WO2021120589A1

WO2021120589A1 - Method and apparatus for abnormal image filtering for use on 3d images, device, and storage medium

Info

Publication number: WO2021120589A1
Application number: PCT/CN2020/099526
Authority: WO
Inventors: 陈凯星; 周鑫; 吕传峰
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-06-17
Filing date: 2020-06-30
Publication date: 2021-06-24
Also published as: CN111754473A

Abstract

A method and an apparatus for abnormal image filtering, a device, and a storage medium, the method comprising: receiving a lesion screening request, the lesion screening request at least carrying a initial image information and a lesion prediction result corresponding to the initial image information (S101); establishing sensitivity curve data corresponding to the lesion prediction result (S102); acquiring high-threshold data corresponding to the sensitivity curve data (S103); acquiring low-threshold data corresponding to the sensitivity curve data (S104); on the basis of the connectedness of the high-threshold data, the low-threshold data, and the initial image information, performing a screening operation on the lesion prediction result to obtain a lesion screening result (S105); and outputting the lesion screening result (S106). In addition, the method further relates to blockchain technology, and the lesion screening result may be stored in a blockchain. The present method effectively addresses both the problem of missed lesions and the problem of false positives, and also uses information of the correlation between lesions in adjacent layers of images, thus being able to supplement and optimize neural network learning.

Description

Abnormal image screening method, device, equipment and storage medium for 3D image

This application is based on the Chinese invention patent application filed on June 17, 2020 with the application number 202010554979.2, titled "Method, device, equipment and storage medium for 3D image screening of abnormal images", and claims its priority .

Technical field

This application relates to the field of artificial intelligence technology, and in particular to an abnormal image screening method, device, computer equipment, and storage medium for 3D images.

Background technique

With the emergence of high-performance computing and the rapid development of information computing, the use of AI technology to achieve intelligent diagnosis aids on medical imaging has become a hot spot.

There is an existing abnormal image screening method. Through the analysis and processing of a single image, it is widely used in 2D image modalities such as DR and fundus color photos. Because each inspection can only produce a single 2D image, it is therefore applicable to this type of data. In the selection of the optimal parameters of the algorithm, the parameters are usually determined by means of manual experience or ROC curve, so as to achieve the purpose of screening and detecting abnormal images.

However, in the process of realizing this application, the inventor realized that traditional abnormal image screening methods are generally not applicable to 3D image modalities such as CT, MRI, and PET.

Summary of the invention

The purpose of the embodiments of the present application is to solve the problem that traditional abnormal image screening methods are generally not applicable to 3D image modalities such as CT, MRI, and PET.

In order to solve the above technical problems, an embodiment of the present application provides an abnormal image screening method for 3D images, which adopts the following technical solutions:

Receiving a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Performing a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information, to obtain a lesion screening result;

Output the results of the lesion screening.

In order to solve the above technical problems, an embodiment of the present application also provides an abnormal image screening device for 3D images, which adopts the following technical solutions:

A request receiving module, configured to receive a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

A curve creation module, which is used to create sensitivity curve data corresponding to the lesion prediction result;

A high-threshold acquisition module for acquiring high-threshold data corresponding to the susceptibility curve data;

A low threshold value acquisition module for acquiring low threshold value data corresponding to the susceptibility curve data;

A screening result acquisition module, configured to perform a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information to obtain the lesion screening result;

The screening result output module is used to output the screening result of the lesion.

In order to solve the above technical problems, the embodiments of the present application also provide a computer device, which adopts the following technical solutions:

Including memory and processor;

A computer process is stored in the memory, and when the processor executes the computer process, the steps of the following abnormal image screening method for 3D images are implemented:

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Output the results of the lesion screening.

In order to solve the above technical problems, the embodiments of the present application also provide a computer-readable storage medium, which adopts the following technical solutions:

The computer-readable storage medium stores a computer process, and when the computer process is executed by a processor, the steps of the method for screening abnormal images for 3D images as described below are implemented:

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Output the results of the lesion screening.

The details of one or more embodiments of the present application are presented in the following drawings and description, and other features and advantages of the present application will become apparent from the description, drawings and claims.

