Automatic selection method for medical image, device, computer-readable storage medium, and program

Abstract

To provide an automatic selection method for a medical image, device, computer-readable storage medium, and program that avoid such a harmful effect that the most suitable medical image for diagnosis is not selected due to human factors, reduce an overlooking rate due to human factors caused by fatigue of a medical worker, and alleviate the burden of repetitive and complex medical image diagnosis and analysis tasks on the medical worker.SOLUTION: A method includes the steps of: acquiring a medical image set in which each medical image includes lesion image information; inputting the medical image set into a classification model and determining, on the basis of the classification model, the target category of the lesion in each medical image within the medical image set and a category score representing the probability that the lesion belongs to the target category; and selecting the target medical image from the medical image set on the basis of the category score of each medical image within the medical image set.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to the field of medical imaging equipment, and specifically to an automatic selection method, device, computer-readable storage medium, and program for medical images.

Background Art

[0002] With the acceleration of the aging trend of society and the increase in the incidence rate of major diseases, higher clinical requirements have been placed on the quality and coverage area of medical examinations. Medical imaging equipment has become a priority disease detection means in clinical medicine due to its non-invasive, wide applicability, fast diagnosis, high accuracy, and other characteristics. Due to factors such as uneven allocation of medical resources and shortage of related personnel, medical institutions are urgently in need of expanding the functions of equipment through intelligentization to improve diagnostic efficiency and accuracy.

[0003] Currently, the selection of medical images is completely manual, which takes a long time, has strong subjectivity, a high miss rate, and requires a large amount of human and material resources from radiologists for repetitive work.

Summary of the Invention

Problems to be Solved by the Invention

[0004] To solve the above problems existing in the prior art, the present invention provides an automatic selection method, device, computer-readable storage medium, and program for medical images.

Means for Solving the Problems

[0005] The method for automatically selecting medical images according to the first aspect of the present invention is applied to a medical imaging device, and includes steps of: obtaining a set of medical images in which each medical image contains lesion image information; inputting the set of medical images into a classification model, and determining, based on the classification model, a target category of a lesion in each medical image in the set of medical images and a category score representing the probability that the lesion belongs to the target category; and selecting a target medical image from the set of medical images based on the category scores of the medical images in the set of medical images.

[0006] The automatic selection method according to the second aspect of the present invention is, in the first aspect, the step of determining the target category and category score of a lesion in each medical image in the set of medical images based on the classification model includes steps of: based on the classification model, displaying lesion prompt information representing the position and range of the lesion in the medical image on the medical image; and determining the target category and category score of the lesion indicated by the lesion prompt information.

[0007] The automatic selection method according to the third aspect of the present invention is, in the first or second aspect, the step of determining the target category and category score of a lesion in each medical image in the set of medical images includes a step of scoring the target category of the lesion based on an S-shaped growth curve to generate the category score of the lesion.

[0008] The automatic selection method according to the fourth aspect of the present invention is, in any one of the first to third aspects, the medical image displays one or more pieces of lesion prompt information, and the target category and category score of the lesion indicated by the one or more pieces of lesion prompt information.

[0009] The automatic selection method according to the fifth aspect of the present invention is, in any one of the first to fourth aspects, the classification model is obtained by a method of obtaining a training set including a plurality of medical images annotated with lesion categories and lesion attributes, and training a lesion detection model for detecting whether the medical image contains lesion information using the training set.

[0010] The medical image automatic selection device according to the sixth aspect of the present invention is applied to medical imaging equipment, and includes an acquisition module that acquires a set of medical images in which each medical image contains lesion image information, an input module that inputs the set of medical images into a classification model, and based on the classification model, determines a target category of the lesion in each medical image in the set of medical images and a category score representing the probability that the lesion belongs to the target category, and a selection module that selects a target medical image from the set of medical images based on the category scores of the medical images in the set of medical images.

[0011] The automatic selection device according to the seventh aspect of the present invention, in the sixth aspect, specifically, based on the classification model, the determination module displays lesion prompt information representing the position and range of the lesion in the medical image on the medical image, and determines the target category and category score of the lesion indicated by the lesion prompt information.

[0012] The automatic selection device according to the eighth aspect of the present invention, in the sixth or seventh aspect, specifically, based on an S-shaped growth curve, the determination module scores the target category of the lesion to generate a category score of the lesion.

