CN114267443B - Pancreatic tumor fibrosis degree prediction method based on deep learning and related device - Google Patents
Pancreatic tumor fibrosis degree prediction method based on deep learning and related device Download PDFInfo
- Publication number
- CN114267443B CN114267443B CN202111314523.XA CN202111314523A CN114267443B CN 114267443 B CN114267443 B CN 114267443B CN 202111314523 A CN202111314523 A CN 202111314523A CN 114267443 B CN114267443 B CN 114267443B
- Authority
- CN
- China
- Prior art keywords
- pancreatic
- image
- pancreatic tumor
- fibrosis
- module
- 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
Abstract
The application discloses a pancreatic tumor fibrosis degree prediction method based on deep learning and a related device. According to the embodiment, the pancreatic images are learned by adopting the prediction network model based on deep learning, the image characteristics in the pancreatic images can be learned, and the pancreatic tumor regions in the pancreatic images and the fibrosis levels corresponding to the pancreatic tumor regions can be predicted based on the learned image characteristics, so that the accuracy of prediction of the pancreatic cancer fibrosis degree can be improved.
Description
Technical Field
The application relates to the technical field of biomedical engineering, in particular to a pancreatic tumor fibrosis degree prediction method based on deep learning and a related device.
Background
Pancreatic cancer is the most common malignant lesion of the pancreas, with an extremely poor prognosis and an overall survival rate of less than 8% for 5 years. There is often varying degrees of fibrosis within pancreatic tumors. Studies have shown that the higher the degree of fibrosis of a tumor tissue, the more sparse the tumor cells are, the greater the resistance to chemoradiation, the higher the possibility of postoperative recurrence and metastasis, and theoretically, when the degree of fibrosis in the tumor tissue is high, chemoradiation is not easy and the margin should be larger at the time of surgical resection and lymph node clearing should be more active. Therefore, the method has important scientific research significance and clinical application value for predicting the fibrosis degree of the pancreatic tumor.
The currently and generally adopted pancreatic tumor fibrosis degree prediction method comprises biopsy and pancreatic functional examination, wherein the biopsy can only be used for local sampling, the actual operation difficulty is high, and complications of patients easily occur; the interventional functional examination process based on the pancreatic endoscope and the like is painful and the sensitivity of the examination result is low.
Thus, the prior art has yet to be improved and enhanced.
Disclosure of Invention
The technical problem to be solved by the present application is to provide a pancreatic tumor fibrosis degree prediction method based on deep learning and a related device, aiming at the defects of the prior art.
In order to solve the above technical problem, a first aspect of the embodiments of the present application provides a method for predicting a fibrosis degree of a pancreatic tumor based on deep learning, the method including:
acquiring a pancreas image to be predicted, wherein the pancreas image carries a pancreas tumor region;
and controlling the prediction network model to determine a pancreatic tumor region in the pancreatic image and a corresponding fibrosis level of the pancreatic tumor region based on the pancreatic image.
The pancreatic tumor fibrosis degree prediction method based on deep learning is characterized in that the prediction network model comprises a coding module, a prediction module, a decoding module and a segmentation module; the controlling the prediction network model based on the pancreas image, and the determining the pancreas tumor region in the pancreas image and the corresponding fibrosis level of the pancreas tumor region specifically includes:
inputting the pancreas image into the coding module, and controlling the coding module to determine a plurality of first feature maps corresponding to the pancreas image;
controlling the decoding module to determine a second feature map corresponding to the pancreas image based on a plurality of first feature maps;
controlling the segmentation module to determine a pancreatic tumor region in the pancreatic image based on the second feature map and a first feature map with a lowest dimension of the plurality of first features;
controlling the prediction module to determine a level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps that is the highest dimension.
The pancreatic tumor fibrosis degree prediction method based on deep learning is characterized in that the coding module comprises a convolution unit and a plurality of residual error units which are sequentially cascaded; the decoding module comprises a plurality of deconvolution units, wherein a plurality of residual error units and a plurality of deconvolution units are connected in a jumping mode, and each residual error unit is connected with the corresponding deconvolution unit in a jumping mode.
