A blockchain-based medical image storage system

Abstract

The present invention discloses a blockchain-based medical image storage system, belonging to the technical field of medical image scheduling. The system includes: an image collection subsystem for collecting medical image data and labeling the data to obtain labeled image data; an attribute analysis subsystem for analyzing the attributes of the labeled image data to obtain image attribute data; a feature calculation subsystem for constructing a self-updating multilayer perceptron that performs feature recognition and calculation on the image attribute data to obtain image feature classification data; and a blockchain storage subsystem for creating a blockchain database based on the image attribute data and storing the image feature classification data in the blockchain database based on image features. By storing medical images through blockchain, the present invention improves the overall system throughput as well as the shared privacy and integrity of image data.

Description

Technical Field

[0001] The present invention belongs to the field of medical image scheduling, and in particular relates to a medical image storage system based on blockchain. Background Art

[0002] Downloading medical examination images is a frequent daily activity on the Medical Imaging Cloud. Radiologists use the Medical Imaging Cloud system to store and access medical examination images. Each time an image is opened, the Medical Imaging Cloud system downloads the image from the client to the server. Each examination has multiple series, each containing hundreds of images, each several hundred KB in size. These images are numerous yet each image is small. With the advancement of medical equipment such as CT and MRI, medical examination images are becoming increasingly clear.

[0003] However, in the current process of storing medical images, different types of images are often stored in the wrong blockchain nodes, resulting in errors when viewing. In addition, during the image sharing process, there are often problems with inadequate privacy protection and incomplete images due to excessive access. Summary of the Invention

[0004] The purpose of the present invention is to provide a blockchain-based medical image storage system to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above objectives, the present invention provides a blockchain-based medical image storage system, comprising:

[0006] An image collection subsystem, configured to collect medical image data and label the data to obtain labeled image data;

[0007] an attribute analysis subsystem connected to the image collection subsystem, for analyzing the attributes of the marked image data to obtain image attribute data;

[0008] A feature calculation subsystem, connected to the attribute analysis subsystem, is used to construct a self-updating multilayer perceptron, and perform feature recognition and calculation on the image attribute data through the multilayer perceptron to obtain image feature classification data;

[0009] The blockchain storage subsystem is connected to the feature calculation subsystem and is used to create a blockchain database based on the image attribute data and store the image feature classification data in the blockchain database based on the image features.

[0010] Preferably, the image collection subsystem includes:

[0011] An image acquisition module, used to acquire medical image data based on the department's diagnostic data;

[0012] The marking module is used to mark the medical image data based on the metadata of the image to obtain the marked image data.

[0013] Preferably, the marking module includes:

[0014] a metadata analysis unit, configured to analyze the medical image data to obtain image metadata;

[0015] A correlation calculation module, configured to perform correlation calculation on the metadata of the image to obtain image correlation data;

[0016] An image marking unit is used to mark the medical image data based on the image correlation data to obtain the marked image data.

[0017] Preferably, the attribute analysis subsystem includes:

[0018] A preprocessing module, configured to remove noise from the labeled image data and perform image enhancement on the image to obtain a preprocessed image;

[0019] An image segmentation module, configured to segment the preprocessed image based on an edge detection algorithm to obtain a segmented image;

[0020] The feature recognition module is used to extract medical morphological features from the segmented image to obtain the image attribute data.

[0021] Preferably, the feature calculation subsystem includes:

[0022] Building blocks for constructing classification perceptrons based on feedforward neural networks;

[0023] A training module, configured to train the classification perceptron based on historical medical image classification data to obtain the self-updating multilayer perceptron;

[0024] A calculation module is used to perform feature recognition and similarity calculation on the image attribute data based on the multi-layer perceptron to obtain the image feature data.

[0025] Preferably, the building blocks include:

[0026] A convolution training parameter acquisition unit, configured to construct a convolutional neural network, train the convolutional neural network based on historical medical image classification data, and obtain convolution training parameters;

[0027] The auxiliary learning unit is used to construct the convolution training parameters as input of a feedforward neural network to obtain the classification perceptron.

