 for downsampling, a convolution kernel of the convolutional layer is 3 × 3, which can be expressed as Conv <3X3> with the value of Relu, which is beneficial to the expression of detail features, the test speed can be effectively guaranteed relative to a deeper residual Network, and the number of feature channels is doubled in each downsampling step; the decoding module also comprises six network blocks (i.e. the second network block mentioned above), each network block comprising a concatenation of layers, from the respective feature layer (i.e. the corresponding first network block mentioned above) and the upsampling (up-sample) (i.e. the corresponding second convolutional layer mentioned above), with a coefficient of 2, followed by the use of two convolutional layers and the side output layer side-output, the convolutional layers using a convolution kernel of 3x3 and using an activation function (Relu). The coding model and the decoding model can use a jump connection mode of Copy and Merge (Copy and Merge), so that the characteristics of different layers can be effectively utilized, and a better segmentation effect is guaranteed; the final output layer may take the level set penalty function as the final penalty.
Optionally, the level set loss function is determined by: determining that the result input to the output layer is a first level set and determining that the result output by the output layer is a second level set; acquiring the shape of the first level set and the shape of the second level set; a level set loss function is derived based on the shape of the first level set and the shape of the second level set.
Optionally, obtaining the shape of the first level set comprises: acquiring a first function value of a first level set and a second function value of a truth level set of a first image; based on the first function value and the second function value, a shape of the first level set is obtained.
Specifically, deriving a shape of the first level set based on the first function value and the second function value includes: obtaining the difference between the second function value and the first function value to obtain a difference value; obtaining an absolute value of the difference value to obtain an absolute value; obtaining the square of the absolute value to obtain a square value; the integral of the squared value is obtained, resulting in the shape of the first level set.
In addition, after obtaining the second function value of the truth level set, the method further comprises the following steps: processing the second function value based on the level set conversion function to obtain a processed second function value; normalizing the processed second function value to obtain a normalized function value; based on the first function value and the normalized function value, a shape of the first level set is obtained.
Similarly, obtaining the shape of the second level set comprises: acquiring a plurality of probability values input to an output layer and a true value corresponding to each probability value; obtaining a product of each probability value based on each probability value and the corresponding true value; and acquiring the sum of the products of the probability values to obtain the shape of the second level set.
Specifically, obtaining a product of each probability value based on each probability value and the corresponding true value includes: obtaining the logarithm of each probability value to obtain the logarithm value of each probability value; and obtaining the product of the logarithm value of each probability value and the corresponding true value to obtain the product of each probability value.
Optionally, deriving a level set loss function based on the shape of the first level set and the shape of the second level set comprises: obtaining a product of the shape of the first level set and the first parameter to obtain a first product; obtaining a product of the shape of the second level set and the second parameter to obtain a second product; obtaining the product of the length of the boundary of the target object and the third parameter to obtain a third product; acquiring a product of an integral value of the area where the target object is located and a fourth parameter to obtain a fourth product, wherein the integral value is used for representing the inside and the outside of the area where the target object is located; and acquiring the sum of the first product, the second product, the third product and the fourth product to obtain a level set loss function.
In an alternative, the level set phi is the input image (i.e. the first image described above), and the value of each pixel (x, y), i.e. the function value of the level set, is the function of the surface
W is the region where the segmented target object is located. The result of the input to the output layer can be taken as the first level set phipreThat is, the output of the U-net network is the level set φ during the training processpreAnd the result output by the output layer, i.e. the result of the image segmentation, is taken as the second level set, H (phi), i.e. the zero level set.
The level set loss function consists essentially of two parts, the first part being a first level set phipreThe second part is the shape of the zero level set H (phi), and specifically, the calculation formula of the first part is as follows:
wherein phi isgtIs the true level set, i.e., the true value of the input image is taken as the level set.
The second part of the calculation formula is as follows:
wherein, H (phi)pre) Probability y of representing U-net network outputi，yi' means probability yiThe corresponding true value.