The pancreatic tumor fibrosis degree prediction method based on deep learning is characterized in that the prediction module comprises a feature dimension reduction unit, a full-connection unit and an activation unit; the controlling the prediction module to determine the level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps having a highest dimension comprises:
controlling the feature dimension reduction unit to map the first feature map with the highest dimension to a low-dimensional space to obtain a third feature map;
control the fully-connected unit to determine a feature vector based on the third feature map, and control the activation unit to determine a level of fibrosis of the pancreatic tumor region based on the feature vector.
The pancreatic tumor fibrosis degree prediction method based on deep learning is characterized in that the feature dimensionality reduction unit comprises an adaptive average pooling layer, and the full-connection unit comprises a flattening layer, a Dropout layer and a full-connection layer.
The pancreatic tumor fibrosis degree prediction method based on deep learning includes:
acquiring a candidate pancreas image acquired by a pancreas image acquisition device, and extracting a region of interest in the candidate pancreas image, wherein the region of interest comprises a pancreas tumor region in the pancreas image;
and adjusting the image size of the region of interest to a preset image size, and taking the adjusted region of interest as a pancreas image to be predicted.
The pancreatic tumor fibrosis degree prediction method based on deep learning is characterized in that the pancreatic image is a pancreatic CT image.
A second aspect of the embodiments of the present application provides a pancreatic tumor fibrosis degree prediction apparatus based on deep learning, the prediction apparatus including:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring a pancreatic image to be predicted, and the pancreatic image carries a pancreatic tumor region;
a control module for controlling the prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image.
A third aspect of embodiments herein provides a computer readable storage medium storing one or more programs, which are executable by one or more processors to implement the steps in the method for predicting a degree of fibrosis in a pancreatic tumor based on deep learning as described in any one of the above.
A fourth aspect of the embodiments of the present application provides a terminal device, including: a processor, a memory, and a communication bus; the memory has stored thereon a computer readable program executable by the processor;
the communication bus realizes connection communication between the processor and the memory;
the processor, when executing the computer readable program, implements the steps of the method for predicting fibrosis degree of pancreatic tumor based on deep learning as described in any one of the above.
Has the beneficial effects that: compared with the prior art, the method comprises the steps of obtaining a pancreatic image to be predicted, controlling the prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image, and performing deep learning-based pancreatic tumor fibrosis degree prediction. According to the embodiment, the pancreatic images are learned by adopting the prediction network model based on deep learning, the image characteristics in the pancreatic images can be learned, and the pancreatic tumor regions in the pancreatic images and the fibrosis levels corresponding to the pancreatic tumor regions can be predicted based on the learned image characteristics, so that the accuracy of prediction of the pancreatic cancer fibrosis degree can be improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without any inventive work.
Fig. 1 is a flowchart of a method for predicting a degree of fibrosis in a pancreatic tumor based on deep learning provided by the present application.
Fig. 2 is a model schematic diagram of a prediction network model in the pancreatic tumor fibrosis degree prediction method based on deep learning provided by the present application.
Fig. 3 is a model schematic diagram of a residual error unit in the deep learning-based pancreatic tumor fibrosis degree prediction method provided by the present application.
Fig. 4 is a model schematic diagram of a deconvolution unit in the deep learning-based pancreatic tumor fibrosis degree prediction method provided by the present application.
Fig. 5 is a model schematic diagram of a prediction module in the deep learning-based pancreatic tumor fibrosis degree prediction method provided by the present application.
Fig. 6 is a schematic structural diagram of a pancreatic tumor fibrosis degree prediction device based on deep learning provided in the present application.
Fig. 7 is a schematic structural diagram of a terminal device provided in the present application.
Detailed Description
The present application provides a method for predicting a fibrosis degree of a pancreatic tumor based on deep learning and a related device, and in order to make the purpose, technical solution and effect of the present application clearer and clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.
It will be understood by those within the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
It should be understood that, the sequence numbers and sizes of the steps in this embodiment do not mean the execution sequence, and the execution sequence of each process is determined by its function and inherent logic, and should not constitute any limitation on the implementation process of this embodiment.