[0013] The automatic selection device according to the ninth aspect of the present invention, in any one of the sixth to eighth aspects, displays one or more pieces of lesion prompt information, and the target category and category score of the lesion indicated by the one or more pieces of lesion prompt information on the medical image.

[0014] The automatic selection device according to the tenth aspect of the present invention, in any one of the sixth to ninth aspects, the classification model is obtained by acquiring a training set including a plurality of medical images annotated with lesion categories and lesion attributes, and training a lesion detection model that detects whether the medical image contains lesion information using the training set.

[0015] The medical image automatic selection device provided by the 11th aspect of the present invention includes a memory for storing instructions, and a processor that calls the instructions stored in the memory and executes the method described in any one of the above 1st to 5th aspects.

[0016] The computer-readable storage medium provided by the 12th aspect of the present invention includes computer program instructions. When the computer program instructions are read by a computer, the computer executes the method described in any one of the above 1st to 5th aspects.

[0017] The computer according to the 13th aspect of the present invention causes the computer to execute the method described in any one of the above 1st to 5th aspects.

Advantages of the Invention

[0018] The technical solution according to the present invention may include at least the following beneficial effects. By inputting a medical image set including lesion image information into a classification model, it is possible to determine the target category and category score to which the lesion included in each medical image in the medical image set belongs. The higher the category score, the higher the probability that the lesion belongs to the target category, indicating that it is suitable for diagnosis. Therefore, by selecting target medical images from the medical image set based on the category scores of each medical image, it is possible to avoid the drawback that in conventional operations, the most optimal medical images for diagnosis are not selected due to human factors, reduce the missed rate due to human factors caused by the overwork of medical staff, and reduce the complicated and repetitive diagnostic and analysis work of medical images for medical staff.

Brief Description of the Drawings

[0019] The above and other objects, features, and advantages of the present invention will be easily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, some embodiments and examples of the present invention are shown illustratively and not limitatively.

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

[0021] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

Embodiments for Carrying Out the Invention

[0022] Hereinafter, the principles and spirit of the present invention will be described with reference to several exemplary embodiments and examples. It should be understood that these embodiments are merely for those skilled in the art to better understand the present invention and do not limit the scope of the present invention in any way.

[0023] In the present text, expressions such as "first" and "second" are used to describe different modules, steps, and data of the embodiments of the present invention. However, the expressions such as "first" and "second" are only for distinguishing different modules, steps, and data, and do not indicate a specific order or importance. In fact, the expressions such as "first" and "second" may be used completely interchangeably.

[0024] Currently, the selection of medical images is completely manual, which takes a long time, has strong subjectivity, a high miss rate, and requires a large amount of human and material resources from radiologists for repetitive work.

[0025] Based on the above problems, the present disclosure provides a method for automatically selecting medical images, avoiding the drawback that in conventional operations, the most suitable medical images for diagnosis are not selected due to human factors, reducing the omission rate caused by human factors resulting from the overwork of medical staff, and alleviating the complicated and repetitive diagnostic and analysis work of medical images for medical staff.

[0026] FIG. 1 shows an exemplary system architecture 10 to which the embodiments of the present disclosure can be applied.

[0027] As shown in FIG. 1, the system architecture 10 may include terminal devices 11, 12, a network 13, and a server 14. The network 13 functions as a medium for providing a communication link between the terminal devices 11, 12 and the server 14. The network 13 may include various connection types such as wired, wireless communication links, or optical fiber cables.

[0028] The user 15 can interact with the server 14 via the network 13 using the terminal devices 11, 12 to send and receive messages, etc. The terminal devices 11, 12 may be various medical imaging devices including, but not limited to, imaging devices such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Computed Radiography (CR), Digital Radiography, and medical ultrasonic diagnostic devices. The server 14 may be a server that provides various services. The server can perform processes such as storage and analysis on the received data and feedback the processing results to the terminal devices.

[0029] It should be noted that the method for automatically selecting medical images according to the embodiments of the present application may be executed by a processor in the terminal devices 11, 12, or may be executed by the server 14, and the terminal devices 11, 12 may obtain the selection result by communicating and interacting with the server 14. The present disclosure does not strictly limit the specific application scenarios of the hardware in the method for selecting the medical images.

