A method, device and medium for quickly realizing identification and classification of medical instrument equipment

Abstract

The present application relates to the field of image recognition, and more particularly, to a method for quickly realizing identification and classification of medical instrument equipment. In the present application, there are multiple edge nodes, each of which is provided with a classifier model based on a convolutional neural network and capable of identifying multiple classifications of medical instrument equipment. The first edge node is the closest location to the terminal for shooting images of medical instrument equipment, and is usually a machine room server of a medical institution where the equipment is located. When the model and type of the equipment are identified, the result can be finally fed back to the terminal. When only the type of the equipment is identified, the image data is transmitted to a second edge node capable of identifying the type of the equipment through a cloud server, wherein the second edge node can be multiple edge nodes. After the second edge node outputs one or more results, the model information of the equipment is obtained, and the result is finally fed back to the terminal.

Description

TECHNICAL FIELD

[0001] The present application generally relates to the field of image recognition and processing. More specifically, the present application relates to a method, device and medium for quickly realizing medical instrument equipment identification classification. BACKGROUND

[0002] Deep learning and convolutional neural networks (CNN) have played a huge role in image recognition and processing. Through deep network structure and large-scale image data training, convolutional neural networks (CNN) can learn more advanced and abstract feature representation, thereby improving the accuracy and efficiency of image recognition. However, because of the variety of medical instrument equipment, a neural network model capable of identifying a large number of medical instrument equipment will require a large-scale image data set to train it, and this training process requires more storage space and computing resources, so the training cost may increase. Moreover, the process of learning and iterative training of a large number of image data by a neural network will take longer to complete. In addition, when the training data set is too large, the neural network may overfit the training data, resulting in poor performance on new unseen images, so the performance of the network on the training set may be good, but it cannot accurately identify unknown images.

[0003] How to reduce the training cost and time of the neural network model for identifying medical instrument equipment and improve the accuracy of the model to identify pictures is an important part of the medical instrument equipment identification classification method being studied. SUMMARY

[0004] To solve one or more of the above technical problems, the present application provides a method for quickly realizing medical instrument equipment identification classification. Meanwhile, the present application also provides a computer device and a computer readable storage medium based on the above method.

[0005] According to a first aspect of the present application, embodiments of the present application provide a method for quickly realizing identification and classification of medical equipment, comprising: acquiring an image of a target equipment captured by a terminal; transmitting the image to a first edge node closest to a current location of the terminal according to the current location; identifying a model or / and a type of the target equipment by the first edge node; in response to the first edge node identifying only the type of the target equipment, transmitting the type information to the cloud server by the first edge node, and simultaneously transmitting the image to the cloud server by the first edge node; filtering a second edge node matching the type information according to the type information by the cloud server; transmitting the image to the second edge node by the cloud server; identifying the image by the second edge node to determine second model information of the target equipment, and feeding back to the cloud server; extracting basic data of the target equipment according to the second model by the cloud server, and transmitting the basic data to the terminal.

[0006] In one embodiment, in response to the first edge node identifying the type and the model, the first edge node transmits the first model information of the target equipment to the cloud server; and the cloud server feeds back the first model information to the terminal.

[0007] In one embodiment, the first edge node deploys a first classifier model for identifying the type or / and the model of the first medical equipment, the first classifier model being used to identify a matching probability of the target equipment and the first medical equipment, wherein the first medical equipment comprises a predetermined number of different models of medical equipment used by a medical institution near the first edge node. The second edge node comprises one or more edge nodes for identifying the model of the target equipment, and the second edge node correspondingly deploys a second classifier model for identifying a corresponding second medical equipment, the second classifier model being used to identify a matching probability of the target equipment and the second medical equipment corresponding to the second classifier model, wherein the second medical equipment comprises a predetermined number of different models of medical equipment used by a predetermined number of medical institutions near the second edge node.

[0008] In one embodiment, the first classifier model is trained using image samples of the first medical equipment; and the second classifier model is trained using image samples of the second medical equipment; wherein the first medical equipment is a medical equipment used by a medical institution near the first edge node, and the second medical equipment is a medical equipment used by a medical institution near the corresponding edge node.