The inventor finds that the pancreatic cancer is the most common malignant lesion of the pancreas through research, the prognosis is very poor, and the 5-year overall survival rate is less than 8%. There is often varying degrees of fibrosis within pancreatic tumors. Studies have shown that the higher the degree of fibrosis of a tumor tissue, the more sparse the tumor cells are, the greater the resistance to chemoradiation, the higher the possibility of postoperative recurrence and metastasis, and theoretically, when the degree of fibrosis in the tumor tissue is high, chemoradiation is not easy and the margin should be larger at the time of surgical resection and lymph node clearing should be more active. Therefore, the method has important scientific research significance and clinical application value for predicting the fibrosis degree of the pancreatic tumor.
The currently and generally adopted pancreatic tumor fibrosis degree prediction method comprises biopsy and pancreatic functional examination, wherein the biopsy can only be used for local sampling, the actual operation difficulty is high, and complications of patients easily occur; the interventional functional examination process based on the pancreatic endoscope and the like is painful and the sensitivity of the examination result is low.
In order to solve the above problem, in the embodiment of the present application, a pancreas image to be predicted is acquired, and the prediction network model is controlled to determine a pancreas tumor region in the pancreas image and a fibrosis level corresponding to the pancreas tumor region based on the pancreas image. According to the embodiment, the pancreatic images are learned by adopting the prediction network model based on deep learning, the image characteristics in the pancreatic images can be learned, and the pancreatic tumor regions in the pancreatic images and the fibrosis levels corresponding to the pancreatic tumor regions can be predicted based on the learned image characteristics, so that the accuracy of prediction of the pancreatic cancer fibrosis degree can be improved.
The following further describes the content of the application by describing the embodiments with reference to the attached drawings.
The embodiment provides a pancreatic tumor fibrosis degree prediction method based on deep learning, as shown in fig. 1, the method includes:
and S10, acquiring a pancreas image to be predicted.
Specifically, the pancreas image carries a pancreas tumor region, that is, the pancreas image is an image obtained by image acquisition of a patient with a pancreas tumor, wherein the pancreas image may be a pancreas CT image. For example, the pancreas image to be predicted is a pancreas CT image formed by CT scanning of a pancreas of a patient having a pancreas tumor by a CT apparatus.
In an implementation manner of this embodiment, the acquiring a pancreatic image to be predicted specifically includes:
candidate pancreas images acquired by a pancreas image acquisition apparatus are acquired, and a region of interest in the candidate pancreas images is extracted.
And adjusting the image size of the region of interest to a preset image size, and taking the adjusted region of interest as a pancreas image to be predicted.
Specifically, the candidate pancreatic image may be a CT image, and the image acquisition device may be a CT device by which a CT scan is performed on a patient having a pancreatic tumor to obtain the candidate pancreatic image. The region of interest includes a pancreatic tumor region in the candidate pancreatic image, and the region of interest removes a portion of redundant information in the candidate pancreatic image. The candidate pancreas image is obtained by scanning the whole pancreas, so that the candidate pancreas image carries image information of the whole pancreas, and a pancreas tumor region possibly occupies only a partial region in the candidate pancreas image, so that in order to reduce redundant information of the candidate pancreas image, after the candidate pancreas image is obtained, a region of interest in the candidate pancreas image can be extracted to remove the redundant information in the pancreas image. In addition, after the region of interest is obtained, the region of interest is cut out of the candidate pancreas image, and the size of the region of interest is adjusted to be a preset image size so as to obtain a pancreas image to be predicted, wherein the preset image size is equal to the image size of an input item of the prediction network model, so that the pancreas image to be predicted can be input into the prediction network model as the input item of the prediction network model.
And S20, controlling the prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image.
Specifically, the prediction network model is a neural network model based on deep learning, the prediction network model is a trained network model, an input item of the prediction network model is a pancreas image, and an output item of the prediction network model is a pancreas tumor region in the pancreas image and a fibrosis level corresponding to the pancreas tumor region. It is understood that the prediction network model is a multitask network model, through which image features in the pancreatic image can be learned, and a pancreatic tumor region in the pancreatic image and a corresponding fibrosis level of the pancreatic tumor region can be predicted based on the learned image features. Wherein, the fibrosis grade is used for reflecting the fibrosis degree of the pancreatic tumor, the higher the fibrosis grade is, the higher the fibrosis degree of the pancreatic tumor is, and conversely, the lower the fibrosis grade is, the lower the fibrosis degree of the pancreatic tumor is. In a particular implementation, the fibrosis level includes a first level and a second level, the first level corresponding to a higher degree of pancreatic tumor fibrosis than the second level.