[0028] Preferably, the training module includes:

[0029] A parameter updating unit, configured to perform adaptive parameter filling on old task parameters and new task parameters of the training perceptron to obtain an adaptive parameter updated perceptron;

[0030] a performance training unit, configured to train the adaptive parameter updating perceptron based on historical medical image classification data to obtain a trained perceptron;

[0031] The optimization unit is used to obtain the gap between the metric network output and the actual label based on the loss function, and optimize the trained perceptron based on the Adam optimization algorithm to obtain the self-updating multilayer perceptron.

[0032] Preferably, the blockchain storage subsystem includes:

[0033] Blockchain building module, used to build a partitioned blockchain based on image attribute data;

[0034] A storage module is used to store the image feature classification data in corresponding partitions in the blockchain based on the image features.

[0035] Preferably, the user accesses the medical image through the storage module, and the storage module generates a key for the user to access the medical image after confirming the user's authorization. The storage module collects the user's access data and stores the access data in the miner node.

[0036] The technical effects of the present invention are:

[0037] By constructing a self-updating classification sensor, this invention can autonomously learn the characteristics of new types of medical images and alleviate the problem of catastrophic forgetting, thereby improving the ability to learn new image features and enhancing the accuracy of medical image classification and storage on the blockchain. Furthermore, by storing medical images on the blockchain, this invention improves the overall system throughput and the shared privacy and integrity of image data. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0039] Figure 1 Schematic diagram of a blockchain-based medical image storage system in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] Example 1

[0042] like Figure 1 As shown, this embodiment provides a blockchain-based medical image storage system, including:

[0043] An image collection subsystem, configured to collect medical image data and label the data to obtain labeled image data;

[0044] an attribute analysis subsystem connected to the image collection subsystem, for analyzing the attributes of the marked image data to obtain image attribute data;

[0045] A feature calculation subsystem, connected to the attribute analysis subsystem, is used to construct a self-updating multilayer perceptron, and perform feature recognition and calculation on the image attribute data through the multilayer perceptron to obtain image feature classification data;

[0046] The blockchain storage subsystem is connected to the feature calculation subsystem and is used to create a blockchain database based on the image attribute data and store the image feature classification data in the blockchain database based on the image features.

[0047] The image collection subsystem is a highly efficient tool designed to gather large amounts of medical image data. This data can come from a variety of sources, including hospitals, laboratories, and research institutions. After collecting the data, the system automatically labels each image correctly for subsequent classification and identification.

[0048] The attribute analysis subsystem is closely linked to the image collection subsystem and is responsible for performing attribute analysis on the collected labeled image data. This involves in-depth analysis of various image features, such as color, shape, and texture, to extract key image attributes. This attribute data provides the basis for subsequent feature calculations.

[0049] The feature computation subsystem receives image attribute data from the attribute analysis subsystem and uses this data to construct a self-updating multilayer perceptron. This perceptron performs in-depth feature recognition and computation on image attribute data, resulting in accurate feature classification data. This data is extremely valuable for medical image recognition, diagnosis, and treatment.

[0050] Finally, the blockchain storage subsystem connects to the feature calculation subsystem and creates a blockchain database based on the image attribute data generated by the feature calculation subsystem. This database leverages the unique advantages of blockchain technology to ensure data immutability and traceability. The image feature classification data generated by the feature calculation subsystem is then stored in this blockchain database, making data storage and use more secure, transparent, and verifiable.

[0051] Further optimizing the solution, the image collection subsystem includes:

[0052] An image acquisition module, used to acquire medical image data based on the department's diagnostic data;

[0053] The marking module is used to mark the medical image data based on the metadata of the image to obtain the marked image data.

[0054] Further optimizing the scheme, the marking module includes:

[0055] a metadata analysis unit, configured to analyze the medical image data to obtain image metadata;

[0056] To effectively manage and retrieve images, each image is usually accompanied by some metadata, such as patient information, examination date, image type, doctor's comments, etc. This helps to associate images with relevant patients and cases and ensure data integrity and privacy protection.

[0057] A correlation calculation module, configured to perform correlation calculation on the metadata of the image to obtain image correlation data;

[0058] An image marking unit is used to mark the medical image data based on the image correlation data to obtain the marked image data.

[0059] An identifier is a label used to identify a patient. It can be a patient number. Each patient number uniquely corresponds to a patient. A patient can have one or two examination items. Each examination item can have one or two sequences, formed by categorizing the patient's medical examination images in a given examination item. Each sequence contains one or more medical examination images. Each medical examination image and its corresponding image information are associated with a patient identifier. Image information is information related to the medical examination image, such as image number, image thickness, and pixel value.

[0060] In one scenario, patient A undergoes a CT scan, and the CT device generates some medical examination images. These images can be classified into n examination sequences, and each examination sequence contains multiple medical examination images. Based on one examination item, multiple examination sequences of medical examination images are generated, and each medical examination image under each examination sequence is associated with the patient identification of patient A. The server manages the medical examination images through a three-layer structure of "examination-examination sequence-image". It should be emphasized that in order to further ensure the privacy and security of the patient information and medical examination images of the above patients, the server that manages the patient information and medical examination images of the above patients can be a node of a blockchain, and the medical examination images are stored in the blockchain. The common information of all images of an examination is used as the examination information. The examination information can be the examination number, examination time, and patient name, etc. The examination number can be used to identify an examination. The common information of all images in each sequence under an examination is used as the examination sequence information corresponding to each sequence. The examination sequence information can be the serial number, sequence description, and equipment type, etc. Each medical examination image under each sequence also has image information corresponding to each medical examination image. The image information can be the image number, image thickness, and pixel value, etc.

[0061] Further optimizing the solution, the attribute analysis subsystem includes:

[0062] A preprocessing module is used to remove noise from the labeled image data and perform image enhancement on the image to obtain a preprocessed image; the preprocessing steps include noise removal, image enhancement, smoothing, etc. to ensure that the quality of the medical image is sufficient for analysis. Noise removal is used in this embodiment.

[0063] An image segmentation module, configured to segment the preprocessed image based on an edge detection algorithm to obtain a segmented image;

[0064] The feature recognition module is used to extract medical morphological features from the segmented image to obtain the image attribute data.

[0065] Further optimizing the solution, the feature calculation subsystem includes:

[0066] Building blocks for constructing classification perceptrons based on feedforward neural networks;

[0067] A training module, configured to train the classification perceptron based on historical medical image classification data to obtain the self-updating multilayer perceptron;

[0068] A calculation module is used to perform feature recognition and similarity calculation on the image attribute data based on the multi-layer perceptron to obtain the image feature data.

[0069] Further optimizing the solution, the building blocks include:

[0070] A convolution training parameter acquisition unit, configured to construct a convolutional neural network, train the convolutional neural network based on historical medical image classification data, and obtain convolution training parameters;

[0071] The auxiliary learning unit is used to construct the convolution training parameters as input of a feedforward neural network to obtain the classification perceptron.

[0072] Further optimizing the solution, the training module includes:

[0073] A parameter updating unit, configured to perform adaptive parameter filling on old task parameters and new task parameters of the training perceptron to obtain an adaptive parameter updated perceptron;

[0074] a performance training unit, configured to train the adaptive parameter updating perceptron based on historical medical image classification data to obtain a trained perceptron;

[0075] The optimization unit is used to obtain the gap between the metric network output and the actual label based on the loss function, and optimize the trained perceptron based on the Adam optimization algorithm to obtain the self-updating multilayer perceptron.

[0076] The Adam algorithm updates the model parameters through the following steps:

[0077] Calculate gradients: Calculate gradients using the current batch of training data.

[0078] Update momentum terms m and v: Update m and v using exponential moving average.

[0079] Perform deviation correction: Perform deviation correction on m and v to reduce the deviation at the initial moment.

[0080] Calculate the learning rate: Calculate the adaptive learning rate based on the modified m and v.

[0081] Update parameters: Use learning rate to update model parameters.

[0082] Further optimizing the solution, the blockchain storage subsystem includes:

[0083] Blockchain building module, used to build a partitioned blockchain based on image attribute data;

[0084] A storage module is used to store the image feature classification data in corresponding partitions in the blockchain based on the image features.

[0085] Establishing a blockchain network involves distributing nodes and implementing communication protocols. Nodes can be full nodes (complete replicas), light nodes (which store only a portion of the data), or miner nodes (responsible for generating new blocks). The network is typically peer-to-peer, using the blockchain protocol to communicate transactions and block data.

[0086] To further optimize the solution, users access medical images through the storage module. After confirming the user's authorization, the storage module generates a key for the user to access the medical image. The storage module collects the user's access data and stores the access data in the miner node.

[0087] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A blockchain-based medical image storage system, characterized in that: include: An image collection subsystem, configured to collect medical image data and mark the medical image data to obtain marked image data; an attribute analysis subsystem connected to the image collection subsystem, for analyzing the attributes of the marked image data to obtain image attribute data; a feature calculation subsystem connected to the attribute analysis subsystem and configured to construct a self-updating multilayer perceptron, and perform feature recognition and calculation on the image attribute data through the self-updating multilayer perceptron to obtain image feature classification data; a blockchain storage subsystem, connected to the feature calculation subsystem, for creating a blockchain database based on the image attribute data, and storing the image feature classification data in the blockchain database based on image features; The feature calculation subsystem includes: Building blocks for constructing classification perceptrons based on feedforward neural networks; A training module, configured to train the classification perceptron based on historical medical image classification data to obtain the self-updating multilayer perceptron; a calculation module, configured to perform feature recognition and similarity calculation on the image attribute data based on the multi-layer perceptron to obtain the image feature classification data; The building blocks include: A convolution training parameter acquisition unit, configured to construct a convolutional neural network, train the convolutional neural network based on historical medical image classification data, and obtain convolution training parameters; An auxiliary learning unit, configured to construct the convolution training parameters as input of a feedforward neural network to obtain the classification perceptron; The training module includes: A parameter updating unit, configured to adaptively fill old task parameters and new task parameters of the trained perceptron to obtain an adaptive parameter-updated perceptron; a performance training unit, configured to train the adaptive parameter updating perceptron based on historical medical image classification data to obtain a trained perceptron; The optimization unit is used to obtain the gap between the metric network output and the actual label based on the loss function, and optimize the trained perceptron based on the Adam optimization algorithm to obtain the self-updating multilayer perceptron.

2. The blockchain-based medical image storage system according to claim 1, characterized in that: The image collection subsystem includes: An image acquisition module, used to acquire medical image data based on the department's diagnostic data; The marking module is used to mark the medical image data based on the metadata of the image to obtain the marked image data.

3. The blockchain-based medical image storage system according to claim 2, characterized in that: The marking module includes: a metadata analysis unit, configured to analyze the medical image data to obtain image metadata; A correlation calculation module, configured to perform correlation calculation on the metadata of the image to obtain image correlation data; An image marking unit is used to mark the medical image data based on the image correlation data to obtain the marked image data.

4. The blockchain-based medical image storage system according to claim 1, characterized in that: The attribute analysis subsystem includes: A preprocessing module, configured to remove noise from the labeled image data and perform image enhancement on the image to obtain a preprocessed image; An image segmentation module, configured to segment the preprocessed image based on an edge detection algorithm to obtain a segmented image; The feature recognition module is used to extract medical morphological features from the segmented image to obtain the image attribute data.

5. The blockchain-based medical image storage system according to claim 1, characterized in that: The blockchain storage subsystem includes: Blockchain building module, used to build a partitioned blockchain based on image attribute data; A storage module is used to store the image feature classification data in corresponding partitions in the blockchain based on the image features.

6. The blockchain-based medical image storage system according to claim 5, characterized in that: The user accesses the medical image through the storage module. After confirming the user's authorization, the storage module generates a key for the user to access the medical image. The storage module collects the user's access data and stores the access data in the miner node.

Images