Colorectal tumor diagnosis auxiliary system based on artificial intelligence

Abstract

The invention discloses a colorectal tumor diagnosis auxiliary system based on artificial intelligence, and relates to the technical field of tumor diagnosis, and the colorectal tumor diagnosis auxiliary system comprises a medical image preprocessing module, a feature extraction and optimization module and a deep learning diagnosis model module; wherein the medical image preprocessing module comprises an image acquisition unit and an image processing unit; the image acquisition unit is used for acquiring medical images of all human body parts with tumor diagnosis results; the image processing unit is used for processing the acquired medical image; the feature extraction and optimization module comprises a feature extraction unit and a feature selection and optimization unit; the feature extraction unit uses a deep convolutional neural network to extract depth features of the image; the deep learning diagnosis model module adopts a hybrid model combining a convolutional neural network and a long-short term memory network to carry out auxiliary diagnosis on colorectal tumors; the technical effects that the time for interpreting the medical image is shortened, and meanwhile the accuracy of interpreting the medical image is improved are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor diagnosis, and in particular to an artificial intelligence-based colorectal tumor diagnosis auxiliary system. Background Art

[0002] Globally, the incidence of colorectal cancer has increased year by year, becoming one of the major diseases threatening human health. At present, the diagnosis of colorectal cancer mainly relies on medical imaging examinations, especially colonoscopy, computed tomography (CT), magnetic resonance imaging (MRI) and other means. These examinations can provide detailed images of the internal structure of the intestine to help doctors detect the presence of tumors. However, existing diagnostic technologies face the following major problems:

[0003] Traditional imaging diagnosis relies on the experience and subjective judgment of radiologists or clinicians. Since doctors’ professional level, experience and interpretation of images may vary, this leads to inconsistent diagnostic results. In addition, manual interpretation of medical images takes a long time, and doctors are prone to visual fatigue when processing large amounts of image data, all of which may affect the accuracy of the diagnosis. Summary of the invention

[0004] The present application provides an artificial intelligence-based colorectal tumor diagnosis assistance system to solve the technical problem in the prior art that manual interpretation of medical images is time-consuming and has low accuracy; it achieves the technical effect of shortening the time for interpreting medical images and improving the accuracy of medical image interpretation.

[0005] The present application provides an artificial intelligence-based colorectal tumor diagnosis auxiliary system, including a medical image preprocessing module, a feature extraction and optimization module, and a deep learning diagnosis model module;

[0006] The medical image preprocessing module includes an image acquisition unit and an image processing unit; the image acquisition unit is used to acquire medical images of various human body parts with tumor diagnosis results; the image processing unit is used to process the acquired medical images;

[0007] The feature extraction and optimization module includes a feature extraction unit and a feature selection and optimization unit; the feature extraction unit uses a deep convolutional neural network to extract deep features of the image;

[0008] The feature selection and optimization unit uses a feature selection method based on a gradient boosting decision tree to select the features that have the greatest impact on the diagnosis results, and uses PCA to reduce the dimension of the features to reduce the computational complexity. The formula is:

[0009]

[0010] Among them, μ represents the mean of all sample feature vectors; X i Represents the i-th sample in the data set;

[0011] The deep learning diagnostic model module uses a hybrid model that combines convolutional neural networks and long short-term memory networks for auxiliary diagnosis of colorectal tumors.

[0012] Furthermore, the image acquisition unit acquires medical images in a variety of image formats, and the image formats of the medical images that can be acquired include at least JPEG and DICOM.

[0013] Furthermore, the processing operation of the image processing unit includes:

[0014] Image preprocessing, unifying image format and size, and providing standard input for subsequent processing;

[0015] De-noising: using wavelet transform denoising algorithm to reduce image noise;

[0016] Contrast enhancement, using histogram equalization technology to improve image contrast and make tumor features clearer;

[0017] Image segmentation uses a deep learning model based on the U-Net architecture for accurate segmentation. The segmentation formula is:

[0018]

[0019] Among them, L is the loss function, P is the predicted segmentation result, G is the true segmentation label, and D ice is the Dice coefficient.

