Medical image processing method for intelligent medical management platform

Abstract

The invention relates to the technical field of medical image processing, and particularly discloses a medical image processing method for an intelligent medical management platform, and the method comprises the following steps: collecting a medical image of a patient, importing the medical image into the intelligent medical management platform, and carrying out the preliminary verification and arrangement; removing image noise, and enhancing the image by using a histogram equalization method based on local contrast enhancement; a segmentation model based on a convolutional neural network is constructed, the marked data set is used for training, the preprocessed image is input into the model for segmentation, and post-processing is carried out; extracting feature points, performing matching by using a feature matching algorithm based on nearest neighbor search, estimating a transformation model according to a matching result, and performing registration; and storing the processed image data in a database, and establishing an index by adopting a content-based image index technology. According to the invention, the image registration accuracy and efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly relates to a medical image processing method for an intelligent medical management platform. Background Art

[0002] Intelligent healthcare is to build a regional healthcare information platform for health records, and use the most advanced Internet of Things technology to realize the interaction between patients, medical staff, medical institutions, and medical devices, and gradually achieve informatization. With the development of the medical industry, intelligent healthcare will integrate more high-tech such as artificial intelligence and sensing technology, making medical services truly intelligent and promoting the prosperous development of the medical cause.

[0003] Medical images such as X-rays, CTs, MRIs, etc. play a crucial role in disease diagnosis, treatment plan formulation, and condition monitoring. With the continuous development of medical technology and the gradual popularization of intelligent medical management platforms, it has become particularly crucial to process medical images efficiently and accurately. Existing medical image processing methods have some deficiencies; in the image registration process, for images collected in different modalities or at different times, the accuracy and efficiency of registration need to be improved, which affects the accuracy and timeliness of medical decisions. Summary of the Invention

[0004] The purpose of the present invention is to provide a medical image processing method for an intelligent medical management platform to solve the above technical problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A medical image processing method for an intelligent medical management platform, As a further solution of the present invention: it includes the following steps: Medical image acquisition and import: Collect medical images of patients through medical imaging devices, import them into the intelligent medical management platform in a standard medical image format, and perform preliminary verification and sorting; Image preprocessing: Use an adaptive filtering algorithm based on multi-scale wavelet transform to remove image noise, and use histogram equalization based on local contrast enhancement to enhance the image; Image segmentation: Build a segmentation model based on a convolutional neural network, train it using a labeled dataset, input the preprocessed image into the model for segmentation, and perform post-processing; Image registration: Use a feature extraction algorithm based on scale-invariant feature transform and local binary pattern to extract feature points, use a feature matching algorithm based on nearest neighbor search for matching, estimate the transformation model according to the matching result and perform registration; Image data management and analysis: Store the processed image data in a database, and establish an index using content-based image indexing technology.

[0006] As a further solution of the present invention: The adaptive filtering algorithm based on multi-scale wavelet transform is specifically as follows: Perform multi-scale wavelet decomposition on the image, adaptively adjust the filtering threshold according to the statistical characteristics of wavelet coefficients, suppress the wavelet coefficients where the noise is located, and reconstruct the image through inverse wavelet transform.

[0007] As a further solution of the present invention: The histogram equalization method based on local contrast enhancement is specifically as follows: Divide the image into multiple local regions, perform histogram equalization on each local region respectively, and adopt an adaptive threshold control strategy to dynamically adjust the equalization degree.

[0008] As a further solution of the present invention: The segmentation model based on convolutional neural network adopts the stochastic gradient descent algorithm to optimize parameters during training, and introduces data augmentation technology to increase the diversity of the dataset.

[0009] As a further solution of the present invention: The feature matching algorithm based on nearest neighbor search adopts the strategies of bidirectional matching and distance threshold screening to remove mis-matched feature point pairs.

[0010] As a further solution of the present invention: The content-based image indexing technology extracts and encodes the feature information of the image, establishes an index structure, and realizes the fast retrieval of image data.

[0011] As a further solution of the present invention: Input the preprocessed image into the model for segmentation, and perform post-processing. The post-processing includes: Remove small isolated regions and noise points in the segmentation result, and smooth the boundaries of objects; Fill small holes inside the objects and connect adjacent objects.

[0012] As a further solution of the present invention: It further includes an interaction function. The interaction function includes operations such as zooming, rotating, and cropping the image, which is convenient for doctors to conduct detailed observation and analysis.

[0013] Advantages of the present invention: By adopting an adaptive filtering algorithm based on multi-scale wavelet transform and a histogram equalization method based on local contrast enhancement, it can effectively remove the noise in medical images, enhance the local contrast of the images, improve the image quality, and provide a better basis for subsequent processing and analysis; the segmentation model based on deep learning can accurately segment the regions of interest in medical images, such as organs, lesions, etc. Through the training with a large amount of data and post-processing operations, the accuracy and reliability of the segmentation are improved, which helps doctors diagnose diseases more accurately; the feature extraction and matching algorithm combining SIFT and LBP features, as well as the transformation model estimated by the least squares method, can achieve accurate registration of medical images acquired in different modalities or at different times, providing strong support for the comparison and analysis of medical images. Brief Description of the Drawings

[0014] The present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 It is a schematic flowchart of a medical image processing method for an intelligent medical management platform according to the present invention. Detailed Embodiments

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Please refer to Figure 1 As shown, the present invention is a medical image processing method for an intelligent medical management platform: Step 1: Medical image acquisition and import: First, medical images of patients are acquired through various medical imaging devices, such as X-ray machines, CT scanners, MRI devices, etc. The acquired images can be image data in different modalities and at different times. These image data are imported into the intelligent medical management platform in a standard medical image format, such as DICOM format. During the import process, the image data are preliminarily verified and sorted to ensure the integrity and accuracy of the data.

[0018] In this embodiment, in the application scenario of the intelligent medical management platform in a certain hospital, a CT scanner from GE and an MRI device from Siemens are used to collect medical images of patients. The collected image data is stored in the local storage of the device in DICOM format. These image data are imported into the server of the intelligent medical management platform through a network interface. During the import process, a special DICOM parsing tool is used to verify the image data, checking whether the format, metadata information, etc. of the images are complete and accurate. For the image data that fails the verification, it is marked and relevant personnel are notified for processing.

[0019] Step 2: Image preprocessing: 1) Noise removal: An adaptive filtering algorithm based on multi-scale wavelet transform is used to remove noise from medical images. This algorithm first performs multi-scale wavelet decomposition on the image to obtain wavelet coefficients of different scales and directions. Then, according to the statistical characteristics of the wavelet coefficients, the filtering threshold is adaptively adjusted to suppress the wavelet coefficients where the noise is located, while retaining the important feature information of the image. Finally, the image is reconstructed through inverse wavelet transform to obtain the denoised image.

[0020] Taking a CT image as an example, an adaptive filtering algorithm based on multi-scale wavelet transform is used for noise removal. The specific steps are as follows: Perform three-layer wavelet decomposition on the CT image to obtain wavelet coefficients of different scales and directions; Calculate the standard deviation of the wavelet coefficients at each scale and direction, and adaptively adjust the filtering threshold according to the standard deviation; Perform threshold processing on the wavelet coefficients where the noise is located, set the coefficients smaller than the threshold to zero, and retain the coefficients larger than the threshold; Reconstruct the image through inverse wavelet transform to obtain the denoised CT image. After the denoising process, the noise in the image is significantly reduced and the clarity of the image is improved.

[0021] 2) Image enhancement: The histogram equalization method based on local contrast enhancement is used to enhance the denoised image. This method divides the image into multiple local regions, and performs histogram equalization on each local region separately to enhance the local contrast of the image. At the same time, in order to avoid image distortion caused by over-enhancement, an adaptive threshold control strategy is adopted to dynamically adjust the degree of equalization according to the statistical characteristics of the local regions.

[0022] Perform histogram equalization based on local contrast enhancement on the denoised CT image. The specific steps are as follows: Divide the image into multiple non-overlapping local regions, and the size of each local region is 32×32 pixels; Histogram equalization is performed on each local region respectively. Calculate the histogram of the local region and map the pixel values according to the histogram to enhance the local contrast. An adaptive threshold control strategy is adopted to dynamically adjust the degree of equalization according to the standard deviation of the local region. For local regions with a small standard deviation, appropriately reduce the intensity of equalization to avoid image distortion caused by over-enhancement. After image enhancement processing, the details of the image are clearer, facilitating subsequent feature extraction and analysis.

[0023] Step Three: Image Segmentation: 1) Training of a segmentation model based on deep learning: Construct a medical image segmentation model based on a convolutional neural network (CNN). Use a large number of labeled medical image datasets to train the model. The datasets include medical images of different modalities and different parts, as well as corresponding segmentation annotation information, such as organs, lesion regions, etc. During the training process, the random gradient descent algorithm is used to optimize the parameters of the model, and at the same time, data augmentation techniques, such as rotation, flipping, scaling, etc., are introduced to increase the diversity of the dataset and improve the generalization ability of the model.

[0024] In specific implementation, construct a convolutional neural network based on the U-Net architecture as the medical image segmentation model. Use a dataset containing 1000 labeled CT and MRI images to train the model. The datasets include images of organs such as the liver and kidneys and corresponding segmentation annotation information. During the training process, the random gradient descent algorithm is adopted, the learning rate is set to 0.001, the batch size is 16, and the number of training epochs is 100. At the same time, data augmentation techniques are introduced to perform random rotation, flipping, and scaling operations on the training images to increase the diversity of the dataset.

[0025] 2) Execution of image segmentation: Input the preprocessed medical image into the trained segmentation model, and the model outputs the segmentation results of each region of interest in the image. For the segmentation results, post-processing operations, such as morphological filtering, connected component analysis, etc., are performed to remove noise and small isolated regions in the segmentation results and improve the accuracy of segmentation.

[0026] For example, input the preprocessed CT image into the trained segmentation model, and the model outputs the segmentation result of the liver region. Perform post-processing operations on the segmentation result, use morphological opening to remove small isolated regions in the segmentation result, and then perform connected component analysis to retain the largest connected component as the final liver segmentation result. By comparing with the manually labeled results, the segmentation accuracy reaches over 95%, indicating that this segmentation method has high precision.

[0027] Step Four: Image Registration: 1) Feature extraction: Adopt a feature extraction algorithm based on Scale - Invariant Feature Transform (SIFT) to extract feature points with scale invariance and rotation invariance from the medical images to be registered. At the same time, in order to improve the matching accuracy of the feature points, local binary pattern (LBP) features are combined to describe the feature points.

[0028] Taking a CT image and an MRI image as an example, adopt the feature extraction algorithm based on SIFT and LBP features. The specific steps are as follows: Perform SIFT feature extraction on the CT image and the MRI image respectively to obtain feature points with scale invariance and rotation invariance; For each feature point, calculate its local LBP feature as the descriptor of the feature point. By combining SIFT and LBP features, the description ability and matching accuracy of the feature points are improved.

[0029] 2) Feature matching: Use a feature matching algorithm based on nearest - neighbor search to match the extracted feature points. During the matching process, adopt the strategies of two - way matching and distance - threshold screening to remove the mismatched feature - point pairs and improve the matching accuracy.

[0030] Taking the above - mentioned CT image and MRI image as an example, first, find the nearest - neighbor feature point in the feature points of the CT image for each feature point of the MRI image, and then find the nearest - neighbor feature point in the feature points of the MRI image for each feature point of the CT image. Only when two feature points are the nearest neighbors to each other in the two - way matching and the distance between them is less than the set threshold, is it considered an effective matching point pair. In this way, a large number of mismatched point pairs are removed, and the matching accuracy is improved.

[0031] 3) Transformation model estimation and registration: According to the matched feature - point pairs, use the least - squares method to estimate the transformation model between the images, such as affine transformation or non - linear transformation. Then, register the images according to the estimated transformation model to obtain the registered images.

[0032] According to the matched feature - point pairs, use the least - squares method to estimate the affine transformation model between the CT image and the MRI image. Then, register the MRI image according to the estimated transformation model to align it with the CT image in space. By calculating the error of the corresponding points in the images before and after registration, the mean square error of registration is less than 1 pixel, indicating that this registration method has high accuracy.

[0033] Step Five: Image data management and analysis: 1) Data storage and indexing: Store the processed medical image data in the database of the intelligent medical management platform. To improve the data retrieval efficiency, content-based image indexing technology is adopted to extract and encode the feature information of the images and establish an index structure. At the same time, the image data is associated and stored with relevant information such as the patient's basic information and diagnosis results to achieve comprehensive data management.

[0034] Store the processed medical image data in the MySQL database of the intelligent medical management platform. Adopt content-based image indexing technology to extract and encode the color features, texture features, etc. of the images and establish an index structure. At the same time, the image data is associated and stored with relevant information such as the patient's basic information and diagnosis results to establish an association table. For example, store the basic information such as the patient's name, age, and gender in the patient information table, store the information such as the image file name, acquisition time, and modality in the image information table, and associate the two tables through the patient ID. In this way, when retrieving image data, it can be quickly retrieved according to the patient's basic information or diagnosis results.

[0035] 2) Image analysis and mining: Use data mining and machine learning algorithms to analyze and mine the stored medical image data. For example, classify the image data through the clustering analysis algorithm to discover the image feature patterns of different types of diseases; through the association rule mining algorithm, find out the association relationships between image features and disease diagnosis and treatment effects to provide support for medical decision-making.

[0036] Use the clustering analysis algorithm to classify the stored liver CT image data. First, extract the texture features and shape features of the images and use these features as the input for clustering. Adopt the K-Means clustering algorithm to divide the liver CT images into three categories. Through the analysis of the clustering results, it is found that there are obvious differences in the morphology, density, etc. of the liver in different categories of images, and these differences may be related to different liver diseases. At the same time, use the association rule mining algorithm to find out the association relationships between image features and disease diagnosis. For example, it is found that there is a strong association between certain texture features in liver images and the diagnosis of liver cirrhosis, providing a reference basis for doctors' diagnosis.

[0037] Among many clustering algorithms, the K-means algorithm is favored for its simplicity and efficiency. The basic idea of the K-means algorithm is to divide the data into K different clusters through an iterative approach, minimizing the sum of the distances between each data point and the centroid (or center point, mean point) of its affiliated cluster. Specifically, the execution process of the K-means algorithm usually includes the following steps: First, randomly select K data points as the initial cluster centroids; then, based on the distances between each data point and the centroids of each cluster, assign it to the nearest cluster; next, recalculate the centroid of each cluster, that is, take the average of all data points within the cluster as the new centroid; repeat the above assignment and update steps until a certain termination condition is met (such as the cluster centroids no longer change significantly or the preset number of iterations is reached).

[0038] Step Six: Result Display and Interaction: Present the processed medical images, segmentation results, registration results, and analysis and mining results to the doctor in an intuitive way. Adopt visualization techniques, such as three-dimensional reconstruction, virtual reality, etc., to provide the doctor with clearer and more comprehensive image information. At the same time, provide an interaction function where the doctor can perform operations such as zooming, rotating, and cropping on the image for detailed observation and analysis.

[0039] Use the Matplotlib library and VTK library in Python to present the processed medical images, segmentation results, registration results, and analysis and mining results to the doctor in an intuitive way. For 2D images, use the Matplotlib library for display, and the doctor can perform operations such as zooming and panning on the image through mouse operations. For 3D images, use the VTK library for three-dimensional reconstruction and visualization, and the doctor can control the rotation and scaling of the 3D model through the keyboard and mouse to observe the image from different angles. At the same time, provide an interaction interface where the doctor can input query conditions to retrieve relevant image data and analysis results.

[0040] The Matplotlib library is a Python plotting library that can be used to create Python-based interactive plots. It can create 2D and 3D graphics and can be used for various different plotting tasks, such as signal processing and image processing. Matplotlib can also manipulate the graphics and supports multiple output modes, such as images, PDFs, SVGs, and various video formats.

[0041] The VTK library (Visual Tool Kit) is a C++ library used to implement 3D computer graphics and can be used for various different graphics tasks. The VTK library provides powerful visualization functions, such as supporting large datasets, multiple data types, and a large number of spatial algorithms and model algorithms. It also supports multiple different data input methods, such as file reading and memory mapping.

[0042] The above has described in detail an embodiment of the present invention, but the above content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention shall still fall within the scope covered by the patent of the present invention.

Claims

1. A medical image processing method for an intelligent medical management platform, characterized in that, It includes the following steps: Medical image acquisition and import: Acquire the medical images of patients through medical imaging devices, import them into the intelligent medical management platform in standard medical image formats, and conduct preliminary verification and collation; Image preprocessing: Use an adaptive filtering algorithm based on multi-scale wavelet transform to remove image noise, and use a histogram equalization method based on local contrast enhancement to enhance the image; Image segmentation: Construct a segmentation model based on a convolutional neural network, train it using a labeled dataset, input the preprocessed image into the model for segmentation, and perform post-processing; Image registration: Use a feature extraction algorithm based on scale-invariant feature transform and local binary pattern to extract feature points, use a feature matching algorithm based on nearest neighbor search for matching, estimate the transformation model according to the matching results and perform registration; Image data management and analysis: Store the processed image data in a database, and establish an index using content-based image indexing technology.

2. The medical image processing method for an intelligent medical management platform according to claim 1, wherein The adaptive filtering algorithm based on multi-scale wavelet transform is specifically as follows: Perform multi-scale wavelet decomposition on the image, adaptively adjust the filtering threshold according to the statistical characteristics of wavelet coefficients, suppress the wavelet coefficients where the noise is located, and reconstruct the image through inverse wavelet transform.

3. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, The histogram equalization method based on local contrast enhancement is specifically as follows: Divide the image into multiple local regions, perform histogram equalization on each local region respectively, and use an adaptive threshold control strategy to dynamically adjust the equalization degree.

4. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, During the training process of the segmentation model based on a convolutional neural network, a stochastic gradient descent algorithm is used to optimize the parameters, and a data augmentation technique is introduced to increase the diversity of the dataset.

5. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, The feature matching algorithm based on nearest neighbor search adopts a strategy of two-way matching and distance threshold screening to remove mis-matched feature point pairs.

6. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, The content-based image indexing technology extracts and encodes the feature information of the image, establishes an index structure, and realizes the fast retrieval of image data.

7. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, Input the preprocessed image into the model for segmentation, and perform post-processing. The post-processing includes: Remove small isolated regions and noise points in the segmentation result, and smooth the boundaries of objects; Fill small holes inside the objects and connect adjacent objects.

8. A medical image processing method for an intelligent medical management platform according to claim 1, characterized in that, It also includes an interaction function, and the interaction function includes operations such as zooming, rotating, and cropping the image, which is convenient for doctors to conduct detailed observation and analysis.

Images