CN117036376B

CN117036376B - Lesion image segmentation method and device based on artificial intelligence and storage medium

Info

Publication number: CN117036376B
Application number: CN202311304131.4A
Authority: CN
Inventors: 牛颢; 章毅; 钟科; 武俊; 徐修远
Original assignee: Sichuan University
Current assignee: Sichuan University
Priority date: 2023-10-10
Filing date: 2023-10-10
Publication date: 2024-01-30
Anticipated expiration: 2043-10-10
Also published as: CN117036376A

Abstract

The application provides a lesion image segmentation method, device and storage medium based on artificial intelligence, which are characterized in that image segmentation is carried out on medical image images to be segmented to obtain image segmentation sets to be identified corresponding to the medical image images to be segmented, a reference image segmentation set is determined according to medical image segmentation to be identified, wherein the medical image segmentation sets to be identified meet identification conditions, the reference image segmentation sets are determined according to suspicious image segmentation in the reference image segmentation sets, identification image segmentation identification is carried out on identification image segmentation blocks in the identification image segmentation sets, and image segmentation marking is carried out when marking critical conditions are deeply identified according to identification results. Therefore, the identification image block set corresponding to the medical image to be segmented can be automatically identified, so that the identification image block identification is carried out on the identification image block set, and the image segmentation efficiency is improved.

Description

Lesion image segmentation method and device based on artificial intelligence and storage medium

Technical Field

The present disclosure relates to the field of image processing and artificial intelligence, and in particular, to a lesion image segmentation method and apparatus based on artificial intelligence, and a storage medium.

Background

Lesion identification is one of the important tasks of medical image analysis, aimed at accurately locating and extracting abnormal areas or lesions of interest from medical images. Where auxiliary collaboration is required by means of various image processing techniques, such as image enhancement, image segmentation, feature extraction, feature selection, classifier construction, object location tracking, etc. In medical image segmentation, the extraction and separation of the region of interest of a medical image, particularly a lesion image, are facilitated for subsequent lesion recognition and positioning based on the segmentation result. Common segmentation methods are only based on threshold segmentation, edge detection, region growing, horizontal line segmentation and other methods, and have limited help for subsequent lesion recognition. In addition, when the existing lesion image is segmented, the segmentation is limited by manual marking, and the accuracy and the efficiency of the segmentation are improved to a certain extent.

Disclosure of Invention

In view of this, embodiments of the present application provide at least a lesion image segmentation method, device and storage medium based on artificial intelligence.

The technical scheme of the embodiment of the application is realized as follows:

in a first aspect, the present application provides an artificial intelligence-based lesion image segmentation method, the method comprising:

Obtaining a medical image to be segmented, performing image blocking operation on the medical image to be segmented to obtain a blocking set of images to be identified corresponding to the medical image to be segmented, wherein the medical image to be segmented is obtained through volume drawing;

determining pathology auxiliary promotion indexes of each to-be-identified medical image block in the to-be-identified image block set, and determining a reference image block set from the to-be-identified image block set through the pathology auxiliary promotion indexes of each to-be-identified medical image block; the medical image blocks in the reference image block set are medical image blocks to be identified, wherein the pathology auxiliary promotion indexes corresponding to the medical image block set to be identified meet identification conditions;

determining an identification image block set according to the reference image block set, wherein identification image blocks in the identification image block set are suspicious image blocks;

performing identification image block identification according to the identification image block set, and performing image block marking when the condition that the marking critical condition is met is deeply identified according to the identification result;

and dividing the medical image to be divided according to the image block marking result of each medical image block to be recognized.

In some embodiments, the determining a pathology assistance facilitation indicator for each medical image patch to be identified in the set of image patches to be identified comprises:

determining an internal relevance evaluation index of the first medical image block to be identified and an external relevance evaluation index of the first medical image block to be identified according to any first medical image block to be identified, wherein the block size of the first medical image block to be identified is larger than a size threshold value, and determining a pathology auxiliary promotion index of the first medical image block to be identified based on the internal relevance evaluation index of the first medical image block to be identified and the external relevance evaluation index of the first medical image block to be identified;

and determining an external relevance evaluation index of the second medical image block to be identified aiming at any second medical image block to be identified, wherein the size of the block in the image block set to be identified is smaller than or equal to the size threshold, and taking the external relevance evaluation index of the second medical image block to be identified as a pathology auxiliary promotion index of the second medical image block to be identified.

In some embodiments, the medical image to be segmented is any slice in the target three-dimensional medical image data; the determining the inter-relevance evaluation index of the first medical image partition to be identified comprises the following steps:

Determining a pixel group corresponding to the first medical image block to be identified, wherein any one pixel group is formed by a plurality of cutting pixel groups obtained by cutting the first medical image block to be identified, and each cutting pixel group is formed by one pixel or a plurality of surrounding pixels in the first medical image block to be identified;

for any pixel group, acquiring the confidence coefficient of each cutting pixel group in the pixel group in the target three-dimensional medical image data;

determining an internal relevance evaluation index corresponding to the pixel group according to the confidence coefficient of each cutting pixel group in the target three-dimensional medical image data and the confidence coefficient of the first medical image block to be identified in the target three-dimensional medical image data;

determining the minimum internal relevance evaluation index in the internal relevance evaluation indexes corresponding to each pixel group as the internal relevance evaluation index of the first medical image block to be identified;

the determining the external relevance evaluation index of the first medical image partition to be identified comprises the following steps:

determining a neighborhood pixel set of the first medical image block to be identified in the target three-dimensional medical image data, wherein the neighborhood pixel set comprises at least one neighborhood pixel;

Acquiring confidence degrees of the fusion medical image blocks to be identified, which are obtained by combining each neighborhood pixel and the first medical image block to be identified, in the target three-dimensional medical image data respectively;

and determining an external relevance evaluation index of the first medical image block to be identified according to the confidence coefficient of each medical image block to be identified in the target three-dimensional medical image data.

In some embodiments, the neighborhood pixels included in the neighborhood pixel set are surrounding pixels of the first medical image segment to be identified in the target three-dimensional medical image data; the obtaining the confidence that each neighborhood pixel and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data, wherein the obtaining the confidence that each neighborhood pixel and the first medical image block to be identified are combined and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data comprises obtaining the confidence that each surrounding pixel in the neighborhood pixel set and the first medical image block to be identified are combined and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data;

the determining the external relevance evaluation index of the first medical image segmentation to be identified according to the confidence coefficient of each medical image segmentation to be identified in the target three-dimensional medical image data comprises the following steps:

Acquiring a plurality of first uncertainty metric values through the confidence coefficient of each first fusion medical image block to be identified in the target three-dimensional medical image data;

and determining the minimum uncertainty measurement value in the first uncertainty measurement values as an external relevance evaluation index of the first medical image partition to be identified.

In some embodiments, the medical image to be segmented is any slice in target three-dimensional medical image data, and the target three-dimensional medical image data is obtained at the current obtaining moment; the determining the identified image block set according to the reference image block set comprises the following steps:

acquiring a reference lesion image set, wherein the reference lesion image set is formed by normal image segmentation and identification image segmentation determined according to three-dimensional medical image data acquired before the current acquisition moment;

and matching each medical image block in the reference image block set in the reference lesion image set, and determining the identification image block set based on medical image blocks with unsuccessful matching.

In some embodiments, the target identification image tile is any one of the set of identification image tiles, the method further comprising:

Matching the matched medical image segments matched with the target identification image segments in the reference lesion image set; if the target identification image block is matched with the matching medical image block, determining a new identification image block through the target identification image block and the matching medical image block, and iterating the identification image block set based on the new identification image block; and iterating the reference lesion image set based on the iterated identification image blocking set.

In some embodiments, the target identification image block is any one identification image block in the identification image block set, and the identifying image block according to the identification image block set includes:

acquiring first occurrence frequencies of the target identification image blocks in the three-dimensional medical image data at R acquisition moments, and acquiring second occurrence frequencies of the target identification image blocks in the three-dimensional medical image data at target acquisition moments, wherein the target acquisition moments are acquisition moments after the R acquisition moments are divided from the S acquisition moments; calculating a quotient between the first frequency of occurrence and the second frequency of occurrence; determining a quotient between the first occurrence frequency and the second occurrence frequency as a recognition result of the target recognition image block;

Or performing the identifying image block identification according to the identifying image block set, including:

acquiring first occurrence frequencies of the target identification image blocks in the three-dimensional medical image data at R acquisition moments, and acquiring third occurrence frequencies of the target identification image blocks in the three-dimensional medical image data at S acquisition moments;

calculating a quotient between the first frequency of occurrence and the third frequency of occurrence;

determining a quotient between the first occurrence frequency and the third occurrence frequency as a recognition result of the target recognition image block;

the S acquisition time points comprise the current acquisition time point and S-1 acquisition time points before the current acquisition time point; the R acquisition moments comprise the current acquisition moment and R-1 acquisition moments before the current acquisition moment, and R is smaller than S.

In some embodiments, the performing image blocking marking when the marking critical condition is identified according to the depth of the identification result includes:

acquiring a preset mark reference value;

and when the quotient corresponding to the identification result of the target identification image block is larger than or equal to the preset mark reference value, determining that the mark critical condition is met, and carrying out corresponding marking on the corresponding target identification image block.

In some embodiments, the obtaining the preset mark reference value includes:

acquiring the occurrence frequency of the target identification image block in the three-dimensional medical image data at the current acquisition time and before the current acquisition time;

determining the identification type corresponding to the target identification image block according to the occurrence frequency;

acquiring a preset mark reference value corresponding to the identification type; the identification image blocks in the identification image block set are classified into at least one identification type according to the occurrence frequency of each identification image block, and one identification type corresponds to a preset mark reference value.

In a second aspect, the present application provides a lesion image segmentation apparatus, comprising:

the image acquisition module is used for acquiring a medical image to be segmented, performing image blocking operation on the medical image to be segmented to obtain a to-be-identified image blocking set corresponding to the medical image to be segmented, wherein the medical image to be segmented is obtained through volume drawing;

the index determining module is used for determining pathology auxiliary promotion indexes of all to-be-identified medical image blocks in the to-be-identified image block set, and determining to obtain a reference image block set from the to-be-identified image block set through the pathology auxiliary promotion indexes of all to-be-identified medical image blocks; the medical image blocks in the reference image block set are medical image blocks to be identified, wherein the pathology auxiliary promotion indexes corresponding to the medical image block set to be identified meet identification conditions;

The block determining module is used for determining an identification image block set according to the reference image block set, wherein identification image blocks in the identification image block set are suspicious image blocks;

the block marking module is used for carrying out block recognition of the recognition image according to the recognition image block set and carrying out block marking of the image when the recognition result depth recognition meets the marking critical condition;

and the image segmentation module is used for segmenting the medical image to be segmented according to the image segmentation marking result of each medical image segmentation to be identified.

In a third aspect, the present application also provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method described above.

The application has at least the beneficial effects that include:

according to the method, image blocking is conducted on the medical image to be segmented to obtain an image blocking set to be identified corresponding to the medical image to be segmented, a reference image blocking set is determined according to the medical image blocking to be identified, which is corresponding to the pathological auxiliary promotion index in the image blocking set to be identified, meeting the identification conditions, the identification image blocking set is determined according to suspicious image blocking in the reference image blocking set, identification image blocking identification is conducted on identification image blocking in the identification image blocking set, and image blocking marking is conducted when the fact that the identification conditions meet the marking critical conditions is deeply identified according to the identification results. In this way, the identification image block set corresponding to the medical image to be segmented can be automatically identified, so that the identification image block identification is carried out on the identification image block set, and the image segmentation efficiency is increased. Meanwhile, the identification image blocks in the identification image block set are suspicious image blocks, the suspicious image blocks can be used for representing newly increased suspicious image blocks which possibly occur, therefore, the suspicious image blocks are identified, and when the fact that the marking critical condition is met is identified according to the depth of the identification result of the suspicious image blocks, the image blocks are marked, and the newly increased suspicious image blocks which possibly occur can be marked in advance, so that the segmentation of lesion images can be accurately and rapidly completed.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the aspects of the present application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the technical aspects of the application.

Fig. 1 is a schematic implementation flow chart of a lesion image segmentation method based on artificial intelligence according to an embodiment of the present application.

Fig. 2 is a schematic structural diagram of a lesion image segmentation device according to an embodiment of the present application.

Fig. 3 is a schematic hardware entity diagram of a computer device according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application are further elaborated below in conjunction with the accompanying drawings and examples, which should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making inventive efforts are within the scope of protection of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. The term "first/second/third" is merely to distinguish similar objects and does not represent a specific ordering of objects, it being understood that the "first/second/third" may be interchanged with a specific order or sequence, as permitted, to enable embodiments of the present application described herein to be practiced otherwise than as illustrated or described herein.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing the present application only and is not intended to be limiting of the present application.

The embodiment of the application provides an artificial intelligence-based lesion image segmentation method which can be executed by a processor of computer equipment. The computer device may refer to a server, a notebook computer, a tablet computer, a desktop computer, a smart television, a mobile device (e.g., a mobile phone, a portable video player, a personal digital assistant), and the like, which have data processing capability.

Fig. 1 is a schematic implementation flow chart of a lesion image segmentation method based on artificial intelligence according to an embodiment of the present application, as shown in fig. 1, the method includes the following operations 101 to 105:

and S101, acquiring a medical image to be segmented, and performing image blocking operation on the medical image to be segmented to obtain a blocking set of images to be identified corresponding to the medical image to be segmented.

In the embodiment of the present application, the medical image to be segmented is obtained by Volume Rendering (Volume Rendering), the medical image to be segmented is any slice in the target three-dimensional medical image data, the target three-dimensional medical image data is obtained at the current obtaining time, and the target three-dimensional medical image data includes at least one medical image to be segmented. After the medical image to be segmented is obtained, image blocking operation can be performed on the medical image to be segmented to obtain an image blocking set to be identified corresponding to the medical image to be segmented, wherein the image blocking set to be identified comprises at least one medical image blocking to be identified obtained by the image blocking operation. In this embodiment, the image segmentation operation may specifically be a process of performing basic image segmentation on the medical image to be segmented, for example, segmentation is performed according to a pixel gray level and a preset gray level threshold, or the image is segmented into different image segments by means of edge detection (such as Canny edge detection and Sobel operator), and in other embodiments, the medical image to be segmented may also be segmented into multiple image segments with the same pixel size according to a mean segmentation manner.

Operation S102 determines a pathology assistance promoting index of each medical image block to be identified in the image block set to be identified, and determines a reference image block set from the image block set to be identified by the pathology assistance promoting index of each medical image block to be identified.

In this embodiment of the present application, the pathology auxiliary promoting index of the medical image segmentation to be identified is an evaluation index for performing secondary image segmentation, and a reference basis of the evaluation index is an identification promoting element for pathology, which may specifically include at least one of the following indexes: the method comprises the steps of (1) evaluating an internal relevance index of a medical image block to be identified and evaluating an external relevance index of the medical image block to be identified; the medical image blocks in the reference image block set are medical image blocks to be identified, wherein the corresponding pathological auxiliary promotion indexes in the image block set to be identified meet the identification conditions, namely, the medical image blocks to be identified, of which the corresponding pathological auxiliary promotion indexes in the image block set to be identified meet the identification conditions, are determined to be the medical image blocks in the reference image block set. For example, for any one of the first medical image segments to be identified in the image segment set, where the segment size is greater than the size threshold, the size threshold is, for example, 0.5mm×0.5mm, and is specifically related to the image modality (such as CT, MRI), the specific disease type, and the lesion feature, which is specifically not limited. Determining a pathology assistance facilitation indicator for a first medical image patch to be identified comprises: determining an inner relevance evaluation index of the first medical image patch to be identified and an outer relevance evaluation index of the first medical image patch to be identified, and determining a pathological auxiliary promotion index of the first medical image patch to be identified based on the inner relevance evaluation index of the first medical image patch to be identified and the outer relevance evaluation index of the first medical image patch to be identified. At this time, the pathological assistance-promoting index of the first medical image patch to be identified satisfies the identification condition, for example, the pathological assistance-promoting index of the first medical image patch to be identified is greater than or equal to the first index value. In other embodiments, for any one of the second medical image segments to be identified in the set of image segments to be identified, the determining the pathological assistance-promoting index for the second medical image segment to be identified includes: determining an external relevance evaluation index of the second medical image block to be identified, taking the external relevance evaluation index of the second medical image block to be identified as a pathology auxiliary promotion index of the second medical image block to be identified, wherein the fact that the pathology auxiliary promotion index of the second medical image block to be identified meets the identification condition means that the pathology auxiliary promotion index of the second medical image block to be identified is larger than or equal to a second index value.

Optionally, the set of image segments to be identified includes at least one first medical image segment to be identified, the first medical image segments to be identified may be arranged according to descending order of pathology assistance promoting indexes of the first medical image segments to be identified, a first medical image segment sequence to be identified is obtained, and the first medical image segments to be identified in the first medical image segment sequence to be identified located before the first distribution order is determined as medical image segments in the set of reference image segments, where the first distribution order is the first M medical image segments to be identified in the first medical image segment sequence to be identified. According to the same thought, the image segmentation set to be identified comprises at least one second medical image segmentation to be identified, the medical image segmentation to be identified is also arranged according to the descending order of the pathology auxiliary promotion indexes of the medical image segmentation to be identified, a second medical image segmentation sequence to be identified is obtained, and the medical image segmentation to be identified, which is positioned before the second distribution sequence in the second medical image segmentation sequence to be identified, is determined as the medical image segmentation in the reference image segmentation set; the second distribution order refers to the first N, M and N specific values of the second medical image blocking sequence to be identified can be set according to actual needs.

In the embodiment of the present application, the intra-relevance evaluation index of the medical image segment to be identified may evaluate whether the medical image segment to be identified is suitable for the degree of independent existence by the involvement of each cut pixel group (i.e., the relevance of each cut pixel group) in the medical image segment to be identified, where each cut pixel group is composed of one pixel or a plurality of surrounding pixels (i.e., a plurality of pixels adjacent to each other) in the medical image segment to be identified. The medical image blocks to be identified are regarded as an independent area, and the larger the internal relevance evaluation index of the medical image blocks to be identified is, the stronger the relevance among all cutting pixel groups in the medical image blocks to be identified is, and the more suitable the medical image blocks to be identified are for independent existence. The external relevance evaluation index of the medical image block to be identified can be measured by an uncertainty metric value (which can be measured by uncertainty of probability distribution, for example, by counting the frequency of each pixel value in the medical image block to be identified, obtaining a histogram of the pixel value, dividing the frequency of each pixel value by the total pixel number of the medical image block to be identified, obtaining the probability of each pixel value, carrying out logarithmic operation on each probability value, multiplying the negative number of the logarithm, and finally summing the results of all the pixel values to obtain the uncertainty metric value of the medical image block to be identified) of the medical image block to be identified. The neighborhood pixels may include surrounding pixels of the medical image block to be identified in the target three-dimensional medical image data; the uncertainty measurement value of the fusion medical image block to be identified is used for representing the unpredictability of the fusion medical image block to be identified; the smaller the uncertainty metric value of the fusion medical image block to be identified, the smaller the unpredictability of the fusion medical image block to be identified, and the higher the possibility that the neighborhood pixels and the medical image block to be identified form part. That is, the smaller the uncertainty metric value of the fused medical image patch to be identified, the smaller the external relevance evaluation index of the medical image patch to be identified, and the less suitable the medical image patch to be identified is for independent existence. The medical image segmentation in the reference image segmentation set is determined based on the medical image segmentation to be identified, which is suitable for independent existence, in the image segmentation set to be identified. For any one of the first medical image segments to be identified in the set of the image segments to be identified, the block size of which is not smaller than the size threshold, determining the pathological assistance-promoting index of the first medical image segment to be identified comprises the following operations:

And firstly, determining an intra-relevance evaluation index of the first medical image partition to be identified.

Specifically, a group of pixel groups corresponding to the first medical image block to be identified may be determined, where any one group of pixel groups is formed by a plurality of cut pixel groups obtained by cutting the first medical image block to be identified, and each cut pixel group is formed by one pixel or a plurality of surrounding pixels in the first medical image block to be identified. And aiming at any pixel group, acquiring the confidence coefficient of each cutting pixel group in the pixel group in the target three-dimensional medical image data, and determining an internal relevance evaluation index corresponding to the pixel group according to the confidence coefficient of each cutting pixel group in the target three-dimensional medical image data and the confidence coefficient of the first medical image block to be identified in the target three-dimensional medical image data. And taking the minimum internal relevance evaluation index of the internal relevance evaluation indexes corresponding to each pixel group as the internal relevance evaluation index of the first medical image block to be identified. The confidence that the cut pixel bolus appears in the target three-dimensional medical image data is the quotient between the number of occurrences of the cut pixel bolus in the target three-dimensional medical image data and the total number of image patches in the target three-dimensional medical image data, or the quotient between the number of occurrences of the cut pixel bolus in the target three-dimensional medical image data and the total number of image patches in the target three-dimensional medical image data. The confidence that the first medical image segmentation to be identified appears in the target three-dimensional medical image data is the quotient between the number of times the first medical image segmentation to be identified is in the target three-dimensional medical image data and the total number of image segmentation in the target three-dimensional medical image data, or the quotient between the number of times the first medical image segmentation to be identified is in the target three-dimensional medical image data and the total number of image segmentation in the target three-dimensional medical image data.

The calculation manner of the intra-relevance evaluation indexes of the pixel group groups can refer to a general PMI (Pointwise Mutual Information) calculation formula, and details are omitted herein, wherein the minimum intra-relevance evaluation index of the intra-relevance evaluation indexes corresponding to the pixel group groups is used as the intra-relevance evaluation index of the first medical image block to be identified.

And secondly, determining an external relevance evaluation index of the first medical image block to be identified.

Specifically, a neighborhood pixel set of the first medical image block to be identified can be determined in the target three-dimensional medical image data, the neighborhood pixel set comprises at least one neighborhood pixel, the neighborhood pixels in the neighborhood pixel set are surrounding pixels of the first medical image block to be identified in the target three-dimensional medical image data, confidence degrees of the occurrence of each neighborhood pixel and the first medical image block to be identified, which are combined, of the fusion medical image blocks to be identified, in the target three-dimensional medical image data are obtained, and external relevance evaluation indexes of the first medical image block to be identified are determined based on the occurrence confidence degrees of the fusion medical image blocks to be identified in the target three-dimensional medical image data. The fusion medical image segmentation to be identified is obtained by combining a neighborhood pixel and a first medical image segmentation to be identified.

For example, the neighborhood pixels included in the neighborhood pixel set are surrounding pixels of the first medical image block to be identified in the target three-dimensional medical image data, confidence levels of the first medical image block to be identified, which is obtained by combining each surrounding pixel in the neighborhood pixel set with the first medical image block to be identified, in the target three-dimensional medical image data are obtained, the first uncertainty metric values are obtained based on the confidence levels of the first medical image blocks to be identified in the target three-dimensional medical image data, and the minimum uncertainty metric value in the first uncertainty metric values is determined to be an external relevance evaluation index of the first medical image block to be identified. The confidence that the first fusion medical image segmentation to be identified appears in the target three-dimensional medical image data is the quotient between the occurrence number of the first fusion medical image segmentation to be identified in the target three-dimensional medical image data and the total number of the image segmentation of the target three-dimensional medical image data or the quotient between the number of the first fusion medical image segmentation to be identified in the target three-dimensional medical image data and the total number of the image segmentation of the target three-dimensional medical image data.

The calculation manner of the first uncertainty metric may refer to a general information entropy calculation formula, which is not described herein, and as another embodiment, the neighborhood pixels included in the neighborhood pixel set are surrounding pixels of the first medical image block to be identified in the target three-dimensional medical image data, a confidence degree that each surrounding pixel in the neighborhood pixel set and the first medical image block to be identified are combined to obtain the first fused medical image block to be identified respectively appear in the target three-dimensional medical image data is obtained, and the first uncertainty metric is obtained based on the confidence degree that each first fused medical image block to be identified appears in the target three-dimensional medical image data, and is used as an external relevance evaluation index of the first medical image block to be identified.

And thirdly, determining a pathological auxiliary promotion index of the first medical image block to be identified according to the internal relevance evaluation index and the external relevance evaluation index of the first medical image block to be identified.

Specifically, a first importance parameter corresponding to an internal relevance evaluation index of a first medical image partition is obtained, a second importance parameter corresponding to an internal relevance evaluation index of the first medical image partition is obtained, and a pathology auxiliary promotion index of the first medical image partition is obtained according to the first importance parameter, the internal relevance evaluation index of the first medical image partition, the second importance parameter and the external relevance evaluation index of the first medical image partition. The calculation formula of the pathological auxiliary promotion index of the first medical image segmentation to be identified is as follows:

G＝αI+βO

G is a pathology auxiliary promotion index of the first medical image block to be identified, alpha is a first importance parameter, I is an internal relevance evaluation index of the first medical image block to be identified, beta is a second importance parameter, and O is an external relevance evaluation index of the first medical image block to be identified.

For any second medical image block to be identified, wherein the block size of the second medical image block to be identified is not larger than the size threshold value, determining the pathology assistance promoting index of the second medical image block to be identified specifically comprises the following steps: and determining an external relevance evaluation index of the second medical image block to be identified, and taking the external relevance evaluation index of the second medical image block to be identified as a pathology auxiliary promotion index of the second medical image block to be identified. The manner of determining the external relevance evaluation index of the second medical image patch to be identified may refer to determining the external relevance evaluation index of the first medical image patch to be identified.

The neighborhood pixels included in the neighborhood pixel set of the second medical image block to be identified are surrounding pixels of the second medical image block to be identified in the target three-dimensional medical image data. The confidence coefficient of each surrounding pixel and each third medical image block to be identified, which is obtained by combining the surrounding pixels with the second medical image block to be identified, in the target three-dimensional medical image data can be obtained, and a third uncertainty measurement value is obtained based on the confidence coefficient of each third medical image block to be identified in the target three-dimensional medical image data; and determining the minimum uncertainty measurement value in the third uncertainty measurement values as an external relevance evaluation index of the second medical image partition to be identified. The confidence that the third fusion medical image segmentation to be identified appears in the target three-dimensional medical image data is a quotient between the occurrence frequency of the third fusion medical image segmentation to be identified in the target three-dimensional medical image data and the total number of image segmentation of the target three-dimensional medical image data or a quotient between the occurrence frequency of the third fusion medical image segmentation to be identified in the target three-dimensional medical image data and the total number of image segmentation of the target three-dimensional medical image data.

In other embodiments, the neighborhood pixels included in the neighborhood pixel set are surrounding pixels of the second medical image block to be identified in the target three-dimensional medical image data, confidence levels of third medical image blocks to be identified, which are obtained by combining each surrounding pixel in the neighborhood pixel set with the second medical image block to be identified, in the target three-dimensional medical image data respectively can be obtained, and third uncertainty measurement values are obtained based on the confidence levels of the third medical image blocks to be identified in the target three-dimensional medical image data, and the calculated third uncertainty measurement values are used as external relevance evaluation indexes of the second medical image blocks to be identified.

Operation S103 determines an identification image block set from the reference image block set.

The identified image patches in the set of identified image patches are medical image patches in the set of reference image patches that do not appear in the set of reference lesion images. The medical image to be segmented is any slice in target three-dimensional medical image data, the target three-dimensional medical image data is obtained at the current obtaining moment, and determining the identification image block set according to the reference image block set comprises: and acquiring a reference lesion image set, wherein the reference lesion image set is formed by normal image segmentation and identification image segmentation determined according to three-dimensional medical image data acquired before the current acquisition time, each medical image segmentation in the reference image segmentation set is matched in the reference lesion image set, and the identification image segmentation set is determined based on medical image segmentation with unsuccessful matching.

Optionally, the reference lesion image set is iterated with the set of identification image patches, i.e. identification image patches in the set of identification image patches are added to the reference lesion image set. And in the iterative process of the reference lesion image set, image blocks in the current reference lesion image set are reserved, and the identification image block set obtained by determining based on the medical image to be segmented each time is added to the reference lesion image set. Based on this, the integrity of the reference lesion image set is improved.

Operation S104, performing recognition image blocking recognition according to the recognition image blocking set, and performing image blocking marking when the depth recognition according to the recognition result satisfies the marking critical condition.

According to the occurrence frequency of each identification image block in the identification image block set in all three-dimensional medical image data, identifying the identification image block in the identification image block set to obtain an identification result of each identification image block, identifying the identification result of each identification image block, and marking the image block when the depth identification meets the marking critical condition. The method comprises the steps that all three-dimensional medical image data are obtained at all obtaining moments, and the occurrence frequency of the identification image blocks in all three-dimensional medical image data is the occurrence frequency of the identification image blocks in all three-dimensional medical image data or the quotient between the occurrence frequency of the identification image blocks in all three-dimensional medical image data and the total image block number in all three-dimensional medical image data.

Operation S105, segmenting the medical image to be segmented by the image segmentation marking result of each medical image segmentation to be identified.

Specifically, each image block in the medical image to be segmented is marked according to the image block marking result to obtain a corresponding marking result, and the marking process is completed to segment the medical image to be segmented.

According to the method, image blocking is conducted on the medical image to be segmented to obtain an image blocking set to be identified corresponding to the medical image to be segmented, a reference image blocking set is determined according to the medical image blocking to be identified, which is corresponding to the pathological auxiliary promotion index in the image blocking set to be identified and meets the identification condition, the medical image blocking in the reference image blocking set is the medical image blocking to be identified, which is suitable for independent existence in the image blocking set to be identified, the identification image blocking set is determined according to the suspicious image blocking in the reference image blocking set, identification image blocking identification is conducted on the identification image blocking in the identification image blocking set, and image blocking marking is conducted when the identification image blocking meets the marking critical condition according to the identification result depth identification. In this way, the identification image block set corresponding to the medical image to be segmented can be automatically identified, so that the identification image block identification is carried out on the identification image block set, and the image segmentation efficiency is increased. Meanwhile, the identification image blocks in the identification image block set are suspicious image blocks, and the suspicious image blocks are used for representing newly increased suspicious image blocks possibly occurring, so that the suspicious image blocks are identified, and when the identification result depth of the suspicious image blocks is identified to meet the marking critical condition, the image blocks are marked, and the newly increased suspicious image blocks possibly occurring can be marked in advance, so that the segmentation of lesion images can be accurately and rapidly completed.

In another embodiment, the artificial intelligence based lesion image segmentation method provided in the embodiments of the present application may include the following operations:

an operation 201, obtaining a medical image to be segmented, and performing image blocking operation on the medical image to be segmented to obtain a to-be-identified image blocking set corresponding to the medical image to be segmented.

Operation 202 determines a pathology assistance promoting index of each medical image block to be identified in the image block set to be identified, and determines a reference image block set from the image block set to be identified by the pathology assistance promoting index of each medical image block to be identified.

An operation 203 determines an identified set of image tiles from the reference set of image tiles.

The process of operations 201 to 203 may refer to operations 101 to 103.

And operation 204, performing identification image block identification according to the identification image block set.

Before operation 204, optionally, the target recognition image segmentation is any one of a set of recognition image segmentation, for example, the image vector cluster is used to match a matching medical image segmentation matched with the target recognition image segmentation in the reference lesion image set, and if so, a new recognition image segmentation is determined based on the target recognition image segmentation and the matching medical image segmentation, for example, the target recognition image segmentation is combined with the matching medical image segmentation to obtain the new recognition image segmentation. The image vector clustering is adopted to match the matched medical image blocks matched with the target identification image blocks in the reference lesion image set, specifically, the medical image blocks to be matched are determined in the reference lesion image set, the image feature vectors of the target identification image blocks and the image feature vectors of the medical image blocks to be matched are obtained through a vector conversion algorithm (when the image feature vectors are obtained, image preprocessing can be firstly carried out, then feature extraction is carried out, for example, extraction is carried out in the modes of LBP, HOG, color histogram or SIFT, etc., then feature encoding is carried out, and finally the feature vectors are normalized, so that the image feature vectors are obtained). And when the vector difference value between the image feature vector of the target identification image block and the image feature vector of the medical image block to be matched is smaller than a threshold value, or when the inner product between the image feature vector of the target identification image block and the image feature vector of the medical image block to be matched is larger than the preset inner product, determining that the medical image block to be matched is the matched medical image block matched with the target identification image block.

In other embodiments, the matching medical image blocks matched with the target recognition image blocks may be matched in the reference lesion image set through image feature vector clustering, if the matching medical image blocks are matched, a new recognition image block is determined according to the target recognition image block and the matching medical image block, for example, the target recognition image block and the matching medical image block are combined to obtain the new recognition image block, the recognition image block set may be iterated according to the new recognition image block, and the reference lesion image set may be iterated based on the iterated recognition image block set, i.e., the recognition image block in the iterated recognition image block set is added to the reference lesion image set. Based on the above, the matching medical image blocks in the reference lesion image set and the target identification image blocks are combined, so that newly increased suspicious image blocks can be obtained, the identification image block set is expanded, and the identification interval is increased.

In operation 204, identifying image tiles from the set of identified image tiles may specifically include one or more of the following policies:

firstly, acquiring the occurrence frequency of each identification image block in an identification image block set in all three-dimensional medical image data, wherein all three-dimensional medical image data comprise all three-dimensional medical image data acquired at all acquisition moments, and the occurrence frequency of the identification image block in all three-dimensional medical image data is the occurrence frequency of the identification image block in all three-dimensional medical image data or the quotient between the occurrence frequency of the identification image block in all three-dimensional medical image data and the total number of image blocks in all three-dimensional medical image data.

The second strategy is that the target identification image block is any identification image block in an identification image block set, when the target identification image block is identified, the first occurrence frequency of the target identification image block in the three-dimensional medical image data is obtained at R obtaining moments, the second occurrence frequency of the target identification image block in the three-dimensional medical image data is obtained at the target obtaining moment, the target obtaining moment is the obtaining moment after dividing R obtaining moments in S obtaining moments, and the quotient between the first occurrence frequency and the second occurrence frequency is calculated; and determining the quotient between the first occurrence frequency and the second occurrence frequency as the identification result of the target identification image block. Identifying object-identifying image segments can refer to the following formula:

θ=（i+t）÷（j+t）

θ is the recognition result of the target recognition image block, i is the first occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained by acquiring R acquisition moments, j is the second occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained by acquiring the target recognition image block at the target acquisition moments, and t is a constant. The S acquisition time points comprise a current acquisition time point and S-1 acquisition time points before the current acquisition time point; the R acquisition time points comprise the current acquisition time point and R-1 acquisition time points before the current acquisition time point, and R is less than S. The recognition result of the target recognition image block passing strategy II can represent the added value of the first occurrence frequency of the target recognition image block at R acquisition moments compared with the second occurrence frequency of the target recognition image block at the target acquisition moments.

The third strategy, the target identification image block is any identification image block in the identification image block set, and the identification of the target identification image block comprises the following steps: the method comprises the steps of obtaining first occurrence frequencies of three-dimensional medical image data by obtaining target identification image blocks at R obtaining moments, obtaining third occurrence frequencies of the three-dimensional medical image data by obtaining target identification image blocks at S obtaining moments, obtaining quotient between the first occurrence frequencies and the third occurrence frequencies, and determining the quotient between the first occurrence frequencies and the third occurrence frequencies as an identification result of the target identification image blocks. The formula for identifying the target identification image blocks is as follows:

λ=i÷ε

λ is an identification result obtained by the target identification image segmentation through the policy three, i is a first occurrence frequency of the target identification image segmentation in the three-dimensional medical image data obtained by acquiring R acquisition moments, and ɛ is a third occurrence frequency of the target identification image segmentation in the three-dimensional medical image data obtained by acquiring S acquisition moments. The recognition result of the target recognition image segmentation passing strategy III is used for representing the proportion of the first occurrence frequency of the target recognition image segmentation in R acquisition moments to the third occurrence frequency of the target recognition image segmentation in S acquisition moments.

In operation 205, a preset mark reference value is obtained.

The frequency of occurrence of the identification image blocks in the identification image block set in all three-dimensional medical image data is classified into at least one identification type by each identification image block, and one identification type corresponds to a preset mark reference value. The obtaining of the preset mark reference value comprises the following steps: obtaining occurrence frequency of the three-dimensional medical image data by obtaining the target identification image blocks at the current obtaining time and the previous time (namely all the obtaining time); and determining the identification type corresponding to the target identification image block according to the occurrence frequency, thereby obtaining a preset mark reference value corresponding to the identification type. For example, the frequency of occurrence of each identification image block in all three-dimensional medical image data in the identification image block set is classified into a multi-data identification type, a general data identification type and a minority data identification type, the multi-data identification type corresponds to a multi-data preset mark reference value, the general data identification type corresponds to a general data preset mark reference value, the minority data identification type corresponds to a minority data preset mark reference value, and when the frequency of occurrence of the target identification image block in all three-dimensional medical image data is determined to be the multi-data type, the acquired preset mark reference value is the multi-data preset mark reference value.

And operation 206, when the quotient corresponding to the identification result of the target identification image block is greater than or equal to the preset mark reference value, determining that the mark critical condition is met, and carrying out corresponding marking on the corresponding target identification image block.

And when the quotient corresponding to the identification result of the target identification image block is not smaller than the preset mark reference value, determining that the mark critical condition is met. Optionally, the identifying image blocks are arranged according to the descending order of the quotient corresponding to the identifying result of each identifying image block, so as to obtain an identifying image block sequence, identifying image blocks located before a third distribution order in the identifying image block sequence as identifying image blocks meeting the marking critical condition, wherein the third distribution order is, for example, the first K identifying image blocks in the identifying image block sequence.

And operation 207, performing image blocking marking according to the image blocking marking strategy to complete image segmentation.

The method comprises the steps that a medical image to be segmented can be obtained, and image blocking operation is carried out on the medical image to be segmented to obtain a to-be-identified image blocking set corresponding to the medical image to be segmented; determining pathology auxiliary promotion indexes of each medical image block to be identified in the image block set to be identified, and determining a reference image block set from the image block set to be identified through the pathology auxiliary promotion indexes of each medical image block to be identified; the medical image blocks in the reference image block set are the medical image blocks to be identified, wherein the pathology auxiliary promotion indexes corresponding to the medical image blocks in the image block set to be identified meet the identification conditions; determining an identification image block set according to the reference image block set, wherein identification image blocks in the identification image block set are suspicious image blocks; and carrying out identification image block identification according to the identification image block set, and marking when the condition is identified to be met according to the identification result.

According to the method, image blocking is conducted on the medical image to be segmented to obtain an image blocking set to be identified corresponding to the medical image to be segmented, a reference image blocking set is determined according to the medical image blocking to be identified, which is corresponding to the pathological auxiliary promotion index in the image blocking set to be identified and meets the identification condition, the medical image blocking in the reference image blocking set is the medical image blocking to be identified, which is suitable for independent existence in the image blocking set to be identified, the identification image blocking set is determined according to the suspicious image blocking in the reference image blocking set, identification image blocking identification is conducted on the identification image blocking in the identification image blocking set, and image blocking marking is conducted when the identification image blocking meets the marking critical condition according to the identification result depth identification. In this way, the identification image block set corresponding to the medical image to be segmented can be automatically identified, so that the identification image block identification is carried out on the identification image block set, and the image segmentation efficiency is increased. Meanwhile, the identification image blocks in the identification image block set are suspicious image blocks, the suspicious image blocks can be used for representing newly increased suspicious image blocks which possibly occur, therefore, the suspicious image blocks are identified, and when the fact that the marking critical condition is met is identified according to the depth of the identification result of the suspicious image blocks, the image blocks are marked, and the newly increased suspicious image blocks which possibly occur can be marked in advance, so that the segmentation of lesion images can be accurately and rapidly completed. Meanwhile, through combining the identification image blocks in the identification image block set with the matched medical image blocks matched with the identification image blocks in the reference lesion image set, suspicious image blocks can be restored, the identification image blocks in the identification image block set and the matched medical image blocks matched with the identification image blocks in the reference lesion image set can be combined to obtain newly increased suspicious image blocks, the identification image block set can be expanded and displayed, an identification interval is increased, the identification image blocks are classified according to the occurrence frequency of the identification image blocks in all three-dimensional medical image data, and a preset mark reference value is matched for each identification type, so that the identification image blocks with different occurrence frequencies can be matched, and the accuracy of image block marks is improved.

Based on the foregoing embodiments, the embodiments of the present application provide a lesion image segmentation device, where each unit included in the device and each module included in each unit may be implemented by a processor in a computer device; of course, the method can also be realized by a specific logic circuit; in practice, the processor may be a central processing unit (Central Processing Unit, CPU), microprocessor (Microprocessor Unit, MPU), digital signal processor (Digital Signal Processor, DSP) or field programmable gate array (Field Programmable Gate Array, FPGA), etc.

Fig. 2 is a schematic structural diagram of a lesion image segmentation device according to an embodiment of the present application, and as shown in fig. 2, a lesion image segmentation device 200 includes:

the image obtaining module 210 is configured to obtain a medical image to be segmented, and perform an image blocking operation on the medical image to be segmented to obtain a set of image blocks to be identified corresponding to the medical image to be segmented, where the medical image to be segmented is obtained by volume rendering;

the index determining module 220 is configured to determine a pathology auxiliary promotion index of each to-be-identified medical image block in the to-be-identified image block set, and determine to obtain a reference image block set from the to-be-identified image block set through the pathology auxiliary promotion index of each to-be-identified medical image block; the medical image blocks in the reference image block set are medical image blocks to be identified, wherein the pathology auxiliary promotion indexes corresponding to the medical image block set to be identified meet identification conditions;

The block determining module 230 is configured to determine an identified image block set according to the reference image block set, where identified image blocks in the identified image block set are suspicious image blocks;

the block marking module 240 is configured to perform recognition image block marking according to the recognition image block set, and perform image block marking when the recognition result depth recognizes that the marking critical condition is satisfied;

the image segmentation module 250 is configured to segment the medical image to be segmented according to an image segmentation marking result of each medical image segmentation to be identified.

The description of the apparatus embodiments above is similar to that of the method embodiments above, with similar advantageous effects as the method embodiments. In some embodiments, functions or modules included in the apparatus provided in the embodiments of the present application may be used to perform the methods described in the embodiments of the methods, and for technical details that are not disclosed in the embodiments of the apparatus of the present application, please refer to the description of the embodiments of the methods of the present application for understanding.

It should be noted that, in the embodiment of the present application, if the above-mentioned lesion image segmentation method based on artificial intelligence is implemented in the form of a software function module, and sold or used as a separate product, the lesion image segmentation method may also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application may be essentially or portions contributing to the related art, and the software product may be stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, an optical disk, or other various media capable of storing program codes. Thus, embodiments of the present application are not limited to any specific hardware, software, or firmware, or to any combination of hardware, software, and firmware.

The embodiment of the application provides a computer device, which comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor executes the program to realize part or all of the steps of the method.

Embodiments of the present application provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs some or all of the steps of the above-described method. The computer readable storage medium may be transitory or non-transitory.

Embodiments of the present application provide a computer program comprising computer readable code which, when run in a computer device, performs some or all of the steps for implementing the above method.

Embodiments of the present application provide a computer program product comprising a non-transitory computer-readable storage medium storing a computer program which, when read and executed by a computer, performs some or all of the steps of the above-described method. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In some embodiments, the computer program product is embodied as a computer storage medium, in other embodiments the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It should be noted here that: the above description of various embodiments is intended to emphasize the differences between the various embodiments, the same or similar features being referred to each other. The above description of apparatus, storage medium, computer program and computer program product embodiments is similar to that of method embodiments described above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the apparatus, storage medium, computer program and computer program product of the present application, please refer to the description of the method embodiments of the present application.

Fig. 3 is a schematic diagram of a hardware entity of a computer device according to an embodiment of the present application, as shown in fig. 3, the hardware entity of the computer device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program executable on the processor 1001, the processor 1001 implementing the steps in the method of any of the embodiments described above when the program is executed.

The memory 1002 stores a computer program executable on a processor, and the memory 1002 is configured to store instructions and applications executable by the processor 1001, and may also cache data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or already processed by each module in the processor 1001 and the computer device 1000, which may be implemented by a FLASH memory (FLASH) or a random access memory (Random Access Memory, RAM).

The processor 1001 performs the steps of the artificial intelligence-based lesion image segmentation method according to any one of the above. The processor 1001 generally controls the overall operation of the computer device 1000.

Embodiments of the present application provide a computer storage medium storing one or more programs executable by one or more processors to implement the steps of the artificial intelligence based lesion image segmentation method according to any of the embodiments above.

It should be noted here that: the description of the storage medium and apparatus embodiments above is similar to that of the method embodiments described above, with similar benefits as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and the apparatus of the present application, please refer to the description of the method embodiments of the present application for understanding. The processor may be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronic device implementing the above-mentioned processor function may be other, and embodiments of the present application are not specifically limited.

The computer storage medium/Memory may be a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable programmable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable Read Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic random access Memory (Ferromagnetic Random Access Memory, FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Read Only optical disk (Compact Disc Read-Only Memory, CD-ROM); but may also be various terminals such as mobile phones, computers, tablet devices, personal digital assistants, etc., that include one or any combination of the above-mentioned memories.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present application, the sequence number of each step/process described above does not mean that the execution sequence of each step/process should be determined by the function and the internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

Alternatively, the integrated units described above may be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the related art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.

The foregoing is merely an embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes and substitutions are intended to be covered in the protection scope of the present application.

Claims

1. An artificial intelligence-based lesion image segmentation method, comprising:

Wherein the determining the pathology assistance promoting index of each medical image patch to be identified in the image patch set to be identified comprises: determining an internal relevance evaluation index of the first medical image block to be identified and an external relevance evaluation index of the first medical image block to be identified according to any first medical image block to be identified, wherein the block size of the first medical image block to be identified is larger than a size threshold value, and determining a pathology auxiliary promotion index of the first medical image block to be identified based on the internal relevance evaluation index of the first medical image block to be identified and the external relevance evaluation index of the first medical image block to be identified; wherein determining a pathology assistance promoting index of the first medical image patch to be identified based on the internal relevance evaluating index and the external relevance evaluating index of the first medical image patch to be identified includes: acquiring a first importance parameter corresponding to an internal relevance evaluation index of a first medical image partition, and acquiring a second importance parameter corresponding to an internal relevance evaluation index of the first medical image partition, wherein the pathological auxiliary promotion index of the first medical image partition is acquired according to the first importance parameter, the internal relevance evaluation index of the first medical image partition, the second importance parameter and the external relevance evaluation index of the first medical image partition; the calculation formula of the pathological auxiliary promotion index of the first medical image segmentation to be identified is as follows:

G＝αI+βO

Wherein G is a pathology auxiliary promotion index of the first medical image block to be identified, alpha is a first importance parameter, I is an internal relevance evaluation index of the first medical image block to be identified, beta is a second importance parameter, and O is an external relevance evaluation index of the first medical image block to be identified;

determining an external relevance evaluation index of a second medical image block to be identified aiming at any second medical image block to be identified, wherein the size of the block in the image block set to be identified is smaller than or equal to the size threshold, and taking the external relevance evaluation index of the second medical image block to be identified as a pathology auxiliary promotion index of the second medical image block to be identified;

determining an identification image block set according to the reference image block set, wherein identification image blocks in the identification image block set are suspicious image blocks; the medical image to be segmented is any slice in target three-dimensional medical image data, and the target three-dimensional medical image data is obtained at the current obtaining moment; the determining the identified image block set according to the reference image block set comprises the following steps: acquiring a reference lesion image set, wherein the reference lesion image set is formed by normal image segmentation and identification image segmentation determined according to three-dimensional medical image data acquired before the current acquisition moment; matching each medical image block in the reference image block set in the reference lesion image set, and determining the identification image block set based on medical image blocks which are not successfully matched;

Performing identification image block identification according to the identification image block set, and performing image block marking when the condition that the marking critical condition is met is deeply identified according to the identification result; the identifying image block identification is carried out according to the identifying image block set, and the identifying image block identification comprises at least one of the following strategies:

the method comprises the steps that a first strategy and target identification image segmentation are any identification image segmentation in the identification image segmentation set, first occurrence frequencies of the target identification image segmentation in three-dimensional medical image data are obtained at R obtaining moments, second occurrence frequencies of the target identification image segmentation in the three-dimensional medical image data are obtained at target obtaining moments, and the target obtaining moments are the obtaining moments after the R obtaining moments are divided from S obtaining moments; calculating a quotient between the first frequency of occurrence and the second frequency of occurrence; determining a quotient between the first occurrence frequency and the second occurrence frequency as a recognition result of the target recognition image block;

the formula for identifying the target identification image blocks is as follows:

θ=（i+t）÷（j+t）

wherein θ is the recognition result of the target recognition image block, i is the first occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained at R acquisition moments, j is the second occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained at the target acquisition moments, and t is a constant; the S acquisition time points comprise a current acquisition time point and S-1 acquisition time points before the current acquisition time point; the R acquisition time points comprise the current acquisition time point and R-1 acquisition time points before the current acquisition time point, and R is less than S; the recognition result of the target recognition image block passing through the strategy II can represent the increased value of the first occurrence frequency of the target recognition image block at R acquisition moments compared with the second occurrence frequency of the target recognition image block at the target acquisition moments;

The second strategy is that the first occurrence frequency of the target identification image block in the three-dimensional medical image data is obtained at R obtaining moments, and the third occurrence frequency of the target identification image block in the three-dimensional medical image data is obtained at S obtaining moments; calculating a quotient between the first frequency of occurrence and the third frequency of occurrence; determining a quotient between the first occurrence frequency and the third occurrence frequency as a recognition result of the target recognition image block; the S acquisition time points comprise the current acquisition time point and S-1 acquisition time points before the current acquisition time point; the R acquisition moments comprise the current acquisition moment and R-1 acquisition moments before the current acquisition moment, and R is smaller than S;

λ=i÷ε

λ is a recognition result obtained by a target recognition image block through a policy III, i is a first occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained by obtaining at R obtaining moments, and ε is a third occurrence frequency of the target recognition image block in the three-dimensional medical image data obtained by obtaining at S obtaining moments; the identification result of the target identification image segmentation passing through the strategy III is used for representing the proportion of the first occurrence frequency of the target identification image segmentation in R acquisition moments to the third occurrence frequency of the target identification image segmentation in S acquisition moments;

2. The method according to claim 1, wherein the medical image to be segmented is any slice in the target three-dimensional medical image data; the determining the inter-relevance evaluation index of the first medical image partition to be identified comprises the following steps:

3. The method of claim 2, wherein the set of neighborhood pixels includes neighborhood pixels that are surrounding pixels of the first medical image segment to be identified in the target three-dimensional medical image data; the obtaining the confidence that each neighborhood pixel and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data, wherein the obtaining the confidence that each neighborhood pixel and the first medical image block to be identified are combined and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data comprises obtaining the confidence that each surrounding pixel in the neighborhood pixel set and the first medical image block to be identified are combined and the first medical image block to be identified are respectively generated in the target three-dimensional medical image data;

4. The method of claim 1, wherein the target identification image tile is any one of the set of identification image tiles, the method further comprising:

5. The method according to claim 1, wherein the performing image blocking marking when the marking critical condition is identified according to the depth of the identification result comprises:

acquiring a preset mark reference value;

when the quotient corresponding to the identification result of the target identification image block is larger than or equal to the preset mark reference value, determining that a mark critical condition is met, and carrying out corresponding marking on the corresponding target identification image block;

the obtaining the preset mark reference value includes:

6. A lesion image segmentation device, comprising:

The index determining module is used for determining pathology auxiliary promotion indexes of all to-be-identified medical image blocks in the to-be-identified image block set, and determining to obtain a reference image block set from the to-be-identified image block set through the pathology auxiliary promotion indexes of all to-be-identified medical image blocks; the medical image blocks in the reference image block set are medical image blocks to be identified, wherein the pathology auxiliary promotion indexes corresponding to the medical image block set to be identified meet identification conditions; wherein the determining the pathology assistance promoting index of each medical image patch to be identified in the image patch set to be identified comprises: determining an internal relevance evaluation index of the first medical image block to be identified and an external relevance evaluation index of the first medical image block to be identified according to any first medical image block to be identified, wherein the block size of the first medical image block to be identified is larger than a size threshold value, and determining a pathology auxiliary promotion index of the first medical image block to be identified based on the internal relevance evaluation index of the first medical image block to be identified and the external relevance evaluation index of the first medical image block to be identified; wherein determining a pathology assistance promoting index of the first medical image patch to be identified based on the internal relevance evaluating index and the external relevance evaluating index of the first medical image patch to be identified includes: acquiring a first importance parameter corresponding to an internal relevance evaluation index of a first medical image partition, and acquiring a second importance parameter corresponding to an internal relevance evaluation index of the first medical image partition, wherein the pathological auxiliary promotion index of the first medical image partition is acquired according to the first importance parameter, the internal relevance evaluation index of the first medical image partition, the second importance parameter and the external relevance evaluation index of the first medical image partition; the calculation formula of the pathological auxiliary promotion index of the first medical image segmentation to be identified is as follows:

G＝αI+βO

the block determining module is used for determining an identification image block set according to the reference image block set, wherein identification image blocks in the identification image block set are suspicious image blocks; the medical image to be segmented is any slice in target three-dimensional medical image data, and the target three-dimensional medical image data is obtained at the current obtaining moment; the determining the identified image block set according to the reference image block set comprises the following steps: acquiring a reference lesion image set, wherein the reference lesion image set is formed by normal image segmentation and identification image segmentation determined according to three-dimensional medical image data acquired before the current acquisition moment; matching each medical image block in the reference image block set in the reference lesion image set, and determining the identification image block set based on medical image blocks which are not successfully matched;

The block marking module is used for carrying out block recognition of the recognition image according to the recognition image block set and carrying out block marking of the image when the recognition result depth recognition meets the marking critical condition; the identifying image block identification is carried out according to the identifying image block set, and the identifying image block identification comprises at least one of the following strategies:

θ=（i+t）÷（j+t）

λ=i÷ε

7. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, realizes the steps in the method according to any one of claims 1 to 5.