Lung cancer brain metastasis risk prediction method and system

Abstract

The invention provides a lung cancer brain metastasis risk prediction method and system, and the method comprises the steps: obtaining the brain MRI image data of a user, and carrying out the data preprocessing; whole-brain feature extraction and partition feature extraction based on an AAL template are carried out on the preprocessed data, and whole-brain features and partition features are combined to obtain a feature matrix; according to the preprocessed brain MRI image data of the user suffering from lung cancer brain metastasis, training an auto-encoder to obtain a trained auto-encoder; inputting the feature matrix into the trained auto-encoder, and calculating a reconstruction error; according to the reconstruction error, an isolated forest method is adopted to further screen abnormal areas, and an abnormal score is generated for each sample; according to the reconstruction error and the abnormal score, the abnormal score of the whole brain feature and the abnormal score of the partition feature are calculated, the high-risk region of the brain can be accurately marked by combining the abnormal detection of the whole brain feature and the partition feature, and a basis is provided for personalized PCI treatment.

Description

Technical Field

[0001] The present invention belongs to the field of medical technology, and in particular relates to a method and system for predicting the risk of brain metastasis from lung cancer. Background Art

[0002] Prophylactic cranial irradiation (PCI) can significantly reduce the incidence of brain metastases, but it may cause side effects such as neurocognitive impairment. Currently, PCI is only applicable to some high-risk patients, and due to the lack of individualization of the method, there is a risk of over-treatment or under-treatment.

[0003] Most existing prediction methods rely on clinical pathological features and lung CT imaging data, which makes it difficult to accurately predict the area of ​​brain metastasis and cannot provide a basis for individualized PCI treatment. In the early stages of tumor brain metastasis, changes in the brain microenvironment (such as changes caused by soluble factors or exosomes secreted by the tumor) may have occurred. MRI is an ideal tool for evaluating the brain microenvironment due to its high soft tissue resolution and multi-parameter imaging. Summary of the invention

[0004] Based on this, an embodiment of the present invention provides a method and system for predicting the risk of brain metastasis of lung cancer, aiming to accurately identify high-risk individuals and metastatic areas of brain metastasis of lung cancer, and provide a strong basis for personalized PCI.

[0005] A first aspect of an embodiment of the present invention provides a method for predicting the risk of brain metastasis from lung cancer, the method comprising: Acquire the user's brain MRI image data and perform data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; The preprocessed data are subjected to whole-brain feature extraction and AAL template-based partition feature extraction, and the whole-brain features and partition features are merged to obtain a feature matrix; According to the pre-processed brain MRI image data of the user with lung cancer brain metastasis, the autoencoder is trained to obtain the trained autoencoder to learn the low-dimensional representation of normal brain features; Inputting the feature matrix into the trained autoencoder to calculate the reconstruction error; Based on the reconstruction error, the isolation forest method is used to further screen abnormal areas and generate an abnormal score for each sample; According to the reconstruction error and the abnormality score, the abnormality scores of the whole brain features and the partition features are calculated, and finally the lung cancer brain metastasis identification is completed through the abnormality scores of the whole brain features and the partition features.

[0006] Furthermore, the data preprocessing step includes: The brain MRI image data were subjected to intensity normalization and artifact removal processing to eliminate differences in scanners and imaging conditions; Use a registration algorithm to align the intensity-normalized and artifact-removed brain MRI image data to the MNI template to ensure that all data are in the same anatomical frame; Based on the data aligned to the MNI template, a deep learning segmentation tool was used to segment the whole brain area.

[0007] Furthermore, in the step of performing whole-brain feature extraction and AAL template-based partition feature extraction on the preprocessed data, whole-brain feature extraction includes extraction of shape features, texture features and signal intensity, and partition feature extraction includes extraction of texture features and signal intensity, wherein shape features include volume, surface area and shape descriptors of the whole brain, texture features include features of gray-level co-occurrence matrix and gray-level run length matrix, and signal intensity includes mean, standard deviation and rate of change.

[0008] Furthermore, the autoencoder includes an encoder and a decoder, wherein the encoder is used to compress high-dimensional input features into a low-dimensional latent space representation, and the decoder restores the low-dimensional representation to the original feature space.

[0009] Furthermore, in the step of further screening abnormal areas using the isolation forest method according to the reconstruction error and generating an abnormal score for each sample, the expression of the abnormal score is:

[0010] in, For sample The average path length in the isolation tree, is a normalization constant used to ensure that the result is within within the range.

[0011] Furthermore, in the step of calculating the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score, the abnormality score expression of the whole brain features is:

[0012] The anomaly score expression of the partition feature is:

[0013] in, and is the weight parameter for the whole-brain feature score, For the The reconstruction error of each brain region is is the reconstruction error of the whole brain, For the Isolation Forest abnormality scores in 3 brain regions The Isolation Forest Abnormality Score for the whole brain, Abnormality scores for whole-brain features, Score the anomaly of the partition feature.

[0014] Further, the step of calculating the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score comprises: The abnormal scores of the partitioned features were mapped to the standard brain template to generate a heat map of high risk of brain metastasis of lung cancer.

[0015] A second aspect of an embodiment of the present invention provides a lung cancer brain metastasis risk prediction system, which is used to implement the lung cancer brain metastasis risk prediction method provided in the first aspect, and the system includes: A preprocessing module is used to obtain the user's brain MRI image data and perform data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; The feature extraction module is used to extract whole-brain features and partition features based on the AAL template from the preprocessed data, and merge the whole-brain features and partition features to obtain a feature matrix; A training module is used to train an autoencoder based on the pre-processed brain MRI image data of a user with lung cancer brain metastasis, to obtain a trained autoencoder to learn a low-dimensional representation of normal brain features; A first calculation module, used for inputting the feature matrix into the trained autoencoder to calculate the reconstruction error; A screening module, used to further screen abnormal areas using an isolation forest method according to the reconstruction error, and generate an abnormal score for each sample; The second calculation module is used to calculate the abnormal scores of the whole brain features and the partition features according to the reconstruction error and the abnormal scores, and finally complete the identification of brain metastasis of lung cancer through the abnormal scores of the whole brain features and the partition features.

[0016] A third aspect of an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the risk of brain metastasis of lung cancer provided in the first aspect.

[0017] A fourth aspect of an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the program, the method for predicting the risk of brain metastasis of lung cancer provided in the first aspect is implemented.

[0018] A method and system for predicting the risk of brain metastasis of lung cancer provided in an embodiment of the present invention obtains the user's brain MRI image data and performs data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images, and functional MRI; whole-brain feature extraction and partition feature extraction based on the AAL template are performed on the preprocessed data, and the whole-brain feature and the partition feature are merged to obtain a feature matrix; according to the preprocessed brain MRI image data of the user who has lung cancer brain metastasis, an autoencoder is trained to obtain a trained autoencoder to learn the low-dimensional representation of normal brain features; the feature matrix is ​​input into the trained autoencoder to calculate the reconstruction error; according to the reconstruction error, the isolation forest method is used to further screen the abnormal area, and an abnormal score is generated for each sample; according to the reconstruction error and the abnormal score, the abnormal score of the whole-brain feature and the partition feature is calculated, and by combining the abnormal detection of the whole-brain feature and the partition feature, the high-risk area of ​​the brain can be accurately marked, providing a basis for personalized PCI treatment. Compared with the prior art, the embodiment of the present invention has at least the following advantages: 1. Accurately locate high-risk areas for brain metastasis: By combining abnormal detection of whole-brain and regional features, high-risk areas of the brain can be accurately marked, providing a basis for personalized PCI treatment. Compared with the method of only predicting high-risk individuals, it has stronger clinical guidance significance; 2. Higher model robustness: The unsupervised learning model is built based on non-transferred patient data, avoiding the reliance on transferred labeled data. It has better adaptability and generalization ability and is suitable for the data conditions of different medical institutions. 3. Reduce overtreatment and side effects: Accurately screen high-risk individuals and areas for brain metastasis, effectively avoid unnecessary PCI treatment for low-risk patients, reduce side effects such as cognitive impairment, and reduce the possibility of missed diagnosis for high-risk patients; 4. Strong scalability: This technical solution is not limited to the prediction of brain metastasis of lung cancer, but is also applicable to the risk assessment of brain metastasis of other tumors (such as breast cancer, melanoma, etc.), as well as the early detection and risk assessment of other brain diseases (such as stroke, Alzheimer's disease, etc.); 5. Intelligence and visualization: Provide intuitive brain heat maps and risk scores to help doctors quickly understand prediction results, improve clinical decision-making efficiency, and reduce reliance on the experience of imaging experts. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flowchart of a method for predicting the risk of brain metastasis of lung cancer provided in Example 1 of the present invention; Figure 2 This is a structural block diagram of a lung cancer brain metastasis risk prediction system provided by Embodiment 3 of the present invention; Figure 3This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0020] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0021] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0023] Embodiment 1 See also Figure 1 , Figure 1 A flowchart of an implementation of a method for predicting the risk of brain metastasis of lung cancer provided in Embodiment 1 of the present invention is shown. The method for predicting the risk of brain metastasis of lung cancer specifically includes steps S01 to S06.

[0024] Step S01, obtaining the user's brain MRI image data and performing data preprocessing.

[0025] Among them, brain MRI imaging data include T1-weighted images, T2-weighted images, enhanced scanning images (such as T1-Gd) and functional MRI (such as diffusion tensor imaging DTI and blood oxygen level-dependent imaging BOLD) to ensure comprehensive coverage of brain microenvironment information.

[0026] Specifically, the data preprocessing step includes: The brain MRI image data was subjected to intensity normalization and artifact removal processing to eliminate the differences in scanners and imaging conditions. The intensity normalization processing was to eliminate the intensity deviation between different images and unify the data range, which can be expressed as:

[0027] in, is the original image pixel value, and are the mean and standard deviation of the image respectively. Gaussian filter is used for artifact removal and noise reduction, which can be expressed as:

[0028] in, is the Gaussian kernel function, is the filter window size; Use a registration algorithm (such as ANTs or SPM) to align the intensity normalized and artifact-removed brain MRI image data to the MNI template to ensure that all data are in the same anatomical frame, which can be expressed as:

[0029] in, is the coordinate position of the transformed brain MRI image data in the MNI template coordinate system, is the transformation function of the registration process, is the coordinate position of the input brain MRI image data in its own original coordinate system, is the coordinate position of the MNI template in the standard MNI space.

[0030] Based on the data aligned to the MNI template, a deep learning segmentation tool is used to segment the whole brain area. The deep learning segmentation tool can use a deep learning segmentation model (such as UNet) or a toolkit (such as FSL).

[0031] Step S02, performing whole-brain feature extraction and AAL template-based partition feature extraction on the preprocessed data, and merging the whole-brain features and the partition features to obtain a feature matrix.

[0032] Specifically, Python language and Pyradiomics toolkit are used to extract whole-brain features and partitioned features based on AAL template. Specifically, in order to achieve whole-brain feature extraction, it is necessary to create a whole-brain mask, and then use FeatureExtractor in Pyradiomics to extract features. It can be understood that Pyradiomics is a Python library for extracting radiomic features from medical images (such as CT, MRI, etc.). FeatureExtractor is the core class in Pyradiomics, which can extract multiple features from medical images and corresponding mask images according to specified configuration parameters. In addition, AAL The template divides the brain into multiple regions, each of which has a unique label value. These label values ​​can be traversed to create a mask for each region, and then the features of the region are extracted. In this embodiment, whole-brain feature extraction includes shape feature, texture feature and signal strength extraction. When extracting partition features, the whole brain is divided into 116 brain regions in the AAL template, and texture features and signal strength are extracted for each brain region. Among them, shape features include the volume, surface area and shape descriptor of the whole brain, texture features include features of gray-level co-occurrence matrix (GLCM) and gray-level run length matrix (GLRLM), and signal strength includes mean, standard deviation and rate of change.

[0033] Among them, 116 brain region features are merged into a partition feature matrix:

[0034] Represents brain area The feature vector of .

[0035] In addition, the whole brain feature matrix and the partition feature matrix based on the AAL template Merge to get the final feature matrix:

[0036] in, is the number of samples, .

[0037] Step S03, training an autoencoder based on the pre-processed brain MRI image data of the user with lung cancer brain metastasis to obtain a trained autoencoder to learn a low-dimensional representation of normal brain features.

[0038] It should be noted that the autoencoder is the core part of unsupervised learning. Its main function is to capture the main mode of the data through feature compression and reconstruction, and use the reconstruction error as an abnormality metric. Specifically, the autoencoder includes an encoder and a decoder. The encoder is used to compress high-dimensional input features into a low-dimensional latent space representation, and the expression is:

[0039] , , is the encoder parameter. The encoder usually uses a multi-layer fully connected network, and the activation function uses ReLU or Leaky ReLU:

[0040] in, and are weight and bias parameters; The decoder restores the low-dimensional representation to the original feature space, expressed as:

[0041] in, , Decoder parameters.

[0042] Step S04: input the feature matrix into the trained autoencoder to calculate the reconstruction error.

[0043] Specifically, the reconstruction error is calculated using the difference between the original input features and the reconstructed features. :

[0044] The larger the reconstruction error, the higher the possibility that the sample is abnormal.

[0045] Step S05: Based on the reconstruction error, the isolation forest method is used to further screen abnormal areas and generate an abnormal score for each sample.

[0046] Specifically, the isolation forest method performs anomaly detection on data points in a high-dimensional feature space by constructing a randomly segmented decision tree. In this embodiment, the reconstruction error of each sample is used as the input feature of the isolation forest:

[0047] The path length is then calculated, which is the length of the path that separates the abnormality of a data point from its path in the decision tree in the Isolation Forest method. Inversely proportional:

[0048] in, For sample The average path length in the isolation tree, is a normalization constant used to ensure that the result is within Within the range, it is understandable that anomaly scores are generated for each sample , used to determine the degree of abnormality of whole-brain or regional characteristics.

[0049] Step S06, calculating the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score, and finally completing the identification of brain metastasis of lung cancer through the abnormality scores of the whole brain features and the partition features.

[0050] It should be noted that the abnormal score expression of the whole brain feature is:

[0051] The anomaly score expression of the partition feature is:

[0052] in, and is the weight parameter for the whole-brain feature score, For the The reconstruction error of each brain region is is the reconstruction error of the whole brain, For the Isolation Forest abnormality scores in each brain region The Isolation Forest Abnormality Score for the whole brain, Abnormality scores for whole-brain features, It is the abnormal score of the partition feature. It can be understood that when the abnormal score of the whole brain feature is higher than the threshold, the user is judged as a high-risk individual for brain metastasis of lung cancer. When the abnormal score of the partition feature is higher than the threshold, it means that the corresponding area is a high-risk area.

[0053] In some other embodiments of the present invention, the abnormal scores of the partition features are mapped to the standard brain template to generate a heat map of high risk of lung cancer brain metastasis, which can be expressed as:

[0054] in, Representing voxels Is it a brain area? ,Understandably, the heat map colors from low to high represent low risk to high risk, respectively.

[0055] In summary, the lung cancer brain metastasis risk prediction method and system in the above-mentioned embodiment of the present invention obtains the user's brain MRI image data and performs data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; the preprocessed data is subjected to whole-brain feature extraction and partition feature extraction based on the AAL template, and the whole-brain features and the partition features are merged to obtain a feature matrix; according to the preprocessed brain MRI image data of the user with lung cancer brain metastasis, the autoencoder is trained to obtain the trained autoencoder to learn the low-dimensional representation of normal brain features; the feature matrix is ​​input into the trained autoencoder to calculate the reconstruction error; according to the reconstruction error, the isolation forest method is used to further screen the abnormal area, and the abnormal score is generated for each sample; according to the reconstruction error and the abnormal score, the abnormal score of the whole brain feature and the partition feature is calculated, and by combining the abnormal detection of the whole brain feature and the partition feature, the high-risk area of ​​the brain can be accurately marked, providing a basis for personalized PCI treatment.

[0056] Embodiment 2 The following is an experimental process of the method for predicting the risk of lung cancer brain metastasis in this embodiment: The experimental background is that a patient diagnosed with small cell lung cancer (SCLC) had a tumor reduction and achieved partial remission after initial treatment. However, since small cell lung cancer has a higher risk of brain metastasis, the radiation oncologist hopes to assess the patient's risk of brain metastasis to decide whether to perform prophylactic cranial irradiation (PCI) treatment and try to avoid unnecessary side effects.

[0057] Specifically, the patient's brain MRI imaging data, including T1-weighted images, enhanced scanning images (T1-Gd) and diffusion tensor imaging (DTI), were collected and stored in DICOM format. The brain MRI imaging data were then subjected to intensity normalization and artifact removal to eliminate differences in scanners and imaging conditions; the brain MRI imaging data after intensity normalization and artifact removal were aligned to the MNI template using a registration algorithm to ensure that all data were in the same anatomical frame; based on the data aligned to the MNI template, a deep learning segmentation tool (UNet) was used to segment the entire brain area.

[0058] Furthermore, radiomics feature extraction is performed.

[0059] Whole brain features: The total brain volume is 1200 cm³; The whole brain entropy value is 2.35, indicating that the signal complexity is high.

[0060] AAL template partition features: The whole brain was divided into 116 brain regions according to the AAL template; The signal intensity mean, change rate and texture features were extracted for each brain region. The volume of the left middle frontal gyrus was 45 cm³, the signal mean was 1.32, and the signal change rate was 0.15. The entropy value of the right parietal lobe was higher, indicating that the local texture changed significantly.

[0061] The patient's feature matrix Input into the autoencoder and calculate the reconstruction error:

[0062] Example: The reconstruction error in the left parietal region is high, reaching 0.45 (which is above the threshold of 0.3).

[0063] Use Isolation Forest to further filter abnormal areas and generate abnormal scores : ,Specifically, the left parietal lobe abnormality score was 0.85, indicating that it was a high-risk area.

[0064] Furthermore, the reconstruction error and anomaly score are combined to calculate the comprehensive risk score of the high-risk brain regions: , where the risk scores of the right middle frontal gyrus and left parietal lobe were 0.65 and 0.78, respectively, which were significantly higher than those of other regions. At the same time, a visual heat map of high-risk brain regions was generated, marking the left parietal lobe and right middle frontal gyrus as high-risk regions, and the heat map colors from low to high represent low risk to high risk, respectively.

[0065] Finally, the comprehensive whole brain risk score , the patient was judged to be at high risk of brain metastasis. In addition, based on the heat map and risk score, the treatment focus was placed on the left parietal lobe and right middle frontal gyrus to optimize the PCI dose and target range, thereby reducing the radiation dose in low-risk areas and reducing unnecessary side effects.

[0066] Embodiment 3 See also Figure 2 , Figure 2 1 is a structural block diagram of a lung cancer brain metastasis risk prediction system provided by Embodiment 3 of the present invention. The lung cancer brain metastasis risk prediction system 200 includes: a preprocessing module 21, a feature extraction module 22, a training module 23, a first calculation module 24, a screening module 25 and a second calculation module 26, wherein: The preprocessing module 21 is used to obtain the user's brain MRI image data and perform data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; A feature extraction module 22 is used to extract whole-brain features and partition features based on the AAL template from the preprocessed data, and merge the whole-brain features and the partition features to obtain a feature matrix, wherein the whole-brain feature extraction includes the extraction of shape features, texture features and signal intensity, and the partition feature extraction includes the extraction of texture features and signal intensity, wherein the shape features include the volume, surface area and shape descriptor of the whole brain, the texture features include the features of the gray-level co-occurrence matrix and the gray-level run length matrix, and the signal intensity includes the mean, standard deviation and rate of change; The training module 23 is used to train the autoencoder according to the pre-processed brain MRI image data of the user with lung cancer brain metastasis, to obtain the trained autoencoder, so as to learn the low-dimensional representation of normal brain features, wherein the autoencoder includes an encoder and a decoder, the encoder is used to compress the high-dimensional input features into a low-dimensional latent space representation, and the decoder restores the low-dimensional representation to the original feature space; A first calculation module 24, used for inputting the feature matrix into the trained autoencoder to calculate the reconstruction error; The screening module 25 is used to further screen abnormal areas using the isolation forest method according to the reconstruction error, and generate an abnormal score for each sample. The expression of the abnormal score is:

[0067] in, For sample The average path length in the isolation tree, is a normalization constant used to ensure that the result is within within the scope; The second calculation module 26 is used to calculate the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score, and finally complete the lung cancer brain metastasis identification through the abnormality scores of the whole brain features and the partition features. The abnormality score expression of the whole brain features is:

[0068] The anomaly score expression of the partition feature is:

[0069] in, and is the weight parameter for the whole-brain feature score, For the The reconstruction error of each brain region is is the reconstruction error of the whole brain, For the Isolation Forest abnormality scores in 3 brain regions The Isolation Forest Abnormality Score for the whole brain, Abnormality scores for whole-brain features, Score the anomaly of the partition feature.

[0070] Furthermore, in some other embodiments of the present invention, the preprocessing module 21 includes: a first processing unit for performing intensity normalization and artifact removal processing on the brain MRI image data to eliminate differences in scanners and imaging conditions; An alignment unit, used to align the brain MRI image data after intensity normalization and artifact removal to the MNI template using a registration algorithm to ensure that all data are in the same anatomical frame; The segmentation unit is used to segment the whole brain area using deep learning segmentation tools based on data aligned to the MNI template.

[0071] Furthermore, in some other embodiments of the present invention, the lung cancer brain metastasis risk prediction system 200 further includes: A mapping module is used to map the abnormal scores of the partition features to the standard brain template to generate a heat map of high risk of lung cancer brain metastasis.

[0072] Embodiment 4 Another aspect of the present invention provides an electronic device, see Figure 3 , shown is an electronic device in Embodiment 4 of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor, wherein the processor 10 implements the above-mentioned method for predicting the risk of brain metastasis of lung cancer when executing the computer program 30.

[0073] In some embodiments, the processor 10 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 20, such as executing access restriction programs.

[0074] The memory 20 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 20 may be an internal storage unit of an electronic device, such as a hard disk of the electronic device. In other embodiments, the memory 20 may also be an external storage device of an electronic device, such as a plug-in hard disk equipped on the electronic device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Further, the memory 20 may also include both an internal storage unit and an external storage device of the electronic device. The memory 20 may be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or is to be output.

[0075] It should be pointed out that Figure 3 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than those shown in the figure, or combine certain components, or arrange the components differently.

[0076] The embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for predicting the risk of brain metastasis of lung cancer.

[0077] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0078] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0079] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0080] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0081] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A method for predicting the risk of brain metastasis of lung cancer, characterized in that: The method comprises: Acquire the user's brain MRI image data and perform data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; The preprocessed data are subjected to whole-brain feature extraction and AAL template-based partition feature extraction, and the whole-brain features and partition features are merged to obtain a feature matrix; According to the pre-processed brain MRI image data of the user with lung cancer brain metastasis, the autoencoder is trained to obtain the trained autoencoder to learn the low-dimensional representation of normal brain features; Inputting the feature matrix into the trained autoencoder to calculate the reconstruction error; Based on the reconstruction error, the isolation forest method is used to further screen abnormal areas and generate an abnormal score for each sample; According to the reconstruction error and the abnormality score, the abnormality scores of the whole brain features and the partition features are calculated, and finally the lung cancer brain metastasis identification is completed through the abnormality scores of the whole brain features and the partition features.

2. The method for predicting the risk of brain metastasis of lung cancer according to claim 1, characterized in that: The data preprocessing step includes: The brain MRI image data were subjected to intensity normalization and artifact removal processing to eliminate differences in scanners and imaging conditions; Use a registration algorithm to align the intensity-normalized and artifact-removed brain MRI image data to the MNI template to ensure that all data are in the same anatomical frame; Based on the data aligned to the MNI template, a deep learning segmentation tool was used to segment the whole brain area.

3. The method for predicting the risk of brain metastasis of lung cancer according to claim 2, characterized in that: In the steps of extracting whole-brain features and partitioning features based on the AAL template on the preprocessed data, the whole-brain feature extraction includes the extraction of shape features, texture features and signal intensity, and the partitioning feature extraction includes the extraction of texture features and signal intensity, wherein the shape features include the volume, surface area and shape descriptors of the whole brain, the texture features include the features of the gray-level co-occurrence matrix and the gray-level run length matrix, and the signal intensity includes the mean, standard deviation and rate of change.

4. The method for predicting the risk of brain metastasis of lung cancer according to claim 3, characterized in that: The autoencoder includes an encoder and a decoder, wherein the encoder is used to compress high-dimensional input features into a low-dimensional latent space representation, and the decoder restores the low-dimensional representation to the original feature space.

5. The method for predicting the risk of brain metastasis of lung cancer according to claim 4, characterized in that: In the step of further screening abnormal areas using the isolation forest method according to the reconstruction error and generating an abnormal score for each sample, the expression of the abnormal score is: in, For sample The average path length in the isolation tree, is a normalization constant used to ensure that the result is within within the range.

6. The method for predicting the risk of brain metastasis of lung cancer according to claim 5, characterized in that: In the step of calculating the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score, the abnormality score expression of the whole brain features is: The anomaly score expression of the partition feature is: in, and is the weight parameter for the whole-brain feature score, For the The reconstruction error of each brain region is is the reconstruction error of the whole brain, For the Isolation Forest abnormality scores in each brain region The Isolation Forest Abnormality Score for the whole brain, Abnormality scores for whole-brain features, Score the anomaly of the partition feature.

7. The method for predicting the risk of brain metastasis of lung cancer according to claim 6, characterized in that: The step of calculating the abnormality scores of the whole brain features and the partition features according to the reconstruction error and the abnormality score comprises: The abnormal scores of the partitioned features were mapped to the standard brain template to generate a heat map of high risk of brain metastasis of lung cancer.

8. A lung cancer brain metastasis risk prediction system, characterized in that: The system is used to implement the method for predicting the risk of brain metastasis of lung cancer according to any one of claims 1 to 7, comprising: A preprocessing module is used to obtain the user's brain MRI image data and perform data preprocessing, wherein the brain MRI image data includes T1-weighted images, T2-weighted images, enhanced scanning images and functional MRI; The feature extraction module is used to extract whole-brain features and partition features based on the AAL template from the preprocessed data, and merge the whole-brain features and partition features to obtain a feature matrix; A training module is used to train an autoencoder based on the pre-processed brain MRI image data of a user with lung cancer brain metastasis, to obtain a trained autoencoder to learn a low-dimensional representation of normal brain features; A first calculation module, used for inputting the feature matrix into the trained autoencoder to calculate the reconstruction error; A screening module, used to further screen abnormal areas using an isolation forest method according to the reconstruction error, and generate an abnormal score for each sample; The second calculation module is used to calculate the abnormal scores of the whole brain features and the partition features according to the reconstruction error and the abnormal scores, and finally complete the identification of brain metastasis of lung cancer through the abnormal scores of the whole brain features and the partition features.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for predicting the risk of brain metastasis of lung cancer as described in any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the program, the method for predicting the risk of brain metastasis of lung cancer according to any one of claims 1 to 7 is implemented.

Images