[0020] Furthermore, the steps of the deep learning diagnostic model module for auxiliary diagnosis of colorectal tumors include:

[0021] Build a hybrid neural network model and input the image features extracted by CNN into the LSTM network;

[0022] Model training: using a large amount of colorectal tumor image data with diagnostic labels to train the hybrid neural network model;

[0023] Model verification and optimization: evaluate the performance of the trained model through the verification set and calculate the accuracy, recall, and F1 score indicators;

[0024] Model deployment: deploy the trained model to the user interaction module to provide clinicians with auxiliary diagnosis services for colorectal tumors and generate diagnostic results.

[0025] Furthermore, the diagnosis result includes at least information about the existence, type, location, and size of the tumor.

[0026] Furthermore, the model training uses a cross entropy loss function to optimize model parameters, and the formula is:

[0027] L=-∑(y·log(p)+(1-y)·log(1-p))

[0028] Among them, L is the loss function, y is the true label, and p is the predicted probability.

[0029] Furthermore, it also includes a user interaction module;

[0030] The user interaction module is integrated with an operating system, and the user interaction module includes a display screen and a mouse;

[0031] The user interaction module provides at least image uploading, diagnosis result viewing, report generation, and user feedback functions.

[0032] Furthermore, it also includes a database and analysis and evaluation module;

[0033] The database stores case data and expert experience data;

[0034] The analysis and evaluation module can call the case data within the database. After the patient's basic information and clinical data are uploaded to the database, the analysis and evaluation module retrieves the database to match the reference case data and generates treatment standard recommendations based on the reference case data.

[0035] Furthermore, the case data at least includes the patient's clinical information and follow-up information.

[0036] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0037] The image preprocessing module performs standardization, denoising and contrast enhancement operations to provide high-quality images for subsequent processing; the feature extraction and optimization module uses deep CNN to extract image features, and uses GBDT and PCA to select features and reduce dimensions; the deep learning diagnosis model module combines CNN and LSTM networks to achieve sequence learning and classification of image features. The user interaction module, database and analysis and evaluation module constitute the system's support framework, providing data storage, interactive interface and case matching analysis, ensuring the system's practicality and clinical applicability, and solving the technical problems of time-consuming and low accuracy in manual interpretation of medical images in existing technologies; achieving the technical effect of shortening the time for interpreting medical images and improving the accuracy of medical image interpretation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the structure of the colorectal tumor diagnosis auxiliary system based on artificial intelligence of the present invention;

[0039] Figure 2 A flowchart of the processing operation of the image processing unit of the artificial intelligence-based colorectal tumor diagnosis auxiliary system of the present invention;

[0040] Figure 3 The present invention is a flowchart of the deep learning diagnosis model module of the artificial intelligence-based colorectal tumor diagnosis assistance system for colorectal tumor auxiliary diagnosis. DETAILED DESCRIPTION

[0041] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but 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 enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0042] It should be noted that the terms “vertical”, “horizontal”, “up”, “down”, “left”, “right” and similar expressions used in this document are only for illustrative purposes and do not represent the only implementation method.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used 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 related listed items.

[0044] Example: Figure 1 , Figure 2 and Figure 3 As shown, the artificial intelligence-based colorectal tumor diagnosis assistance system of the present application includes a medical image preprocessing module, a feature extraction and optimization module, a deep learning diagnosis model module, a user interaction module, a database and an analysis and evaluation module.

[0045] The medical image preprocessing module includes an image acquisition unit and an image processing unit; the image acquisition unit is used to acquire medical images of various parts of the human body with tumor diagnosis results; the image processing unit is used to process the acquired medical images, and the processing operations of the image processing unit include:

[0046] 1. Image preprocessing, unifying image format and size, and providing standard input for subsequent processing;

[0047] 2. De-noising: using wavelet transform denoising algorithm to reduce image noise;

[0048] 3. Contrast enhancement: using histogram equalization technology to improve image contrast and make tumor features clearer;

[0049] 4. Image segmentation: Use a deep learning model based on the U-Net architecture for accurate segmentation. The segmentation formula is:

[0050]

[0051] Among them, L is the loss function, P is the predicted segmentation result, G is the true segmentation label, and D ice is the Dice coefficient.

[0052] It should be noted that the image acquisition unit can acquire medical images in a variety of image formats, such as JPEG, DICOM and other image formats.