[0009] In an embodiment, the first edge node identifying the model or / and type of the target device comprises inputting the target device image into the first classifier model; the first classifier model calculating a matching probability of the target device image and the first medical device image sample; according to the matching probability, distinguishing the model of the first medical device into a first determined model or / and a first suspected model; in response to the first classifier model outputting the first determined model and the first suspected model, taking the first determined model as the model of the target device; in response to the first classifier model only outputting the first suspected model, adjudicating the first suspected model and determining the type of the target device, wherein the first suspected model is one or more device models.

[0010] In an embodiment, the second edge node identifying the model information of the target device from the image and feeding back to the cloud server comprises: inputting the image data into one or more second classifier models corresponding to the second edge node; the second classifier model calculating a matching probability of the target device image and the second medical device image sample; according to the matching probability, distinguishing the model of the second medical device into a second determined model or / and a second suspected model, wherein the second determined model is one or more medical device models of the same type; according to the second determined model, adjudicating the second model; and transmitting the information of the second model to the cloud server.

[0011] In an embodiment, according to the second determined model, adjudicating the second model comprises: in response to more than half of the same models being the second determined model, taking the second model as the second determined model; and in response to one second edge node being of the same type of medical institution as the first edge node, taking the identification result of the second edge node as priority.

[0012] In an embodiment, the basic data of the target device comprises: functions of the device, operation steps, precautions, and problems occurred in previous use of the device.

[0013] According to a second aspect of the present application, the embodiments of the present application further provide a device for quickly realizing identification and classification of medical equipment, comprising a processor and a memory, the memory storing a computer program, and the processor executing the computer program to realize the steps of any of the above method embodiments.

[0014] According to a third aspect of the present application, the embodiments of the present application further provide a computer readable storage medium, which stores a computer program, and the computer program is executed to realize the steps of any of the above method embodiments.

[0015] In the present application, there are multiple edge nodes, each of which has a classifier model based on a convolutional neural network and capable of identifying multiple medical instrument devices. Among them, the multiple medical instrument devices capable of being identified are related to the medical institution where the edge node is located, for example, the classifier model deployed by the edge node located in the medical instrument room of the orthopedic medical institution can identify multiple orthopedic medical instrument devices. The first edge node is the closest place to the terminal shooting medical instrument device image, which is usually located in the server room of the target device in the medical institution, when the type and model of the device are identified, the result can be finally fed back to the terminal, when only the type of the device is identified, the image data is transmitted to the second edge node capable of identifying the type of the device through the cloud server, wherein the second edge node can be multiple edge nodes. After the second edge node outputs one or more results, it is determined to obtain the model information of the device, and finally fed back to the terminal.

[0016] Each edge node only stores several medical devices with the highest recognition frequency, which greatly reduces the cost and time of model training. And each model identifies the type of the medical instrument device, thereby increasing the accuracy of identification. BRIEF DESCRIPTION OF DRAWINGS

[0017] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description read in conjunction with the accompanying drawings, in which several embodiments of the application are shown. In the drawings, the same or corresponding parts are denoted by the same or corresponding reference numerals, and embodiments of the present application are shown by way of example and not limitation, in which:

[0018] Figure 1 is a sequence diagram schematically showing a method for quickly implementing medical instrument device identification classification according to an embodiment of the present application;

[0019] Figure 2 is a flowchart schematically showing a method for quickly implementing medical instrument device identification classification according to an embodiment of the present application;

[0020] Figure 3 is a flowchart schematically showing step S3 according to an embodiment of the present application;

[0021] Figure 4 is a flowchart schematically showing step S7 according to an embodiment of the present application;

[0022] Figure 5 is a block diagram schematically showing a structure of a device for quickly implementing medical instrument device identification classification according to the present embodiment. DETAILED DESCRIPTION

[0023] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0024] The specific implementation of the present application will be described in detail below with reference to the drawings.

[0025] Embodiments of a method for rapid implementation of medical device identification and classification

[0026] Figure 1 is a sequence diagram schematically showing a method for quickly realizing identification and classification of medical instrument equipment according to an embodiment of the present application. As shown in Figure 1 , in one embodiment, a terminal obtains an image of a target equipment and transmits the image to a first edge node; the first edge node obtains a type and an image of the target equipment through the image of the target equipment and transmits them to a cloud server; the cloud server transmits the image of the target equipment to a second edge node according to the type of the target equipment; the second edge node obtains a model and a category of the target equipment and transmits them to the cloud server, and the cloud server transmits basic data of the target equipment to the terminal according to the model of the target equipment. In another embodiment, the terminal transmits an image of a target equipment to a first edge node; the first edge node obtains a model and a category of the target equipment and transmits them to a cloud server, and the cloud server transmits basic data of the target equipment to the terminal according to the model of the target equipment.

[0027] Figure 2 is a flowchart schematically showing a method for quickly realizing identification and classification of medical instrument equipment according to an embodiment of the present application; as shown in Figure 2 , the identification method comprises the following steps

[0028] Step S1, obtaining an image of a target equipment taken by a terminal.

[0029] The terminal obtains image data of the target equipment through a camera, and the terminal can detect position information when the target equipment is taken.

[0030] Step S2, transmitting the image to a first edge node closest to a current position of the terminal according to the current position.

[0031] The image data of the target equipment obtained by the terminal is generally a medical equipment of a medical institution where the equipment is located, based on which the first edge node is an edge node closest to the terminal, which can be arranged in a machine room server of the medical institution where the terminal is located.

[0032] Furthermore, the terminal transmits the image to the first edge node through a gateway.

[0033] Step S3, the first edge node identifies the model and / or type of the target device.

[0034] wherein the first edge node deploys a first classifier model for identifying the type or / and model of the first medical device, the first classifier model is used to identify the matching probability of the target device and the first medical device, wherein the first medical device includes a predetermined number of different models of medical devices in the medical institution near the first edge node.

[0035] Further, the first classifier model is a deep learning model trained using a convolutional neural network, wherein the first data set includes image samples of the first medical device, and the image samples are added with corresponding first medical device model information and class labels.

[0036] In addition, the first medical device includes a plurality of devices in the medical institution where the target device is located, and the plurality of devices are identified more frequently than other medical devices.

[0037] Referring to Figure 3 , step S3 includes steps S301-S304, as follows:

[0038] Step S301, input the image data into the first classifier model.

[0039] Step S302, the first classifier model calculates the matching probability of the target device image and the first medical device image sample.

[0040] wherein the first classifier model uses the similarity of texture, edge shape and color of the target device image and the first medical device image as the calculation parameter of the matching probability.

[0041] Step S303, according to the matching probability, the model of the first medical device is divided into a first determined model or / and a first suspected model.

[0042] wherein in the probability vector output by the first classifier model, if the matching probability of the image of a type of device exceeds 95%, the type is the first determined model; if the matching probability of one or more types of devices is between 70% and 95%, the one or more types of devices are the first suspected devices.

[0043] Step S304, in response to the first classifier model outputting the first determined model and the first suspected model, determining the first determined model as the model of the target device

[0044] Further, in another embodiment, there is a step S304' instead of step S304. Step S304', in response to the first classifier model only outputting the first suspected model number, adjudicating the first suspected model number and determining the type of the target device, wherein the first suspected model number is one or more device model numbers.

[0045] Step S4, in response to the first edge node only identifying the type of the target device, the first edge node transmits the type information and the image to the cloud server.

[0046] Wherein, the first determined model number does not appear, and one or more first suspected model numbers appear. At this time, the type of the target device is adjudicated according to the first suspected model number.

[0047] Further, the adjudication method can be that the medical device with the most first suspected model number type is regarded as the type of the target device. For example: when the first suspected model number has A11 model number, A12 model number and A13 model number of A type, and the first suspected model number also has B11 model number of B type, the target device is regarded as a device of A type; when the first suspected model number only has A11 model number of A type, the target device is regarded as a device of A type.