In an implementation manner of this embodiment, as shown in fig. 2, the prediction network model includes an encoding module, a decoding module, a prediction module, and a segmentation module, the encoding module is connected to the prediction module, the encoding module is connected to the decoding module and provides a plurality of first feature maps for the decoding module, and the decoding module is connected to the segmentation module. Correspondingly, the controlling the prediction network model to determine, based on the pancreatic image, a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region specifically includes:
inputting the pancreas image into the coding module, and controlling the coding module to determine a plurality of first feature maps corresponding to the pancreas image;
controlling the decoding module to determine a second feature map corresponding to the pancreas image based on a plurality of first feature maps;
controlling the segmentation module to determine a pancreatic tumor region in the pancreatic image based on the second feature map and a first feature map with a lowest dimension of the plurality of first features;
controlling the prediction module to determine a level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps that is the highest dimension.
Specifically, the input item of the encoding module is a pancreas image, the output item is a plurality of first feature maps, and feature dimensions of the first feature maps in the first feature maps are different, wherein the encoding module is configured to down-sample the pancreas image to obtain a plurality of first feature maps with different feature dimensions. The input of the decoding module is a plurality of first feature maps, and a second feature map is determined based on the plurality of first feature maps, wherein the decoding module is used for performing up-sampling to learn the features of the pancreatic tumor region in the pancreatic image, so that the segmentation module determines the pancreatic tumor region in the pancreatic image based on the second feature map and the first feature map with the lowest dimension in the plurality of first features. Therefore, in the training process of the prediction network model, the prediction pancreatic tumor region and the pancreatic tumor label can be adopted to carry out segmentation training on the prediction network model, so that the model performance of the prediction network model obtained by training can be improved, and the accuracy of the fibrosis grade predicted by the prediction network model can be improved.
In an implementation manner of this embodiment, as shown in fig. 2, the encoding module includes a convolution unit and a plurality of residual error units that are cascaded in sequence; the decoding module comprises a plurality of deconvolution units, wherein a plurality of residual error units and a plurality of deconvolution units are connected in a jumping mode, and each residual error unit is connected with the corresponding deconvolution unit in a jumping mode. In the embodiment, the image characteristics of high resolution and low resolution can be fused by adopting a jumping connection mode for the residual error units and the deconvolution units, so that the pancreatic tumor region can be finely divided, the prediction network model can be ensured to extract more image characteristics, and the model performance of the prediction network model can be improved. In addition, the convolution unit is connected with the segmentation module, and the output item of the convolution unit is input into the segmentation module, so that the segmentation module receives the second feature map input by the decoding unit and the first feature map input by the convolution unit, and determines the pancreatic tumor region based on a fusion feature map obtained by fusing the second feature map and the first feature map.
In a typical implementation manner, the number of residual error units is 4, and as shown in fig. 3, each residual error unit is a pooling layer output and a residual error block, the residual error block includes a first branch, a second branch, and an activation layer relu, the input items of the first branch and the second branch are output items of the pooling layer, and the output item of the first branch and the output item of the second branch are merged and then input into the activation layer relu. The first branch comprises a convolution layer conv, and the second branch comprises a convolution layer conv, a batch normalization layer BN, an activation layer relu, a convolution layer conv and a batch normalization layer BN which are sequentially cascaded. In addition, the number of deconvolution units is 4, and as shown in fig. 4, each deconvolution unit includes an upsampling layer and a residual block, where a model structure of the residual block in the deconvolution unit is the same as a model structure of the residual block in the residual unit, and a model structure of the convolution unit is the same as a model structure of the residual block in the residual unit, and specifically, reference may be made to the description of the residual block in the residual unit, which is not repeated here.
In one implementation of this embodiment, the first feature map with the highest dimension is an output item of a residual unit located at the last among the residual units, that is, the residual unit located at the last among the dry residual units is connected to the prediction module, and the first feature map output by the prediction module is input to the prediction module, so that the prediction module determines the fibrosis level of the tumor region in the pancreatic image based on the first feature map. The first feature map with the highest dimension (namely, the first feature map output by the last residual error unit) is used as an input item of the prediction module in the implementation manner, and the difference of the tumor region can be more effectively displayed by the high-dimensional image features, so that the prediction accuracy of the prediction module can be improved. In addition, the output item of each residual unit in the residual units is the input item of the corresponding deconvolution unit, and is also the input item of the network layer positioned behind the residual unit. For example, the output items of the frontmost residual unit in the several residual units are the input items of the residual unit positioned behind and the input item of the deconvolution unit jump-connected with the residual unit.
In an implementation manner of this embodiment, the prediction module includes a feature dimension reduction unit, a full connection unit, and an activation unit; the controlling the prediction module to determine the level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps having a highest dimension comprises:
controlling the feature dimension reduction unit to map the first feature map with the highest dimension to a low-dimensional space to obtain a third feature map;
control the fully-connected unit to determine a feature vector based on the third feature map, and control the activation unit to determine a level of fibrosis of the pancreatic tumor region based on the feature vector.
Specifically, the feature dimension reduction unit is configured to map the feature map from a high-dimensional space to a low-dimensional space, that is, the dimension space where the third feature map is located is lower than the dimension space where the first feature map is located, so that on one hand, redundant features in the first feature map are removed, and the prediction performance of the prediction network model is improved, and on the other hand, the time complexity and the space complexity of the feature map can be reduced, so that model parameters of the prediction network model can be reduced, and the prediction speed of the prediction network model can be improved. The feature dimension reduction unit can be configured with a linear or nonlinear mapping function, and the first feature map with the highest dimension is mapped to a low-dimension space through the linear or nonlinear mapping function to obtain a third feature map. In a specific implementation manner, as shown in fig. 5, the feature dimension reduction unit may include an adaptive average Pooling layer Pooling, and the dimension of the first feature map with the highest dimension may be reduced by the adaptive average Pooling layer to reduce the number of features, so as to reduce the model parameters of the predicted network model, and on the other hand, the feature invariance in the first feature map with the highest dimension may be maintained by the adaptive average Pooling function, so as to ensure the prediction accuracy of the predicted network model. For example, the first feature map with the highest dimension is the feature map of 2 × 1024 shown in fig. 2, and the feature map can be reduced to the feature map of 2 × 1024 by the feature dimension reduction unit. Furthermore, as shown in fig. 5, the activation unit includes a simoid activation layer; the full-connection unit comprises a flattening layer View, a Dropout layer Dropout and a full-connection layer Linear, wherein the flattening layer is connected with the Dropout layer, the Dropout layer is connected with the full-connection layer, and the Dropout layer is used for randomly abandoning a part of parameters to enhance the generalization of the model, so that the model performance of the prediction network model can be improved.
In an implementation manner of this embodiment, in a training process of the prediction network model, the prediction network model may be trained based on the predicted fibrosis level determined by the prediction module and the predicted pancreatic tumor region determined by the segmentation model, so that model performance of the prediction network model obtained by training may be improved, where a loss term corresponding to the predicted fibrosis level may adopt a cross entropy loss function, and a loss term corresponding to the predicted pancreatic tumor region may adopt a Dice loss function.
In summary, the present embodiment provides a method for predicting a fibrosis degree of a pancreatic tumor based on deep learning, the method includes obtaining a pancreatic image to be predicted, and controlling the prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image. According to the embodiment, the pancreatic images are learned by adopting the prediction network model based on deep learning, the image characteristics in the pancreatic images can be learned, and the pancreatic tumor regions in the pancreatic images and the fibrosis levels corresponding to the pancreatic tumor regions can be predicted based on the learned image characteristics, so that the accuracy of prediction of the pancreatic cancer fibrosis degree can be improved.
Based on the above method for predicting fibrosis degree of pancreatic tumor based on deep learning, the present embodiment provides a device for predicting fibrosis degree of pancreatic tumor based on deep learning, as shown in fig. 6, the device comprising:
an obtaining module 100, configured to obtain a pancreatic image to be predicted, where the pancreatic image carries a pancreatic tumor region;
a control module 200, configured to control the prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image.
Based on the above-described deep learning-based pancreatic tumor fibrosis degree prediction method, the present embodiment provides a computer-readable storage medium storing one or more programs, which are executable by one or more processors, to implement the steps in the deep learning-based pancreatic tumor fibrosis degree prediction method according to the above-described embodiment.
