Lung image processing method, device, equipment and readable storage medium

Abstract

The application discloses a lung image processing method, device and equipment and a readable storage medium. The method comprises the following steps: acquiring a lung image; inputting the lung image into a texture enhancement network to segment the lung image and obtain a segmented image feature with enhanced texture, wherein the texture enhancement network comprises a fine-grained texture enhancement unit. According to the application, the lung image can be subjected to fine-grained texture enhancement based on the texture enhancement network, so that the segmented image feature with enhanced texture is obtained, which is beneficial to subsequent identification of the segmented image feature, and the identification accuracy of lung lesions is improved.

Description

Technical Field

[0001] This invention relates to the field of lung image processing technology, and in particular to a lung image processing method, apparatus, device, and readable storage medium. Background Technology

[0002] CT imaging of the human lungs is considered an important tool for diagnosing and monitoring lung infections. Studies have shown that lesion size and severity can be assessed from chest CT images to evaluate disease progression and subsequent treatment. However, manually identifying these infected areas is highly inefficient, often requiring several hours to complete the identification of a single patient, resulting in low efficiency. Therefore, developing reliable AI-assisted annotation tools to improve identification efficiency is essential. However, for some highly heterogeneous lung lesions with indistinct borders, AI-based identification also performs poorly.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a lung image processing method, apparatus, device, and readable storage medium, which aims to address the poor performance of lung lesion identification based on artificial intelligence.

[0005] To achieve the above objectives, the present invention provides a lung image processing method, the lung image processing method comprising the following steps:

[0006] Obtain lung images;

[0007] The lung image is input into a texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture. The texture enhancement network includes fine-grained texture enhancement units.

[0008] Furthermore, the texture enhancement network further includes an encoder and a decoder, and the step of inputting the lung image into the texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture includes:

[0009] The lung image is input into a fine-grained texture enhancement unit to obtain first image features;

[0010] The segmented image features are determined based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder.

[0011] Furthermore, the fine-grained texture enhancement unit includes a first convolutional layer, a PReLU layer, channels, and an attention mechanism module connected in sequence. The fine-grained texture enhancement unit also includes a second convolutional layer. The step of inputting the lung image into the fine-grained texture enhancement unit to obtain the first image features includes:

[0012] The lung image features are input into a first convolutional layer, a PReLU layer, a channel, and an attention mechanism module connected in sequence to obtain second image features, wherein the first convolutional layer is initialized with a high-pass filter;

[0013] The second image feature is stitched together with the lung image feature to obtain the third image feature;

[0014] The third image feature is input into the second convolutional layer to obtain the first image feature.

[0015] Further, the encoder includes: a first residual unit, a second residual unit, a third residual unit, a fourth residual unit, and a fifth residual unit. The step of determining the segmented image features based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder includes:

[0016] The first image feature is input into the first residual unit to obtain the fourth image feature;

[0017] The fourth image feature is input into the fine-grained texture enhancement unit to obtain the fifth image feature;

[0018] The fifth image feature is input into the second residual unit to obtain the sixth image feature;

[0019] The sixth image feature is input into the third residual unit to obtain the seventh image feature;

[0020] The seventh image feature is input into the fourth residual unit to obtain the eighth image feature;

[0021] The eighth image feature is input into the fifth residual unit to obtain the ninth image feature;

[0022] The segmented image features are determined based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder.

[0023] Further, the first residual unit includes a third convolutional layer and a fourth convolutional layer, and the step of inputting the first image features into the first residual unit to obtain the fourth image features includes:

[0024] The first image features are input into the third convolutional layer to obtain the first output result, wherein the third convolutional layer includes a convolution with a stride of a first preset value, an instance normalization layer, and a PReLU layer;

[0025] The first output result is input into the fourth convolutional layer to obtain the second output result, wherein the fourth convolutional layer includes a convolution with a stride of a second preset value, an instance normalization layer, and a PReLU layer;

[0026] The first image feature is input into a convolution with a stride of a third preset value to obtain a third output result, and the second output result is added to the third output result to obtain a fourth image feature.

[0027] Further, the decoder includes a first upsampling unit, a second upsampling unit, a third upsampling unit, and a fourth upsampling unit. The step of determining the segmented image features based on the fifth image features, the sixth image features, the seventh image features, the eighth image features, the ninth image features, and the decoder includes:

[0028] The ninth image feature is concatenated with the eighth image feature to obtain the tenth image feature, and the tenth image feature is input into the first upsampling unit to obtain the eleventh image feature;

[0029] The eleventh graphic feature is concatenated with the seventh image feature to obtain the twelfth image feature, and the twelfth image feature is input into the second upsampling unit to obtain the thirteenth image feature;

[0030] The thirteenth image feature is concatenated with the sixth image feature to obtain the fourteenth image feature, and the fourteenth image feature is input into the third upsampling unit to obtain the fifteenth image feature;

[0031] The fifteenth image feature is concatenated with the fifth image feature to obtain the sixteenth image feature, and the sixteenth image feature is input into the fourth upsampling unit to obtain the segmented image feature.

[0032] Further, the first upsampling unit includes a fifth convolutional layer and a sixth convolutional layer, and the step of inputting the tenth feature into the first upsampling unit to obtain the eleventh image feature includes:

[0033] The first image features are input into the fifth convolutional layer to obtain the fourth output result. The fifth convolutional layer includes a deconvolution layer with a stride of a fourth preset value, an instance normalization layer, and a PReLU layer.

[0034] The fourth output result is input into the sixth convolutional layer to obtain the fifth output result. The sixth convolutional layer includes a convolutional layer with a stride of the fifth preset value, an instance normalization layer, and a PReLU layer.

[0035] The sixth output result of the deconvolution in the fifth convolutional layer is determined by inputting the first image feature. The sixth output result is then added to the fifth output result to obtain the eleventh image feature.

[0036] In addition, to achieve the above objectives, the present invention also provides a lung image processing device, the lung image processing device comprising: an acquisition module for acquiring lung images;

[0037] The segmentation module is used to input the lung image into the texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture, wherein the texture enhancement network includes fine-grained texture enhancement units.

[0038] In addition, to achieve the above objectives, the present invention also provides a lung image processing device, the lung image processing device comprising: a memory, a processor, and a lung image processing program stored in the memory and executable on the processor, wherein the lung image processing program, when executed by the processor, implements the steps of the aforementioned lung image processing method.

[0039] This invention acquires lung images and then inputs them into a texture enhancement network to segment the lung images, obtaining segmented image features with enhanced texture. The texture enhancement network includes fine-grained texture enhancement units. This allows for fine-grained texture enhancement of the lung images using the texture enhancement network, thereby obtaining segmented image features with enhanced texture, which facilitates subsequent recognition of the segmented image features and improves the accuracy of lung lesion identification. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of the lung image processing device in the hardware operating environment involved in the embodiments of the present invention;

[0041] Figure 2 This is a schematic flowchart of the first embodiment of the lung image processing method of the present invention;

[0042] Figure 3 This is a schematic diagram of the functional modules of an embodiment of the lung image processing device of the present invention;

[0043] Figure 4 This is a schematic diagram of a fine-grained texture enhancement unit according to an embodiment of the lung image processing method of the present invention;

[0044] Figure 5This is a schematic diagram of the high-pass filter initialization convolutional layer in an embodiment of the lung image processing method of the present invention;

[0045] Figure 6 This is a schematic diagram of a texture enhancement network according to an embodiment of the lung image processing method of the present invention.

[0046] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0047] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0048] like Figure 1 As shown, Figure 1 This is a schematic diagram of the structure of a lung image processing device in the hardware operating environment involved in the embodiments of the present invention.

[0049] The lung imaging processing device in this embodiment of the invention can be a PC, or a mobile terminal device with display function such as a smartphone, tablet computer, or portable computer.

[0050] like Figure 1 As shown, the lung imaging processing device may include: a processor 1001, such as a CPU; a network interface 1004; a user interface 1003; a memory 1005; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0051] Optionally, the lung imaging processing device may also include a camera, RF (Radio Frequency) circuitry, sensors, audio circuitry, a WiFi module, and so on. These sensors may include light sensors, motion sensors, and other sensors. Specifically, the light sensors may include ambient light sensors and proximity sensors. The ambient light sensor can adjust the brightness of the display screen according to the ambient light level, while the proximity sensor can turn off the display screen and / or backlight when the lung imaging processing device is moved to the ear. As a type of motion sensor, a gravity accelerometer can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity, and can be used for applications that identify the posture of the lung imaging processing device (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition functions (such as pedometers, tapping), etc. Of course, the lung imaging processing device may also be equipped with other sensors such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, which will not be elaborated here.

[0052] Those skilled in the art will understand that Figure 1 The terminal structure shown does not constitute a limitation on the terminal and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0053] like Figure 1 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a lung image processing program.

[0054] exist Figure 1 In the terminal shown, the network interface 1004 is mainly used to connect to the backend server and communicate data with the backend server; the user interface 1003 is mainly used to connect to the client (user terminal) and communicate data with the client; and the processor 1001 can be used to call the lung image processing program stored in the memory 1005.

[0055] In this embodiment, the lung image processing device includes: a memory 1005, a processor 1001, and a lung image processing program stored in the memory 1005 and executable on the processor 1001. When the processor 1001 calls the lung image processing program stored in the memory 1005, it executes the steps of the lung image processing methods in the following embodiments.

[0056] The present invention also provides a method for processing lung images, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the method of the present invention.

[0057] In this embodiment, the lung image processing method includes the following steps:

[0058] Step S101: Acquire lung images;

[0059] In this embodiment, lung images can first be obtained using CT imaging technology.

[0060] Step S102: Input the lung image into the texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture, wherein the texture enhancement network includes fine-grained texture enhancement units.

[0061] In this embodiment, lung image features are input into a texture enhancement network, and the fine-grained texture enhancement unit in the texture enhancement network enhances the texture of the lung image features to obtain segmented image features with enhanced texture.

[0062] Further, in one embodiment, step S102 includes:

[0063] Step a: Input the lung image into the fine-grained texture enhancement unit to obtain the first image features;

[0064] Step b: Determine the segmented image features based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder.

[0065] In this implementation, fine-grained texture enhancement units can be used to enhance...

[0066] The lung image is input into the fine-grained texture enhancement unit, and the output result of the fine-grained texture enhancement unit is obtained. The output result of the fine-grained texture enhancement unit is the first image feature.

[0067] Then, based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder, segmented image features with enhanced fine-grained texture are obtained.

[0068] Furthermore, in one embodiment, step a further includes:

[0069] Step a1: The lung image features are input into the first convolutional layer, PReLU layer, channel and attention mechanism module connected in sequence to obtain the second image features, wherein the first convolutional layer is initialized with a high-pass filter;

[0070] Step a2: The second image feature is stitched together with the lung image feature to obtain the third image feature;

[0071] Step a3: Input the third image features into the second convolutional layer to obtain the first image features.

[0072] The fine-grained texture enhancement unit includes a first convolutional layer, a PReLU layer, channels, and an attention mechanism module connected in sequence. The fine-grained texture enhancement unit also includes a second convolutional layer, and the first convolutional layer is initialized using a high-pass filter.

[0073] Specifically, such as Figure 4 As shown, the overall structure of the fine-grained texture enhancement unit is illustrated. Figure 5 The diagram illustrates the process of initializing a convolutional layer using a high-pass filter. The weights of the shadow channel are assigned by the generated filter, while the transparent portion is initialized to 0. Taking novel coronavirus infection as an example, the small inter-class variance between infected and normal regions makes fine-grained texture extraction crucial. Since this fine-grained information is often stored in the high-frequency components of the image, a fine-grained texture enhancement unit utilizing a high-pass filter is introduced. Figure 4 The structure of the fine-grained texture enhancement unit is shown. The first block of this unit is a convolutional layer initialized with a high-pass filter, i.e., the first convolutional layer, which can be a 7×7 convolution. This first convolutional layer is used to extract fine-grained texture, and its weights are still trainable, so they can be further adapted to infected features. PReLU helps retain more information from the previous layer. The channel and spatial attention modules provide channel attention, which allows the model to decide for itself which extracted texture features to pay more attention to. The second convolutional layer can be a 1×1 convolution, and the high-pass filter can be calculated using the following formula:

[0074]

[0075] Here, the output of F represents the value of a specific pixel. x and y represent the pixel positions. K represents the filter size, which can be set to 7.

[0076] In this embodiment, lung image features are input into a first convolutional layer, a PReLU layer, a channel, and an attention mechanism module connected in sequence to obtain second image features. The first convolutional layer is initialized with a high-pass filter, and the second image features are concatenated with the lung image features to obtain third image features. The third image features are then input into the second convolutional layer to obtain first image features.

[0077] Furthermore, in one embodiment, step b further includes:

[0078] Step b1: Input the first image feature into the first residual unit to obtain the fourth image feature;

[0079] Step b2: Input the fourth image feature into the fine-grained texture enhancement unit to obtain the fifth image feature;

[0080] Step b3: Input the fifth image feature into the second residual unit to obtain the sixth image feature;

[0081] Step b4: Input the sixth image feature into the third residual unit to obtain the seventh image feature;

[0082] Step b5: Input the seventh image feature into the fourth residual unit to obtain the eighth image feature;

[0083] Step b6: Input the eighth image feature into the fifth residual unit to obtain the ninth image feature;

[0084] Step b7: Determine the segmented image features based on the fifth image features, the sixth image features, the seventh image features, the eighth image features, the ninth image features, and the decoder.

[0085] In this embodiment, the encoder includes: a first residual unit, a second residual unit, a third residual unit, a fourth residual unit, and a fifth residual unit.

[0086] It's important to note that a typical U-Net has 5 layers, with the channel counts usually being 32, 64, 128, 256, and 512 respectively. As the number of layers increases, the spatial resolution decreases, and semantic information accumulates across the channels, leading to an increase in the number of channels. Shallow features in neural networks primarily capture edge, corner, and texture information, which play a crucial role in infection segmentation. Therefore, channel reweighting techniques enhance the model's ability to identify lung lesions by increasing the number of channels in the shallow layers of the neural network. On the other hand, to maintain the same model size while reducing the number of channels in deeper layers, channel reweighting does not increase model parameters, mitigating the risk of overfitting. Furthermore, without reducing the number of model layers, the model still requires high-level semantic features to identify lung lobe regions.

[0087] As an example, the number of channels in the first residual unit can be 248, the number of channels in the second residual unit can be 248, the number of channels in the third residual unit can be 112, the number of channels in the fourth residual unit can be 112, and the number of channels in the fifth residual unit can be 112.

[0088] In this embodiment, for example, Figure 6As shown, the first image feature is input into the first residual unit to obtain the fourth image feature; the fourth image feature is input into the fine-grained texture enhancement unit to obtain the fifth image feature; the fifth image feature is input into the second residual unit to obtain the sixth image feature; the sixth image feature is input into the third residual unit to obtain the seventh image feature; the seventh image feature is input into the fourth residual unit to obtain the eighth image feature; the eighth image feature is input into the fifth residual unit to obtain the ninth image feature; and the segmented image feature is determined based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder.

[0089] Furthermore, in one embodiment, step b1 further includes:

[0090] Step b11: Input the first image features into the third convolutional layer to obtain the first output result, wherein the third convolutional layer includes a convolution with a stride of a first preset value, an instance normalization layer, and a PReLU layer;

[0091] Step b12: Input the first output result into the fourth convolutional layer to obtain the second output result, wherein the fourth convolutional layer includes a convolution with a stride of a second preset value, an instance normalization layer, and a PReLU layer;

[0092] Step b13: Input the first image feature into a convolution with a stride of a third preset value to obtain a third output result, and perform matrix addition operation between the second output result and the third output result to obtain a fourth image feature.

[0093] The first residual unit includes two convolutional layers, such as... Figure 6 As shown, these are the third and fourth convolutional layers, respectively. The third convolutional layer includes a convolutional layer with a stride of a first preset value, an instance normalization layer, and a PReLU layer, where the first preset value can be 2. The fourth convolutional layer includes a convolutional layer with a stride of a second preset value, an instance normalization layer, and a PReLU layer, where the second preset value can be 1. The first residual unit also includes a convolutional layer with a stride of a third preset value, where the third preset value can be 2. It should also be noted that the structure of the first residual unit is similar to that of the second, third, fourth, and fifth residual units, which will not be described in detail here.

[0094] In this embodiment, firstly, the first image feature is input into the third convolutional layer to obtain the first output result. Then, the first output result is input into the fourth convolutional layer to obtain the second output result. Next, the first image feature is input into a convolution with a stride of a third preset value to obtain the third output result. Finally, the second output result and the third output result are subjected to matrix addition to obtain the fourth image feature.

[0095] Furthermore, in one embodiment, step b7 further includes:

[0096] Step b71: The ninth image feature is concatenated with the eighth image feature to obtain the tenth image feature, and the tenth feature is input into the first upsampling unit to obtain the eleventh image feature;

[0097] Step b72: The eleventh graphic feature is concatenated with the seventh image feature to obtain the twelfth image feature, and the twelfth image feature is input into the second upsampling unit to obtain the thirteenth image feature;

[0098] Step b73: The thirteenth image feature is concatenated with the sixth image feature to obtain the fourteenth image feature, and the fourteenth image feature is input into the third upsampling unit to obtain the fifteenth image feature;

[0099] Step b74: The fifteenth image feature is concatenated with the fifth image feature to obtain the sixteenth image feature, and the sixteenth image feature is input into the fourth upsampling unit to obtain the segmented image feature.

[0100] In this embodiment, the decoder includes a first upsampling unit, a second upsampling unit, a third upsampling unit, and a fourth upsampling unit, such as... Figure 6 As shown.

[0101] In this embodiment, firstly, the ninth image feature is concatenated with the eighth image feature to obtain the tenth image feature, and the tenth image feature is input into the first upsampling unit to obtain the eleventh image feature; secondly, the eleventh image feature is concatenated with the seventh image feature to obtain the twelfth image feature, and the twelfth image feature is input into the second upsampling unit to obtain the thirteenth image feature; thirdly, the thirteenth image feature is concatenated with the sixth image feature to obtain the fourteenth image feature, and the fourteenth image feature is input into the third upsampling unit to obtain the fifteenth image feature; fourthly, the fifteenth image feature is concatenated with the fifth image feature to obtain the sixteenth image feature, and the sixteenth image feature is input into the fourth upsampling unit to obtain the segmented image feature. The structures of the first, second, third, and fourth upsampling units are similar. As an example, the number of channels in the first, second, third, and fourth upsampling units can be 112, 248, 248, and 2, respectively.

[0102] Furthermore, in one embodiment, the step of inputting the tenth feature into the first upsampling unit to obtain the eleventh image feature further includes:

[0103] Step c, the first image features are input into the fifth convolutional layer to obtain the fourth output result, wherein the fifth convolutional layer includes a deconvolution layer with a stride of a fourth preset value, an instance normalization layer and a PReLU layer;

[0104] Step d: Input the fourth output result into the sixth convolutional layer to obtain the fifth output result. The sixth convolutional layer includes a convolutional layer with a stride of the fifth preset value, an instance normalization layer, and a PReLU layer.

[0105] Step e: Determine the sixth output result of the deconvolution of the first image feature input into the fifth convolutional layer, and perform matrix addition operation on the sixth output result and the fifth output result to obtain the eleventh image feature.

[0106] like Figure 6 As shown, the first upsampling unit includes a fifth convolutional layer and a sixth convolutional layer. The fifth convolutional layer includes a deconvolution layer with a stride of a fourth preset value, an instance normalization layer, and a PReLU layer. The fourth preset value can be 2, and the deconvolution size can be 3×3. The sixth convolutional layer includes a convolution layer with a stride of a fifth preset value, an instance normalization layer, and a PReLU layer. The fifth preset value can be 1, and the convolution size can be 3×3.

[0107] In this embodiment, the first image feature is input into the fifth convolutional layer to obtain the fourth output result, and the fourth output result is input into the sixth convolutional layer to obtain the fifth output result. Finally, the sixth output result of the deconvolution of the first image feature input into the fifth convolutional layer is determined, and the sixth output result and the fifth output result are subjected to matrix addition to obtain the eleventh image feature. The processing of image features by the second upsampling unit, the third upsampling unit and the fourth upsampling unit is similar to that of the first upsampling unit, and will not be described again here.

[0108] The lung image processing method proposed in this embodiment acquires lung images and then inputs the lung images into a texture enhancement network to segment the lung images and obtain segmented image features with enhanced texture. The texture enhancement network includes fine-grained texture enhancement units, which can perform fine-grained texture enhancement on the lung images according to the texture enhancement network, thereby obtaining segmented image features with enhanced texture, which is beneficial for subsequent recognition of segmented image features and improves the recognition accuracy of lung lesions.

[0109] The present invention also provides a lung image processing device, with reference to Figure 3 The lung imaging processing device includes:

[0110] Acquisition module 10 is used for: acquiring lung images;

[0111] The segmentation module 20 is used to: input the lung image into a texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture, wherein the texture enhancement network includes fine-grained texture enhancement units.

[0112] Furthermore, the segmentation module 20 is also used for:

[0113] The lung image is input into a fine-grained texture enhancement unit to obtain first image features;

[0114] The segmented image features are determined based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder.

[0115] Furthermore, the segmentation module 20 is also used for:

[0116] The lung image features are input into a first convolutional layer, a PReLU layer, a channel, and an attention mechanism module connected in sequence to obtain second image features, wherein the first convolutional layer is initialized with a high-pass filter;

[0117] The second image feature is stitched together with the lung image feature to obtain the third image feature;

[0118] The third image feature is input into the second convolutional layer to obtain the first image feature.

[0119] Furthermore, the segmentation module 20 is also used for:

[0120] The first image feature is input into the first residual unit to obtain the fourth image feature;

[0121] The fourth image feature is input into the fine-grained texture enhancement unit to obtain the fifth image feature;

[0122] The fifth image feature is input into the second residual unit to obtain the sixth image feature;

[0123] The sixth image feature is input into the third residual unit to obtain the seventh image feature;

[0124] The seventh image feature is input into the fourth residual unit to obtain the eighth image feature;

[0125] The eighth image feature is input into the fifth residual unit to obtain the ninth image feature;

[0126] The segmented image features are determined based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder.

[0127] Furthermore, the segmentation module 20 is also used for:

[0128] The first image features are input into the third convolutional layer to obtain the first output result, wherein the third convolutional layer includes a convolution with a stride of a first preset value, an instance normalization layer, and a PReLU layer;

[0129] The first output result is input into the fourth convolutional layer to obtain the second output result, wherein the fourth convolutional layer includes a convolution with a stride of a second preset value, an instance normalization layer, and a PReLU layer;

[0130] The first image feature is input into a convolution with a stride of a third preset value to obtain a third output result, and the second output result is added to the third output result to obtain a fourth image feature.

[0131] Furthermore, the segmentation module 20 is also used for:

[0132] The ninth image feature is concatenated with the eighth image feature to obtain the tenth image feature, and the tenth image feature is input into the first upsampling unit to obtain the eleventh image feature;

[0133] The eleventh graphic feature is concatenated with the seventh image feature to obtain the twelfth image feature, and the twelfth image feature is input into the second upsampling unit to obtain the thirteenth image feature;

[0134] The thirteenth image feature is concatenated with the sixth image feature to obtain the fourteenth image feature, and the fourteenth image feature is input into the third upsampling unit to obtain the fifteenth image feature;

[0135] The fifteenth image feature is concatenated with the fifth image feature to obtain the sixteenth image feature, and the sixteenth image feature is input into the fourth upsampling unit to obtain the segmented image feature.

[0136] Furthermore, the segmentation module 20 is also used for:

[0137] The first image features are input into the fifth convolutional layer to obtain the fourth output result. The fifth convolutional layer includes a deconvolution layer with a stride of a fourth preset value, an instance normalization layer, and a PReLU layer.

[0138] The fourth output result is input into the sixth convolutional layer to obtain the fifth output result. The sixth convolutional layer includes a convolutional layer with a stride of the fifth preset value, an instance normalization layer, and a PReLU layer.

[0139] The sixth output result of the deconvolution in the fifth convolutional layer is determined by inputting the first image feature. The sixth output result is then added to the fifth output result to obtain the eleventh image feature.

[0140] The methods executed by the above-mentioned program units can be referred to in the various embodiments of the lung image processing method of the present invention, and will not be repeated here.

[0141] Furthermore, this invention also proposes a lung image processing device, which includes: a memory, a processor, and a lung image processing program stored in the memory and executable on the processor. When the lung image processing program is executed by the processor, it implements the steps of the lung image processing method described above.

[0142] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a lung image processing program, which, when executed by a processor, implements the steps of the lung image processing method described above.

[0143] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0144] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0145] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0146] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A lung image processing method, characterized in that, The lung image processing method includes the following steps: Obtain lung images; The lung image is input into a texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture. The texture enhancement network includes fine-grained texture enhancement units. The texture enhancement network further includes an encoder and a decoder. The step of inputting the lung image into the texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture includes: The lung image is input into a fine-grained texture enhancement unit to obtain first image features; The segmented image features are determined based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder; The fine-grained texture enhancement unit includes a first convolutional layer, a PReLU layer, channels, and an attention mechanism module connected in sequence. The fine-grained texture enhancement unit also includes a second convolutional layer. The step of inputting the lung image into the fine-grained texture enhancement unit to obtain the first image features includes: Lung image features are input into a first convolutional layer, a PReLU layer, channels, and an attention mechanism module connected in sequence to obtain second image features. The first convolutional layer is initialized with a high-pass filter and uses a 7x7 convolution. The calculation formula for the high-pass filter is as follows: Where F's output represents the value of a specific pixel, x and y represent the pixel's position, and K represents the filter size, which is set to 7; The second image feature is stitched together with the lung image feature to obtain the third image feature; The third image feature is input into the second convolutional layer to obtain the first image feature; The encoder includes a first residual unit, a second residual unit, a third residual unit, a fourth residual unit, and a fifth residual unit. The step of determining the segmented image features based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder includes: The first image feature is input into the first residual unit to obtain the fourth image feature; The fourth image feature is input into the fine-grained texture enhancement unit to obtain the fifth image feature; The fifth image feature is input into the second residual unit to obtain the sixth image feature; The sixth image feature is input into the third residual unit to obtain the seventh image feature; The seventh image feature is input into the fourth residual unit to obtain the eighth image feature; The eighth image feature is input into the fifth residual unit to obtain the ninth image feature; The segmented image features are determined based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder. The first residual unit has 248 channels, the second residual unit has 248 channels, the third residual unit has 112 channels, the fourth residual unit has 112 channels, and the fifth residual unit has 112 channels.

2. The lung image processing method as described in claim 1, characterized in that, The first residual unit includes a third convolutional layer and a fourth convolutional layer. The step of inputting the first image features into the first residual unit to obtain the fourth image features includes: The first image features are input into the third convolutional layer to obtain the first output result, wherein the third convolutional layer includes a convolution with a stride of a first preset value, an instance normalization layer, and a PReLU layer; The first output result is input into the fourth convolutional layer to obtain the second output result, wherein the fourth convolutional layer includes a convolution with a stride of a second preset value, an instance normalization layer, and a PReLU layer; The first image feature is input into a convolution with a stride of a third preset value to obtain a third output result, and the second output result is added to the third output result to obtain a fourth image feature.