Further, the formula for the minimized level set loss function is calculated as follows:
where C is the active contour (i.e., the boundary of the target object), the last two terms are the inside and outside of the connected region w (i.e., the region where the target object is located), and C1And c2Is constant and can be calculated by the following formula:
in addition, the U-net network training process includes two truth values, i.e., a truth level set and a truth label, as shown in fig. 4a and 4b, respectively. The truth level set can be obtained by fast distance conversion, since the Euclidean distance is used, the image becomes larger, the overall distance becomes larger, and if the Euclidean distance is directly normalized to [ -1,1], the response of the corresponding boundary is reduced, so that the truth level set can be converted by the following method:
wherein phi isNgt(x, y) is the converted level set (i.e., the processed second function value described above).
That is, the function value of the truth level set may be compared with preset values, and if the function value is greater than or equal to one preset value (i.e., the first preset value), the function value may be converted into the preset value; if the value is smaller than another preset value (namely the first preset value), the function value can be converted into the preset value; for other cases, it remains unchanged. The converted level set may further be normalized to [ -1,1], with the normalized result shown in fig. 4 c.
Optionally, before the step S104, performing image segmentation on the first image by using a U-shaped full convolution neural network model to obtain the boundary of the target object, the method further includes the following steps: processing the first image by using an edge detection algorithm to obtain a second image of the region where the target object is located; and carrying out image segmentation on the second image by utilizing the U-shaped full convolution neural network model to obtain the boundary of the target object.
Specifically, in order to reduce redundant interference information, an edge detection algorithm may be used to extract an image of the lens region, obtain a second image with a size of 1024 × 1024, and input the extracted image to a U-shaped full convolution neural network for image segmentation.
Optionally, processing the first image by using an edge detection algorithm to obtain a second image of the region where the target object is located, including: and processing the first image by utilizing a multi-level edge detection algorithm to obtain a second image.
In particular, the multi-level edge detection algorithm canny operator can be used to extract the lens region as a pre-process.
Optionally, acquiring a first image containing the target object comprises: and scanning the target object by utilizing a front-segment optical coherence tomography technology to obtain a first image.
Specifically, the lens can be photographed by the AS-OCT technique, resulting in an AS-OCT fundus image (i.e., the first image described above).
Through the scheme, compared with the prior art, the method directly models the edge and has more forward control force on the edge to be learned; constraint conditions are easy to add, and a plurality of constraint conditions such as shape or area consistency are easy to express by a horizontal set, so that the effect is obviously improved; the generalization is strong, and the migration to other tasks is easy.
Example 2
According to an embodiment of the present invention, an embodiment of an apparatus for identifying a target object is provided.
Fig. 5 is a schematic diagram of an apparatus for identifying a target object according to an embodiment of the present invention, as shown in fig. 5, the apparatus including: an acquisition module 52 and an image segmentation module 54.
The acquiring module 52 is configured to acquire a first image containing a target object; the image segmentation module 54 is configured to perform image segmentation on the first image by using a U-shaped full convolution neural network model, so as to obtain a boundary of the target object, where the U-shaped full convolution neural network model uses a level set loss function.
Alternatively, the target object may be a lens nucleus.
Specifically, the first image may be an AS-OCT fundus image, AS shown in fig. 2a, in which the image includes a lens structure, and the area where the white box is located is a lens area. Due to the blurring between the boundaries of the lens nucleus and the cortex, the accurate segmentation of the lens structure can be realized by accurately segmenting the boundaries of the lens nucleus and the cortex, and therefore, the target object can be determined to be the lens nucleus, and the region where the target object is located is as shown in fig. 2b, wherein the black part is the cortex, and the white part in the middle of the black part is the nucleus.
In the field of medical image segmentation, the image segmentation algorithm based on deep learning is developed rapidly at present, and particularly, a U-shaped full convolution neural network model, namely a U-net network, is developed greatly in the field of medical images such as lung nodule, tumor and blood vessel thickness segmentation. Thus, the lens nucleus region can be determined AS a segmentation of the initial contour of the lens structure in the AS-OCT image through the U-net network, the final segmentation result being shown in FIG. 2 c.