[0053] Furthermore, the feature extraction and optimization module includes a feature extraction unit and a feature selection and optimization unit; the feature extraction unit uses a deep convolutional neural network (CNN) to extract deep features of the image, including a convolutional layer, a pooling layer, and a fully connected layer to capture key information such as the shape, texture, and edge of the tumor;

[0054] Among them, the network structure of the convolutional neural network (CNN) includes 3 convolutional layers, 2 pooling layers and 2 fully connected layers.

[0055] The feature selection and optimization unit uses a feature selection method based on gradient boosted decision tree (GBDT) to select the features that have the greatest impact on the diagnosis results;

[0056] Use PCA to reduce feature dimensionality and reduce computational complexity. The formula is:

[0057]

[0058] Among them, μ represents the mean of all sample feature vectors; X i Represents the i-th sample in the dataset.

[0059] The features are standardized and normalized to eliminate the dimensional differences and numerical range differences between different features.

[0060] Furthermore, based on the feature extraction and optimization module, the deep learning diagnosis model module uses a hybrid model combining convolutional neural network (CNN) and long short-term memory network (LSTM) for auxiliary diagnosis of colorectal tumors. The specific steps are as follows:

[0061] 1. Build a hybrid neural network model: Input the image features extracted by CNN into the LSTM network to make full use of the advantages of CNN in image feature extraction and the advantages of LSTM in sequence data learning;

[0062] 2. Model training: A large amount of colorectal tumor image data with diagnostic labels is used to train the hybrid neural network model. The cross entropy loss function is used to optimize the model parameters. The formula is:

[0063] L=-∑(y·log(p)+(1-y)·log(1-p))

[0064] Among them, L is the loss function, y is the true label, and p is the predicted probability.

[0065] 3. Model verification and optimization: Evaluate the performance of the trained model through the verification set, and calculate indicators such as accuracy, recall, and F1 score; adjust the model parameters based on the evaluation results to improve the diagnostic performance;

[0066] 4. Model deployment: Deploy the trained model to the user interaction module to provide clinicians with auxiliary diagnosis services for colorectal tumors and generate diagnostic results.

[0067] Among them, accuracy, recall, F1 score and AUC are used as performance indicators, and the target accuracy is set to at least 95%.

[0068] It should be noted that the diagnosis results include key information such as the presence, type, location, and size of the tumor; the diagnosis results are displayed on the user interface, usually in the form of image tags, text descriptions, or charts.

[0069] Furthermore, an operating system is integrated into the user interaction module, and the user interaction module includes a display screen and a mouse.

[0070] It should be noted that the user interaction module provides functions such as image uploading, diagnosis result viewing, report generation, and user feedback.

[0071] Furthermore, a large amount of case data and expert experience data are stored in the database, wherein the case data at least includes the patient's clinical information and follow-up information.

[0072] The database uses MySQL database, and the design includes patient basic information table, image data table, diagnosis result table, etc.

[0073] The analysis and evaluation module can call the case data within the database. After the patient's basic information and clinical data are uploaded to the database, the analysis and evaluation module retrieves the database to match the reference case data and generates treatment standard recommendations based on the reference case data.

[0074] Among them, the patient's basic information includes at least name, age, and gender; the clinical data includes at least the generated diagnosis results.

[0075] It should be noted that the expert experience is encoded as a rule set and combined with the system suggestions through a logical reasoning engine.

[0076] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0077] 1. Through deep learning technology and optimized feature selection methods, the target accuracy can be improved, which is of great significance for clinical diagnosis and helps to reduce misdiagnosis and missed diagnosis;

[0078] 2. The medical image preprocessing module can effectively process medical images in various formats and improve image quality through technologies such as denoising, contrast enhancement and precise segmentation, making tumor features clearer and facilitating subsequent feature extraction and diagnosis;

[0079] 3. Use deep convolutional neural networks (CNNs) to extract deep features of images, which can capture key information such as the shape, texture, and edges of tumors, which are crucial for tumor diagnosis;

[0080] 4. Feature selection and dimensionality reduction through gradient boosting decision tree (GBDT) and PCA not only improve the accuracy of diagnosis, but also reduce the computational complexity and improve the operating efficiency of the system;