Based on the above method for predicting degree of fibrosis of pancreatic tumor based on deep learning, the present application further provides a terminal device, as shown in fig. 7, including at least one processor (processor) 20; a display screen 21; and a memory (memory) 22, and may further include a communication Interface (Communications Interface) 23 and a bus 24. The processor 20, the display 21, the memory 22 and the communication interface 23 can communicate with each other through the bus 24. The display screen 21 is configured to display a user guidance interface preset in the initial setting mode. The communication interface 23 may transmit information. The processor 20 may call logic instructions in the memory 22 to perform the methods in the embodiments described above.
Furthermore, the logic instructions in the memory 22 may be implemented in software functional units and stored in a computer readable storage medium when sold or used as a stand-alone product.
The memory 22, which is a computer-readable storage medium, may be configured to store a software program, a computer-executable program, such as program instructions or modules corresponding to the methods in the embodiments of the present disclosure. The processor 20 executes the functional application and data processing, i.e. implements the method in the above-described embodiments, by executing the software program, instructions or modules stored in the memory 22.
The memory 22 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal device, and the like. Further, the memory 22 may include a high speed random access memory and may also include a non-volatile memory. For example, a variety of media that can store program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, may also be transient storage media.
In addition, the specific processes loaded and executed by the storage medium and the instruction processors in the terminal device are described in detail in the method, and are not stated herein.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (9)
1. A pancreatic tumor fibrosis degree prediction method based on deep learning is characterized by comprising the following steps:
acquiring a pancreas image to be predicted, wherein the pancreas image carries a pancreas tumor region;
controlling a predictive network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image;
the prediction network model comprises a coding module, a prediction module, a decoding module and a segmentation module; the determining, by the control prediction network model, a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image specifically includes:
inputting the pancreas image into the coding module, and controlling the coding module to determine a plurality of first feature maps corresponding to the pancreas image;
controlling the decoding module to determine a second feature map corresponding to the pancreas image based on a plurality of first feature maps;
controlling the segmentation module to determine a pancreatic tumor region in the pancreatic image based on the second feature map and a first feature map with a lowest dimension of the plurality of first features;
controlling the prediction module to determine a level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps that is the highest dimension.
2. The deep learning-based pancreatic tumor fibrosis degree prediction method according to claim 1, wherein the coding module comprises a convolution unit and several residual units which are cascaded in sequence; the decoding module comprises a plurality of deconvolution units, wherein a plurality of residual error units and a plurality of deconvolution units are connected in a jumping mode, and each residual error unit is connected with the corresponding deconvolution unit in a jumping mode.
3. The method for predicting fibrosis degree of pancreatic tumor based on deep learning of claim 1, wherein said prediction module comprises a feature dimension reduction unit, a full connection unit and an activation unit; the controlling the prediction module to determine the fibrosis level of the pancreatic tumor region based on a first feature map of the plurality of first feature maps having a highest dimension specifically comprises:
controlling the feature dimension reduction unit to map the first feature map with the highest dimension to a low-dimensional space to obtain a third feature map;
control the fully-connected unit to determine a feature vector based on the third feature map, and control the activation unit to determine a level of fibrosis of the pancreatic tumor region based on the feature vector.
4. The deep learning-based pancreatic tumor fibrosis prediction method of claim 3, wherein the feature dimension reduction unit comprises an adaptive mean pooling layer, and the fully-connected unit comprises a flattening layer, a Dropout layer, and a fully-connected layer.
5. The method for predicting fibrosis degree of pancreatic tumor based on deep learning of claim 1, wherein the obtaining of the pancreatic image to be predicted specifically comprises:
acquiring a candidate pancreas image acquired by a pancreas image acquisition device, and extracting a region of interest in the candidate pancreas image, wherein the region of interest comprises a pancreatic tumor region in the pancreas image;
and adjusting the image size of the region of interest to a preset image size, and taking the adjusted region of interest as a pancreas image to be predicted.
6. The method for predicting fibrosis degree of pancreatic tumor based on deep learning according to any one of claims 1 to 5, wherein said pancreatic image is a pancreatic CT image.