[0030] The models mentioned in the following description may be provided in the terminal devices 11 and 12, or may be provided in the server 14. In some embodiments, the model is trained in the server 14, and the trained model may be stored in the server 14 for screening medical images.

[0031] It should be understood that the numbers of the terminal devices, network, and server in FIG. 1 are merely exemplary. Any number of terminal devices, network, and server may be provided according to actual needs.

[0032] FIG. 2 is a flowchart of automatic screening of medical images according to an exemplary embodiment. As shown in FIG. 2, the following steps S1 to S3 are included.

[0033] In step S1, a medical image set in which each medical image includes lesion image information is obtained.

[0034] The medical image set may be medical images from computed tomography (CT), magnetic resonance imaging (MRI), computed radiography (CR), digital radiography (DR), medical ultrasonic diagnostic devices, etc., and the present disclosure is not specifically limited. The medical image set may be directly extracted from corresponding devices such as CT, MRI, and medical ultrasonic diagnostic devices, or may be obtained from a hospital database. In addition, the content of the medical image itself can be adjusted according to the needs of the actual application scenario. For example, it may be a transmission image in the scene of brain examination in neurosurgery or a transmission image in the scene of lung examination in thoracic surgery.

[0035] In step S2, the medical image set is input into a classification model, and based on the classification model, the target category and category score of the lesion in each medical image in the medical image set are determined.

[0036] The category score represents the probability that the lesion belongs to the target category.

[0037] It should be understood that the higher the category score, the higher the probability that the lesion belongs to the target category, and the more suitable the medical image is for diagnosis.

[0038] Preferably, there may be lesions of multiple categories, such as nodules, calcifications, scabs, etc. in one medical image, and based on the classification model, the target category and category score of each lesion can be determined.

[0039] It should be understood that the lesion categories corresponding to each part are not necessarily exactly the same. For example, the lesion categories corresponding to the thyroid include nodules, tumors, etc., and the lesion categories corresponding to the breast include calcifications, scabs, tumors, etc.

[0040] In step S3, based on the category scores of each medical image in the medical image set, the target medical image is selected from the medical image set.

[0041] Preferably, the category scores of each medical image are sorted in descending order, and the medical image with the highest category score is selected as the target medical image. Or, the N medical images with the highest category scores are selected as the target medical images.

[0042] In addition, when the medical image input into the classification model does not contain lesion image information, the target category of the lesion output by the classification model is "none", and the category score is 0.

[0043] By inputting a medical image set including lesion image information into a classification model, it is possible to determine the target category and category score to which the lesion included in each medical image in the medical image set belongs. The higher the category score, the higher the probability that the lesion belongs to the target category, indicating that it is suitable for diagnosis. Therefore, by selecting target medical images from the medical image set based on the category scores of each medical image, it is possible to avoid the drawback that in conventional operations, the most optimal medical images for diagnosis are not selected due to human factors, reduce the overlooking rate due to human factors caused by the overwork of medical staff, and reduce the complicated and repetitive diagnostic and analysis work of medical images for medical staff.

[0044] In some embodiments, based on the classification model, lesion presentation information is displayed on the medical image, and the target category and category score of the lesion indicated by the lesion presentation information are determined.

[0045] The lesion presentation information represents the position and range of the lesion in the medical image. Preferably, the lesion presentation information includes first presentation information and second presentation information, where the first presentation information represents the presence of the lesion, the approximate position, and the approximate range of the lesion in the medical image, and the second presentation information may represent the specific position and specific range of the lesion in the medical image.

[0046] The first presentation information may be represented by a regular shape such as a square or a rectangle, and the second presentation information may be represented by an irregular polygon.

[0047] Furthermore, after determining the specific position and specific range of the lesion in the medical image, the target category of the lesion is determined based on the classification model, and the target category of the lesion is scored based on the S-shaped growth curve (Sigmoid function) to generate the category score of the lesion.

[0048] Preferably, when a single medical image includes a plurality of lesions, the target categories of each lesion are scored based on the S-shaped growth curve to generate the category scores of each lesion.

[0049] The Sigmoid function is a non-linear activation function of neurons and is widely used in neural networks.

[0050] The learning of a neural network is performed based on a set of samples, including inputs and outputs (represented here as desired outputs), and the number of components of the inputs and outputs corresponds to the number of neurons in the inputs and outputs. The weights and thresholds of the first neural network are arbitrarily given, and the learning is to gradually adjust the weights and thresholds so that the actual output of the network matches the desired output.