It should be noted that, by image segmentation of the AS-OCT fundus map through the U-net network, not only the nucleus region of the crystalline lens but also the cornea and cortex regions of the crystalline lens can be determined.
Because the boundary segmented by the U-net network is irregular, particularly the segmentation of the lens nucleus, and the level set method obtains better performance under the constraints of shape, noise, resolution, occlusion and the like, the level set loss function can be learned through the network, and the finally obtained zero level set represents the final segmentation result, namely, the other level set identifies the boundary of the region where the target object is located.
According to the embodiment of the invention, after the first image containing the target object is acquired by the acquisition module, the image segmentation module can be used for carrying out image segmentation on the first image by using the U-shaped full convolution neural network model adopting the level set loss function to obtain the boundary of the target object, namely, the final segmentation result. It is easy to notice that the full-automatic lens structure segmentation based on deep learning is realized by combining the U-shaped full-convolution neural network model and the level set algorithm to perform the lens structure segmentation, so that the technical effect of effectively improving the accuracy and repeatability of the lens structure segmentation is achieved, and the technical problems of low identification accuracy and poor repeatability of a target object in the prior art are solved.
Optionally, the U-shaped full convolution neural network model includes: an encoding model and a decoding model, the encoding model comprising: a plurality of first network blocks, the first network blocks comprising: at least two first convolution layers connected in sequence, wherein the first convolution layers adopt a modified linear unit activation function and a pooling operation, and the last first convolution layer in a first network block is connected with the first convolution layer in the next first network block; the decoding model includes: a plurality of second network blocks and an output layer, the number of the first network blocks and the number of the second network blocks being the same, the second network blocks comprising: the cascade layer is connected with the corresponding first network block and the last second convolution layer in the last second network block in a copying and merging jump connection mode, the output layer is connected with the side output layer of the last second in the last second network block, and the output layer adopts a level set loss function.
Optionally, the pooling operation described above may include one of: a maximum pooling operation and a mean pooling operation.
Optionally, the level set loss function is determined by: the device comprises a determining submodule, an obtaining submodule and a processing submodule.
The determining submodule is used for determining that a result input to the output layer is a first level set and determining that a result output by the output layer is a second level set; the obtaining submodule is used for obtaining the shape of the first level set and the shape of the second level set; the processing submodule is used for obtaining a level set loss function based on the shape of the first level set and the shape of the second level set.
Optionally, the obtaining sub-module includes: the device comprises a first acquisition unit and a first processing unit.
The first obtaining unit is used for obtaining a first function value of a first level set and a second function value of a truth level set of the first image; the first processing unit is used for obtaining the shape of the first level set based on the first function value and the second function value.
Specifically, the first processing unit includes: a first acquisition subunit, a second acquisition subunit, a third acquisition subunit, and a fourth acquisition subunit.
The first obtaining subunit is configured to obtain a difference between the second function value and the first function value, so as to obtain a difference value; the second obtaining subunit is configured to obtain an absolute value of the difference value to obtain an absolute value; the third acquisition subunit is used for acquiring the square of the absolute value to obtain a square value; the fourth obtaining subunit is configured to obtain an integral of the square value, and obtain a shape of the first level set.
In addition, the apparatus further comprises: the device comprises a first processing subunit, a second processing subunit and a third processing subunit.
The first processing subunit is configured to process the second function value based on the level set conversion function to obtain a processed second function value; the second processing subunit is used for carrying out normalization processing on the processed second function value to obtain a normalized function value; the third processing subunit is configured to obtain a shape of the first level set based on the first function value and the normalized function value.
Similarly, the obtaining sub-module includes: the device comprises a second acquisition unit, a second processing unit and a third acquisition unit.
The second acquisition unit is used for acquiring a plurality of probability values input to the output layer and a true value corresponding to each probability value; the second processing unit is used for obtaining the product of each probability value based on each probability value and the corresponding true value; the third obtaining unit is used for obtaining the sum of the products of the probability values to obtain the shape of the second level set.