7. A pancreatic tumor fibrosis degree prediction apparatus based on deep learning, the prediction apparatus comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring a pancreatic image to be predicted, and the pancreatic image carries a pancreatic tumor region;
a control module for controlling a prediction network model to determine a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image;
the prediction network model comprises a coding module, a prediction module, a decoding module and a segmentation module; the determining, by the control prediction network model, a pancreatic tumor region in the pancreatic image and a fibrosis level corresponding to the pancreatic tumor region based on the pancreatic image specifically includes:
inputting the pancreas image into the coding module, and controlling the coding module to determine a plurality of first feature maps corresponding to the pancreas image;
controlling the decoding module to determine a second feature map corresponding to the pancreas image based on a plurality of first feature maps;
controlling the segmentation module to determine a pancreatic tumor region in the pancreatic image based on the second feature map and a first feature map with a lowest dimension of the plurality of first features;
controlling the prediction module to determine a level of fibrosis for the pancreatic tumor region based on a first feature map of the plurality of first feature maps that is the highest dimension.
8. A computer readable storage medium storing one or more programs, the one or more programs being executable by one or more processors to perform the steps of the method for predicting the degree of fibrosis in a pancreatic tumor based on deep learning according to any one of claims 1 to 6.
9. A terminal device, comprising: a processor, a memory, and a communication bus; the memory has stored thereon a computer readable program executable by the processor;
the communication bus realizes connection communication between the processor and the memory;
the processor, when executing the computer readable program, implements the steps of the deep learning based pancreatic tumor fibrosis degree prediction method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111314523.XA CN114267443B (en) | 2021-11-08 | 2021-11-08 | Pancreatic tumor fibrosis degree prediction method based on deep learning and related device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111314523.XA CN114267443B (en) | 2021-11-08 | 2021-11-08 | Pancreatic tumor fibrosis degree prediction method based on deep learning and related device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114267443A CN114267443A (en) | 2022-04-01 |
| CN114267443B true CN114267443B (en) | 2022-10-04 |
Family
ID=80824897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111314523.XA Active CN114267443B (en) | 2021-11-08 | 2021-11-08 | Pancreatic tumor fibrosis degree prediction method based on deep learning and related device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114267443B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109242844A (en) * | 2018-09-04 | 2019-01-18 | 青岛大学附属医院 | Pancreatic tumour automatic recognition system based on deep learning, computer equipment, storage medium |
| CN109598728A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技（深圳）有限公司 | Image partition method, device, diagnostic system and storage medium |
| TWI681406B (en) * | 2018-12-20 | 2020-01-01 | 中國醫藥大學附設醫院 | Deep learning of tumor image-aided prediction of prognosis of patients with uterine cervical cancer system, method and computer program product thereof |
| CN111310764A (en) * | 2020-01-20 | 2020-06-19 | 上海商汤智能科技有限公司 | Network training method and device, image processing method and device, electronic equipment and storage medium |
| CN112233112A (en) * | 2020-12-07 | 2021-01-15 | 深圳大学 | Prediction method of CK19 of hepatocellular carcinoma based on deep learning analysis |
| CN112562855A (en) * | 2020-12-18 | 2021-03-26 | 深圳大学 | Hepatocellular carcinoma postoperative early recurrence risk prediction method |
| CN113299385A (en) * | 2021-04-30 | 2021-08-24 | 北京深睿博联科技有限责任公司 | Pancreatic cystic lesion clinical decision method and system based on deep learning |
| KR102321487B1 (en) * | 2021-04-16 | 2021-11-04 | 주식회사 휴런 | artificial intelligence-based diagnostic device for shapes and characteristics of detected tumors |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20190105461A (en) * | 2018-03-05 | 2019-09-17 | 주식회사 인공지능연구원 | Computer assisted diagnosis system and method for providing evidence of malignancy and malignancy inference of tumor |
| CN112784849B (en) * | 2021-02-05 | 2023-10-10 | 苏州大学 | Glandular segmentation method based on multi-scale attention selection |
-
2021