[0051] In some embodiments, one or more lesion presentation information and the target category and category score of the lesion indicated by the one or more lesion presentation information are displayed on a medical image.

[0052] Preferably, a score threshold is preset, and only category scores greater than the score threshold are displayed on the medical image.

[0053] Preferably, in the form of a comment box, the target category and category score of the lesion are displayed on the medical image. Specifically, a frame is added outside the lesion position, and more specifically, a rectangular frame can be added, and the category and category score of the lesion are represented by the color of the rectangular frame corresponding to the category. Alternatively, the position of the lesion is indicated by increasing the luminance of the lesion area, the category score is represented by a specific luminance change, and the target category of the lesion is represented by text presentation.

[0054] Furthermore, when multiple lesion presentation information is displayed on a medical image, when selecting a target medical image, the target image information is selected according to the target category of each lesion. For example, there are three medical images. The first medical image contains one lesion with a target category of nodule and a category score of 85. The second medical image contains two lesions, with target categories of nodule and tumor respectively. The category score of the nodule is 90, and the category score of the tumor is 60. The third medical image contains two lesions, with target categories of nodule and tumor respectively. The category score of the nodule is 20, and the category score of the tumor is 70. If only the medical image with the highest category score is used as the target medical image, both the second and third medical images will be the target medical images. The second medical image is the target medical image for the nodule category, and the third medical image is the target medical image for the tumor category.

[0055] Hereinafter, the training process of the classification model will be described. As shown in FIG. 3, the training process of the classification model includes the following steps S21 to S22.

[0056] In step S21, a training set is obtained.

[0057] The training set includes a plurality of medical images annotated with lesion categories and lesion attributes.

[0058] In step S22, a lesion detection model is trained using the training set to obtain a classification model.

[0059] The lesion detection model detects whether lesion information is included in a medical image.

[0060] Preferably, in the training process, first, the position of the lesion included in the medical image is determined, the coordinates of the position of the lesion are obtained, and the coordinates of the position of the lesion and the lesion attributes are associated with the corresponding lesion category in the medical image.

[0061] Preferably, the lesion detection model can use a convolutional neural network architecture, a fully connected neural network architecture, etc. The specific type and architecture can be adjusted according to the needs of the actual application scenario, and the present application does not strictly limit the specific type and architecture of the lesion detection model.

[0062] Taking the convolutional neural network model as an example, the classification model in the embodiments of the present disclosure is based on a convolutional neural network. The convolutional neural network has a weight sharing characteristic, and the weight sharing refers to a convolutional kernel. The same feature at different positions of the image data can be extracted by the operation of one convolutional kernel. In other words, even at different positions of one piece of image data, the features of the same object are basically the same. Although only some features can be obtained by using one convolutional kernel, it will be understood that by setting multi-kernel convolution, different features can be learned using each convolutional kernel to extract the features of the image. In image classification, the convolutional layer functions to extract and analyze low-level features and turn them into high-level features. The low-level features are basic features, such as features like texture and edge, and the high-level features are, for example, the shape of a face or an object, etc., which can more specifically represent the attributes of the sample. This process is the hierarchy of the convolutional neural network.

[0063] Furthermore, after training the lesion detection model to obtain a classification model, a test set can be set to test the classification model. If the error of the test result is not within the allowable range, retrain the lesion detection model until the error of the test result is within the allowable range.

[0064] Preferably, after determining the category of the lesion in the medical image using the classification model, the classification model can also be retrained using the medical image in which the target category of the lesion is determined.

[0065] Embodiments of the present disclosure can obtain the target category to which a lesion belongs and the category score by training a classification model, avoiding the drawback that in conventional operations, medical images optimal for diagnosis are not selected due to human factors, reducing the omission rate due to human factors caused by the overwork of medical staff, and alleviating the complicated and repetitive diagnostic and analysis work of medical images by medical staff.

[0066] Based on the same concept, embodiments of the present disclosure further provide an automatic selection device for medical images.

[0067] It will be understood that the automatic selection device for medical images according to embodiments of the present disclosure includes corresponding hardware structures and / or software modules that execute each function in order to implement the above functions. Based on each example of the units and algorithm steps disclosed in the embodiments of the present disclosure, the embodiments of the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software drives the hardware is determined by the specific application of the technical solution and the design constraints. A person skilled in the art can implement the functions described in different ways for each specific application, but such implementation should not be understood as exceeding the scope of the technical solution of the embodiments of the present disclosure.