Specifically, the second processing unit includes: a fifth acquisition subunit and a sixth acquisition subunit.
The fifth obtaining subunit is configured to obtain a logarithm of each probability value to obtain a logarithm value of each probability value; the sixth obtaining subunit is configured to obtain a product of the logarithm value of each probability value and the corresponding true value, and obtain a product of each probability value.
Optionally, the processing submodule includes: a fourth acquisition unit, a fifth acquisition unit, a sixth acquisition unit, a seventh acquisition unit, and an eighth acquisition unit.
The fourth acquiring unit is used for acquiring the product of the shape of the first level set and the first parameter to obtain a first product; the fifth acquiring unit is used for acquiring the product of the shape of the second level set and the second parameter to obtain a second product; the sixth obtaining unit is used for obtaining the product of the length of the boundary of the target object and the third parameter to obtain a third product; the seventh acquiring unit is used for acquiring a product of an integral value of the area where the target object is located and a fourth parameter to obtain a fourth product, wherein the integral value is used for representing the inside and the outside of the area where the target object is located; the eighth obtaining unit is configured to obtain a sum of the first product, the second product, the third product, and the fourth product to obtain a level set loss function.
Optionally, the apparatus comprises: and a processing module.
The processing module is used for processing the first image by utilizing an edge detection algorithm to obtain a second image of the area where the target object is located; the image segmentation module is further used for carrying out image segmentation on the second image by using the U-shaped full convolution neural network model to obtain the boundary of the target object.
Optionally, the processing module comprises: and a processing submodule.
The processing submodule is further used for processing the first image by utilizing a multi-level edge detection algorithm to obtain a second image.
Optionally, the obtaining module includes: the sub-modules are scanned.
The scanning sub-module is used for scanning the target object by utilizing a front-segment optical coherence tomography technology to obtain a first image.
Example 3
According to an embodiment of the present invention, an embodiment of a storage medium is provided, the storage medium including a stored program, wherein when the program runs, a device on which the storage medium is located is controlled to execute the method for identifying a target object in the above-described embodiment 1.
Example 4
According to an embodiment of the present invention, an embodiment of a processor for running a program is provided, where the program executes the method for identifying a target object in embodiment 1.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (14)
1. A method for identifying a target object, comprising:
acquiring a first image containing a target object;
carrying out image segmentation on the first image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function;
wherein the level set loss function is determined by:
determining a result input to an output layer of the U-shaped full convolution neural network model as a first level set, and determining a result output by the output layer as a second level set;
obtaining a shape of the first level set and a shape of the second level set;
deriving the level set loss function based on the shape of the first level set and the shape of the second level set;
obtaining shapes of the first level set, comprising:
obtaining a first function value of the first level set and a second function value of a truth level set of the first image;
deriving a shape of the first level set based on the first function value and the second function value;
obtaining shapes of the second level set, comprising:
acquiring a plurality of probability values input to the output layer and a true value corresponding to each probability value;
obtaining a product of each probability value based on each probability value and the corresponding true value;
and acquiring the sum of the products of the probability values to obtain the shape of the second level set.
2. The method of claim 1, wherein the U-shaped full convolution neural network model comprises: a coding model and a decoding model, and,
the coding model comprises: a plurality of first network blocks, the first network blocks comprising: at least two first convolution layers connected in sequence, wherein the first convolution layers adopt a modified linear unit activation function and a pooling operation, and the last first convolution layer in the first network block is connected with the first convolution layer in the next first network block;
the decoding model comprises: a plurality of second network blocks and an output layer, the first network blocks and the second network blocks being the same in number, the second network blocks including: the cascade layer and two at least second convolution layers that connect gradually, the cascade layer is connected with the last second convolution layer in the first network piece that corresponds and the last second network piece, the cascade layer with the jump connection mode that corresponds first network piece adoption duplication and amalgamation is connected, the output layer is connected with the side output layer of last second in the last second network piece, the output layer adopts level set loss function.