[0048] Step S5, the cloud server screens out a second edge node that meets the type information according to the type information.

[0049] Wherein, the screened out second edge node is one or more edge nodes that can identify the type of the medical device.

[0050] Further, the second edge node includes one or more edge nodes for identifying the type of the target device, and the second edge node corresponds to the deployment of a second classifier model for identifying the corresponding second medical device. The second classifier model is used to identify the matching probability of the target device and the second medical device corresponding to the second classifier model, wherein the second medical device includes a predetermined number of different types of medical devices corresponding to a predetermined number of medical institutions near the second edge node.

[0051] In addition, each second classifier model is a deep learning model trained using a corresponding second data set, and each second data set includes image samples of the corresponding second medical device, and the image samples are also added with corresponding second medical device model information and class labels.

[0052] Further, the second medical device includes a plurality of devices of the medical institution where the target device is located, and the plurality of devices are identified more frequently than other medical devices.

[0053] Step S6, the cloud server transmits the image to the second edge node.

[0054] The cloud server transmits the image to the second edge node through the gateway.

[0055] Step S7, the second edge node identifies the image to determine the model information of the target device, and feeds back to the cloud server. Referring to Figure 4 Step S7 includes steps S701-S705, as follows:

[0056] Step S701, input the image data to the second classifier model corresponding to the second edge node.

[0057] The terminal transmits the image data to each second edge node through the gateway.

[0058] Step S702, the second classifier model calculates the matching probability of the target device image and the second medical device image sample.

[0059] Step S703, according to the matching probability, the model of the second medical device is divided into a second determined model or / and a second suspected model.

[0060] One or more second classifier models correspond to output one or more probability vectors, and each probability vector determines at most one second determined model. When the image matching probability of one or more model devices exceeds 95%, the one or more models are the second determined model; when the matching probability of one or more model devices is between 70% and 95%, the one or more model devices are considered as the second suspected device.

[0061] Furthermore, each second classifier model obtains at most one second determined model, so one or more classifier models obtain one or more second determined models. Subsequently, only the second determined model is used to determine the second model, and the second suspected device (model) is not used.

[0062] Step S704, according to the second determined model, the second model is determined.

[0063] In response to more than half of the same type of second determined type being the second type, and in response to a second edge node being of the same type of medical institution as the first edge node, the identification result of the second edge node is prioritized. For example, a target device located in a tumor A medical institution, and the corresponding second determined type includes a K11 type edge node located in a people A medical institution, a K11 type edge node located in a people B medical institution, and a K12 type edge node located in a people C medical institution, and the type of the target device is considered to be the K11 type. Different medical institutions can set different weights, and the same type number and its corresponding weight are used to determine. For another example, a target device located in a tumor A medical institution, and the corresponding second determined type includes a K11 type edge node located in a tumor B medical institution and a K12 type edge node located in a people A medical institution, and the type of the target device is considered to be the K11 type.

[0064] In step S705, the information of the second type is transmitted to the cloud server.

[0065] In the register of the cloud server, the type of the medical device corresponding to all the second type information is stored, and the basic information of the medical device corresponding to the type of the medical device is also stored. The basic data of the medical device includes the function, operation steps, precautions, and problems that have occurred in the past use of the device.

[0066] Further, the second edge node transmits the information of the second type to the cloud server through wireless communication.

[0067] In step S8, the cloud server analyzes and processes the information and transmits it to the terminal.

[0068] Further, in another embodiment, steps S4` and S5` replace steps S4 to S8: in step S4`, in response to the first edge node identifying the type and the type, the first edge node transmits the type information to the cloud server; in step S5`, the cloud server analyzes and processes the information and feeds back to the terminal.

[0069] Embodiments of a device for rapid implementation of medical device identification and classification

[0070] In one embodiment, the present application provides a computer device, the internal structure of which can be as shown in Figure 5The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the medical instrument device identification and classification device is used to provide computing and control capabilities, and can be selected from various types such as CPU, single-chip microcomputer, DSP or FPGA. The memory of the medical instrument device identification and classification device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. When the computer program is executed, the steps described in the above method embodiments can be completed, for example, steps S1 to S8. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0071] Those skilled in the art can understand that, Figure 5 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device of the present application. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0072] Embodiments of a computer-readable storage medium

[0073] In one embodiment, the present application provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the steps described in the above method embodiments.

[0074] Those skilled in the art can understand that all or part of the processes in the above-mentioned method embodiments can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the method can be included. Any reference to memory, storage, database or other medium used in each embodiment of the present application can include non-volatile or / and volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0075] Each technical feature of the above-mentioned embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of each technical feature in the above-mentioned embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0076] The above-mentioned embodiments only express several embodiments of the present application, and the description is more specific and detailed, however, it should not be understood as a limitation on the scope of the patent application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

Claims

1. A method for quickly implementing a medical instrument device identification classification, characterized by, The method comprises: acquiring an image of a target device taken by a terminal; transmitting the image to a first edge node closest to a current location of the terminal according to the current location; the first edge node identifying a model or / and a type of the target device, comprising: inputting the image of the target device into a first classifier model; the first classifier model calculating a matching probability of the image of the target device and a first medical device image sample; according to the matching probability, distinguishing a first medical device model into a first determined model or / and a first suspected model; in response to the first classifier model outputting the first determined model and the first suspected model, taking the first determined model as the model of the target device; in response to the first classifier model only outputting the first suspected model, adjudicating the first suspected model and determining the type of the target device, wherein the first suspected model is one or more device models; in response to the first edge node only identifying the type of the target device, the first edge node transmitting the type information to a cloud server, and the first edge node simultaneously transmitting the image to the cloud server; the cloud server screening a second edge node adapting to the type information according to the type information; the cloud server transmitting the image to the second edge node; the second edge node identifying the image to determine second model information of the target device and feeding back to the cloud server, comprising: inputting image data into one or more second classifier models corresponding to the second edge node; the second classifier model calculating a matching probability of the image of the target device and a second medical device image sample; according to the matching probability, distinguishing a second medical device model into a second determined model or / and a second suspected model, wherein the second determined model is one or more medical device models of the same type; adjudicating the second model according to the second determined model; transmitting information of the second model to the cloud server, comprising: in response to more than half of the same models, the second determined model being the second model; in response to one second edge node being of the same type of medical institution as the first edge node, taking the recognition result of the second edge node as priority; the cloud server extracting basic data of the target device according to the second model and transmitting the basic data to the terminal.

2. The method of claim 1, wherein: in response to the first edge node identifying the type and the model, the first edge node transmits first model information of the target device to the cloud server; the cloud server feeds back to the terminal according to the first model information.

3. The method of claim 1, wherein, The first edge node deploys a first classifier model for identifying the type or / and model of the first medical device, the first classifier model being used for identifying the matching probability of the target device with the first medical device, wherein the first medical device comprises a predetermined number of different models of medical devices used in the medical institution near the first edge node; and the second edge node comprises one or more edge nodes for identifying the model of the target device, and the second edge node correspondingly deploys a second classifier model for identifying the corresponding second medical device, the second classifier model being used for identifying the matching probability of the target device with the second medical device corresponding to the second classifier model, wherein the second medical device comprises a predetermined number of different models of medical devices used in the medical institution near the second edge node.

4. The method of claim 3, wherein, The first classifier model is trained using image samples of the first medical device; and the second classifier model is trained using image samples of the second medical device; wherein the first medical device is a medical device used in the medical institution near the first edge node, and the second medical device is a medical device used in the medical institution near the corresponding edge node.

5. The method of claim 1, wherein, The basic data of the target device comprises: the function, operation steps, matters needing attention, and problems occurred in the past use of the device.

6. A device for quickly implementing identification classification of medical instrument equipment, comprising a processor and a memory, wherein the memory stores a computer program, and the computer program comprises the following steps of: The processor executes the computer program to implement the steps of any one of the methods of claims 1-5. ​ 7. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program, when executed, implements the steps of any one of the methods of claims 1-5.

Images