[0081] 5. The hybrid model combining CNN and LSTM can fully utilize the advantages of CNN in image feature extraction and LSTM in sequence data learning to improve the accuracy and robustness of diagnosis;

[0082] 6. The user interaction module provides a convenient operation interface, allowing clinicians to easily upload images, view diagnostic results, generate reports, etc., improving user experience;

[0083] 7. The system uses MySQL database to store a large amount of case data and expert experience data, which facilitates data retrieval and analysis and provides reference for clinical decision-making;

[0084] 8. The analysis and evaluation module can match the reference case data based on the patient's basic information and clinical data, and provide doctors with treatment standard suggestions, which helps to improve the efficiency of diagnosis and treatment;

[0085] 9. Encode expert experience into a set of rules and combine it with system recommendations through a logical reasoning engine, so that the system can simulate the expert's diagnosis and treatment ideas to a certain extent and improve the reliability of diagnosis.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An artificial intelligence-based colorectal tumor diagnosis auxiliary system, characterized in that: It includes medical image preprocessing module, feature extraction and optimization module and deep learning diagnosis model module; The medical image preprocessing module includes an image acquisition unit and an image processing unit; the image acquisition unit is used to acquire medical images of various human body parts with tumor diagnosis results; the image processing unit is used to process the acquired medical images; The feature extraction and optimization module includes a feature extraction unit and a feature selection and optimization unit; the feature extraction unit uses a deep convolutional neural network to extract deep features of the image; The feature selection and optimization unit uses a feature selection method based on a gradient boosting decision tree to select the features that have the greatest impact on the diagnosis results, and uses PCA to reduce the dimension of the features to reduce the computational complexity. The formula is: Among them, μ represents the mean of all sample feature vectors; X i Represents the i-th sample in the data set; The deep learning diagnostic model module uses a hybrid model that combines convolutional neural networks and long short-term memory networks for auxiliary diagnosis of colorectal tumors.

2. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 1, characterized in that: The image acquisition unit acquires medical images in a variety of image formats, and the image formats of the medical images that can be acquired include at least JPEG and DICOM.

3. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 1, characterized in that: The processing operation of the image processing unit includes: Image preprocessing, unifying image format and size, and providing standard input for subsequent processing; De-noising: using wavelet transform denoising algorithm to reduce image noise; Contrast enhancement, using histogram equalization technology to improve image contrast and make tumor features clearer; Image segmentation uses a deep learning model based on the U-Net architecture for accurate segmentation. The segmentation formula is: Among them, L is the loss function, P is the predicted segmentation result, G is the true segmentation label, and D ice is the Dice coefficient.

4. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 3, characterized in that: The steps of the deep learning diagnosis model module for auxiliary diagnosis of colorectal tumors include: Build a hybrid neural network model and input the image features extracted by CNN into the LSTM network; Model training: using a large amount of colorectal tumor image data with diagnostic labels to train the hybrid neural network model; model validation and optimization: using the validation set to evaluate the performance of the trained model and calculate the accuracy, recall, and F1 score indicators; Model deployment: deploy the trained model to the user interaction module to provide clinicians with auxiliary diagnosis services for colorectal tumors and generate diagnostic results.

5. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 4, characterized in that: The diagnosis result includes at least information on the existence, type, location, and size of the tumor.

6. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 4, characterized in that: The model training uses the cross entropy loss function to optimize the model parameters, and the formula is: L=-∑(y·log(p)+(1-y)·log(1-p)) Among them, L is the loss function, y is the true label, and p is the predicted probability.

7. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 1, characterized in that: Also includes a user interaction module; The user interaction module is integrated with an operating system, and the user interaction module includes a display screen and a mouse; The user interaction module provides at least image uploading, diagnosis result viewing, report generation, and user feedback functions.

8. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 1, characterized in that: Also includes database and analytical evaluation modules; The database stores case data and expert experience data; The analysis and evaluation module can call the case data within the database. After the patient's basic information and clinical data are uploaded to the database, the analysis and evaluation module retrieves the database to match the reference case data and generates treatment standard recommendations based on the reference case data.

9. The artificial intelligence-based colorectal tumor diagnosis auxiliary system according to claim 8, characterized in that: The case data at least includes the patient's clinical information and follow-up information.

Images