3. The method of claim 2, wherein the pooling operation comprises one of: a maximum pooling operation and a mean pooling operation.
4. The method of claim 2, wherein deriving the shape of the first level set based on the first function value and the second function value comprises:
obtaining the difference between the second function value and the first function value to obtain a difference value;
obtaining an absolute value of the difference value to obtain an absolute value;
obtaining the square of the absolute value to obtain a square value;
and acquiring the integral of the square value to obtain the shape of the first level set.
5. The method of claim 2, wherein after obtaining the second function value for the truth level set, the method further comprises:
processing the second function value based on a level set conversion function to obtain a processed second function value;
normalizing the processed second function value to obtain a normalized function value;
deriving a shape of the first level set based on the first function value and the normalized function value.
6. The method of claim 2, wherein deriving a product of each probability value based on the each probability value and a corresponding true value comprises:
obtaining the logarithm of each probability value to obtain the logarithm value of each probability value;
and obtaining the product of the logarithm value of each probability value and the corresponding true value to obtain the product of each probability value.
7. The method of claim 2, wherein deriving the level set loss function based on the shape of the first level set and the shape of the second level set comprises:
obtaining a product of the shape of the first level set and a first parameter to obtain a first product;
obtaining a product of the shape of the second level set and a second parameter to obtain a second product;
obtaining a product of the length of the boundary of the target object and a third parameter to obtain a third product;
acquiring a product of an integral value of the area where the target object is located and a fourth parameter to obtain a fourth product, wherein the integral value is used for representing the inside and the outside of the area where the target object is located;
and acquiring the sum of the first product, the second product, the third product and the fourth product to obtain the level set loss function.
8. The method of claim 1, wherein prior to image segmenting the first image using a U-shaped full convolution neural network model to obtain the boundary of the target object, the method further comprises:
processing the first image by utilizing an edge detection algorithm to obtain a second image of the area where the target object is located;
and carrying out image segmentation on the second image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object.
9. The method of claim 8, wherein processing the first image using an edge detection algorithm to obtain a second image of the region where the target object is located comprises:
and processing the first image by utilizing a multi-level edge detection algorithm to obtain the second image.
10. The method of claim 1, wherein acquiring a first image containing a target object comprises:
and scanning the target object by utilizing a front-segment optical coherence tomography technology to obtain the first image.
11. The method of claim 1, wherein the target object is a lens nucleus.
12. An apparatus for identifying a target object, comprising:
an acquisition module for acquiring a first image containing a target object;
the image segmentation module is used for carrying out image segmentation on the first image by utilizing a U-shaped full convolution neural network model to obtain the boundary of the target object, wherein the U-shaped full convolution neural network model adopts a level set loss function;
wherein the level set loss function is determined by:
the determining submodule is used for determining that a result input to an output layer of the U-shaped full convolution neural network model is a first level set and determining that a result output by the output layer is a second level set;
an obtaining sub-module for obtaining a shape of the first level set and a shape of the second level set;
a processing sub-module for deriving the level set loss function based on a shape of the first level set and a shape of the second level set;
the acquisition sub-module includes:
a first obtaining unit, configured to obtain a first function value of the first level set and a second function value of a truth level set of the first image;
a first processing unit configured to obtain a shape of the first level set based on the first function value and the second function value;
the acquisition sub-module includes:
the second acquisition unit is used for acquiring a plurality of probability values input to the output layer and a true value corresponding to each probability value;
the second processing unit is used for obtaining the product of each probability value based on each probability value and the corresponding true value;
and the third acquisition unit is used for acquiring the sum of the products of the probability values to obtain the shape of the second level set.
13. A storage medium, characterized in that the storage medium comprises a stored program, wherein when the program runs, a device where the storage medium is located is controlled to execute the identification method of the target object according to any one of claims 1 to 11.
