CN115662635A - Method for establishing combined regression prediction model for radiation pneumonitis - Google Patents
Method for establishing combined regression prediction model for radiation pneumonitis Download PDFInfo
- Publication number
- CN115662635A CN115662635A CN202211260800.8A CN202211260800A CN115662635A CN 115662635 A CN115662635 A CN 115662635A CN 202211260800 A CN202211260800 A CN 202211260800A CN 115662635 A CN115662635 A CN 115662635A
- Authority
- CN
- China
- Prior art keywords
- radiation pneumonitis
- model
- image
- network
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a method for establishing a combined regression prediction model for radiation pneumonitis, which is characterized in that a multi-feature fusion regression prediction model for machine learning and deep learning combined prediction is established, methods such as image omics features, expert models, transfer learning and the like are introduced on the basis of realizing classification of radiation pneumonitis by deep learning, and the prediction model is established through a machine learning algorithm, so that different features are fused to jointly predict prognosis prediction of the radiation pneumonitis. The invention adopts a method combining theoretical analysis, multiple regression, neural network modeling and experimental verification, closely associates the prediction research and the actual diagnosis of the radiation pneumonitis, establishes rule constraint on the diagnosis experience and excavates implicit rules on historical data, and mutually verifies and promotes the prediction result and the actual diagnosis.
Description
Technical Field
The invention belongs to the technical field of radiation pneumonitis prediction, and particularly relates to a method for establishing a combined regression prediction model for radiation pneumonitis.
Background
There are two main directions in the current phase of research on the prediction of radiation pneumonitis: extracting the imaging characteristics in the lung radiological image, and establishing a classification model by using a machine learning algorithm to predict prognosis; and (3) learning and training the image data by using a deep learning algorithm, and classifying and predicting according to a training result.
(1) Machine learning-based prediction study of radiation pneumonitis
Imaging plays a key role in the diagnosis of lung cancer, breast cancer and other malignant tumors in the breast, including early diagnosis, efficacy monitoring and prognosis evaluation, which are indistinguishable from medical imaging. With the development of computer science, many computer-aided analysis techniques have appeared in recent years, and the maturity of related techniques has also gradually fused computer-aided analysis and medical diagnosis, wherein imaging group is the hot spot of current research. The imaging group means high-throughput extraction of quantitative image characteristics, and medical images are converted into data resources which can be further mined, and the correlation between imaging and diseases is explored through a specific method and a specific program. The rapid development of the artificial intelligence technology further promotes the application of the imaging omics in the actual clinical diagnosis. The main mode of combining artificial intelligence and image omics is to deeply mine the internal relation between image features extracted by the image omics and diseases through a machine learning method, screen effective and reliable features from the extracted high-dimensional features, eliminate redundant features and construct a machine learning model for detecting different disease features. In order to establish the model, many students make many attempts in the direction, zhang Zhen and the like screen radiologic characteristics related to the occurrence of the radiation pneumonitis on the basis of breast positioning CT images of lung cancer patients, construct a machine learning model and discuss the value of the radiologic in predicting the occurrence of the radiation pneumonitis; guo Meiying, etc., which combines the image omics with the artificial intelligence method, and screens and constructs image markers for solving different medical problems by using the artificial intelligence method while deeply mining the image characteristics; chen Wentao and the like illustrate the mechanism and clinical image expression of the radiation pneumonitis, grading and relevant factor analysis, and then introduce several conventional machine learning algorithm principles to construct and predict a radiation pneumonitis model through an image omics and machine learning method. In addition, the hypothesis of imaging group is that the change of microscopic gene or protein pattern is expressed on macroscopic image, so that the potential molecular biological mechanism reflected by the image marker can be further mined by artificial intelligence method.
(2) Deep learning-based prediction study of radiation pneumonitis
Deep learning belongs to an important branch of machine learning, and the deep learning learns the intrinsic rules and the representation levels of sample data by combining low-level features to form more abstract high-level representation attribute categories or features, so that a computer has human-like thinking and analysis capability, and identifies things and even predicts the development forms of the things. The most important basis in the field of deep learning is the neural network, which is a computer model abstracted by simulating the operation mode of the human cranial nerve system. The neural network is formed into a network structure model by connecting neurons, and each neuron has a single output corresponding to a weight coefficient which is connected with other similar neurons and is used as an input of the neuron. The neural network obtains information through storage and processing, and the neural network generates a corresponding model structure after training. The neural network can mine hidden relations among data, the comprehensive analysis and processing capacity of the data is superior to that of a traditional statistical method, the neural network has a great application value in medicine, and the neural network can be used for assisting doctors in diagnosing diseases, predicting prognosis and the like. At present, the research in the deep learning direction is still in a starting stage, and old and other people respectively construct a model for predicting the radiation pneumonitis by using a neural network, judge the discrimination capability of the model and screen the advantage characteristics of the model; wu Yunfeng and the like provide a lung CT classification model based on improved addition-ResNet and a semantic segmentation model based on an improved U-Net full convolution network image for classification detection of radiation pneumonitis; wang B and the like establish a novel coronavirus pneumonia model according to CT image data of 723 positive pneumonia and 413 negative pneumonia. All the above studies have the problems of small data volume, shallow research and study depth in the deep learning direction, and the like, so the problem of the prediction of the radiation pneumonitis is further studied by combining a deep learning algorithm.
The primary features and derived features extracted from the medical CT image are key data referred to for predicting the radiation pneumonitis, and the key for accurately predicting the radiation pneumonitis is how to extract the primary CT image omics features and mine various derived image features. The characteristics of the single native CT image omics can well fit the data curve of the image omics, but cannot be combined with the disease development of a patient area, and the real objective law is difficult to fit under the condition of fewer data sets. The deep learning algorithm can realize accurate detection of the target through a multi-size characteristic channel, but the highly abstract characteristics lose local detail information, attention is paid to class attribution degree of the target, classification capability is strengthened, and regression prediction capability is weak. Therefore, how to establish a fusion strategy on the imaging omics characteristics and the deep learning abstract characteristics of the CT image and establish a combined regression prediction model by a machine learning method is a core problem for realizing accurate prediction of the radiation pneumonitis.
In order to solve the above problems, a method for establishing a combined regression prediction model for radiation pneumonitis is proposed.
Disclosure of Invention
In order to solve the technical problems, the invention designs a method for establishing a combined regression prediction model for the radiation pneumonitis, the invention establishes a multi-feature fusion regression prediction model for the combined prediction of machine learning and deep learning, introduces methods such as image omics features, expert models, transfer learning and the like on the basis of realizing classification of the radiation pneumonitis by the deep learning, establishes a prediction model by a machine learning algorithm, fuses different features and jointly predicts the prognosis prediction of the radiation pneumonitis.
In order to achieve the technical effects, the invention is realized by the following technical scheme: a method for establishing a combined regression prediction model for radiation pneumonitis comprises the following steps:
s1, guiding the extraction of the characteristics of an image omics by a radiation pneumonitis image omics mapping theory, cleaning and preprocessing original CT image data, and extracting high-relevance characteristics;
s2, establishing a classification algorithm based on a deep learning neural network model, modifying the network model according to the mapping property of a 3D image of the radiation pneumonitis CT in a 2D image, strengthening the retention of the network on the feature integrity, improving the detection capability of the network on a small target, establishing a transfer learning training flow, training the network by using diseases similar to the characteristics of the novel coronary pneumonia, obstructive pneumonia and the like on the CT image, and realizing the network model with high precision, high stability and high generalization capability;
s3, establishing a machine learning-based radiation pneumonitis prediction model: the method comprises the steps of constructing a decision tree type expert model according to the experience of professional doctors, selecting rule experience according to the principle of 'information entropy', constructing an elastic network regression algorithm according to the score, the image omics characteristics and the deep learning output type confidence coefficient of the expert model, continuously optimizing the regression algorithm according to historical data, and finally achieving prediction of the radiation pneumonitis.
Further, the S2 specifically includes: firstly, preprocessing a medical CT image of radiation pneumonitis and a medical CT image of similar lung diseases, reducing the dimension of 3D medical CT image data, and unifying the data into a neural network which can be used for carrying out feature extraction files; secondly, modifying a network model of a YOLOX algorithm, modifying the convolutional layer, the normalization layer, the upper sampling layer, the lower sampling layer and the classification head based on the characteristic expression property of the radiation pneumonitis, and continuously debugging and verifying the network performance; secondly, performing transfer learning on the modified neural network, using the modified network model to pre-train a similar lung disease data set, then transferring the weight of the pre-trained model into the model to adjust the radiation pneumonitis data set, and continuously adjusting a loss function solving strategy to achieve an ideal fitting state; and finally, carrying out classification detection on the untrained CT image, verifying the performance of the network model, and saving the classification result and the confidence as output.
Further, the S3 specifically includes: firstly, carrying out deep communication with professional doctors engaged in the radiation pneumonitis diagnosis industry, recording reliable experience in the aspect of radiation pneumonitis diagnosis, and establishing an expert model by utilizing a decision tree algorithm; secondly, the information of classification detection of image omics and deep learning is used as the input of an elastic regression network, and the network is trained; and finally, checking the prediction accuracy of the model on the radiation pneumonitis by using a K-turn verification method, and giving the sequence of the correlation degree of different characteristics and the radiation pneumonitis.
The invention has the beneficial effects that: the method establishes a multi-feature fusion regression prediction model of machine learning and deep learning combined prediction, introduces methods such as image omics features, expert models, transfer learning and the like on the basis of realizing classification of the radiation pneumonitis by deep learning, establishes a prediction model by a machine learning algorithm, fuses different features and jointly predicts prognosis prediction of the radiation pneumonitis; by adopting a method combining theoretical analysis, multiple regression, neural network modeling and experimental verification, the method closely associates the prediction research and the actual diagnosis of the radiation pneumonitis, establishes rule constraint on diagnosis experience and excavates implicit rules on historical data, so that the prediction result and the actual diagnosis are mutually verified and mutually promoted.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of the present invention;
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
As shown in fig. 1, a method for building a joint regression prediction model for radiation pneumonitis includes the following steps:
s1, guiding image omics feature extraction by a radioactive pneumonia image omics mapping theory, cleaning and preprocessing original CT image data, and extracting high-relevance features;
s2, establishing a classification algorithm based on a deep learning neural network model, modifying the network model according to the mapping property of a 3D image of the radiation pneumonitis CT in a 2D image, strengthening the retention of the network on the feature integrity, improving the detection capability of the network on a small target, establishing a transfer learning training flow, training the network by using diseases similar to the characteristics of the novel coronary pneumonia, obstructive pneumonia and the like on the CT image, and realizing the network model with high precision, high stability and high generalization capability;
s3, establishing a machine learning-based radiation pneumonitis prediction model: the method comprises the steps of constructing a decision tree type expert model according to the experience of professional doctors, selecting rule experience according to the principle of 'information entropy', constructing an elastic network regression algorithm according to the score, the image omics characteristics and the deep learning output type confidence coefficient of the expert model, continuously optimizing the regression algorithm according to historical data, and finally achieving prediction of the radiation pneumonitis.
Example 2
As shown in fig. 1, the specific contents of S2 are: firstly, preprocessing a medical CT image of radiation pneumonitis and a medical CT image of similar lung diseases, reducing the dimension of 3D medical CT image data, and unifying the data into a neural network which can be used for carrying out feature extraction files;
secondly, modifying a network model of a YOLOX algorithm, modifying the convolutional layer, the normalization layer, the upper sampling layer, the lower sampling layer and the classification head based on the characteristic expression property of the radiation pneumonitis, and continuously debugging and verifying the network performance;
secondly, performing transfer learning on the modified neural network, using the modified network model to pre-train a similar lung disease data set, then transferring the weight of the pre-trained model into the model to adjust the radiation pneumonitis data set, and continuously adjusting a loss function solving strategy to achieve an ideal fitting state;
and finally, carrying out classification detection on the untrained CT image, verifying the performance of the network model, and saving the classification result and the confidence as output.
Example 3
As shown in fig. 1, the specific contents of S3 are: firstly, carrying out deep communication with professional doctors engaged in the radiation pneumonitis diagnosis industry, recording reliable experience in the aspect of radiation pneumonitis diagnosis, and establishing an expert model by utilizing a decision tree algorithm;
secondly, the information of classification detection of image omics and deep learning is used as the input of an elastic regression network, and the network is trained;
and finally, checking the prediction accuracy of the model on the radiation pneumonitis by using a K-turn verification method, and giving the sequence of the correlation degree of different characteristics and the radiation pneumonitis.
In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand the invention for and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (3)
1. A method for establishing a combined regression prediction model for radiation pneumonitis is characterized by comprising the following steps:
s1, guiding image omics feature extraction by a radioactive pneumonia image omics mapping theory, cleaning and preprocessing original CT image data, and extracting high-relevance features;
s2, establishing a classification algorithm based on a deep learning neural network model, modifying the network model according to the mapping property of a 3D image of the radiation pneumonitis CT in a 2D image, strengthening the retention of the network on the feature integrity, improving the detection capability of the network on a small target, establishing a transfer learning training flow, training the network by using diseases similar to the characteristics of the novel coronary pneumonia, obstructive pneumonia and the like on the CT image, and realizing the network model with high precision, high stability and high generalization capability;
s3, establishing a machine learning-based radiation pneumonitis prediction model: the method comprises the steps of constructing a decision tree type expert model according to the experience of professional doctors, selecting rule experience according to the principle of 'information entropy', constructing an elastic network regression algorithm according to the score, the image omics characteristics and the deep learning output type confidence coefficient of the expert model, continuously optimizing the regression algorithm according to historical data, and finally achieving prediction of the radiation pneumonitis.
2. The method of claim 1, wherein the method comprises the steps of: the S2 specifically comprises: firstly, preprocessing a medical CT image of radiation pneumonitis and a medical CT image of similar lung diseases, reducing the dimension of 3D medical CT image data, and unifying the data into a neural network for feature extraction; secondly, modifying a network model of a YOLOX algorithm, modifying the convolutional layer, the normalization layer, the upper sampling layer, the lower sampling layer and the classification head based on the characteristic expression property of the radiation pneumonitis, and continuously debugging and verifying the network performance; secondly, performing transfer learning on the modified neural network, using the modified network model to pre-train a similar lung disease data set, then transferring the weight of the pre-trained model into the model to adjust the radiation pneumonitis data set, and continuously adjusting a loss function solving strategy to achieve an ideal fitting state; and finally, carrying out classification detection on the untrained CT image, verifying the performance of the network model, and saving the classification result and the confidence as output.
3. The method of claim 1, wherein the method comprises the steps of: the S3 specifically comprises: firstly, carrying out deep communication with professional doctors engaged in the radiation pneumonitis diagnosis industry, recording reliable experience in the aspect of radiation pneumonitis diagnosis, and establishing an expert model by utilizing a decision tree algorithm; secondly, the information of classification detection of image omics and deep learning is used as the input of an elastic regression network, and the network is trained; and finally, checking the prediction accuracy of the model on the radiation pneumonitis by using a K-turn verification method, and giving the sequence of the correlation degree of different characteristics and the radiation pneumonitis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211260800.8A CN115662635A (en) | 2022-10-14 | 2022-10-14 | Method for establishing combined regression prediction model for radiation pneumonitis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211260800.8A CN115662635A (en) | 2022-10-14 | 2022-10-14 | Method for establishing combined regression prediction model for radiation pneumonitis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115662635A true CN115662635A (en) | 2023-01-31 |
Family
ID=84987212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211260800.8A Pending CN115662635A (en) | 2022-10-14 | 2022-10-14 | Method for establishing combined regression prediction model for radiation pneumonitis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115662635A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309368A (en) * | 2023-02-21 | 2023-06-23 | 北京透彻未来科技有限公司 | Lung cancer pathological diagnosis system based on deep migration learning |
-
2022
- 2022-10-14 CN CN202211260800.8A patent/CN115662635A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309368A (en) * | 2023-02-21 | 2023-06-23 | 北京透彻未来科技有限公司 | Lung cancer pathological diagnosis system based on deep migration learning |
| CN116309368B (en) * | 2023-02-21 | 2023-11-14 | 北京透彻未来科技有限公司 | Lung cancer pathological diagnosis system based on deep migration learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fuhrman et al. | A review of explainable and interpretable AI with applications in COVID‐19 imaging | |
| CN111223553B (en) | Two-stage deep transfer learning traditional Chinese medicine tongue diagnosis model | |
| CN113205504B (en) | Artificial intelligence kidney tumor prediction system based on knowledge graph | |
| CN111767952B (en) | Interpretable lung nodule benign and malignant classification method | |
| CN112381178B (en) | Medical image classification method based on multi-loss feature learning | |
| WO2024065987A1 (en) | Lung cancer prognosis prediction system based on multi-omics of radiomics, pathomics and genomics | |
| CN109994151A (en) | Predictive genes system is driven based on the tumour of complex network and machine learning method | |
| CN113744873A (en) | Heating to-be-checked auxiliary differential diagnosis system based on task decomposition strategy | |
| CN113113130A (en) | Tumor individualized diagnosis and treatment scheme recommendation method | |
| US20230230241A1 (en) | System and method for detecting lung abnormalities | |
| Yu et al. | A diagnostic prediction framework on auxiliary medical system for breast cancer in developing countries | |
| Hou et al. | Multi-label learning with visual-semantic embedded knowledge graph for diagnosis of radiology imaging | |
| Feng et al. | Deep learning for chest radiology: a review | |
| CN115662635A (en) | Method for establishing combined regression prediction model for radiation pneumonitis | |
| CN116883768A (en) | Lung nodule intelligent grading method and system based on multi-modal feature fusion | |
| Meng et al. | Clinical applications of graph neural networks in computational histopathology: A review | |
| Tian et al. | Radiomics and its clinical application: artificial intelligence and medical big data | |
| Zhang et al. | Trustworthy learning with (un) sure annotation for lung nodule diagnosis with CT | |
| Hu et al. | A Hybrid Convolutional Neural Network Model Based on Different Evolution for Medical Image Classification. | |
| Sudharson et al. | Enhancing the Efficiency of Lung Disease Prediction using CatBoost and Expectation Maximization Algorithms | |
| JP7365747B1 (en) | Disease treatment process abnormality identification system based on hierarchical neural network | |
| CN113342973A (en) | Diagnosis method of auxiliary diagnosis model based on disease two-classifier | |
| CN117174238A (en) | Automatic pathology report generation method based on artificial intelligence | |
| Estépar | Artificial intelligence in COPD: New venues to study a complex disease | |
| Al-qaness et al. | Chest X-ray Images for Lung Disease Detection Using Deep Learning Techniques: A Comprehensive Survey |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |