Intelligent analysis method, system, device and medium based on ultrasonic examination data

Abstract

The application provides an intelligent analysis method, system, device and medium based on ultrasonic examination data, and relates to the field of ultrasonic examination data analysis.The method comprises the following steps: acquiring an ultrasonic examination video of a finger; determining a plurality of bone information of the finger based on the ultrasonic examination video of the finger by using a video processing model; constructing graph structure data by using a convolutional neural network model based on the plurality of bone information of the finger, wherein the graph structure data comprises a plurality of nodes and a plurality of edges between the plurality of nodes, each node in the plurality of nodes represents each bone, each node comprises a plurality of node features, and the node features of each node comprise the shape of each bone, the position of each bone, the size of each bone and echo intensity distribution information of each bone; and determining whether the finger is fractured by processing the graph structure data based on a graph neural network model.The method can quickly and accurately determine whether the finger is fractured.

Description

Technical Field

[0001] This invention relates to the field of ultrasound examination data analysis technology, specifically to an intelligent analysis method, system, device, and medium based on ultrasound examination data. Background Technology

[0002] In traditional medical imaging, X-rays are a common tool for diagnosing and assessing finger fractures. However, X-rays have some limitations, such as radiation exposure, high cost, and the need for specialized equipment and technicians.

[0003] Ultrasound measurement is a non-invasive method that uses high-frequency sound waves to generate images to assess tissue structure and abnormalities. In the case of finger fractures, ultrasound measurement can provide information about the location and type of fracture and can be used to guide treatment and assess the fracture healing process. However, there are some specific challenges and drawbacks in the field of ultrasound measurement. Due to the low image resolution of ultrasound measurements, details of finger fractures are not clearly displayed, often requiring careful interpretation by specialized technicians, leading to low analytical efficiency and a high risk of errors during interpretation.

[0004] Therefore, how to quickly and accurately determine whether a finger is fractured is an urgent problem to be solved. Summary of the Invention

[0005] The main technical problem this invention addresses is how to quickly and accurately determine whether a finger is fractured.

[0006] According to a first aspect, the present invention provides an intelligent analysis method based on ultrasound examination data, comprising: acquiring an ultrasound examination video of a finger; determining multiple skeletal information of the finger using a video processing model based on the ultrasound examination video of the finger, the multiple skeletal information of the finger including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone; constructing graph structure data using a convolutional neural network model based on the multiple skeletal information of the finger, the graph structure data including multiple nodes and multiple edges between the multiple nodes, each node representing each bone, each node including multiple node features, the node features of each node including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone; and processing the graph structure data based on the graph neural network model to determine whether the finger is fractured.

[0007] Furthermore, the video processing model is a long short-term neural network model, the input of the video processing model is the ultrasound examination video of the finger, and the output of the video processing model is multiple skeletal information of the finger.

[0008] Furthermore, the input to the convolutional neural network model is multiple skeletal information of the finger, and the output of the convolutional neural network model is the graph structure data.

[0009] Furthermore, the input to the graph neural network model is the graph structure data, and the output of the graph neural network model is either a fracture or no fracture.

[0010] According to a second aspect, the present invention provides an intelligent analysis system based on ultrasound examination data, comprising: a first acquisition module for acquiring ultrasound examination video of a finger;

[0011] The video processing module is used to determine multiple bone information of the finger based on the ultrasound examination video of the finger using a video processing model. The multiple bone information of the finger includes the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0012] The construction module is used to construct graph structure data using a convolutional neural network model based on multiple bone information of the finger. The graph structure data includes multiple nodes and multiple edges between the nodes. Each node represents a bone. Each node includes multiple node features, including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0013] The fracture determination module is used to process the graph structure data based on a graph neural network model to determine whether a finger is fractured.

[0014] Furthermore, the video processing model is a long short-term neural network model, the input of the video processing model is the ultrasound examination video of the finger, and the output of the video processing model is multiple skeletal information of the finger.

[0015] Furthermore, the input to the convolutional neural network model is multiple skeletal information of the finger, and the output of the convolutional neural network model is the graph structure data.

[0016] Furthermore, the input to the graph neural network model is the graph structure data, and the output of the graph neural network model is either a fracture or no fracture.

[0017] According to a third aspect, the present invention provides an electronic device comprising: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method described above.

[0018] According to a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method as described in any of the preceding aspects.

[0019] This invention provides an intelligent analysis method, system, device, and medium based on ultrasound examination data. The method includes acquiring an ultrasound examination video of a finger; using a video processing model to determine multiple skeletal information of the finger based on the ultrasound examination video, the multiple skeletal information including the shape, position, size, and echo intensity distribution of each bone; constructing graph structure data using a convolutional neural network model based on the multiple skeletal information of the finger, the graph structure data including multiple nodes and multiple edges between the nodes, each node representing a bone, each node including multiple node features, the node features including the shape, position, size, and echo intensity distribution of each bone; and processing the graph structure data based on the graph neural network model to determine whether the finger is fractured. This method can quickly and accurately determine whether a finger is fractured. Attached Figure Description

[0020] Figure 1 A flowchart illustrating an intelligent analysis method based on ultrasound examination data provided in an embodiment of the present invention;

[0021] Figure 2 A schematic diagram of an intelligent analysis system based on ultrasound examination data provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0023] In this embodiment of the invention, the following are provided: Figure 1 The above describes an intelligent analysis method based on ultrasound examination data, which includes steps S1 to S4:

[0024] Step S1: Obtain an ultrasound examination video of the finger.

[0025] Ultrasound examination is a non-invasive medical imaging technique that uses high-frequency sound waves to generate images for observing the internal structures of the human body. Ultrasound examination can be used to examine the skeletal structure of the fingers to diagnose fractures.

[0026] As an example, doctors use an ultrasound machine to transmit ultrasound waves to a finger, and use the echoes to generate an ultrasound examination video of the finger.

[0027] Ultrasound examination video refers to the video obtained through ultrasound examination. Ultrasound examination video shows the internal structure of the examined area, including information on bones, soft tissues, and blood flow. Ultrasound examination videos can be processed to determine the presence of finger fractures.

[0028] Step S2: Based on the ultrasound examination video of the finger, a video processing model is used to determine multiple bone information of the finger. The multiple bone information of the finger includes the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0029] The video processing model is a long short-term neural network model. The input of the video processing model is the ultrasound examination video of the finger, and the output of the video processing model is multiple skeletal information of the finger.

[0030] Long Short-Term Memory (LSTM) neural network models are a type of Recurrent Neural Network (RNN). LSTM models can process sequences of arbitrary length, capture sequence information, and output results based on the relationships between preceding and following data within the sequence. By processing continuous ultrasound examination videos of the finger using LSTM models, it is possible to output features that comprehensively consider the relationships between ultrasound examination videos of the finger at different time points, making the output features more accurate and comprehensive.

[0031] The finger contains multiple bones.

[0032] Information about multiple bones in the finger includes the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution of each bone.

[0033] The shape of each bone refers to the outer contour or geometry of the finger bones.

[0034] The position of a bone refers to its relative or absolute location within the overall structure of the finger. In some embodiments, the position of a bone is determined by a coordinate system. The position of each bone in the finger can be obtained by locating it within a pre-defined coordinate system. For example, by establishing a coordinate system beforehand and placing the output positions of multiple finger bones within that system, the bone positions can be determined.

[0035] The size of a skeleton includes information such as its length, width, and height.

[0036] The echo intensity distribution information of bones refers to the echo reflection of bones in an ultrasound examination video of a finger. When ultrasound waves propagate through bone tissue, they are reflected and scattered. Different tissue densities and structures affect the propagation and reflection of ultrasound waves differently, resulting in varying echo intensities on the ultrasound image. Echo intensity distribution information includes the brightness variations of pixels at different locations within the bone. By processing ultrasound examination videos of the finger, the echo intensity distribution information of the bones can be obtained, allowing for the assessment of bone quality, bone density, and abnormalities at fracture sites, and determining whether a fracture has occurred.

[0037] In some embodiments, the video processing model includes a background segmentation sub-model, a skeleton segmentation sub-model, and a skeleton information output sub-model. The background segmentation sub-model, skeleton segmentation sub-model, and skeleton information output sub-model are all long short-term neural network (LSN) models. The input to the background segmentation sub-model is an ultrasound examination video of a finger, and the output of the background segmentation sub-model is an ultrasound examination video of the finger after removing the background. The input to the skeleton segmentation sub-model is the ultrasound examination video of the finger after removing the background, and the output of the skeleton segmentation sub-model is an ultrasound examination video of each bone in the finger. The input to the skeleton information output sub-model is the ultrasound examination video of each bone in the finger, and the output of the skeleton information output sub-model is multiple bone information of the finger.

[0038] Removing the background from the ultrasound video of a finger by using a background segmentation sub-model allows the focus to be on the finger itself, improving the accuracy of subsequent model processing. Background removal also reduces noise and interference, making it easier for the model to analyze and extract skeletal information from the finger.

[0039] The skeletal segmentation sub-model, based on background-removed ultrasound examination videos of the finger, accurately extracts the ultrasound images of each bone in the finger. This facilitates more precise localization and segmentation of each bone, providing accurate input for the subsequent skeletal information output sub-model.

[0040] The skeletal information output sub-model obtains multiple skeletal information of the finger based on the ultrasound examination video output of each bone.

[0041] By progressively processing finger ultrasound examination videos using background segmentation, bone segmentation, and bone information output sub-models, the system can extract background-removed finger images, segment each bone image, and obtain detailed bone information. This progressive processing approach helps preserve key information, improves the accuracy of bone localization, and provides a more comprehensive basis for fracture diagnosis, thereby increasing the accuracy rate of finger fracture diagnosis.

[0042] Step S3: Based on the multiple bone information of the finger, a convolutional neural network model is used to construct graph structure data. The graph structure data includes multiple nodes and multiple edges between the multiple nodes. Each node in the multiple nodes represents a bone. Each node includes multiple node features. The node features of each node include the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0043] Convolutional Neural Network (CNN) models include CNNs. A CNN can be a multi-layer neural network (e.g., comprising at least two layers). The at least two layers can include at least one of a convolutional layer (CONV), a rectified linear unit (ReLU) layer, a pooling layer (POOL), or a fully connected layer (FC). The at least two layers of a CNN can correspond to neurons arranged in three dimensions: width, height, and depth. In some embodiments, a CNN can have an architecture of [input layer - convolutional layer - rectified linear unit layer - pooling layer - fully connected layer]. The convolutional layer can compute the output of neurons connected to local regions in the input, and compute the dot product between the weight of each neuron and the small region connected to it in the input volume.

[0044] The input to the convolutional neural network model is multiple skeletal information of the finger, and the output of the convolutional neural network model is the graph structure data.

[0045] Graph-structured data refers to data structures represented in the form of a graph, consisting of multiple nodes and edges. Nodes represent data elements, and edges represent relationships between nodes.

[0046] The graph structure data includes multiple nodes and multiple edges between the nodes. Each node represents a bone, and each node includes multiple node features, including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0047] In some embodiments, the characteristics of an edge among the plurality of edges include the distance and orientation between the plurality of nodes.

[0048] By converting multiple skeletal information of the finger into graph-structured data and extracting node features, further processing and analysis can be performed by subsequent graph neural network models. Graph-structured data can reflect the positional relationships between bones, and by comprehensively considering the shape, position, size, and echo intensity distribution of each bone, it helps to fully understand the characteristics and relationships of finger bones and improve the accuracy of skeletal abnormalities.

[0049] Step S4: Process the graph structure data based on the graph neural network model to determine whether the finger is fractured.

[0050] Graph Neural Network (GNN) models consist of a Graph Neural Network (GNN) and fully connected layers. A GNN is a neural network that operates directly on graph-structured data, which is a data structure composed of nodes and edges. The GNN model processes the graph-structured data of a finger, learning the features and relationships of each node (i.e., finger bones) and using this information to determine if the finger is fractured. Through information transmission and aggregation, the model can comprehensively consider features such as the shape, position, size, and echo intensity distribution of the bones to determine whether the finger is fractured.

[0051] Compared to traditional methods, graph neural network models are better able to capture the relationships between skeletons and global features, improving prediction accuracy and reliability.

[0052] Based on the same inventive concept Figure 2 A schematic diagram of an intelligent analysis system based on ultrasound examination data is provided as an embodiment of the present invention. The intelligent analysis system based on ultrasound examination data includes:

[0053] The first acquisition module 21 is used to acquire ultrasound examination video of the finger;

[0054] The video processing module 22 is used to determine multiple bone information of the finger based on the ultrasound examination video of the finger using a video processing model. The multiple bone information of the finger includes the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0055] The construction module 23 is used to construct graph structure data using a convolutional neural network model based on multiple bone information of the finger. The graph structure data includes multiple nodes and multiple edges between the multiple nodes. Each node in the multiple nodes represents a bone. Each node includes multiple node features. The node features of each node include the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone.

[0056] The fracture determination module 24 is used to process the graph structure data based on a graph neural network model to determine whether a finger is fractured.

[0057] Based on the same inventive concept, embodiments of the present invention provide an electronic device, such as... Figure 3 As shown, it includes:

[0058] The system includes: a processor 31; a memory 32; and a computer program; wherein the computer program is stored in the memory 32 and configured to be executed by the processor 31 to implement the intelligent analysis method based on ultrasound examination data as described above, the method comprising: acquiring an ultrasound examination video of a finger; determining multiple bone information of the finger using a video processing model based on the ultrasound examination video of the finger, the multiple bone information of the finger including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone; constructing graph structure data using a convolutional neural network model based on the multiple bone information of the finger, the graph structure data including multiple nodes and multiple edges between the multiple nodes, each node representing each bone, each node including multiple node features, the node features of each node including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone; and processing the graph structure data based on the graph neural network model to determine whether the finger is fractured.

[0059] Based on the same inventive concept, this embodiment provides a computer-readable storage medium storing a computer program. When executed by processor 31, the program implements the aforementioned intelligent analysis method based on ultrasound examination data. The method includes: acquiring an ultrasound examination video of a finger; determining multiple bone information of the finger using a video processing model based on the ultrasound examination video, the multiple bone information including the shape, position, size, and echo intensity distribution information of each bone; constructing graph structure data using a convolutional neural network model based on the multiple bone information of the finger, the graph structure data including multiple nodes and multiple edges between the multiple nodes, each node representing a bone, each node including multiple node features, the node features including the shape, position, size, and echo intensity distribution information of each bone; and processing the graph structure data based on the graph neural network model to determine whether the finger is fractured.

Claims

1. An intelligent analysis method based on ultrasound examination data, characterized in that, include: Obtain an ultrasound video of the finger; Based on the ultrasound examination video of the finger, a video processing model is used to determine multiple bone information of the finger, including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone. Based on the multiple bone information of the finger, a graph structure data is constructed using a convolutional neural network model. The graph structure data includes multiple nodes and multiple edges between the nodes. Each node represents a bone, and each node includes multiple node features, including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone. The graph structure data is processed using a graph neural network model to determine whether a finger is fractured.

2. The intelligent analysis method based on ultrasound examination data as described in claim 1, characterized in that, The video processing model is a long short-term neural network model. The input of the video processing model is the ultrasound examination video of the finger, and the output of the video processing model is multiple skeletal information of the finger.

3. The intelligent analysis method based on ultrasound examination data as described in claim 1, characterized in that, The input to the convolutional neural network model is multiple skeletal information of the finger, and the output of the convolutional neural network model is the graph structure data.

4. The intelligent analysis method based on ultrasound examination data as described in claim 1, characterized in that, The input to the graph neural network model is the graph structure data, and the output of the graph neural network model is either a fracture or no fracture.

5. An intelligent analysis system based on ultrasound examination data, characterized in that, include: The first acquisition module is used to acquire ultrasound examination videos of the fingers; The video processing module is used to determine multiple bone information of the finger based on the ultrasound examination video of the finger using a video processing model. The multiple bone information of the finger includes the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone. The construction module is used to construct graph structure data using a convolutional neural network model based on multiple bone information of the finger. The graph structure data includes multiple nodes and multiple edges between the nodes. Each node represents a bone. Each node includes multiple node features, including the shape of each bone, the position of each bone, the size of each bone, and the echo intensity distribution information of each bone. The fracture determination module is used to process the graph structure data based on a graph neural network model to determine whether a finger is fractured.

6. The intelligent analysis system based on ultrasound examination data as described in claim 5, characterized in that, The video processing model is a long short-term neural network model. The input of the video processing model is the ultrasound examination video of the finger, and the output of the video processing model is multiple skeletal information of the finger.

7. The intelligent analysis system based on ultrasound examination data as described in claim 5, characterized in that, The input to the convolutional neural network model is multiple skeletal information of the finger, and the output of the convolutional neural network model is the graph structure data.

8. The intelligent analysis system based on ultrasound examination data as described in claim 5, characterized in that, The input to the graph neural network model is the graph structure data, and the output of the graph neural network model is either a fracture or no fracture.

9. An electronic device, characterized in that, include: Memory; processor; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the intelligent analysis method based on ultrasound examination data as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the intelligent analysis method based on ultrasound examination data as described in any one of claims 1 to 4.

Images