- 2021-11-08 CN CN202111314523.XA patent/CN114267443B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109242844A (en) * | 2018-09-04 | 2019-01-18 | 青岛大学附属医院 | Pancreatic tumour automatic recognition system based on deep learning, computer equipment, storage medium |
| CN109598728A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技（深圳）有限公司 | Image partition method, device, diagnostic system and storage medium |
| TWI681406B (en) * | 2018-12-20 | 2020-01-01 | 中國醫藥大學附設醫院 | Deep learning of tumor image-aided prediction of prognosis of patients with uterine cervical cancer system, method and computer program product thereof |
| CN111310764A (en) * | 2020-01-20 | 2020-06-19 | 上海商汤智能科技有限公司 | Network training method and device, image processing method and device, electronic equipment and storage medium |
| CN112233112A (en) * | 2020-12-07 | 2021-01-15 | 深圳大学 | Prediction method of CK19 of hepatocellular carcinoma based on deep learning analysis |
| CN112562855A (en) * | 2020-12-18 | 2021-03-26 | 深圳大学 | Hepatocellular carcinoma postoperative early recurrence risk prediction method |
| KR102321487B1 (en) * | 2021-04-16 | 2021-11-04 | 주식회사 휴런 | artificial intelligence-based diagnostic device for shapes and characteristics of detected tumors |
| CN113299385A (en) * | 2021-04-30 | 2021-08-24 | 北京深睿博联科技有限责任公司 | Pancreatic cystic lesion clinical decision method and system based on deep learning |
Non-Patent Citations (3)
| Title |
|---|
| Accurate and Feasible Deep Learning Based Semi-Automatic Segmentation in CT for Radiomics Analysis in Pancreatic Neuroendocrine Neoplasms;Bingsheng Huang 等;《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS》;20210402;第25卷(第9期);第3498-3506页 * |
| 基于深度卷积神经网络的胰腺器官CT影像分割算法研究;杨争争;《中国优秀博硕士学位论文全文数据库(硕士) 医药卫生科技辑》;20210215(第02期);第E072-1381页 * |
| 基于深度学习的胰腺肿瘤计算机辅助检测与病理分级研究;肖焕辉;《中国优秀博硕士学位论文全文数据库(硕士) 医药卫生科技辑》;20211015(第10期);第E072-33页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114267443A (en) | 2022-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166130B (en) | Image processing method and image processing device | |
| CN107492099B (en) | Medical image analysis method, medical image analysis system, and storage medium | |
| CN110189323B (en) | Breast ultrasound image lesion segmentation method based on semi-supervised learning | |
| CN109410216B (en) | Ischemic stroke image region segmentation method and device | |
| US11816870B2 (en) | Image processing method and device, neural network and training method thereof, storage medium | |
| CN112102321A (en) | Focal image segmentation method and system based on deep convolutional neural network | |
| US20210158524A1 (en) | Deep learning-based multi-site, multi-primitive segmentation for nephropathology using renal biopsy whole slide images | |
| JP2023520846A (en) | Image processing method, image processing apparatus, computer program and computer equipment based on artificial intelligence | |
| CN111145147B (en) | Multi-mode medical image segmentation method and terminal equipment | |
| CN114171197B (en) | Breast cancer HER2 state prediction method and related equipment | |
| CN112614144A (en) | Image segmentation method, device, equipment and storage medium | |
| CN115375711A (en) | Image segmentation method of global context attention network based on multi-scale fusion | |
| CN111968137A (en) | Head CT image segmentation method and device, electronic device and storage medium | |
| WO2023014789A1 (en) | System and method for pathology image analysis using a trained neural network and active learning framework | |
| CN112396605B (en) | Network training method and device, image recognition method and electronic equipment | |
| CN113205488B (en) | Blood flow characteristic prediction method, device, electronic equipment and storage medium | |
| CN114267443B (en) | Pancreatic tumor fibrosis degree prediction method based on deep learning and related device | |
| CN111915626B (en) | Automatic segmentation method, device and storage medium for heart ultrasonic image ventricular region | |
| CN115359066B (en) | Focus detection method and device for endoscope, electronic device and storage medium | |
| CN115908363B (en) | Tumor cell statistics method, device, equipment and storage medium | |
| Aldemir et al. | Chain code strategy for lossless storage and transfer of segmented binary medical data | |
| CN115965785A (en) | Image segmentation method, device, equipment, program product and medium | |
| CN115439470A (en) | Polyp image segmentation method, computer-readable storage medium, and computer device | |
| Ni et al. | SFA-Net: Scale and Feature Aggregate Network for Retinal Vessel Segmentation | |
| Singla et al. | Speckle and shadows: ultrasound-specific physics-based data augmentation applied to kidney segmentation |
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 |