TWI696441B - Smart marking system for surgical video and method thereof - Google Patents

Abstract

A smart marking system for surgical video and method thereof is disclosed. By performing image recognition for each of the surgical videos, so that generates a plurality of object as a feature data corresponding to each of the surgical videos. After loading a video to be analyzed, performing eigenvalue analysis and statistics according to the video to be analyzed and the feature data to calculate a similarity rate between the video to be analyzed and each of the surgical videos, and selecting a surgical name corresponding to the surgical videos with the highest similarity rate to make the video to be analyzed, and storing the video to be analyzed and the surgical name in a database as one of the surgical videos. The mechanism is help to improve the efficiency of marking the surgical videos.

Description

Intelligent marking system and method for surgical images

The invention relates to a marking system and method, in particular to a smart marking system and method for surgical images.

In recent years, with the popularization and vigorous development of medical technology, various related applications have sprung up, among which the value-added applications of surgical images have attracted the most attention.

Generally speaking, surgical images can be provided to doctors, students or related practitioners to learn, and can even be used as surgical records to avoid medical disputes. However, due to the variety of operations, in order to facilitate browsing, it is necessary to mark the surgical images, such as: records Surgery, organs, surgeons, etc. In this way, all relevant surgical images can be quickly queried based on the marked information. However, in the past, the methods of manual judgment and marking are prone to problems such as a large increase in error rate and poor efficiency due to fatigue.

In view of this, some manufacturers have proposed a technology that combines image recognition, which uses organ recognition technology to automatically classify and mark the organs that appear in the surgical image, thereby improving the labeling error rate and improving the labeling efficiency. However, this method can only roughly classify the organs, and cannot further identify the surgical methods used. Therefore, it still needs to rely on professional medical personnel to judge the surgical methods and perform supplementary marking, so there is still an inefficiency in labeling surgical images. Good question.

In summary, it can be seen that the prior art has long had the problem of poor efficiency in marking surgical images, so it is necessary to propose improved technical means to solve this problem.

The invention discloses a smart marking system and method for surgical images.

First, the present invention discloses a smart marking system for surgical images. The system includes: a database, a feature generation module, an analysis module, and a storage module. Among them, the database is used to store surgical images, and each surgical image corresponds to the name of the operation method; the feature generation module is connected to the database to initially load the surgical images of the database for image recognition, and according to The image recognition result generates corresponding object information, and stores the object information as the feature data of the corresponding surgical image in the database, wherein each object information includes the coordinates, scale and time point of the image object existing in the surgical image; analysis The module connection database is used to perform feature value analysis and statistics based on the image to be analyzed and the feature data after loading the image to be analyzed, to calculate the similarity rate between the image to be analyzed and each surgical image; store the module connection analysis The module is used to select the operation name corresponding to the operation image with the highest similarity rate and mark it as the operation name corresponding to the analysis image, and store the analysis image and the marked operation name together in the database as the operation image one of them.

In addition, the present invention discloses a method for intelligently marking surgical images. The steps include: providing surgical images in a database, each surgical image corresponding to a surgical procedure name; at the beginning, the surgical images in the database are separately loaded for image recognition , And generate corresponding object information according to the image recognition results, and store the object information as the feature data of the corresponding surgical image and store it in the database, where each object information includes the coordinates, scale and time of the image object existing in the surgical image After loading the image to be analyzed, perform feature value analysis and statistics based on the image to be analyzed and feature data to calculate the similarity rate between the image to be analyzed and each surgical image; and select the surgical image corresponding to the highest similarity rate The operation name is marked as the operation name corresponding to the image to be analyzed, and the operation name and the image to be analyzed are stored in the database as one of the operation images.

The system and method disclosed in the present invention are as above, and the difference from the prior art is that the present invention performs image recognition to generate object information corresponding to the surgical image, and uses the generated object information as the feature data corresponding to the surgical image for use in After loading the image to be analyzed, perform feature value analysis and statistics based on the image to be analyzed and the feature data, so as to calculate the similarity rate between the image to be analyzed and each surgical image, and select the surgical procedure name corresponding to the surgical image with the highest similarity rate to It is marked as the name of the operation corresponding to the image to be analyzed, and the image to be analyzed and the name of the corresponding operation are stored in the database as the operation image.

Through the above technical means, the present invention can achieve the technical effect of improving the efficiency of marking surgical images.

The embodiments of the present invention will be described in detail below in conjunction with the drawings and examples, so as to fully understand and implement the implementation process of how the present invention uses technical means to solve technical problems and achieve technical effects.

Before describing the intelligent marking system and method of the surgical images disclosed by the present invention, the nouns defined by the present invention will be described first. The “surgical name” in the present invention refers to the name of the surgical method, for example: “uterus "Resection", "Myomectomy" and so on. As for the "object information", the information is generated by image recognition technology such as organs and surgical instruments. For example, the organ recognition in the surgical image is recognized by organ recognition technology, and the image of this organ is captured as The image object, and the image object and its coordinates, scale, and time point existing in the surgical image are used as the object information. In other words, an object that can be recognized by an image is called an object, and the object information records detailed information about the object, such as the coordinates, scale, and time point of occurrence.

The smart labeling system and method of the surgical image of the present invention will be further described below with reference to the drawings. Please refer to "Figure 1", which is a system block diagram of the smart labeling system of the surgical image of the present invention. This system includes : Database 110, feature generation module 120, analysis module 130 and storage module 140. The database 110 is used to store surgical images, and each surgical image corresponds to a surgical name. In actual implementation, the database 110 can be implemented using a relational database (Relational Database), a NoSQL database, and so on. In addition, the database 110 can also store organ characteristic values and surgical instrument characteristic values for loading when performing image recognition on the surgical images, and extract the operation organ images or surgical instrument characteristic values that match the loaded organ characteristic values or surgical instrument characteristic values The image is used as the image object of the object information, and the coordinates, scale and time point of the image object existing in the surgical image are recorded in the object information. Among them, the organ characteristic value and the surgical instrument characteristic value can be generated through a neural network (Neural Network, ANN), deep learning (Deep Learning), etc. For example, a large number of images of organs and surgical instruments are used as training in advance Data, and then generate corresponding characteristic values, namely: organ characteristic value and surgical instrument characteristic value.

The feature generation module 120 is connected to the database 110 to load the surgical images of the database 110 for image recognition at the initial stage, and generate corresponding object information according to the image recognition results, and use the object information as the corresponding surgical image The feature data is stored in the database 110, wherein each object information includes the coordinates, scale, and time point of the image object existing in the surgical image. In actual implementation, the generated object information can only retain the parts related to the key organs, and the object information of other non-key organs is deleted. The key organs are determined according to the coordinates and scale of the image objects, for example: The image object is located in the center of the surgical image or the proportion occupied is greater than the preset value, which represents the key organ. In addition, the feature generation module 120 can also load all the surgical images with the same surgical name to identify the type of surgical instruments and the type, shape and size of abnormal organs, and generate according to the recognition result The starting-hand feature value corresponding to the operation name, and the starting-hand feature value and the corresponding operation name are stored in the database 110 together. In actual implementation, the generation method of the starting-hand characteristic value is the same as the generation method of the organ characteristic value and the surgical instrument characteristic value described above.

The analysis module 130 is connected to the database 110, and is used to perform feature value analysis and statistics based on the image to be analyzed and feature data after loading the image to be analyzed to calculate the similarity rate between the image to be analyzed and each surgical image. The image to be analyzed refers to a surgical image that has not been marked with the name of the surgical procedure. In actual implementation, since the feature generation module 120 has generated corresponding feature data based on each surgical image, the analysis module 130 analyzes and counts the feature image and the feature data to calculate the pending The similarity rate between the analysis image and each surgical image, for example: the similarity rate between the image to be analyzed and the first surgical image is "90%", the similarity rate between the image to be analyzed and the second surgical image is "40%", etc. . In addition, when the database 110 stores the hand-held feature value, the analysis module 130 can also compare the loaded image to be analyzed with the hand-held feature value in the database 110, and load the matching The operation name corresponding to the starting-hand characteristic value. When the loaded operation name is different from the operation name selected according to the similarity ratio, a prompt message is generated and one of them can be selected as the operation name corresponding to the image to be analyzed. For example, suppose the name of the loaded operation is "total hysterectomy", and the name of the operation selected according to the similarity rate is "secondary hysterectomy". At this time, a prompt message is generated for display and the user is allowed to One of them can be selected as the name of the operation corresponding to the image to be analyzed, and the user can even input text to specify the operation name. In particular, in practical implementation, the analysis module 130 can also determine whether there are solid tissues or tumor-edge cells based on the key organs (such as the organ occupying the largest proportion of the image or the organ with the central coordinate position, etc.) Adjust the similarity rate based on the abnormality. Take the image to be analyzed, the first surgical image and the second surgical image as examples, suppose that the key organ of the image to be analyzed does not have solid tissue, and the key organ of the first surgical image has solid Organization, then the similarity rate between the image to be analyzed and the first surgical image can be reduced, for example: from "90%" to "80%". Conversely, if the abnormalities of all solid tissues or tumor marginal cells are similar, the similarity rate can be increased. Although the present invention illustrates the adjustment method of the similarity rate as described above, the present invention is not limited to this. Any method for further adjusting the similarity rate according to the differences in organs, tissues, cells or surgical instruments appearing in the image will not deviate from this. The scope of application of the invention. For example, the type of surgical instruments, environmental judgment (eg: ovaries), abnormal organs (eg: cysts), medical statistical information (eg: about 80% to 80% of ovarian tumors are benign), and instant images of fluid outflow The proportion and liquid color are used as the analysis parameters.

The storage module 140 is connected to the analysis module to select the surgical procedure name corresponding to the surgical image with the highest similarity rate to be marked as the surgical procedure name corresponding to the to-be-analyzed image, and to store the to-be-analyzed image and the marked surgical procedure name together To the database 110 as one of the surgical images. For example, if the image to be analyzed has the highest similarity to the third surgical image, and the name of the surgical procedure corresponding to the third surgical image is "myomectomy", then the storage module 140 will refer to the third The operation name corresponding to the surgical image "myomectomy" is marked as the operation name corresponding to the image to be analyzed, and the image to be analyzed and its corresponding operation name "myomectomy" are stored in the data The library 110 is one of the surgical images. In this way, after inputting the image to be analyzed, the corresponding image name of the image to be analyzed can be quickly marked, and can even be classified and stored according to the marked operation name, for example: the same operation name is the same type, save In the same folder.

In addition, the system of the present invention may further include a segmentation module 150 for detecting that the organ characteristic value and the surgical instrument characteristic value exist in the surgical image when the organ characteristic value and the surgical instrument characteristic value are stored in the database 110 at least At one of the time points, and when the time interval between adjacent time points is greater than the preset interval, the frame corresponding to the time interval in the surgical image is deleted to form a segmented image and stored in the database 110. In other words, the images of organs, surgical instruments, or both are kept, otherwise, they are deleted. In actual implementation, the image in which both exist is usually retained as the key image. With the above method, unimportant images can be deleted, and even multiple divided images can be reorganized into a single image to achieve the effect of cutting out the best fragments.

In particular, in actual implementation, each module described in the present invention can be implemented in various ways, including software, hardware, or any combination thereof. For example, in some embodiments, each module may Using software and hardware or one of them, in addition to this, the present invention can also be implemented partially or completely based on hardware, for example, one or more modules in the system can be through integrated circuit chips, System on Chip (SoC), Complex Programmable Logic Device (CPLD), Field Programmable Gate Array (FPGA), etc. are implemented. The invention may be a system, method and/or computer program. The computer program may include a computer-readable storage medium loaded with computer-readable program instructions for causing the processor to implement various aspects of the present invention. The computer-readable storage medium may be a tangible form that can hold and store instructions used by the instruction execution device equipment. The computer-readable storage medium may be, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (non-exhaustive lists) of computer-readable storage media include hard disks, random access memory, read-only memory, flash memory, optical disks, floppy disks, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, optical signals through fiber optic cables), or through wires The transmitted electrical signal. In addition, the computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing/processing devices, or via a network, such as the Internet, regional networks, wide area networks, and/or wireless networks To external computer equipment or external storage devices. The network may include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, hubs, and/or gateways. The network card or network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for computer-readable storage media stored in each computing/processing device in. The computer program instructions to perform the operations of the present invention may be combined language instructions, instruction set architecture instructions, machine instructions, machine-related instructions, microinstructions, firmware instructions, or source code or object code written in any combination of one or more programming languages (Object Code), the programming language includes object-oriented programming languages, such as: Common Lisp, Python, C++, Objective-C, Smalltalk, Delphi, Java, Swift, C#, Perl, Ruby, PHP, etc., as well as conventional programs Procedural programming language, such as: C language or similar programming language. Computer readable program instructions can be executed entirely on the computer, partly on the computer, as a stand-alone software, partly on the client computer and partly on the remote computer, or entirely on the remote computer or server To execute.

Please refer to "Picture 2A" to "Picture 2C". "Picture 2A" to "Picture 2C" are flow charts of the method for intelligently labeling surgical images of the present invention. The steps include: providing surgical images in the database 110 , Each surgical image corresponds to a surgical procedure name (step 210); at the beginning, the surgical images of the database 110 are loaded for image recognition, and the corresponding object information is generated according to the image recognition result, and the object information is used as the corresponding The characteristic data of the surgical image is stored in the database 110, wherein each object information includes the coordinates, scale and time point of the image object existing in the surgical image (step 220); when the image to be analyzed is loaded, according to this Analyze images and feature data to perform feature value analysis and statistics to calculate the similarity rate between the image to be analyzed and each surgical image (step 230); select the surgical procedure name corresponding to the surgical image with the highest similarity rate and mark it as the image to be analyzed The name of the operation, and store the image to be analyzed and the name of the operation in the database 110 as one of the operation images (step 240). Through the above steps, you can perform image recognition to generate object information corresponding to the surgical image, and use the generated object information as the feature data corresponding to the surgical image, used to load the image to be analyzed, according to the image and feature to be analyzed Perform eigenvalue analysis and statistics on the data to calculate the similarity between the image to be analyzed and each surgical image, and select the name of the surgical method corresponding to the surgical image with the highest similarity to mark the name of the surgical method corresponding to the image to be analyzed, and The image to be analyzed and the name of the corresponding operation are stored in the database 110 as the operation image.

In addition, as shown in "Figure 2B", after step 220, all the surgical images of the same surgical name can be loaded to identify the type of surgical instruments and the type, shape and size of abnormal organs, and according to The recognition result generates a starting-hand feature value corresponding to the surgical name, and stores the starting-hand feature value and the corresponding surgical name together in the database 110 (step 221); and after step 240, the loaded The image to be analyzed is compared with the starting hand feature value, and the operation name corresponding to the matching starting hand feature value is loaded. When the loaded operation name is different from the operation name selected according to the similarity rate, Prompt messages are generated and one of them is allowed to be selected as the name of the operation corresponding to the image to be analyzed (step 241). In addition, as shown in "Figure 2C", after step 210, the database 110 can further provide organ characteristic values and surgical instrument characteristic values (step 211); and detect the presence of organ characteristic values and surgery in the surgical image At least one of the instrument feature values, and when the time interval between adjacent time points is greater than the preset interval, delete the frame corresponding to the time interval in the surgical image to form a segmented image and store it in the database 110 (step 212).

The following description will be made in conjunction with "Figure 3" to "Figure 5" by way of example. Please refer to "Figure 3" first. "Figure 3" is the name of the surgical method for judging the image to be analyzed and applied by the present invention. Schematic diagram of the mark. It is assumed that at the initial stage, corresponding feature data has been generated for each surgical image of the database 110. After the user opens the display window 300 and clicks on the loading element 321 with the cursor, the image to be marked can be selected for loading The image to be analyzed, and display the image to be analyzed in the playback block 310, and then, the user can click on the analysis component 322 to perform feature value analysis and statistics on the image to be analyzed and the feature data in the database 110, so that Calculate the similarity rate between the image to be analyzed and each surgical image. Suppose that two surgical images have been stored in the database 110, of which the name of the surgical procedure corresponding to the first surgical image is "myomectomy" and the name of the surgical procedure corresponding to the second surgical image is "lipoma" "Shaving", after calculating the similarity rate, the similarity rate between the image to be analyzed and the first surgical image is "90%", and the similarity rate between the image to be analyzed and the second surgical image is "10%" . At this time, as shown in "Figure 3", according to the level of the similarity rate, the surgical procedure name corresponding to the surgical image and the corresponding similarity rate are sequentially displayed in the surgical name block 330, and the similarity rate is directly The name of the surgical procedure corresponding to the highest surgical image, "Myomectomy" is marked as the name of the surgical procedure corresponding to the image to be analyzed. In particular, assuming that the similarity ratio is very close, the user can also click the playback component 323 to play the image to be analyzed, and then directly select the appropriate operation in the operation name block 330 according to the browsing result of the image to be analyzed The style name is marked as the name of the surgical style corresponding to the image to be analyzed. After the marking is completed, the image to be analyzed and the name of the corresponding surgical procedure can be stored in the database 110 as a surgical image. In addition, in order to improve the efficiency of labeling, the above process can also be processed using batch operations. For example, multiple images to be analyzed are simultaneously loaded and labeled one by one. The labeling method is to directly place the surgical image with the highest similarity rate. The corresponding operation name is marked as the operation name corresponding to the current image to be analyzed, and it is not necessary to provide the user to select the operation name and browse the image to be analyzed, so the marking efficiency can be greatly improved.

As shown in "Figure 4", "Figure 4" is a schematic diagram of applying the present invention to generate a hand-held feature value. Assuming that the database 110 pre-stores a plurality of surgical images, and each surgical image corresponds to a surgical name, the user can first load all surgical images with the same surgical name to identify the surgical instruments in the image The type and the shape, size and size of the abnormal organ, and generate a starting characteristic value corresponding to the name of the operation according to the recognition result, and store the starting characteristic value and the corresponding operation name together in the database 110. Taking the name of the surgical procedure as "shaving of uterine fibroids" as an example, the user can first load the beginning of all surgical images (410, 420 and 430) corresponding to the name of the surgical procedure, and use these loaded parts as Training samples in order to use neural network-like or deep learning methods to generate corresponding starting-hand feature values. During the training process, all surgical images have the same kind of first surgical instrument, namely: first surgical instrument (412, 422, 432) and the same kind of second surgical instrument, namely: second surgical instrument (413, 423, 433), and also has uterus (414, 424, and 434) and fibroids (411, 421, and 431). Therefore, even if there are slight differences in position, size, and shape, after training with neural-like networks or deep learning, it is still possible to generate the starting-hand eigenvalues corresponding to uterine fibroids shaving. In other words, in the surgical image, under the premise that the starting hand of the same operation is similar, the corresponding starting hand feature value can be established by neural network or deep learning, and the same can be used to establish each A surgical method and its corresponding starting hand characteristic value.

Please refer to "Figure 5". "Figure 5" is a schematic diagram of applying the present invention to determine the name of the operation type in combination with the hand-type characteristic value. After all the surgical procedure names have generated corresponding starting-hand feature values, the analysis module 130 performs feature value analysis and statistics on the image to be analyzed and the feature data to calculate the similarity rate between the image to be analyzed and the surgical image In addition, you can compare the loaded image to be analyzed with all the hands-on feature values, and load the operation name corresponding to the hands-on feature value that matches (for example: the highest similarity rate). In other words, at this time, in addition to selecting the name of the surgical method corresponding to the surgical image with the highest similarity rate (for example: "Uterine fibroids shaving"), it will also load the matching starting hand feature value. The name of the corresponding operation (for example: "shaving of uterine fibroids"). Generally speaking, the name of the loaded surgical method will be the same as the name of the selected surgical method according to the similarity rate of the surgical image, so the surgical name can be directly marked as the surgical name of the image to be analyzed. However, assuming that the two are not the same (for example: the former is a total hysterectomy; the latter is a secondary hysterectomy), at this time, the prompt message 500 can be generated as shown in "Figure 5", and the user is allowed to "Resection" and "Sub-hysterectomy" can be selected as the name of the operation corresponding to the image to be analyzed, and in order to help the user to select the appropriate operation name, you can also play the image to be analyzed in the playback block 520 for The user browses, in this case, since the presence of the cervix 521 can be seen in the image, it is more appropriate to choose a secondary hysterectomy. In addition, if there is no proper surgical name, the user can also directly enter the appropriate surgical name in the input block 510 to correspond to the image to be analyzed, and then complete the labeling.

In summary, it can be seen that the difference between the present invention and the prior art lies in that the object information corresponding to the surgical image is generated by performing image recognition, and the generated object information is used as the feature data corresponding to the surgical image to be used in loading After analyzing the images, perform feature value analysis and statistics based on the images and feature data to be analyzed, in order to calculate the similarity between the image to be analyzed and each surgical image, and select the surgical procedure name corresponding to the surgical image with the highest similarity rate to be marked as pending The name of the operation corresponding to the analysis image, and the image to be analyzed and the name of the operation corresponding to it are stored in the database as the operation image, by which a technical method can solve the problems existing in the previous technology, and thus achieve an improved mark Technical efficiency of the efficiency of surgical imaging.

Although the present invention is disclosed as the foregoing embodiments, it is not intended to limit the present invention. Any person who is familiar with similar arts can make some changes and modifications without departing from the spirit and scope of the present invention. The scope of patent protection shall be subject to the definition of the scope of patent application attached to this specification.

110: database 120: Feature generation module 130: Analysis module 140: storage module 150: Split module 300: display window 310: Play block 321: Load component 322: Analysis components 323: Play component 330: spell name block 410, 420, 430: surgical images 411, 421, 431: fibroids 412, 422, 432: the first surgical instrument 413, 423, 433: Second surgical instruments 414, 424, 434: uterus Step 210: Provide multiple surgical images in a database, each surgical image corresponds to a surgical name Step 211: The database further provides multiple organ characteristic values and multiple surgical instrument characteristic values Step 212: Detect multiple time points in which at least one of the organ characteristic value and the surgical instrument characteristic value exists in the surgical image, and a time interval between adjacent time points is greater than one At a preset interval, multiple frames corresponding to the time interval in the surgical image are deleted to form multiple segmented images and stored in the database Step 220: Initially, load the surgical images of the database separately for image recognition, and generate at least one object information corresponding to the image recognition result, and use the object information as a corresponding to the surgical image Feature data is stored in the database, where each object information includes the coordinates, scale and time point of an image object existing in the surgical image Step 221: Load all the surgical images with the same name of the operation method to identify the type of surgical instrument and the type, shape and size of the abnormal organ, and generate a hand corresponding to the name of the operation method according to the recognition result Characteristic value, and the starting-type characteristic value and the corresponding name of the operation type are stored in the database together Step 230: After loading an image to be analyzed, perform feature value analysis and statistics based on the image to be analyzed and the feature data to calculate a similarity rate between the image to be analyzed and each surgical image Step 240: Select the surgical procedure name corresponding to the surgical image with the highest similarity rate to mark it as the surgical procedure name corresponding to the to-be-analyzed image, and combine the to-be-analyzed image and the marked surgical procedure name together Save to the database as one of the surgical images Step 241: Compare the loaded image to be analyzed with the starting-hand feature value, and load the name of the operation method corresponding to the starting-hand feature value matching the comparison. When the name of the method is different from the name of the method selected according to the similarity ratio, a prompt message is generated and one of them can be selected as the name of the method corresponding to the image to be analyzed

FIG. 1 is a system block diagram of a smart marking system for surgical images of the present invention. 2A to 2C are flowcharts of the method for intelligently marking surgical images of the present invention. FIG. 3 is a schematic diagram of using the present invention to determine the name of a surgical method to be analyzed and mark it. FIG. 4 is a schematic diagram of generating a starting-type feature value by applying the present invention. FIG. 5 is a schematic diagram of applying the present invention to determine the name of a surgical method in combination with a starting feature value.

110: database

120: Feature generation module

130: Analysis module

140: storage module

150: Split module

Claims (10)

A smart marking system for surgical images, the system includes: a database for storing a plurality of surgical images, each surgical image corresponding to a surgical name; a feature generation module connected to the database for initial , Respectively load the surgical images of the database for image recognition, and generate corresponding at least one object information according to the image recognition results, and store the object information as a feature data corresponding to the surgical image and store it in the Database, wherein each object information includes the coordinates, scale and time point of an image object existing in the surgical image; an analysis module is connected to the database for loading an image to be analyzed, Perform feature value analysis and statistics based on the image to be analyzed and the feature data to calculate a similarity ratio between the image to be analyzed and each surgical image; and a storage module connected to the analysis module for selection and analysis The name of the operation corresponding to the surgical image with the highest similarity rate is marked as the name of the operation corresponding to the image to be analyzed, and the image to be analyzed and the name of the operation are stored in the database as One of the surgical images. The intelligent marking system for surgical images according to item 1 of the patent application scope, wherein the feature generation module loads all the surgical images with the same name of the surgical method to identify the types of surgical instruments and abnormal organs , Shape and size, and according to the recognition result, generate a hand-type feature value corresponding to the name of the operation type, and store the hand-type feature value and its corresponding name of the operation type together in the database. The intelligent marking system for surgical images according to item 2 of the patent application scope, wherein the analysis module compares the loaded image to be analyzed with the starting-hand characteristic value, and loads the matching The name of the method corresponding to the characteristic value of the hand type. When the name of the method loaded is different from the name of the method selected according to the similarity ratio, a prompt message is generated and one of them can be selected as the corresponding image to be analyzed. The name of the operation. A smart marking system for surgical images according to item 1 of the patent application scope, wherein the database further stores multiple organ characteristic values and multiple surgical instrument characteristic values for loading when the surgical images are image-recognized, and in Capturing the image object matching the loaded organ characteristic value or the surgical instrument characteristic value from the surgical image as the image object of the object information, and storing the image object in the surgical image The coordinates, scale and time point of the are recorded in the object information. A smart marking system for surgical images according to item 4 of the patent application scope, wherein the system further includes a segmentation module for detecting the presence of at least one of the organ characteristic value and the surgical instrument characteristic value in the surgical image A plurality of time points, and when a time interval between adjacent time points is greater than a preset interval, delete a plurality of frames corresponding to the time interval in the surgical image to form a plurality of Split the image and store it in the database. A method for intelligently marking surgical images, the steps of which include: providing a plurality of surgical images in a database, each surgical image corresponding to a surgical procedure name; at the initial stage, separately loading the surgical images in the database for imaging Identify and generate at least one object message corresponding to the image recognition result, and store the object message as a feature data corresponding to the surgical image and store it in the database, wherein each object message includes an image object Coordinates, proportions and time points in the surgical images; after loading an image to be analyzed, perform feature value analysis and statistics based on the image to be analyzed and the feature data to calculate the image to be analyzed and each operation A similarity ratio of the images; and selecting the surgical procedure name corresponding to the surgical image with the highest similarity ratio to be marked as the surgical procedure name corresponding to the image to be analyzed, and the image to be analyzed and the marked surgical procedure The formula name is also stored in the database as one of the surgical images. According to item 6 of the patent application scope, a method for intelligent labeling of surgical images, wherein the method further includes loading all of the same surgical images with the same surgical name to identify the types of surgical instruments and abnormal organs, The shape and size, and according to the recognition result, generate a hand-type feature value corresponding to the name of the operation type, and store the starting-hand type characteristic value and the corresponding operation type name together in the database. The smart marking method of the surgical image according to item 7 of the patent application scope, wherein the method further includes comparing the loaded image to be analyzed with the starting-hand characteristic value, and loading the matching The name of the method corresponding to the characteristic value of the hand type. When the name of the method loaded is different from the name of the method selected according to the similarity ratio, a prompt message is generated and one of them can be selected as the corresponding image to be analyzed. The steps of the spell name. According to the intelligent marking method of the surgical image item 6, the database further provides multiple organ characteristic values and multiple surgical instrument characteristic values for loading when the surgical images are image-recognized, and in Capturing the image object matching the loaded organ characteristic value or the surgical instrument characteristic value from the surgical image as the image object of the object information, and storing the image object in the surgical image The coordinates, scale and time point of the are recorded in the object information. According to the smart marking method of the surgical image of claim 9, the method further includes detecting a plurality of time points in which at least one of the organ characteristic value and the surgical instrument characteristic value exists in the surgical image, And when a time interval between the adjacent time points is greater than a preset interval, a plurality of frames corresponding to the time interval in the surgical image are deleted to form a plurality of divided images and stored in the data Library.

Images