14. A processor, configured to execute a program, wherein the program executes the method for identifying a target object according to any one of claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910176768.7A CN109919098B (en) | 2019-03-08 | 2019-03-08 | Target object identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910176768.7A CN109919098B (en) | 2019-03-08 | 2019-03-08 | Target object identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109919098A CN109919098A (en) | 2019-06-21 |
| CN109919098B true CN109919098B (en) | 2021-06-15 |
Family
ID=66963937
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910176768.7A Active CN109919098B (en) | 2019-03-08 | 2019-03-08 | Target object identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109919098B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111539306B (en) * | 2020-04-21 | 2021-07-06 | 中南大学 | Remote sensing image building identification method based on activation expression replaceability |
| CN117422880B (en) * | 2023-12-18 | 2024-03-22 | 齐鲁工业大学(山东省科学院) | Segmentation method and system combining improved attention mechanism and CV model |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831608A (en) * | 2012-08-06 | 2012-12-19 | 哈尔滨工业大学 | Unsteady measurement algorithm based image segmentation method of improved rule distance level set |
| EP3365870B1 (en) * | 2015-10-19 | 2020-08-26 | The Charles Stark Draper Laboratory, Inc. | System and method for the segmentation of optical coherence tomography slices |
| CN108492286B (en) * | 2018-03-13 | 2020-05-05 | 成都大学 | Medical image segmentation method based on dual-channel U-shaped convolutional neural network |
| CN109035252B (en) * | 2018-06-29 | 2019-09-24 | 山东财经大学 | A kind of super-pixel method towards medical image segmentation |
| CN109087327B (en) * | 2018-07-13 | 2021-07-06 | 天津大学 | Thyroid nodule ultrasonic image segmentation method of cascaded full convolution neural network |
-
2019
- 2019-03-08 CN CN201910176768.7A patent/CN109919098B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109919098A (en) | 2019-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110232383B (en) | Focus image recognition method and focus image recognition system based on deep learning model | |
| Math et al. | Adaptive machine learning classification for diabetic retinopathy | |
| CN107369160B (en) | Choroid neogenesis blood vessel segmentation algorithm in OCT image | |
| CN110276356B (en) | Fundus image microaneurysm identification method based on R-CNN | |
| CN111292338B (en) | Method and system for segmenting choroidal neovascularization from fundus OCT image | |
| US20210390696A1 (en) | Medical image processing apparatus, medical image processing method and computer-readable storage medium | |
| CN109919954B (en) | Target object identification method and device | |
| CN112132817A (en) | Retina blood vessel segmentation method for fundus image based on mixed attention mechanism | |
| CN112017185B (en) | Focus segmentation method, device and storage medium | |
| CA3104562A1 (en) | Method and computer program for segmentation of optical coherence tomography images of the retina | |
| CN113643353B (en) | Measurement method for enhancing resolution of vascular caliber of fundus image | |
| CN113870270B (en) | Fundus image cup and optic disc segmentation method under unified frame | |
| CN109754388B (en) | Carotid artery stenosis degree calculation method and device and storage medium | |
| CN109919098B (en) | Target object identification method and device | |
| US11883099B2 (en) | Noninvasive techniques for identifying choroidal neovascularization in retinal scans | |
| CN114881968A (en) | OCTA image vessel segmentation method, device and medium based on deep convolutional neural network | |
| CN115393239A (en) | Multi-mode fundus image registration and fusion method and system | |
| CN115018863A (en) | Image segmentation method and device based on deep learning | |
| CN116030042B (en) | Diagnostic device, method, equipment and storage medium for doctor's diagnosis | |
| CN112862850A (en) | Digital silhouette angiography image processing method and device | |
| CN113077418A (en) | CT image skeleton segmentation method and device based on convolutional neural network | |
| CN116407080A (en) | Evolution identification and 3D visualization system and method for fundus structure of myopic patient | |
| CN111462067B (en) | Image segmentation method and device | |
| CN114359308A (en) | Aortic dissection method based on edge response and nonlinear loss | |
| Sousa et al. | Learning to segment the lung volume from CT scans based on semi-automatic ground-truth |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |