A method for predicting gene mutation information from lung cancer histopathology images

Abstract

The present invention discloses a method for predicting gene mutation information from lung cancer histopathology images, comprising the following steps: S1, collecting lung cancer histopathology whole slide images and corresponding gene test reports; S2, performing image preprocessing to construct a pathology omics dataset; S3, constructing a NAVF-Bio model, which includes a pre-training module, a pathology space topology representation learning module, a multi-scale feature fusion module, and an adaptive cross-view knowledge supplementation module; S4, training the NAVF-Bio model, constraining the NAVF-Bio model based on multi-scale loss and weighted fusion loss, to obtain the final NAVF-Bio model; S5, using the final NAVF-Bio model to predict gene mutation information from lung cancer histopathology whole slide images. This invention solves the technical problem that current AI models cannot effectively predict gene mutation subtypes and mutant exon locations.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and biomedical engineering technology, and specifically relates to a method for predicting gene mutations, mutant subtypes, and mutant exons from lung cancer tissue pathology images. Background Art

[0002] Lung cancer is one of the most lethal malignancies worldwide, ranking first in both morbidity and mortality. Accurately predicting lung cancer gene mutations, including mutation subtypes and their exon locations, is crucial for personalized, targeted treatment and prognosis for lung cancer patients. Mutations in genes such as TP53, EGFR, KRAS, and ALK, as well as tumor mutation burden (TMB), are closely associated with tumorigenesis, treatment response, and prognosis. Traditional gene mutation identification relies primarily on polymerase chain reaction (PCR), Sanger sequencing, fluorescence in situ hybridization (FISH), and next-generation sequencing (NGS) for measurement and assessment. However, current genetic testing for lung cancer has two drawbacks: First, NGS, the most commonly used genetic testing method, involves complex steps, requires expensive equipment and reagents, and typically takes days to weeks to complete the complex genetic testing process and analyze the results. Therefore, NGS is both expensive and time-consuming. Second, primary hospitals or regions with limited medical resources are unable to meet patients' needs for personalized, targeted treatment. Whole-slide imaging (WSI) has revolutionized traditional pathology, providing comprehensive visual data at the cellular and tissue levels. The rich information provided in WSI, including multi-scale features from microscopic to macroscopic, provides a unique opportunity to explore the complex relationships within tumors.

[0003] Currently, several deep learning-based models have been applied to tumor biomarker prediction research, including ABMIL, CLAM, DSMIL, TransMIL, DTFT-MIL, IBMIL, HIPT, MHIM, R2T_MIL, and Patch-GCN. However, these models are not specifically designed for lung cancer gene prediction and can only provide a superficial prediction of the presence or absence of gene mutations in lung cancer, failing to delve deeper into the detailed information of gene variants (such as mutation subtype and mutation exon site). The AUC values ​​of these models for predicting the presence or absence of mutations in lung cancer genes (TP53, EGFR, KRAS) based on WSI are approximately 0.733-0.856, which does not meet the requirements for accurate prediction. Furthermore, traditional methods often rely on handcrafted features or limited deep learning architectures, which cannot fully exploit the complex spatial relationships inherent in WSI. Finally, the interpretability of the models has only been analyzed visually at the patch level.

[0004] Therefore, the limitations of existing models lead to unsatisfactory lung cancer gene prediction functions, and it is necessary to design a more advanced, comprehensive and targeted method to accurately predict lung cancer gene mutation information. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for predicting gene mutation information from lung cancer tissue pathology images, so as to solve the problem that the current deep learning model proposed in the background art still has certain limitations in the prediction of gene mutations, mutant subtypes, and mutant exons.

[0006] To achieve the above object, the present invention provides a method for predicting gene mutation information from lung cancer tissue pathology images, comprising the following steps:

[0007] S1. Collect histopathological whole-slide images and corresponding genetic testing reports of lung cancer patients;

[0008] S2. Perform image preprocessing on the histopathology whole slide images to construct the pathology dataset;

[0009] S3. Build a NAVF-Bio model for predicting gene mutation information from lung cancer histopathology whole slide images. The NAVF-Bio model includes a pre-training module, a pathology space topology representation learning module, a multi-scale feature fusion module, and an adaptive cross-view knowledge supplementation module.

[0010] The pre-training module is used to extract features from pre-processed images; the pathological spatial topology representation learning module is used to further process the extracted features to form a one-dimensional vector; the multi-scale feature fusion module is used to further extract and fuse features of different scales; and the adaptive cross-view knowledge supplementation module is used to incorporate features from multiple views to improve the prediction performance of the model.

[0011] S4. Using the pathology dataset to train the constructed NAVF-Bio model, constraining the NAVF-Bio model based on multi-scale loss and weighted fusion loss to obtain the final NAVF-Bio model;

[0012] S5. The final NAVF-Bio model is used to predict gene mutation information from lung cancer tissue pathology whole slide images, wherein the gene mutation information includes gene mutation, tumor mutation load, gene mutation subtype and protein functional domain.

[0013] In a specific embodiment, in step S2, preprocessing the histopathology whole slide image includes digitizing the histopathology whole slide image, segmenting tissue regions, and detecting background and blurred regions.

[0014] In a specific embodiment, in step S3, the pre-training module uses the Otsu algorithm to distinguish the background area of ​​the whole slide image, and uses a sliding window strategy to divide the whole slide image into large, medium and small scale patches, and uses the CTransPath pre-training model to extract a feature representation with a dimension of 768 for each patch.

[0015] In a specific embodiment, the pathological spatial topology map represents a learning module in which a KNN algorithm is used to construct points and edges of spatial topology maps of patches of different scales; an HD-Yolo algorithm is used to segment and classify cells in whole slide images, and to quantify tumor microenvironment indicators, and a KNN algorithm is also used to construct a cell spatial topology map of the tumor microenvironment. The nodes of the cell spatial topology map include cell labels, classification probabilities, and the areas of cell nuclei, and the relationships between cells of the same type serve as edges.

[0016] In a specific embodiment, the pathological space topology representation learning module includes an SGAEConv module and a multi-layer perception module;

[0017] The SGAEConv module uses a linear transformation to combine the node's own features and neighbor features, expressed as:

[0018]

[0019] Among them, W m ,W n are all learnable weight matrices, is the embedding of node v at layer k, is the aggregated feature of neighbor nodes, σ is the RuLU activation function;

[0020] The multi-layer perception module includes ReLU activation function, layer normalization and information dropout regularization; uses ReLU activation function to perform nonlinear transformation on input features; based on layer normalization, it stabilizes model training and accelerates model convergence; the multi-layer perception module uses information dropout regularization to drop a part of neurons with a certain probability in each training, thereby helping the model to train better and prevent overfitting.

[0021] In a specific embodiment, the multi-scale feature fusion module includes an adaptive pooling layer, a splicing layer and a self-attention feature importance extractor; the adaptive pooling layer is used to dynamically adjust the length of the input sequence so that the output length matches the given target size; the splicing layer is used to perform splicing in a unified feature space; the self-attention feature importance extractor uses an attention mechanism to obtain features that are relatively important to the model based on feature importance.

[0022] In a specific embodiment, the adaptive cross-view knowledge supplementation module includes an adaptive feature fusion module, an adaptive weight module and a classifier; the adaptive feature fusion module uses an attention mechanism to process the input data and generate a global representation of the input features; the adaptive weight module includes two fully connected layers, and the adaptive weight module is used to perform feature scaling to form adaptive weights; the classifier is used to obtain the representation of each label.

[0023] In a specific embodiment, the multi-scale loss is optimized separately at different scales; the weighted fusion loss is used to integrate losses from different scales and reflect the importance of losses of different scales to the overall task in a weighted manner.

[0024] In a specific embodiment, the predicted gene mutation information includes:

[0025] Predict whether there is TP53 gene mutation, EGFR mutation, KRAS mutation, or ALK mutation;

[0026] Predicting the level of tumor mutation burden;

[0027] Predict TP53 gene mutation subtypes;

[0028] Predict TP53, EGFR, KRAS, and ALK mutant exons.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] The present invention aims to simulate the pathologist's film reading steps to observe the multi-scale features of WSI and TME features to predict key gene mutations in lung cancer, including gene mutation subtypes and exons. The present invention constructs a deep learning model for adaptive multi-view feature fusion for the first time. By adopting multi-scale feature fusion and adaptive cross-view knowledge complementation strategy to extract multi-view information such as TME from WSI to predict key gene mutations and tumor mutation burden (TMB) status, the model is named NAVF-Bio model. The process of multi-scale feature fusion completely simulates the pathologist's film reading mode to perform feature extraction and fusion, that is, obtaining tumor distribution, boundary and other information under low magnification, obtaining tumor tissue structure framework and cell morphological characteristics under medium magnification, and obtaining tumor cell karyotype changes under high magnification; the adaptive cross-view knowledge supplementation (ACKC) strategy uses the attention mechanism and adaptive weight module to flexibly incorporate multi-view features to improve the prediction performance of the model. NAVF-Bio addresses the following challenges: First, in predicting the presence of gene mutations and TMB status, the NAVF-Bio model achieved an AUC of 0.9162-0.9293 in the LCSXH-CSU and The Cancer Genome Atlas-Lung Adenocarcinoma (TCGA-LUAD) datasets, demonstrating exceptionally high accuracy. Second, the NAVF-Bio model, for the first time, predicted TP53 mutation subtypes and mutated exons for genes (TP53, EGFR, KRAS, ALK), achieving clinical-grade performance and providing a reference for precise targeted drug treatment in lung cancer patients whose gene mutations could not be detected for various reasons.

[0031] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention is further described in detail below. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0033] Figure 1 is a training flow chart of the NAVF-Bio model according to an embodiment of the present invention;

[0034] Figure 2 is a schematic diagram of an attention mechanism according to an embodiment of the present invention, wherein: The Kronecker product represents an operation between two matrices of arbitrary size;

[0035] Figure 3 is an architectural diagram of a NAVF-Bio model according to an embodiment of the present invention;

[0036] Figure 4 1 is an architectural diagram of the ACKC module of the NAVF-Bio model according to an embodiment of the present invention;

[0037] Figure 5 is a ROC curve diagram of the NAVF-Bio model according to one embodiment of the present invention for predicting TP53 mutation subtypes based on the LCSXH-CSU dataset;

[0038] Figure 6 This is a confusion matrix diagram of the NAVF-Bio model according to one embodiment of the present invention for predicting TP53 mutation subtypes based on the LCSXH-CSU dataset;

[0039] Figure 7 1 is a ROC curve diagram of the NAVF-Bio model according to an embodiment of the present invention for predicting TP53 mutation subtypes based on the TCGA-LUAD dataset;

[0040] Figure 8 This is a confusion matrix diagram of the NAVF-Bio model according to one embodiment of the present invention for predicting TP53 mutation subtypes based on the TCGA-LUAD dataset;

[0041] Figure 9 This is an AUCROC curve diagram of the NAVF-Bio model according to an embodiment of the present invention predicting TP53 mutant exons based on the LCSXH-CSU dataset;

[0042] Figure 10 This is a confusion matrix diagram of the NAVF-Bio model according to an embodiment of the present invention predicting TP53 mutant exons based on the LCSXH-CSU dataset;

[0043] Figure 11 This is an AUCROC curve diagram of the NAVF-Bio model according to one embodiment of the present invention predicting EGFR mutant exons based on the LCSXH-CSU dataset;

[0044] Figure 12 This is a confusion matrix diagram of the NAVF-Bio model according to an embodiment of the present invention predicting EGFR mutant exons based on the LCSXH-CSU dataset;

[0045] Figure 13 This is an AUCROC curve diagram of the NAVF-Bio model according to one embodiment of the present invention predicting KRAS mutant exons based on the LCSXH-CSU dataset;

[0046] Figure 14 This is a confusion matrix diagram of the NAVF-Bio model according to an embodiment of the present invention predicting KRAS mutant exons based on the LCSXH-CSU dataset;

[0047] Figure 15This is an AUCROC curve diagram of the NAVF-Bio model according to one embodiment of the present invention predicting ALK mutant exons based on the LCSXH-CSU dataset;

[0048] Figure 16 This is a confusion matrix diagram of the NAVF-Bio model according to an embodiment of the present invention for predicting ALK mutant exons based on the LCSXH-CSU dataset. DETAILED DESCRIPTION

[0049] The embodiments of the present invention are described in detail below. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] The present invention provides a method for predicting gene mutation information from lung cancer tissue pathology images, comprising the following steps:

[0051] S1. Collect histopathological whole slide images (WSI) and corresponding genetic testing reports of lung cancer patients;

[0052] S2. Perform image preprocessing on the histopathology whole slide images to construct the pathology dataset;

[0053] S3. Build a NAVF-Bio model for predicting gene mutation information from lung cancer histopathology whole slide images. The NAVF-Bio model includes a pre-training module, a pathology space topology representation learning module, a multi-scale feature fusion module, and an adaptive cross-view knowledge supplementation module.

[0054] The pre-training module is used to extract features from pre-processed images; the pathological spatial topology representation learning module is used to further process the extracted features to form a one-dimensional vector; the multi-scale feature fusion module is used to further extract and fuse features of different scales; and the adaptive cross-view knowledge supplementation module is used to incorporate features from multiple views to improve the prediction performance of the model.

[0055] S4. Using the pathology dataset to train the constructed NAVF-Bio model, constraining the NAVF-Bio model based on multi-scale loss and weighted fusion loss to obtain the final NAVF-Bio model;

[0056] S5. The final NAVF-Bio model is used to predict gene mutation information from lung cancer tissue pathology whole slide images, wherein the gene mutation information includes gene mutation, tumor mutation load, gene mutation subtype and protein functional domain.

[0057] Example 1

[0058] A method for predicting biomarkers from lung cancer histopathology images, comprising the following steps:

[0059] S1. Collect histopathological whole-slide images and corresponding genetic testing reports of lung cancer patients.

[0060] S2. Perform image preprocessing on histopathology whole slide images and construct a pathology dataset.

[0061] The present invention collected histopathological images and high-throughput sequencing gene detection reports of 1576 lung cancer patients and constructed a large histopathological image dataset.

[0062] The step S2 specifically includes: digitizing the histopathology whole slide image, segmenting the tissue area, detecting the background and blurred areas, and constructing a large pathology dataset.

[0063] S3. Build a NAVF-Bio model for predicting gene mutation information from lung cancer histopathology whole slide images. The NAVF-Bio model includes a pre-training module, a pathology space topology representation learning module, a multi-scale feature fusion module, and an adaptive cross-view knowledge supplementation module.

[0064] The pre-training module uses the Otsu algorithm to distinguish the background area of ​​the WSI and uses a sliding window strategy to divide the WSI into many large, medium and small scale patches. The pixel size of the large-scale patch is preferably 1024×1024, the pixel size of the medium-scale patch is 512×512, and the pixel size of the small-scale patch is 256×256. The spatial position of the patch is marked. The pre-trained model of CTransPath is used to extract one-dimensional features for each patch at multiple scales to enrich the diverse feature representation of the image. CTransPath is based on the Swin Transformer architecture and consists of three convolutional layers and four Swin Transformer modules. It can extract local features from image data and capture global features and long-range dependencies. Based on CTransPath, the embedded features of the patches are extracted, and each patch embedding is unified into a 768-dimensional vector representation. Therefore, all patches at each scale in the WSI are represented in the form of feature vectors.

[0065] However, the patches in WSI have spatial position relationships. The present invention uses the KNN algorithm to construct the spatial topology graph of patches of different scales (G l ,G m ,G s ) points and edges. Secondly, in order to construct the tumor microenvironment (TME) of WSI, the present invention uses the HD-Yolo algorithm to segment and classify the cells in WSI and quantify the TME indicators. The present invention also uses the KNN algorithm to construct the cell space topology map (G T), where the nodes of the graph contain cell labels, classification probabilities, and cell nucleus areas, and the relationships between cells of the same species serve as edges. Therefore, the multi-scale features and TMEs under large, medium, and small fields of view are represented as G l =(V l ,E l ), G m =(V m ,E m ), G s =(V s ,E s ), G T =(V T ,E T ).

[0066] The spatial topology graph represents learning. The spatial topology graphs under large, medium and small fields of view and the TME spatial topology graphs in the present invention are represented as G l =(V l ,E l ),G m =(V m ,E m ),G s =(V s ,E s ),G T =(V T ,E T ). The present invention proposes a pathological spatial topology representation learning module to learn, including the SGAEConv module and the Multilayer Reception module (multi-layer perception module). Each node v∈V of the multi-scale G has a feature vector Denotes the node embedding at the kth layer. The purpose of SGAEConv is to generate the node embedding at the k+1th layer. First, the SGAEConv module samples a subset N from its neighbor node set N(v) S (v) Then, these neighbor node features are averaged and aggregated. The process is as follows:

[0067]

[0068] Among them, N S (v) is the neighbor aggregation of node v, the features of the sampled neighbor nodes Secondly, the features of node v itself are combined with the aggregated neighbor features to obtain a new embedding representation of node v. The SAGEConv module uses a linear transformation to combine the node's own features and neighbor features, expressed as:

[0069]

[0070] Among them, W m ,Wn are all learnable weight matrices, is the embedding of node v at layer k, is the aggregated feature of neighboring nodes, and σ is the RuLU activation function. To stabilize the training process, it is preferred to normalize the aggregated features so that the embedding of each node has a stable numerical range. Here, to prevent gradient explosion or vanishing, the model performs L2 normalization on the output of the graph neural network:

[0071]

[0072] Among them, ||.|| represents the norm of the input feature vector.

[0073] The features processed by the SGAEConv module need to be further processed by the multi-layer perception module (MPB). The multi-layer perception module includes ReLU activation function, layer normalization (LayerNormalization) and information dropout (Dropout) regularization. In order to allow the model to learn more complex feature representations, the present invention uses the ReLU activation function to perform nonlinear transformation on the input features. Secondly, based on the function of layer normalization to stabilize the model training and accelerate the convergence of the model, the model will make the input features more robust. Finally, the multi-layer perception module uses information dropout regularization to drop a part of the neurons with a probability of 20% in each training, thereby helping the model to train better and prevent overfitting. Especially when processing very large graphs, information dropout regularization forces the model to learn more robustly and improve generalization ability.

[0074] The features extracted by the pathological space topology representation learning module are super-long one-dimensional vectors The process of multi-scale feature fusion completely simulates the pathologist’s experience in reading films to extract and fuse features. Small field of view characteristics and mid-field features Therefore, the present invention proposes a feature fusion module to guide the pairwise fusion of different features. The module consists of an adaptive pooling layer, a splicing layer, and a self-attentive feature importance extractor.

[0075] Unlike ordinary pooling layers, Adaptive Average Pooling can automatically adjust the size and stride of the pooling window according to the given target output size, so that the output feature map size is consistent with the desired one. The purpose of this layer is to dynamically adjust the length of the input sequence so that the output length matches the given target size. Its core idea is to divide the input sequence into several intervals and then apply the maximum pooling operation to each interval. Secondly, the present invention uses the splicing layer to splice in a unified feature space. and and

[0076] Given the huge number of patches and cells in WSI, the fused features need to further extract key information. Therefore, the key step in feature fusion is to use the attention mechanism to obtain the features that are relatively important to the model based on the feature importance. and and The merged features are and In order to save computing resources, this paper proposes an attention mechanism with an algorithm complexity of o(N) to learn and The present invention first and Mapped to query, key and value vectors respectively, the process is:

[0077]

[0078] Among them, W Q ∈R d×d , W K ∈R d×d , W V ∈R d×d are trainable parameters. The present invention uses L2 normalization to process the key and query vectors. The present invention reduces the complexity of attention by swapping the order of matrix products. With the help of the Taylor expansion of the exponential function, it can be used to stimulate the new attention function:

[0079]

[0080] Among them Ι N×1 Is an all-one vector. In the above calculation, the present invention realizes the aggregation of full attention, avoids the N×N attention matrix, and only requires linear complexity o(Nd 2 ). For large vectors and N is usually several orders of magnitude larger than d, so in practice it can significantly improve computational efficiency.

[0081] The adaptive cross-view knowledge supplement module can flexibly incorporate features from multiple views to improve the prediction performance of the model.

[0082] The Adaptive Cross-View Knowledge Complementation (ACKC) module consists of an Adaptive Feature Fusion module, an Adaptive Weight module, and a classifier. The function of the Adaptive Feature Fusion module is to adaptively fuse features from different sources or different levels to generate a more effective feature representation. The goal of this module is to dynamically select and combine features to achieve better performance in downstream tasks, including classification and regression. The large, medium, and small field of view features and TME features input to the ACKC module are represented as follows: and Multilayer Perceptron (MLP) first maps the features of different branches to the same dimensional space, and then uses the splicing layer to combine them and

[0083]

[0084] Then, the adaptive feature fusion module uses the attention mechanism to process the input data and generate a global representation of the input features. Given that the features of different branches are in the same dimensional space, the attention mechanism can capture the global features of all features and the dependencies of the context, and further extract important features. The idea is to use one-dimensional features to perform linear mapping and position encoding of features. The attention mechanism can capture the dependencies between each feature and the global in long sequence data. Stacked multi-layer attention mechanism layers can extract multi-level inputs of features, from low-level features to high-level global semantic features. Secondly, the attention mechanism can dynamically assign weights to input features during feature extraction, allowing the model to focus on important features, thereby effectively extracting key information. Specifically, the input feature X first undergoes linear projection, and the process is:

[0085] X p =W emb ·X+b emb

[0086] in is the weight matrix, b emb is the bias, X p is the result of linear mapping of feature X. In order to adapt to the calculation of the attention mechanism module, the present invention converts the one-dimensional feature X pThe length S is divided into multiple blocks, each block has a length of C, assuming X p Can be divided evenly, then the number of blocks N = S / C. Secondly, add position encoding for each input sequence block This enables the model to understand the order information of each position in the sequence. Next, the Transformer encoder is used to encode X p For processing, each layer of transformer includes an attention mechanism and a feedforward neural network. The process is as follows:

[0087]

[0088] FFN(X)=W2·ReLU(W1·X+b1)+b2

[0089] In order to enhance the expressive power of key features and balance the contributions of different features, the features after feature importance screening need to be further processed by the adaptive weight module. This step is also to suppress irrelevant and redundant features, supplement the problem of insufficient attention mechanism, and improve the generalization performance of the model. Specifically, the adaptive weight module consists of two fully connected layers, which perform feature scaling to form adaptive weights. Assume that the weight matrix of the fully connected layer is Where r is the scaling factor, then the features after full connection are:

[0090] s=W m ReLU(W n x)

[0091] Here, ReLU represents the ReLU activation function. The features are then weighted using the Sigmoid activation function, which limits the output to the range [0, 1]. Finally, the generated weights are used to perform element-wise multiplication of the input features. Therefore, the workflow of the adaptive weight module can be summarized as follows:

[0092] y=x⊙σ(W m ReLU(W n ·x))

[0093] Where σ is the Sigmoid activation function and ⊙ is the element-wise multiplication. To obtain the final output result, the present invention uses the Linear layer as the backbone of the feature classifier to obtain the representation of each label.

[0094] Based on multi-view loss and weighted fusion loss, the model is constrained and multiple biomarkers are predicted, including gene mutations, tumor mutation burden, gene mutation subtypes and protein functional domains.

[0095] In view of the fact that the model obtains features from the multi-scale images of WSI, the present invention proposes a weighted multi-scale and fused Loss to optimize the performance of the model. Multi-scale loss is to fully explore the diverse information in the data by optimizing separately at different scales. The weighted fusion Loss is used to integrate these losses from different scales and reflect their importance to the overall task in a weighted manner. The combination of the two can help the model understand data from multiple scales during the learning process, improve the comprehensiveness of feature learning, and ensure that the contribution of each scale is treated reasonably, thereby improving the overall performance, robustness and generalization of the model. For the features at each scale, the present invention defines the corresponding loss function Loss ms In the present invention, the biomarker prediction task can be summarized as a classification task, and the cross-entropy loss function can be used to optimize the performance of the model. Secondly, the present invention assigns different weights to the loss function according to the contribution of each scale. Assuming that there are N∈[1,4], the process can be expressed as:

[0096]

[0097] Among them, λ is the weight coefficient. x and y represent the input data and labels of different scales respectively. Each Loss adp is the loss after multi-scale fusion. These weights can dynamically adjust the impact of each scale loss on the overall optimization process.

[0098] All models were trained on a Linux host, based on the PyTorch 2.1 platform and on four V100s and one H800. The model hyperparameters used the Adam optimizer and weight decay 1*10. -5 The learning rate is set to 1*10 -4 The training step size is 100. The present invention uses a five-fold cross-validation strategy to train the model. When the model achieves a good prediction effect in each fold, the model with the best performance in that fold is saved.

[0099] This paper uses the Area Under the Receiver Operating Characteristic Curve (AUCROC) curve, the Precision-Recall Curve (PRC) curve, and the Accuracy, Recall, Precision, F1-score, and AUCROC as model evaluation indicators. The larger the area under the AUCROC and PRC curves, the larger the evaluation indicator value and the better the model performance.

[0100] Experimental results:

[0101] Currently, there are 11 AI models that can perform rough predictions of cancer gene mutations, namely: ABMIL, CLAM-SB, CLAM-MB, DSMIL, TransMIL, DTFT-MIL, IBMIL, HIPT, MHIM, RRT_MIL and Patch-GCN. The NAVF-Bio model of the present invention has opened up a new track in the field of lung cancer gene prediction, successfully improving the prediction accuracy of key lung cancer gene mutation subtypes and mutated exons to the clinical level. Specifically, in the two datasets of LCSXH-CSU and TCGA-LUAD, the NAVF-Bio model of the present invention, compared with other models, showed ultra-high accuracy in predicting the presence or absence of TP53 gene mutations (Table 1 and Table 6), EGFR mutations (Table 2 and Table 7), KRAS mutations (Table 3 and Table 8), ALK mutations (Table 4 and Table 9), and distinguishing between high and low TMB (Table 5 and Table 10), reaching the clinical application level, and has strong generalization performance, which can be widely used in different datasets.

[0102] In addition, the NAVF-Bio model of the present invention can achieve good results in the classification of TP53 gene mutation subtypes. The results of ROC curve analysis showed that the NAVF-Bio model had good predictive performance for TP53 gene mutation subtypes in both the LCSXH-CSU dataset and the TCGA-LUAD dataset. The confusion matrix showed that NAVF-Bio achieved good accuracy in each type of TP53 mutation subtype, especially in the expression prediction of the wild type, with accuracy rates reaching 0.8728 and 0.9312, respectively. The NAVF-Bio model's prediction results for TP53, EGFR, KRAS, and ALK mutated exons showed that the ROC curves of the mutated exons of each key gene showed high predictive performance, with AUCs reaching 0.9313, 0.8772, 0.9476, and 0.9467, respectively; the confusion matrix showed that NAVF-Bio achieved good accuracy in the positioning of the mutated exons of each gene.

[0103] Table 1. Performance comparison of the present invention and 11 models on the TP53 binary classification task based on the LCSXH-CSU dataset.

[0104] AUCROC ACC Precision Recall F1-Score ABMIL 80.9±7.28 71.16±9.23 78.38±5.87 79.05±3.39 78.06±4.59 CLAM-SB 88.53±8.82 85.61±8.89 86.43±8.43 86.5±8.31 85.47±9.2 CLAM-MB 88.47±8.47 85.74±8.5 86.37±8.33 86.86±8.52 85.65±9.07 DSMIL 83.96±8.13 78.02±6.07 81.3±6 81.26±5.41 79.03±7.44 TransMIL 92.15±7.02 91.61±7.43 91.95±7.13 91.61±7.43 91.73±7.32 DTFD-MIL 66.96±4.35 65.15±2.44 65.18±2.43 65.22±2.42 61.61±4.71 IBMIL 64.72±3.88 62.25±1.46 62.46±1.45 62.69±1.43 59±4.32 HIPT 79.35±6.46 72.23±4.61 73.08±4.73 74.82±4.99 73.66±4.75 MHIM 51.4±1.29 45.09±2.19 58.49±0.57 52.39±2.29 46.43±3.53 RRT_MIL 52.69±1.85 52.23±1.53 58.05±1.12 62.28±2.75 55.72±2.4 Patch-GCN 87.98±6.78 83.15±5.67 83.3±5.69 83.59±5.65 83.08±5.77 The present invention 92.93±6.27 92.3±6.24 92.3±6.23 92.38±6.18 92.3±6.24

[0105] Table 2. Performance comparison of the present invention and 11 models on the EGFR binary classification task based on the LCSXH-CSU dataset.

[0106] AUCROC ACC Precision Recall F1-Score ABMIL 83.71±5.1 79.04±4.74 79.7±4.49 81.07±4.15 79.72±4.35 CLAM-SB 82.35±5.54 73.09±4.53 74.42±4.46 76.64±4.74 75.87±4.64 CLAM-MB 69.22±13.64 76±4.85 77.12±4.33 79.33±5.15 76.82±4.6 DSMIL 85.84±6.37 78.6±5.55 81.51±5.08 82.27±5.21 81.18±5.34 TransMIL 90.44±7.15 88.16±6.19 88.2±6.15 88.16±6.19 87.42±6.66 DTFD-MIL 66.76±2.88 64.54±1.99 64.6±1.97 64.6±1.97 63.38±2.02 IBMIL 66.1±2.79 63.97±1.14 64.07±1.13 64.07±1.13 62.47±1.61 HIPT 82.02±5.75 74.36±4.59 75.58±4.49 74.55±4.46 74.74±4.41 MHIM 51.98±0.76 47.06±2.57 60.51±0.8 54.52±1.88 46.82±1.6 RRT_MIL 53.69±1.85 52.25±0.98 57.79±0.87 58.28±1.49 54.38±1.3 Patch-GCN 75.43±5.55 71.45±4.01 72.12±3.99 72.12±3.99 70.18±4.45 The present invention 92.32±6.69 91.15±6.71 91.46±6.43 91.32±6.55 91.61±6.3

[0107] Table 3. Performance comparison of the present invention and 11 models on the KRAS binary classification task based on the LCSXH-CSU dataset.

[0108] AUCROC ACC Precision Recall F1-Score ABMIL 77.96±8.69 59.8±9.43 62.93±7.75 60.66±9.46 52.09±6.6 CLAM-SB 87.7±9.61 94.63±2.06 94.99±1.8 95.01±1.98 84.23±7.58 CLAM-MB 87.41±9.69 90.72±4.56 91.44±4.38 91.73±4 82.83±7.68 DSMIL 80.8±8.91 77.53±3.34 84.16±2.78 83.71±3.93 64.6±3.4 TransMIL 88.43±10.35 97.98±1.6 97.98±1.6 98.11±1.62 89.75±8.71 DTFD-MIL 69.03±5.93 69.44±3.41 70.79±3.14 72.65±2.82 55.79±2.37 IBMIL 66.14±7.12 76.82±2.66 77.1±2.61 77.11±2.67 58.8±3 HIPT 74.43±7.48 58.22±10.21 58.24±10.19 58.26±10.21 48.64±7.8 MHIM 53.07±2.86 65..06±2.68 78.69±2.82 79.84±3.89 54.31±0.87 RRT_MIL 53.7±1.83 65.83±5.76 77.98±4.07 71.07±6.06 51.65±1.82 Patch-GCN 72.57±6.73 78.79±4.05 79.49±4.09 79.85±4.4 60.13±3.15 The present invention 90.6±7.74 96.7±2.32 96.64±2.41 96.54±2.48 96.77±2.34

[0109] Table 4. Performance comparison of the present invention and 11 models on the ALK binary classification task based on the LCSXH-CSU dataset.

[0110] AUCROC ACC Precision Recall F1-Score ABMIL 79±5.5 71.51±1.42 73.76±1.17 72.2±1.52 54.99±2.01 CLAM-SB 90.53±8.08 94.5±2.92 94.77±2.87 95.07±2.54 86.18±7.54 CLAM-MB 90.58±8.01 94.5±2.19 94.53±2.18 94.75±2.09 83.08±6.78 DSMIL 84.74±5.98 77.51±3.5 84.81±2.14 84.01±4.35 64.75±3.8 TransMIL 91.21±7.22 96.02±1.97 96.08±1.97 96.02±1.97 87.08±7.04 DTFD-MIL 71.58±3.19 75.49±1.72 75.9±1.83 75.55±1.77 55.8±1.95 IBMIL 69.51±3.11 76.56±3.4 77.26±3.56 76.69±3.4 56.55±3.2 HIPT 78±3.65 78.9±2.35 78.95±2.34 78.9±2.35 58.77±2.21 MHIM 48.88±1.86 81.83±4.38 88.06±3.95 87.46±1.12 55.53±1.06 RRT_MIL 49.89±3.29 84.71±4.48 87.06±3.72 86.16±4.64 52.69±1.13 Patch-GCN 72.78±4.36 78.6±4.98 79.67±4.57 78.66±4.94 58.12±3.26 The present invention 92.58±7.32 98.83±0.97 98.41±1.35 98.4±1.36 98.83±0.97

[0111] Table 5. Performance comparison of the present invention and 11 models on the TMB binary classification task based on the LCSXH-CSU dataset.

[0112] AUCROC ACC Precision Recall F1-Score ABMIL 81.31±6.11 74.84±4.75 76.61±4.45 80.43±5.08 66.07±4.78 CLAM-SB 90.43±8.48 95.99±2.87 95.99±2.87 95.99±2.87 89.78±7.35 CLAM-MB 90.49±8.47 95.87±3.02 96.12±3.05 95.87±3.02 90.31±7.6 DSMIL 86.44±7.66 80.71±6.04 83.33±4.92 90.18±6.35 74.96±5.41 TransMIL 91.1±7.96 95.74±3.81 95.93±3.64 96.65±3 92.08±7.08 DTFD-MIL 77.28±5.54 77.7±2.84 77.76±2.82 77.95±2.78 64.57±2.52 IBMIL 73.94±4.95 72.5±5.12 72.85±4.97 72.65±5.18 59.98±2.97 HIPT 79.41±6.1 78.87±3.94 79.05±3.83 78.87±3.94 66.04±3.36 MHIM 50.98±2.61 69.8±5.6 77.38±4.41 83.01±0.85 57.1±0.92 RRT_MIL 49.8±1.92 73.63±2.25 80.13±2.1 78.7±3.33 54.53±1.86 Patch-GCN 80.42±7.67 86.91±2.58 86.91±2.58 86.91±2.58 71.58±6.02 The present invention 91.62±7.75 96.6±2.5 96.67±2.42 96.59±2.5 96.77±2.33

[0113] Table 6. Performance comparison of the present invention and 11 models on the TP53 binary classification task based on the TCGA-LUAD dataset.

[0114] AUCROC ACC Precision Recall F1-Score ABMIL 80.9±16.3 71.16±20.65 65.86±2.76 66.93±4 78.06±10.27 CLAM-SB 88.53±19.72 85.61±19.88 82.92±7.37 83.57±6.87 85.47±20.57 CLAM-MB 88.47±18.95 85.74±19 80.46±7 80.99±6.86 85.65±20.3 DSMIL 71.45±5.99 63.37±3.33 68.76±3.88 69.32±4.46 68.29±3.84 TransMIL 91.19±7.39 89.51±8.07 90.64±7.06 92.28±5.6 91.34±6.43 DTFD-MIL 58.88±2.12 62.78±0.35 63.97±0.55 63.57±0.178 60.72±1.85 IBMIL 59.1±2.29 61.79±0.53 63.22±0.78 63.57±0.52 59.38±1.96 HIPT 69.09±2.68 65.35±2.08 67.26±2.33 65.35±2.08 65.99±2.23 MHIM 55.82±0.55 38.28±2.49 60.2±1.21 60.74±4.99 59.68±1.99 RRT_MIL 56.59±0.85 51.6±1.6 62.06±1.12 59.13±3.18 57.19±2.47 Patch-GCN 64.28±2.93 63.96±2.29 64.91±1.71 65.94±0.99 63.09±2.54 The present invention 90.18±8.78 92.2±6.98 92.34±6.85 92.28±6.91 92.4±6.8

[0115] Table 7. Performance comparison of the present invention and 11 models on the EGFR binary classification task based on the TCGA-LUAD dataset.

[0116] AUCROC ACC Precision Recall F1-Score ABMIL 72.21±1.35 71.09±2.85 79.11±2.44 74.06±3.58 66.26±1.02 CLAM-SB 68.64±2.11 80.2±1.96 82.23±1.45 81.59±1.76 66.25±1.48 CLAM-MB 72.21±1.35 71.09±2.85 79.11±2.44 74.05±3.57 66.26±1.02 DSMIL 73.39±5.32 78.02±2.25 82.8±2.06 85.35±1.57 69.44±2.93 TransMIL 94.84±4.55 92.87±5.51 93.13±5.28 93.27±5.16 91.1±6.44 DTFD-MIL 58.89±1.99 57.23±5.3 58.25±5.21 59.61±5.21 51.62±3.53 IBMIL 60.16±3.33 51.73±7.35 52.15±7.37 51.98±7.55 46.79±4.61 HIPT 63.71±4.51 73.47±5 75.32±4.33 75.84±3.47 62.48±3.12 MHIM 56.15±1.53 60.51±5.99 76.74±3.85 71.23±5.28 56.51±2.31 RRT_MIL 61.71±2.68 71.86±4.39 78.17±2.22 85.28±4.39 66.56±1.24 Patch-GCN 60.76±3.09 77.63±3.81 77.77±3.78 77.63±3.81 61.81±1.86 The present invention 87.72±7.49 77.44±11.02 86.69±8.46 86.18±8.23 88.43±8.2

[0117] Table 8. Performance comparison of the present invention and 11 models on the KRAS binary classification task based on the TCGA-LUAD dataset.

[0118] AUCROC ACC Precision Recall F1-Score ABMIL 90.33±2.83 85.35±3.16 86.23±3.3 85.35±3.16 65.2±4.85 CLAM-SB 96.29±0.53 96.61±0.87 96.61±0.87 96.61±0.87 92.08±5.07 CLAM-MB 96.54±0.21 96.4±0.22 96.41±0.22 96.41±0.22 94.74±2.05 DSMIL 90.15±2.62 91.19±4.73 84.94±4.17 83.76±5.03 64.32±5.49 TransMIL 96.3±1.32 96.23±2.05 96.23±2.05 96.23±2.05 89.97±6.94 DTFD-MIL 83.42±5.02 78.02±4.89 78.19±4.95 78.02±4.89 58.28±5.27 IBMIL 83.87±5 77.43±5.69 77.6±5.69 77.63±5.78 58.59±5.65 HIPT 87.93±1.86 81.59±2.97 81.68±2.98 82.18±3.1 60.08±3.04 MHIM 88.77±2.08 83.74±0.94 85.6±1.07 87.72±2.4 62.64±0.9 RRT_MIL 69.97±3.94 69.48±4.46 75.86±4.4 77.26±3.92 57.15±2.86 Patch-GCN 65.25±5.65 69.9±15.64 71.69±15.59 71.88±15.8 51.83±11.06 The present invention 96.67±2.85 97.09±2.39 97.03±2.48 96.92±2.55 97.15±2.41

[0119] Table 9. Performance comparison of the present invention and 11 models in the ALK binary classification task based on the TCGA-LUAD dataset.

[0120] AUCROC ACC Precision Recall F1-Score ABMIL 74.94±3.98 89.11±2.79 89.54±2.78 89.51±2.64 71.54±3.92 CLAM-SB 75.88±4.7 79.21±4.41 83.28±4.57 85.35±4.54 62.74±4.29 CLAM-MB 83.9±5.06 81.98±4.65 82.04±3.96 80±4.47 70.03±5.5 DSMIL 80.65±6.73 84.36±3.58 88.01±3.23 90.3±3.4 71.96±4.88 TransMIL 89.68±7.41 97.43±1.67 97.44±1.67 98.02±1.77 91.07±5.72 DTFD-MIL 64.32±3.8 80.4±2.99 83.13±3.42 80.4±3 57.03±1.47 IBMIL 65.74±3.45 81.19±3.35 81.77±3.51 81.59±3.78 59.97±3.52 HIPT 62.15±3.95 80.6±4.5 80.98±4.25 81.78±3.77 59.4±4.2 MHIM 66.43±4.36 73.79±5.05 83.96±3.42 78.95±4.87 59.07±1.56 RRT_MIL 69.97±3.94 69.49±4.46 75.86±4.4 77.26±3.92 57.15±2.86 Patch-GCN 64.23±4.65 78.22±6.37 78.58±6.27 78.42±6.26 60.34±6.28 The present invention 89.86±8.64 98.83±0.97 98.41±1.35 98.83±0.97 98.4±1.36

[0121] Table 10. Performance comparison of the present invention and 11 models in the TMB binary classification task based on the TCGA-LUAD dataset.

[0122] AUCROC ACC Precision Recall F1-Score ABMIL 65.1±3.87 67.54±3.5 69±3.13 69.14±2.68 66.67±2.57 CLAM-SB 86.1±8.04 85.47±6.06 85.47±6.06 85.47±6.06 85.16±6.14 CLAM-MB 86.98±7.29 84.64±5.46 85.24±5.01 85.85±4.82 85.25±4.9 DSMIL 80.41±6.95 72.82±4.1 81.91±5.09 76.89±4.39 78.09±4.19 TransMIL 90.76±8.19 91.58±6.63 91.73±6.49 91.98±6.27 91.83±6.39 DTFD-MIL 63.29±2.53 67.09±1.39 67.89±1.01 67.09±1.39 65.23±1.28 IBMIL 63.05±2.76 65.07±1.72 66.79±1.19 65.48±1.77 65.17±1.29 HIPT 71.44±5.02 70.78±2.73 71.02±2.72 71.59±2.75 70.7±2.61 MHIM 53.97±1.71 45.33±1.97 57.7±1.59 59.59±3.22 53.38±3.45 RRT_MIL 55.6±2.3 51.87±1.69 61.25±1.07 66.47±4.29 61.22±3.27 Patch-GCN 80.42±7.67 86.91±2.58 86.91±2.58 86.91±2.58 71.58±6.02 The present invention 92.52±6.69 94.61±4.82 95.02±4.46 95.61±3.93 94.61±4.82

[0123] Table 11. Performance comparison of the present invention and 11 models on the TP53 multi-classification task based on the LCSXH-CSU dataset.

[0124] AUCROC ACC Precision Recall F1-Score ABMIL 59.97±2.59 25±2.12 57.81±0.65 46.15±2.47 45.24±3.82 CLAM-SB 86.35±7.95 64.34±7.27 74.96±6.44 84.54±4.18 75.3±7.48 CLAM-MB 86.37±7.95 64.21±7.23 74.51±6.25 85.04±4.24 75.23±7.46 DSMIL 59.58±2.3 25.95±1.75 50.21±3.72 51.65±4.99 43.08±3.85 TransMIL 86.84±7.18 77.72±10.22 82.0±8.32 86.21±5.76 83.62±8.27 DTFD-MIL 61.76±5.1 36.2±1.13 47.71±2.61 62.07±5.25 47.03±4.88 IBMIL 63.43±4.03 37±3.98 49.02±1.65 64.2±2.2 48.83±2.91 HIPT 81.13±6.76 59.08±5.69 62.99±5.4 72.86±6.49 64.04±6.2 MHIM 49.76±1.27 33.58±3.18 48.94±2.04 56.54±3.53 36.55±2.68 RRT_MIL 50.95±1.86 39.15±4.39 51.45±1.68 55.44±3.6 37.11±2.55 Patch-GCN 76.59±5.22 49.81±3.35 57.93±2.93 71.91±6.78 59.36±5.1 The present invention 87.33±6.45 79.28±6.95 82.41±6.06 86.95±5.98 85.06±5.76

[0125] Table 12. Performance comparison of the present invention and 11 models on the TP53 multi-classification task based on the TCGA-LUAD dataset.

[0126] AUCROC ACC Precision Recall F1-Score ABMIL 51.39±1.27 13.12±4.13 26.24±2.42 63.87±8.29 30.44±1.96 CLAM-SB 49.22±0.81 8.51±3.69 28.42±8.97 27.33±11.84 18.27±5.92 CLAM-MB 48.31±1.71 1.78±0.59 35.25±15.22 33.86±12.23 20.41±7.46 DSMIL 47.80±1.53 9.31±5.74 24.8±2.58 71.09±6.01 32.25±1.68 TransMIL 90.96±8.05 80.99±15.48 84.08±13.02 91.49±7.4 85.13±12.16 DTFD-MIL 61.99±3.58 12.12±2.57 30.09±2.8 61.52±5.07 36.02±1.9 [[ID= 62.78±3.8 8.33±2.31 31.31±2.75 66.09±5.61 36.59±2.33 ​ 53.95±3.09 26.18±1.81 35.27±1.22 55.34±4.3 32.47±1.46 ​ 56.81±1.07 11.73±4.48 32.75±3.23 62.24±9.04 36.04±1.63 ​ 90.95±8.05 80.2±15.29 83.49±12.88 91.29±7.35 84.84±12.09 ​ 65.29±5.08 23.03±3.8 27.31±2.64 55.78±2.89 34.37±1.86 ​ 91.21±8.74 89.6±9.08 91.08±7.84 89.55±9.07 93.07±6.2

[0127] Table 13. Performance comparison of the present invention and 11 models in the TP53 functional domain prediction task based on the LCSXH-CSU dataset.

[0128] ​ ​ ​ ​ ​ ​ 84.79±7.99 38.02±8.37 52.25±7.03 83.26±8.07 63.46±7.81 ​ 89.94±8.78 78.93±16.34 82.81±12.99 91.43±7.33 84.9±12.17 ​ 89.91±8.89 78.78±16.67 82.68±13.07 91.43±7.17 84.75±12.19 ​ 89.02±8.1 67.11±12.95 74.49±10.9 87.03±6.07 78.57±9.99 ​ 90.04±8.91 81.8±16.27 84.91±13.5 88.97±9.86 86.27±12.28 ​ 80.66±6.39 27.26±5.93 44.73±4.46 77.06±8.28 55.96±5.48 ​ 75.38±5.7 16.95±3.49 38.22±3.16 77.95±6.17 49.23±5.17 ​ 88.04±8.46 61.06±13.51 66.6±10.34 88.68±4.26 74.18±9.45 ​ 51.51±1.29 4.09±1.47 24.28±0.58 65.84±2.71 33.45±1.01 ​ 53.11±0.9 13.18±1.31 27.89±0.68 41.36±2.72 31.59±1.52 ​ 86.2±7.4 47.87±11.31 56.88±8.27 85.96±4.51 67.1±7.73 ​ 93.13±6.8 27.99±8.01 68.05±7.79 59.16±6.92 87.72±7.67

[0129] Table 14. Performance comparison of the present invention and 11 models in the EGFR functional domain prediction task based on the LCSXH-CSU dataset.

[0130] ​ ​ ​ ​ ​ ​ 67.99±3.55 41.89±4.83 71.33±8.29 56.73±3.57 57.88±4.96 ​ 56.31±2.78 27.38±1.98 56.45±2.91 49.02±1.02 41.45±1.03 ​ 59.72±2.49 23.7±4.44 53.71±3.98 50.96±5.03 41.99±2.51 ​ 65.88±3.41 35.44±2.52 66.1±5.83 59.15±1.59 54.77±4.05 ​ 89.94±8.66 82.18±11.71 85.96±9.41 88.15±7.4 85.37±10.11 ​ 74.2±4.21 53.28±0.81 61.75±1.28 69.52±4.66 56.48±2.5 ​ 72.43±3.68 47.65±3.85 54.77±2.22 67.78±2.43 52.98±1.92 ​ 70.54±3.19 33.49±2.38 53.83±3.07 81.11±6.05 61.42±4.49 ​ 50.08±0.98 23.98±4.33 39.74±2.99 54.02±1.8 38.37±0.7 ​ 53.3±1.26 34.91±5.09 49.46±2.31 59.8±3.88 42.6±2.25 ​ 79.99±5.58 56.29±5.45 60.99±4.62 71.96±4.32 58.51±4.49 ​ 90.72±7.49 77.44±11.02 86.69±8.46 86.18±8.43 88.43±8.2

[0131] Table 15. Performance comparison of the present invention and 11 models in the KRAS functional domain prediction task based on the LCSXH-CSU dataset.

[0132] ​ ​ ​ ​ ​ ​ 94.11±4.48 86.41±10.69 96.99±1.42 98.67±0.73 97.5±1 ​ 93.62±5.71 86.45±12.12 94.94±4.53 91.92±7.22 94.94±4.53 ​ 82.59±15.58 81.94±16.16 83.51±14.75 98.71±1.15 81.94±16.16 ​ 89.14±9.72 87.74±10.96 92.97±6.29 91.16±7.9 95.48±4.04 ​ 85.69±12..8 81.29±16.73 80.27±17.64 80.15±17.75 81.29±16.73 ​ 89.31±9.56 87.74±10.96 89.82±9.11 98.83±1.04 87.74±10.96 ​ 89.64±8.87 87.72±10.25 89.33±8.81 91.71±6.69 87.72±10.25 ​ 93.64±0.32 93.33±0.6 93.39±0.54 93.56±0.4 93.33±0.6 ​ 48.46±4.81 43.69±9.49 61.07±11.07 91.77±0.88 55.07±9.75 ​ 60.51±7.7 64.3±7.52 93.97±1.27 77.13±5.09 69.01±6.65 ​ 80.61±8.6 90.71±2.57 92.27±2.03 95.21±1.34 90.71±2.57 ​ 94.69±4.53 97.86±1.92 98.05±1.74 98.3±1.52 97.86±1.92

[0133] Table 16. Performance comparison of the present invention and 11 models in the ALK functional domain prediction task based on the LCSXH-CSU dataset.

[0134] ​ ​ ​ ​ ​ ​ 91.39±7.1 95±3.46 95.67±2.7 97.21±1.84 95±3.46 ​ 94.67±2.98 95±4.47 96.59±3.05 96.19±3.41 97±2.68 ​ 92.5±6.71 89±9.84 90±8.94 91.25±7.83 89±9.84 ​ 94.17±5.22 90±8.94 95±4.47 92.86±6.39 98±1.79 ​ 93.19±6.09 91±8.05 91.71±7.42 91.43±7.67 92±7.16 ​ 93.61±5.71 98±1.79 98.51±1.33 98.05±1.74 99±0.89 ​ 94.44±4.97 92±7.16 92±7.16 92±7.16 92±7.16 ​ 93.89±5.47 90±8.94 92.9±6.35 98.18±1.63 90±8.94 ​ 85.6±4.54 87.07±5.55 92.16±2.74 96.68±1.4 90.4±3.54 ​ 87.14±4.56 86.12±4.69 90.59±3.61 95.41±1.07 89.23±5.04 ​ 94.67±2.98 94±3.58 94±3.58 94±3.58 94±3.58 ​ 94.72±4.72 99±0.89 99±0.89 99±0.89 99±0.89

[0135] The NAVF-Bio model of the present invention extracts multi-view information from whole slide images (WSI) by adopting multi-scale feature fusion and adaptive cross-view knowledge supplementation modules to predict a variety of gene mutation information, including gene mutation subtypes and gene mutation exon predictions. Multi-scale feature fusion simulates the pathologist's film reading steps to perform feature extraction and fusion, and the adaptive cross-view knowledge supplementation module can flexibly incorporate multi-view features to improve the predictive performance of the model. In the task of predicting the presence or absence of mutations in the key lung cancer genes TP53, EGFR, KRAS, and ALK, NAVF-Bio achieved an AUCROC value of over 90 in a five-fold cross-validation of the data set of the Second Xiangya Hospital of Central South University (AUCROC values ​​were 92.93±6.27, 92.32±6.69, 90.6±7.74, and 92.58±7.32, respectively), which is better than the most advanced methods currently available. NAVF-Bio predicts that this task can also be generalized to the TCGA_LUAD data set. More importantly, NAVF-Bio achieved AUCROC values ​​of 90.72±7.49 and 87.72±7.49, respectively, in predicting TP53 mutation subtypes using the Second Xiangya Hospital of Central South University and TCGA_LUAD datasets. It also predicted the exons and tumor mutation burden (TMB) of TP53, EGFR, KRAS, and ALK mutations for the first time, achieving clinical-grade performance. The NAVF-Bio model addresses the technical challenges of current AI models in effectively predicting gene mutation subtypes and mutated exon locations.

[0136] The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for predicting gene mutation information from lung cancer tissue pathology images, characterized in that: The following steps are involved: S1. Collect histopathological whole-slide images and corresponding genetic testing reports of lung cancer patients; S2. Perform image preprocessing on the histopathology whole slide images to construct the pathology dataset; S3. Build a NAVF-Bio model for predicting gene mutation information from lung cancer histopathology whole slide images. The NAVF-Bio model includes a pre-training module, a pathology space topology representation learning module, a multi-scale feature fusion module, and an adaptive cross-view knowledge supplementation module. The pre-training module is used to extract features from the pre-processed images; the pathological space topology representation learning module is used to further process the extracted features to form a one-dimensional vector; The multi-scale feature fusion module is used to further extract and fuse features of different scales; the adaptive cross-view knowledge supplement module is used to incorporate features from multiple views to improve the prediction performance of the model; S4. Using the pathology dataset to train the constructed NAVF-Bio model, constraining the NAVF-Bio model based on multi-scale loss and weighted fusion loss to obtain the final NAVF-Bio model; S5. Predicting gene mutation information from lung cancer histopathology whole slide images using the final NAVF-Bio model, wherein the gene mutation information includes gene mutation, tumor mutation burden, gene mutation subtype, and protein functional domain; In the pathology spatial topology representation learning module, the KNN algorithm is used to construct points and edges in the spatial topology of patches of different scales. The HD-Yolo algorithm is used to segment and classify cells in whole-slide images and quantify tumor microenvironment indicators. The KNN algorithm is also used to construct a cellular spatial topology map of the tumor microenvironment. The nodes of the cellular spatial topology map include cell labels, classification probabilities, and cell nucleus areas, and the relationships between cells of the same type serve as edges. The pathological space topology representation learning module includes an SGAEConv module and a multi-layer perception module; The SGAEConv module uses a linear transformation to combine the node's own features and neighbor features, expressed as: Among them, W m , W n are all learnable weight matrices, is the embedding of node v at layer k, is the aggregated feature of neighbor nodes, σ is the RuLU activation function; The multi-layer perception module includes ReLU activation function, layer normalization and information dropout regularization; the ReLU activation function is used to perform nonlinear transformation on the input features; The layer normalization-based model training stabilizes and accelerates model convergence; the multi-layer perception module uses information dropout regularization to drop a portion of neurons with a certain probability in each training, thereby helping the model to train better and prevent overfitting.

2. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: In step S2, preprocessing the histopathology whole slide image includes digitizing the histopathology whole slide image, segmenting tissue regions, and detecting background and blurred regions.

3. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: In step S3, the pre-training module uses the Otsu algorithm to distinguish the background area of ​​the whole slide image, and uses a sliding window strategy to divide the whole slide image into large, medium and small scale patches, and uses the CTransPath pre-training model to extract a feature representation with a dimension of 768 for each patch.

4. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: The multi-scale feature fusion module includes an adaptive pooling layer, a splicing layer and a self-attention feature importance extractor; The adaptive pooling layer is used to dynamically adjust the length of the input sequence so that the output length matches the given target size; the splicing layer is used to perform splicing in a unified feature space; the self-attention feature importance extractor uses the attention mechanism to obtain features that are relatively important to the model based on feature importance.

5. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: The adaptive cross-view knowledge supplement module includes an adaptive feature fusion module, an adaptive weight module and a classifier; The adaptive feature fusion module uses the attention mechanism to process the input data and generate a global representation of the input features; The adaptive weight module includes two fully connected layers. The adaptive weight module is used to perform feature scaling to form adaptive weights; the classifier is used to obtain the representation of each label.

6. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: The multi-scale loss is optimized separately at different scales; the weighted fusion loss is used to integrate losses from different scales and reflect the importance of losses of different scales to the overall task in a weighted manner.

7. The method for predicting gene mutation information from lung cancer tissue pathology images according to claim 1, characterized in that: Predicted gene mutation information includes: Predict whether there is TP53 gene mutation, EGFR mutation, KRAS mutation, or ALK mutation; Predicting the level of tumor mutation burden; Predict TP53 gene mutation subtypes; Predict TP53, EGFR, KRAS, and ALK mutant exons.

Images