[0068] FIG. 4 is a block diagram of an automatic selection device for medical images according to an exemplary embodiment. As shown in FIG. 4, the device includes an acquisition module 401, an input module 402, and a selection module 403.

[0069] The acquisition module 401 acquires a set of medical images in which each medical image includes lesion image information.

[0070] The input module 402 inputs the set of medical images into the classification model and determines, based on the classification model, the target category of the lesion in each medical image in the set of medical images and the category score representing the probability that the lesion belongs to the target category.

[0071] The sorting module 403 sorts the target medical images from the set of medical images based on the category scores of each medical image in the set of medical images.

[0072] In one embodiment, specifically, the acquisition module 401 displays, on the medical image, lesion presentation information representing the location and scope of a lesion in the medical image based on a classification model, and determines the target category and category score of the lesion indicated by the lesion presentation information.

[0073] In another embodiment, specifically, the acquisition module 401 scores the target category of the lesion based on an S-shaped growth curve to generate a category score of the lesion.

[0074] In yet another embodiment, one or more pieces of lesion presentation information and the target category and category score of the lesion indicated by the one or more pieces of lesion presentation information are displayed on the medical image.

[0075] In yet another embodiment, the classification model is obtained by acquiring a training set including a plurality of medical images annotated with lesion categories and lesion attributes, and training a lesion detection model that detects whether the medical image contains lesion information using the training set.

[0076] FIG. 5 is a block diagram of an automatic sorting device 100 for medical images according to an exemplary embodiment. For example, the device 100 may be a medical imaging device.

[0077] Specifically, the medical imaging device 100 may include components such as a processor 101, a communication device 102, a display 103, a power supply 104, a memory 105, an audio circuit 106, a multimedia device 107, a sensor 108, and a peripheral interface 109. Those skilled in the art will understand that the hardware structure shown in FIG. 5 does not limit the medical imaging device, and the medical imaging device may include more or fewer components than shown in the figure, some components may be combined, and different component arrangements may be provided.

[0078] Processor 101 is the control center of the medical imaging device. It is connected to each part of the mobile phone via various interfaces and lines, and operates or executes the application programs (hereinafter may be abbreviated as Apps) stored in memory 105, and calls the data stored in memory 105, thereby executing various functions and data processing of the medical imaging device. In some embodiments, processor 101 may include one or more processing units.

[0079] Communication device 102 transmits and receives wireless signals during information transmission / reception or a call. In particular, after receiving the downlink data from the base station, communication device 102 can cause processor 101 to process it. Also, it transmits uplink data to the base station. Usually, the radio frequency circuit includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc. Also, communication device 102 can communicate with other devices via wireless communication. The above wireless communication can use any communication standard or protocol including, but not limited to, global mobile communication system, general packet radio service, code division multiple access, wideband code division multiple access, long term evolution, email, short message service, etc.

[0080] Display 103 may include a liquid crystal display (LCD) and a touch panel (TP). When the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The above touch sensors can not only sense the boundaries of touch or swipe operations, but also detect the duration and pressure associated with the above touch or swipe operations.

[0081] Power supply 104 is logically connected to processor 101 via a power management chip to realize functions such as charging, discharging, and power consumption management.

[0082] Memory 105 stores application programs and data, and processor 101 executes the application programs and data stored in memory 105 to perform various functions and data processing of the mobile phone. Memory 105 mainly includes a program storage area and a data storage area. The program storage area can store an operating system and application programs required for at least one function (such as a voice playback function, an image playback function, etc.). The data storage area can store data created during the use of the mobile phone (such as audio data, phone book, etc.). Also, memory 105 may include a high-speed random access memory, and may further include a non-volatile memory such as a magnetic disk memory device, a flash memory device, or other volatile solid memory devices. Memory 105 can store various operating systems such as the iOS operating system developed by Apple Inc. and the Android operating system developed by Google Inc.

[0083] The audio circuit 106 is configured to output and / or input audio signals.

[0084] The multimedia device 107 may include a Wi-Fi device, a Bluetooth (registered trademark) assembly, etc. The Wi-Fi device provides network access to the mobile phone according to Wi-Fi-related standard protocols. The mobile phone accesses a Wi-Fi access point via the Wi-Fi device to assist in sending and receiving the user's e-mails, browsing web pages, accessing streaming media, etc., and provides the user with wireless broadband Internet access. In some other embodiments, the Wi-Fi device can also provide Wi-Fi network access to other terminals as a Wi-Fi wireless access point.

[0085] The medical imaging device may further include at least one sensor 108 such as an optical sensor, a motion sensor, and other sensors. Specifically, the optical sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display 103 according to the brightness of the ambient light, and the proximity sensor can turn off the power of the display when the medical imaging device moves close to the ear. As a type of motion sensor, the acceleration sensor can detect the magnitude of acceleration in each direction (generally three axes), can detect the magnitude and direction of gravity at rest, and can be used for applications that identify the posture of a mobile phone (for example, switching between portrait / landscape, related games, attitude calibration of a magnetometer), functions related to vibration identification (for example, pedometer, tapping), etc. Other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, an infrared sensor, etc. may be further arranged in the mobile phone, and the description thereof is omitted here.

[0086] The peripheral interface 109 provides various interfaces to external input / output devices (for example, a keyboard, a mouse, an external display, an external memory, a subscriber identification module card, etc.). For example, it is connected to a mouse or a display via a Universal Serial Bus (USB) interface, connected to a Subscriber Identity Module (SIM) card provided by an electric communication operator via the metal contacts of a subscriber identification module card slot, and realizes a communication function with other terminals via the interface of a Wi-Fi device, the interface of a Near Field Communication (NFC) device, the interface of a Bluetooth (registered trademark) module, etc. The peripheral interface 109 can couple the above external input / output devices to the processor 101 and the memory 105.

[0087] Although not shown in FIG. 5, the medical imaging device may further include a camera, a flash, a micro projection device, a Near Field Communication (NFC) device, etc., and the description thereof is omitted here.

[0088] FIG. 6 is a block diagram of an automatic selection device 200 for medical images according to an exemplary embodiment. For example, the device 200 may be provided as a server. As shown in FIG. 6, the device 200 includes a memory 201, a processor 202, and an input / output (I / O) interface 203. The memory 201 stores instructions. The processor 202 calls the instructions stored in the memory 201 to execute the automatic selection method for medical images according to the embodiments of the present disclosure. The processor 202 is connected to the memory 201 and the I / O interface 203 respectively, and may be connected via, for example, a bus system and / or other forms of connection mechanisms (not shown). The memory 201 stores a program including a program of the footprint trajectory display method according to the embodiments of the present disclosure and data. The processor 202 executes the various functional applications and data processing of the device 200 by executing the program stored in the memory 201.

[0089] In an embodiment of the present disclosure, the processor 202 may be implemented in at least one hardware form of a digital signal processor (DSP), a field-programmable gate array (FPGA), or a programmable logic array (PLA). The processor 202 may be one or a combination of a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities.

[0090] The memory 201 in the embodiments of the present disclosure may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, Random Access Memory (RAM) and / or cache memory. The non-volatile memory may include, for example, Read-Only Memory (ROM), Flash Memory, Hard Disk Drive (HDD), or Solid-State Drive (SSD).

[0091] In the embodiments of the present disclosure, the I / O interface 203 can receive the input instructions (such as numerical or character information) and generate the input of key signals related to the user settings and function control of the device 200, and can also output various information (such as images or sounds) externally. In the embodiments of the present disclosure, the I / O interface 203 may include one or more of a physical keyboard, function buttons (such as volume control buttons, switch buttons, etc.), a mouse, a joystick, a trackball, a microphone, a speaker, and a touch panel.

[0092] The embodiments of the present disclosure further provide a computer-readable storage medium, which includes computer-executable instructions. When the computer-executable instructions are executed by a computer, the computer executes the method for automatically selecting medical images according to the above embodiments.

[0093] In the embodiments of the present invention, although the operations are described in a specific order in the drawings, it should be understood that it is not required to execute these operations in the specific order or serial order shown, or to execute all the operations shown, in order to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0094] The method and apparatus according to embodiments of the present invention can be implemented using standard programming techniques, and various method steps can be realized using rule-based logic or other logic. It should be noted that the terms "apparatus" and "module" used in this specification and the claims are intended to include implementations using one or more lines of software code and / or hardware implementations, and / or devices that receive inputs.

[0095] Any step, operation, or program described in this specification may be executed or realized using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software module is implemented using a computer program product including a computer-readable medium containing computer program code, and via a computer processor, any or all of the described steps, operations, or programs can be executed.

[0096] The above description regarding the implementation of the present invention is provided for purposes of illustration and explanation. The foregoing description is not exhaustive, nor is it intended to limit the present invention to the exact form disclosed, and various modifications and variations are possible based on the above teachings, or various modifications and variations can be obtained from the implementation of the present invention. These examples have been selected and described to explain the principles of the present invention and its practical applications, whereby those skilled in the art can utilize the present invention with various embodiments and various changes suitable for the intended specific uses.

Description of Reference Numerals

[0097] 10 System architecture 11, 12 Terminal devices 13 Network 14 Server 15 User 100 Medical imaging device, automatic sorting device 101 Processor 102 Communication device 103 Display 104 Power supply 105 Memory 106 Audio circuit 107 Multimedia device 108 Sensor 109 Peripheral interface 200 Automatic sorting device 201 Memory 202 Processor 203 Interface 401 Acquisition module 402 Input module 403 Sorting module

Claims

1. Applied to a medical imaging device, obtaining a set of medical images in which each medical image contains lesion image information; inputting the set of medical images into a classification model, and based on the classification model, determining a target category of a lesion in each medical image in the set of medical images and a category score representing the probability that the lesion belongs to the target category; selecting a target medical image from the set of medical images based on the category scores of each medical image in the set of medical images, characterized in that it is an automatic selection method for medical images.

2. Based on the classification model, the step of determining the target category and category score of a lesion in each medical image in the set of medical images is: displaying lesion presentation information representing the position and range of the lesion in the medical image on the medical image based on the classification model; determining the target category and category score of the lesion indicated by the lesion presentation information, characterized in that it is the method according to claim 1.

3. Based on the classification model, the step of determining the target category and category score of a lesion in each medical image in the set of medical images is: including the step of scoring the target category of the lesion based on an S-shaped growth curve to generate the category score of the lesion, characterized in that it is the method according to claim 1 or 2.

4. displaying one or more pieces of lesion presentation information and the target category and category score of the lesion indicated by the one or more pieces of lesion presentation information on the medical image, characterized in that it is the method according to claim 3.

5. The classification model is: obtaining a training set including a plurality of medical images annotated with lesion categories and lesion attributes; obtained by training a lesion detection model for detecting whether the medical image contains lesion information using the training set, characterized in that it is the method according to claim 1.

6. Applied to a medical imaging device, an acquisition module for obtaining a set of medical images in which each medical image contains lesion image information; an input module for inputting the set of medical images into a classification model and, based on the classification model, determining a target category of a lesion in each medical image in the set of medical images and a category score representing the probability that the lesion belongs to the target category; A selection module that selects a target medical image from the medical image set based on the category score of each medical image in the medical image set. A medical image automatic selection device, characterized in that it includes the above.

7. Specifically, the acquisition module displays lesion presentation information representing the position and range of a lesion in the medical image on the medical image based on the classification model, and determines the target category and category score of the lesion indicated by the lesion presentation information. The device according to claim 6, characterized in that.

8. Specifically, the acquisition module scores the target category of the lesion based on an S-shaped growth curve to generate the category score of the lesion. The device according to claim 6 or 7, characterized in that.

9. Displaying one or more pieces of lesion presentation information and the target category and category score of the lesion indicated by the one or more pieces of lesion presentation information on the medical image. The device according to claim 8, characterized in that.

10. The classification model is obtained by acquiring a training set including a plurality of medical images annotated with lesion categories and lesion attributes, and training a lesion detection model that detects whether the medical image contains lesion information using the training set. The device according to claim 6, characterized in that.

11. A memory for storing instructions; A processor that calls instructions stored in the memory and executes the method according to any one of claims 1 to 5. A medical image automatic selection device, characterized in that it includes the above.

12. Including computer program instructions, when the computer program instructions are read by a computer, the computer executes the method according to any one of claims 1 to 5. A computer-readable storage medium, characterized in that.

13. A program for causing a computer to execute the method according to any one of claims 1 to 5.

Images