Based on the correlation information of the lesions between adjacent layers of the image in the 3D image modal, the abnormal image screening method for 3D images provided by the embodiments of the present application not only effectively takes into account the problems of missed lesion detection and false positive introduction, but also uses images The correlation information of lesions between adjacent layers can supplement and optimize the network learning method; compared to learning the correlation information of lesions between adjacent layers through a 3D neural network -+, this application is not limited to video memory, operating speed , Scanning layer thickness, and doctor’s usage habits, and other factors, have good generalizability and usability; the embodiment of this application can be connected to any lesion detection network as a simple supplement to the network output result, so it has universal Adaptability and plug-and-play advantages.

Description of the drawings

In order to explain the solution in this application more clearly, the following will briefly introduce the drawings used in the description of the embodiments of the application. Obviously, the drawings in the following description are some embodiments of the application. Ordinary technicians can obtain other drawings based on these drawings without creative work.

FIG. 1 is an implementation flowchart of an abnormal image screening method for 3D images according to Embodiment 1 of the present application;

FIG. 2 is an implementation flowchart of obtaining a lesion prediction result provided by Embodiment 1 of the present application;

FIG. 3 is a flowchart of a specific implementation of step S103 in FIG. 1;

FIG. 4 is a flowchart of a specific implementation of step S104 in FIG. 1;

FIG. 5 is a schematic structural diagram of an abnormal image screening device for 3D images provided in Embodiment 2 of the present application;

FIG. 6 is a schematic structural diagram of a lesion prediction result acquisition module provided in Embodiment 2 of the present application;

FIG. 7 is a schematic structural diagram of a specific implementation of the high threshold acquisition module in FIG. 5;

FIG. 8 is a schematic structural diagram of a specific implementation of the low threshold value acquisition module in FIG. 5;

FIG. 9 is a schematic structural diagram of a computer device provided in Embodiment 3 of the present application.

Detailed ways

Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the technical field of the application; the terms used in the specification of the application herein are only for describing specific embodiments. The purpose is not to limit the application; the terms "including" and "having" in the specification and claims of the application and the above-mentioned description of the drawings and any variations thereof are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific sequence.

The reference to "embodiments" herein means that a specific feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art clearly and implicitly understand that the embodiments described herein can be combined with other embodiments.

In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings.

Example one

Referring to FIG. 1, there is shown a flow chart of the method for screening abnormal images for 3D images provided in the first embodiment of the present application. For ease of description, only the parts related to the present application are shown.

In step S101, a lesion screening request is received, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information.

In the embodiments of the present application, the original image information is used to represent the medical images of the examples. In the embodiments of the present application, it mainly refers to medical images in the form of 3D image modalities, as examples, such as CT, MRI, PET, etc., Each inspection can produce a sequence of images, the different levels of image information not only have continuity, but also have high relevance in content.

In the examples of this application, taking the subarachnoid hemorrhage in brain CT as an example, the lesions mainly appear in the subarachnoid space, and the subarachnoid space is distributed on different levels in the CT sequence. Therefore, in the actual reading In the process, doctors usually do not make a conclusion based on the appearance of suspicious lesions on a certain image level, but often make further diagnosis by viewing and analyzing the information of adjacent layers to distinguish true and false lesions. This also further illustrates the importance of the correlation information of the lesions between adjacent layers of the image in the 3D image modality for the diagnosis of the lesions.

In the embodiments of the present application, the focus prediction result refers to the above-mentioned original image information is input into the trained focus prediction model for focus prediction, and the prediction result obtained thereby specifically includes true positive (TP) and false positive (FP) , True Negative (TN), False Negative (FN) and their corresponding confidence.

In step S102, susceptibility curve data corresponding to the lesion prediction result is created.

In the examples of this application, the sensitivity curve, that is, the receiver's operating characteristic curve, is currently used in clinical diagnostic experiments for the reasonable selection of the normal threshold. Its abscissa is represented by FPR, which represents the actual prediction of the positive case. The proportion of the number of negative instances to the number of all negative instances is shown in formula (1); its ordinate is represented by TPR, which represents the proportion of the number of positive instances predicted to be actually positive in all positive instances, as shown in formula (2 ) Shown.

FPR=FP/(FP+TN) (1)

TPR=TP/(TP+FN) (2)

Among them, FP represents the number of predicted positive instances that are actually negative, that is, the number of false positive instances, TN represents the number of predicted negative instances that are actually negative, and TP represents the number of predicted positive instances that are actually positive. FN represents the actual number of positive cases among the predicted negative cases, that is, the number of missed cases.

In step S103, high threshold data corresponding to the sensitivity curve data is acquired.

In the examples of this application, the high-threshold data is used to screen whether there are real lesions in each of the above-mentioned images, and the detected various types of lesions are classified according to their confidence. If the confidence of the existence of a certain type of lesion in the detection result is greater than or equal to the high threshold For example, it can be determined that the image has such a lesion.

In the examples of this application, according to the output results of the model and using the above formulas, we calculate the TPR and FPR values of various types of lesions at the case level under different confidence levels, and determine the best critical points of various types of lesions. The value is used as the high threshold of the lesion.

In step S104, low threshold data corresponding to the sensitivity curve data is acquired.

In the embodiment of this application, the low threshold data is used to screen whether there are no lesions in each of the above images, and the detected various types of lesions are classified according to their confidence. If the confidence of the existence of a certain type of lesion in the detection result is less than the low threshold data , It can be determined that this image does not have such lesions.

In the embodiment of this application, the method of obtaining low-threshold data is different from the above-mentioned method of obtaining high-threshold data. The low-threshold is based on the prerequisite that the detected lesions are prioritized. Through the observation and analysis of the ROC curve, it can be found that such points are often located at the position where the curvature of the curve tends to zero or the curvature change tends to zero. This position is the best balance point based on the ROC of the lesion.

In step S105, a screening operation is performed on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information to obtain a lesion screening result.

In the examples of this application, first, based on the high and low thresholds obtained in the above steps, we will classify the detected various types of lesions according to their confidence. If there is an instance of a certain type of lesion with a confidence greater than or equal to the high threshold in the detection result , It can be determined that this type of lesions exist in this case, and the lesions with confidence of such lesions less than the high threshold and greater than the low threshold are set as candidates for the lesions, and lesions less than the low threshold are directly deleted from the results; then we use the relative The feature of correlation between adjacent-level lesions, around the confidence level greater than the high-threshold lesions, we can analyze whether there is a candidate lesion in the adjacent layer through connected domain analysis or simply judge whether there is a candidate lesion in the adjacent layer, and if it exists, the lesion will be included in the ranks of high-threshold lesions, and then iterate Find, and finally use high-threshold lesions as the model's lesion screening results.

In the embodiment of this application, the result of the lesion screening can be stored in a blockchain. The blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information for verification. The validity of the information (anti-counterfeiting) and the generation of the next block. The blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

In step S106, the result of the lesion screening is output.

In an embodiment of the present application, an abnormal image screening method for 3D images is provided, and a lesion screening request is received, and the lesion screening request carries at least original image information and information corresponding to the original image information. Lesion prediction result; creating sensitivity curve data corresponding to the lesion prediction result; acquiring high threshold data corresponding to the sensitivity curve data; acquiring low threshold data corresponding to the sensitivity curve data; based on the high The threshold data, the low threshold data, and the connectivity of the original image information perform a screening operation on the lesion prediction result to obtain the lesion screening result; and output the lesion screening result. Based on the correlation information of the lesions between adjacent layers of the image in the 3D image modal, the abnormal image screening method for 3D images provided by the embodiments of the present application not only effectively takes into account the problems of missed lesion detection and false positive introduction, but also uses images The correlation information of lesions between adjacent layers can supplement and optimize the network learning method; compared to learning the correlation information of lesions between adjacent layers through a 3D neural network -+, this application is not limited to video memory, operating speed , Scanning layer thickness, and doctor’s usage habits, and other factors, have good generalizability and usability; the embodiment of this application can be connected to any lesion detection network as a simple supplement to the network output result, so it has universal Adaptability and plug-and-play advantages.

Continuing to refer to FIG. 2, there is shown a flow chart for obtaining a focus prediction result provided in the first embodiment of the present application. For ease of description, only the parts related to the present application are shown.

In step S201, a system database is read, and training image information and training prediction results corresponding to the training image information are acquired in the system database.

In the embodiment of the present application, the system database is mainly used to pre-store training image information and training prediction results, and the training image information and the training prediction results have a corresponding relationship.

In step S202, the training image information and the training prediction result are input to the deep neural network model to perform a model training operation to obtain a lesion prediction model.

In the embodiment of the present application, the deep neural network model can perform model training based on the predicted training image information and the training prediction result, so that the prediction result of the lesion prediction model is closer to the original target.

In step S203, a lesion prediction request sent by a user terminal is received, where the lesion prediction request carries at least the original image information.

In the embodiments of this application, the user terminal may be, for example, a mobile phone, a smart phone, a notebook computer, a digital broadcast receiver, a PDA (personal digital assistant), a PAD (tablet computer), a PMP (portable multimedia player), a navigation device, etc. It should be understood that the examples of user terminals here are only for ease of understanding, and are not used to limit this application.

In step S204, the original image information is input to the lesion prediction model to perform a lesion prediction operation to obtain the lesion prediction result.

Continuing to refer to FIG. 3, a flowchart of a specific implementation of step S103 in FIG. 1 is shown. For ease of description, only the parts related to the present application are shown.

In step S301, the best critical point corresponding to the susceptibility curve data is obtained.

In the embodiment of this application, the optimal critical point is expressed as:

Wherein, P is expressed as the nearest upper left corner point on the susceptibility curve.

In step S302, the optimal critical point is used as the high threshold data.

Continuing to refer to FIG. 4, a flowchart of a specific implementation of step S104 in FIG. 1 is shown. For ease of description, only the parts related to the present application are shown.

In step S401, an equation fitting operation is performed on the susceptibility curve based on the least square method to obtain a curve coordinate equation.

In the embodiments of the present application, the least square method refers to finding the best function match of the data by minimizing the square sum of the error. The least square method can be used to easily obtain unknown data, and minimize the sum of squares of errors between the obtained data and the actual data. The least squares method can also be used for curve fitting. Some other optimization problems can also be expressed by the least square method by minimizing energy or maximizing entropy.

In the embodiment of this application, the curve coordinate equation is expressed as:

y=f(x)

In step S402, the curvature corresponding to the curve coordinate equation is obtained.

In the embodiment of this application, the curvature is expressed as:

In the embodiment of the present application, the formula is the curvature calculation formula of the "ROC curve", where y=f(x), and K represents the curvature of the ROC curve at the X position on the abscissa.

In step S403, the point at which the curvature approaches 0 or the curvature change is small is used as the low threshold data.

In the embodiment of the present application, the curvature of each point on the curve can be calculated according to the above-mentioned curvature formula (5), and then the point where the curvature approaches 0 or the position where the curvature changes little is selected as the best balance point, that is, Determine the low threshold.

In summary, the abnormal image screening method for 3D images provided by the embodiments of the present application receives a lesion screening request, and the lesion screening request carries at least original image information and information corresponding to the original image information. Lesion prediction result; creating sensitivity curve data corresponding to the lesion prediction result; acquiring high threshold data corresponding to the sensitivity curve data; acquiring low threshold data corresponding to the sensitivity curve data; based on the high The threshold data, the low threshold data, and the connectivity of the original image information perform a screening operation on the lesion prediction result to obtain the lesion screening result; and output the lesion screening result. Based on the correlation information of the lesions between adjacent layers of the image in the 3D image modal, the abnormal image screening method for 3D images provided by the embodiments of the present application not only effectively takes into account the problems of missed lesion detection and false positive introduction, but also uses images The correlation information of lesions between adjacent layers can supplement and optimize the network learning method; compared to learning the correlation information of lesions between adjacent layers through a 3D neural network -+, this application is not limited to video memory, operating speed , Scanning layer thickness, and doctor’s usage habits, and other factors, have good generalizability and usability; the embodiment of this application can be connected to any lesion detection network as a simple supplement to the network output result, so it has universal Adaptability and plug-and-play advantages.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer process. The computer process can be stored in a computer readable storage medium. When executed, it may include the procedures of the above-mentioned method embodiments. Among them, the aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

It should be understood that although the various steps in the flowchart of the drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless explicitly stated in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least part of the steps in the flowchart of the drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the order of execution is also It is not necessarily performed sequentially, but may be performed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

Example two

With further reference to FIG. 5, as an implementation of the method shown in FIG. 1, this application provides an embodiment of an abnormal image screening device for 3D images, which is similar to the method embodiment shown in FIG. Correspondingly, the device can be specifically applied to various electronic devices.

As shown in FIG. 5, the abnormal image screening device 500 for 3D images in this embodiment includes: a request receiving module 501, a curve creation module 502, a high threshold acquisition module 503, a low threshold acquisition module 504, and a result acquisition module 505 and the result output module 506. among them:

The request receiving module 501 is configured to receive a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

The curve creation module 502 is configured to create sensitivity curve data corresponding to the lesion prediction result;

The high-threshold acquisition module 503 is configured to acquire high-threshold data corresponding to the susceptibility curve data;

The low threshold acquisition module 504 is configured to acquire low threshold data corresponding to the susceptibility curve data;

The screening result acquisition module 505 is configured to perform a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information to obtain the lesion screening result;

The screening result output module 506 is used to output the lesion screening result.

In the examples of this application, taking the subarachnoid hemorrhage in brain CT as an example, the lesions mainly appear in the subarachnoid space, and the subarachnoid space is distributed on different levels in the CT sequence. Therefore, in real image reading In the process, doctors usually do not make a conclusion based on the appearance of suspicious lesions on a certain image level, but often make further diagnosis by viewing and analyzing the information of adjacent layers to distinguish true and false lesions. This also further illustrates the importance of the correlation information of the lesions between adjacent layers of the image in the 3D image modality for the diagnosis of the lesions.

FPR=FP/(FP+TN) (1)

TPR=TP/(TP+FN) (2)

In the examples of this application, according to the output results of the model and using the above formula, we calculate the TPR and FPR values of various types of lesions at the case level under different confidence levels, and determine the best critical points of various types of lesions from them. The value is used as the high threshold of the lesion.

In the examples of this application, first, based on the high and low thresholds obtained in the above steps, we will classify the detected various types of lesions according to their confidence. If there is an instance of a certain type of lesion with a confidence greater than or equal to the high threshold in the detection result , It can be determined that this type of lesion exists in this case, and the lesions with confidence of such lesions less than the high threshold and greater than the low threshold are set as candidates for the lesions, and the lesions less than the low threshold are directly deleted from the results; then we use the relative The feature of correlation between adjacent layer lesions, around the confidence level greater than the high threshold lesion, we can analyze whether there is a candidate lesion in the adjacent layer through connected domain analysis or simply judge whether there is a candidate lesion in the adjacent layer, and if it exists, the lesion will be included in the ranks of high threshold lesions, and iterative Find, and finally use high-threshold lesions as the model's lesion screening results.

In an embodiment of the present application, an abnormal image screening device for 3D images is provided. Based on the correlation information of the lesions between adjacent layers of the image of the 3D image modal, the abnormality of the 3D image provided by the embodiment of the present application is The image screening method not only effectively takes into account the problems of missed lesion detection and false positive introduction, but also uses the correlation information of the lesion between adjacent layers of the image, which can supplement and optimize the network learning method; compared to the 3D neural network- + To learn the correlation information of the lesions between adjacent layers, this application is not limited to various factors such as video memory, operating speed, scanning layer thickness, and doctors' habits, and has good generalizability and usability; examples of this application It can be connected to any lesion detection network as a simple supplement to the output of the network, so it has the advantages of universality and plug-and-play.

Continuing to refer to FIG. 6, there is shown a schematic structural diagram of the lesion prediction result acquisition module provided in the second embodiment of the present application. For ease of description, only the parts related to the present application are shown.

In some optional implementations of the second embodiment of the present application, the above-mentioned abnormal image screening device 500 for 3D images further includes: a training data acquisition sub-module 507, a prediction model acquisition sub-module 508, a request receiving sub-module 509, and The prediction result obtaining sub-module 510. among them:

The training data acquisition sub-module 507 is used to read a system database, and acquire training image information and training prediction results corresponding to the training image information in the system database;

The prediction model acquisition sub-module 508 is configured to input the training image information and the training prediction result into the deep neural network model to perform model training operations to obtain the lesion prediction model;

The request receiving submodule 509 is configured to receive a lesion prediction request sent by a user terminal, where the lesion prediction request carries at least the original image information;

The prediction result obtaining sub-module 510 is configured to input the original image information into the lesion prediction model to perform a lesion prediction operation, and obtain the lesion prediction result.

In the embodiment of the present application, the deep neural network model may perform model training based on the predicted training image information and the training prediction result, so that the prediction result of the lesion prediction model is closer to the original target.

Continuing to refer to FIG. 7, it shows a schematic structural diagram of the high-threshold acquisition module. For ease of description, only the parts related to the present application are shown.

In some optional implementation manners of the second embodiment of the present application, the above-mentioned high threshold value obtaining module 503 includes: a critical point obtaining submodule 5031 and a high threshold value determining submodule 5032. among them:

The critical point acquisition submodule 5031 is used to acquire the optimal critical point corresponding to the susceptibility curve data, and the optimal critical point is expressed as:

Wherein, P represents the nearest upper left corner point on the susceptibility curve;

The high threshold determination sub-module 5032 is configured to use the optimal critical point as the high threshold data.

Continuing to refer to FIG. 8, it shows a schematic structural diagram of the low-threshold acquisition module. For ease of description, only the parts related to the present application are shown.

In some optional implementation manners of the second embodiment of the present application, the aforementioned low threshold value acquisition module 504 includes: a curve acquisition sub-module 5041, a curvature acquisition sub-module 5042, and a low threshold value determination sub-module 5043. among them:

The curve acquisition sub-module 5041 is used to perform an equation fitting operation on the susceptibility curve based on the least square method to obtain the curve coordinate equation:

y=f(x);

The curvature acquisition sub-module 5042 is configured to acquire the curvature corresponding to the curve coordinate equation, and the curvature is expressed as:

The low threshold determination sub-module 5043 is configured to use a point with a curvature approaching 0 or a small change in curvature as the low threshold data.

In order to solve the above technical problems, the embodiments of the present application also provide computer equipment. Please refer to FIG. 9 for details. FIG. 9 is a block diagram of the basic structure of the computer device in this embodiment.

The computer device 9 includes a memory 91, a processor 92, and a network interface 93 that communicate with each other through a system bus. It should be pointed out that the figure only shows the computer device 9 with components 91-93, but it should be understood that it is not required to implement all the illustrated components, and more or fewer components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing in accordance with pre-set or stored instructions. Its hardware includes, but is not limited to, a microprocessor, a dedicated Integrated Circuit (Application Specific Integrated Circuit, ASIC), Programmable Gate Array (Field-Programmable Gate Array, FPGA), Digital Processor (Digital Signal Processor, DSP), embedded equipment, etc.

The computer device may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The computer device can interact with the user through a keyboard, a mouse, a remote control, a touch panel, or a voice control device.

The memory 91 includes at least one type of readable storage medium. The readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), and static memory. Random access memory (SRAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), programmable read only memory (PROM), magnetic memory, magnetic disks, optical disks, etc., the computer readable storage The medium can be non-volatile or volatile. In some embodiments, the memory 91 may be an internal storage unit of the computer device 9, for example, a hard disk or a memory of the computer device 9. In other embodiments, the memory 91 may also be an external storage device of the computer device 9, for example, a plug-in hard disk, a smart media card (SMC), and a secure digital device equipped on the computer device 9. (Secure Digital, SD) card, Flash Card, etc. Of course, the memory 91 may also include both the internal storage unit of the computer device 9 and its external storage device. In this embodiment, the memory 91 is generally used to store an operating system and various application software installed in the computer device 9, such as computer readable instructions for a 3D image screening method for abnormal images. In addition, the memory 91 can also be used to temporarily store various types of data that have been output or will be output.

The processor 92 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 92 is generally used to control the overall operation of the computer device 9. In this embodiment, the processor 92 is configured to run computer-readable instructions or processed data stored in the memory 91, for example, run the computer-readable instructions of the abnormal image screening method for 3D images.

The network interface 93 may include a wireless network interface or a wired network interface, and the network interface 93 is generally used to establish a communication connection between the computer device 9 and other electronic devices.

This application also provides another embodiment, that is, a computer-readable storage medium that stores an abnormal image screening process for 3D images, and the abnormal image screening process for 3D images The inspection process may be executed by at least one processor, so that the at least one processor executes the steps of the abnormal image screening method for 3D images as described above.

Through the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better.的实施方式。 Based on this understanding, the technical solution of this application essentially or the part that contributes to the existing technology can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, The optical disc) includes several instructions to make a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of the present application.

Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The drawings show preferred embodiments of the present application, but do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these examples is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although this application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it is still possible for those skilled in the art to modify the technical solutions described in each of the foregoing specific embodiments, or equivalently replace some of the technical features. . All equivalent structures made by using the contents of the description and drawings of this application, directly or indirectly used in other related technical fields, are similarly within the scope of patent protection of this application.

Claims

An abnormal image screening method for 3D images, which includes the following steps:

Receiving a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Performing a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information, to obtain a lesion screening result;

Output the results of the lesion screening.
The abnormal image screening method for 3D images according to claim 1, wherein before the step of receiving a lesion screening request, the method further comprises the following steps:

Reading a system database, and obtaining training image information and training prediction results corresponding to the training image information in the system database;

Input the training image information and the training prediction result into the deep neural network model to perform model training operations to obtain the lesion prediction model;

Receiving a lesion prediction request sent by a user terminal, where the lesion prediction request carries at least the original image information;

The original image information is input to the lesion prediction model to perform a lesion prediction operation to obtain the lesion prediction result.
The abnormal image screening method for 3D images according to claim 1, wherein the step of obtaining high threshold data corresponding to the sensitivity curve data includes the following steps:

Obtain the optimal critical point corresponding to the susceptibility curve data, and the optimal critical point is expressed as:

Wherein, P represents the nearest upper left corner of the susceptibility curve, TPR represents the ratio of the number of predicted positive instances that are actually positive to all positive instances, and FPR represents the actual negative instances of the predicted positive instances The ratio of the number to the number of all negative instances;

The optimal critical point is used as the high threshold data.
The abnormal image screening method for 3D images according to claim 1, wherein the step of obtaining low threshold data corresponding to the sensitivity curve data includes the following steps:

Perform an equation fitting operation on the susceptibility curve based on the least square method to obtain the curve coordinate equation:

y=f(x);

Among them, f(x) represents the curve coordinate equation after fitting;

Obtain the curvature corresponding to the curve coordinate equation, and the curvature is expressed as:

Among them, y=f(x), K represents the curvature of the ROC curve at the X position on the abscissa;

The point at which the curvature approaches 0 or the curvature change is small is used as the low threshold data.
An abnormal image screening device for 3D images, wherein the device includes:

A request receiving module, configured to receive a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

A curve creation module, which is used to create sensitivity curve data corresponding to the lesion prediction result;

A high-threshold acquisition module for acquiring high-threshold data corresponding to the susceptibility curve data;

A low threshold value acquisition module for acquiring low threshold value data corresponding to the susceptibility curve data;

A screening result acquisition module, configured to perform a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information to obtain the lesion screening result;

The screening result output module is used to output the screening result of the lesion.
The abnormal image screening device for 3D images according to claim 5, wherein the device further comprises:

The training data acquisition sub-module is used to read the system database, and acquire training image information and training prediction results corresponding to the training image information in the system database;

The prediction model acquisition sub-module is used to input the training image information and the training prediction result into the deep neural network model to perform model training operations to obtain the lesion prediction model;

The request receiving submodule is configured to receive a lesion prediction request sent by a user terminal, where the lesion prediction request carries at least the original image information;

The prediction result obtaining submodule is configured to input the original image information into the lesion prediction model to perform a lesion prediction operation, and obtain the lesion prediction result.
The abnormal image screening device for 3D images according to claim 5, wherein the high threshold value acquisition module comprises:

The critical point acquisition sub-module is used to acquire the optimal critical point corresponding to the susceptibility curve data, and the optimal critical point is expressed as:

Wherein, P represents the nearest upper left point on the susceptibility curve;

The high threshold value determination sub-module is configured to use the optimal critical point as the high threshold value data.
The abnormal image screening device for 3D images according to claim 5, wherein the low threshold acquisition module comprises:

The curve acquisition sub-module is used to perform an equation fitting operation on the susceptibility curve based on the least square method to obtain the curve coordinate equation:

y=f(x);

The curvature acquisition sub-module is used to acquire the curvature corresponding to the curve coordinate equation, and the curvature is expressed as:

The low threshold determination sub-module is used to use a point with a curvature approaching 0 or a small change in curvature as the low threshold data.
A computer device includes a memory, a processor, and computer readable instructions stored in the memory and capable of running on the processor, wherein the processor executes the computer readable instructions as follows The steps of the abnormal image screening method for 3D images:

Receiving a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Performing a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information, to obtain a lesion screening result;

Output the results of the lesion screening.
9. The computer device according to claim 9, wherein before the step of receiving a lesion screening request, it further comprises the following steps:

Reading a system database, and obtaining training image information and training prediction results corresponding to the training image information in the system database;

Input the training image information and the training prediction result into the deep neural network model to perform model training operations to obtain the lesion prediction model;

Receiving a lesion prediction request sent by a user terminal, where the lesion prediction request carries at least the original image information;

The original image information is input to the lesion prediction model to perform a lesion prediction operation to obtain the lesion prediction result.
9. The computer device according to claim 9, wherein the step of obtaining high threshold data corresponding to the sensitivity curve data comprises the following steps:

Obtain the optimal critical point corresponding to the susceptibility curve data, and the optimal critical point is expressed as:

Wherein, P represents the nearest upper left corner of the susceptibility curve, TPR represents the ratio of the number of predicted positive instances that are actually positive to all positive instances, and FPR represents the actual negative instances of the predicted positive instances The ratio of the number to the number of all negative instances;

The optimal critical point is used as the high threshold data.
9. The computer device of claim 9, wherein the step of obtaining low threshold data corresponding to the sensitivity curve data comprises the following steps:

Perform an equation fitting operation on the susceptibility curve based on the least square method to obtain the curve coordinate equation:

y=f(x);

Among them, f(x) represents the curve coordinate equation after fitting;

Obtain the curvature corresponding to the curve coordinate equation, and the curvature is expressed as:

Among them, y=f(x), K represents the curvature of the ROC curve at the X position on the abscissa;

The point at which the curvature approaches 0 or the curvature change is small is used as the low threshold data.
A computer-readable storage medium, wherein, when the computer-readable instruction is executed by a processor, the process is caused to execute the steps of the abnormal image screening method for 3D images:

Receiving a lesion screening request, the lesion screening request carrying at least original image information and a lesion prediction result corresponding to the original image information;

Create sensitivity curve data corresponding to the lesion prediction result;

Acquiring high threshold data corresponding to the susceptibility curve data;

Acquiring low-threshold data corresponding to the susceptibility curve data;

Performing a screening operation on the lesion prediction result based on the connectivity of the high threshold data, the low threshold data, and the original image information, to obtain a lesion screening result;

Output the results of the lesion screening.
The computer-readable storage medium according to claim 13, wherein before the step of receiving a lesion screening request, the method further comprises the following step:

Reading a system database, and obtaining training image information and training prediction results corresponding to the training image information in the system database;

Input the training image information and the training prediction result into the deep neural network model to perform model training operations to obtain the lesion prediction model;

Receiving a lesion prediction request sent by a user terminal, where the lesion prediction request carries at least the original image information;

The original image information is input to the lesion prediction model to perform a lesion prediction operation to obtain the lesion prediction result.
15. The computer-readable storage medium according to claim 13, wherein the step of obtaining high threshold data corresponding to the sensitivity curve data comprises the following steps:

Obtain the optimal critical point corresponding to the susceptibility curve data, and the optimal critical point is expressed as:

Wherein, P represents the nearest upper left corner of the susceptibility curve, TPR represents the ratio of the number of predicted positive instances that are actually positive to all positive instances, and FPR represents the actual negative instances of the predicted positive instances The ratio of the number to the number of all negative instances;

The optimal critical point is used as the high threshold data.
15. The computer-readable storage medium according to claim 13, wherein the step of obtaining low-threshold data corresponding to the susceptibility curve data comprises the following steps:

Perform an equation fitting operation on the susceptibility curve based on the least square method to obtain the curve coordinate equation:

y=f(x);

Among them, f(x) represents the curve coordinate equation after fitting;

Obtain the curvature corresponding to the curve coordinate equation, and the curvature is expressed as:

Among them, y=f(x), K represents the curvature of the ROC curve at the X position on the abscissa;

The point at which the curvature approaches 0 or the curvature change is small is used as the low threshold data.