3. The lung image processing method as described in claim 1, characterized in that, The decoder includes a first upsampling unit, a second upsampling unit, a third upsampling unit, and a fourth upsampling unit. The step of determining the segmented image features based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder includes: The ninth image feature is concatenated with the eighth image feature to obtain the tenth image feature, and the tenth image feature is input into the first upsampling unit to obtain the eleventh image feature; The eleventh image feature is concatenated with the seventh image feature to obtain the twelfth image feature, and the twelfth image feature is input into the second upsampling unit to obtain the thirteenth image feature; The thirteenth image feature is concatenated with the sixth image feature to obtain the fourteenth image feature, and the fourteenth image feature is input into the third upsampling unit to obtain the fifteenth image feature; The fifteenth image feature is concatenated with the fifth image feature to obtain the sixteenth image feature, and the sixteenth image feature is input into the fourth upsampling unit to obtain the segmented image feature.

4. The lung image processing method as described in claim 3, characterized in that, The first upsampling unit includes a fifth convolutional layer and a sixth convolutional layer. The step of inputting the tenth image feature into the first upsampling unit to obtain the eleventh image feature includes: The tenth image feature is input into the fifth convolutional layer to obtain the fourth output result. The fifth convolutional layer includes a deconvolution layer with a stride of a fourth preset value, an instance normalization layer, and a PReLU layer. The fourth output result is input into the sixth convolutional layer to obtain the fifth output result. The sixth convolutional layer includes a convolutional layer with a stride of the fifth preset value, an instance normalization layer, and a PReLU layer. The sixth output result of the deconvolution in the fifth convolutional layer is determined by inputting the tenth image feature. The sixth output result and the fifth output result are then subjected to matrix addition to obtain the eleventh image feature.

5. A lung image processing device, characterized in that, The lung imaging processing device includes: The acquisition module is used to acquire lung images; A segmentation module is used to input the lung image into a texture enhancement network to segment the lung image and obtain segmented image features with enhanced texture, wherein the texture enhancement network includes fine-grained texture enhancement units; The texture enhancement network further includes an encoder and a decoder. The segmentation module is also used to input the lung image into the fine-grained texture enhancement unit to obtain first image features. The segmented image features are determined based on the first image features, the encoder, the fine-grained texture enhancement unit, and the decoder; The fine-grained texture enhancement unit includes a first convolutional layer, a PReLU layer, channels, and an attention mechanism module connected in sequence. The fine-grained texture enhancement unit also includes a second convolutional layer. The segmentation module is further used to input lung image features into the first convolutional layer, PReLU layer, channels, and attention mechanism module connected in sequence to obtain second image features. The first convolutional layer is initialized with a high-pass filter and uses a 7x7 convolution. The calculation formula for the high-pass filter is: Where F's output represents the value of a specific pixel, x and y represent the pixel's position, and K represents the filter size, which is set to 7; The second image feature is stitched together with the lung image feature to obtain the third image feature; The third image feature is input into the second convolutional layer to obtain the first image feature; The encoder includes a first residual unit, a second residual unit, a third residual unit, a fourth residual unit, and a fifth residual unit. The segmentation module is further configured to input the first image feature into the first residual unit to obtain the fourth image feature. The fourth image feature is input into the fine-grained texture enhancement unit to obtain the fifth image feature; The fifth image feature is input into the second residual unit to obtain the sixth image feature; The sixth image feature is input into the third residual unit to obtain the seventh image feature; The seventh image feature is input into the fourth residual unit to obtain the eighth image feature; The eighth image feature is input into the fifth residual unit to obtain the ninth image feature; The segmented image features are determined based on the fifth, sixth, seventh, eighth, and ninth image features and the decoder. The first residual unit has 248 channels, the second residual unit has 248 channels, the third residual unit has 112 channels, the fourth residual unit has 112 channels, and the fifth residual unit has 112 channels.

6. A lung imaging processing device, characterized in that, The lung imaging processing device includes: a memory, a processor, and a lung imaging processing program stored in the memory and executable on the processor, wherein the lung imaging processing program, when executed by the processor, implements the steps of the lung imaging processing method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a lung image processing program, which, when executed by a processor, implements the steps of the lung image processing method as described in any one of claims 1 to 4.

Images