CN114119645B

CN114119645B - Method, system, device and medium for determining image segmentation quality

Info

Publication number: CN114119645B
Application number: CN202111413135.7A
Authority: CN
Inventors: 邹彤; 王瑜; 张欢; 金鸽; 钏兴炳; 王少康; 陈宽
Original assignee: Infervision Medical Technology Co Ltd
Current assignee: Infervision Medical Technology Co Ltd
Priority date: 2021-11-25
Filing date: 2021-11-25
Publication date: 2022-10-21
Anticipated expiration: 2041-11-25
Also published as: CN114119645A

Abstract

The embodiment of the invention discloses a method, a system, equipment and a medium for determining image segmentation quality. The method can comprise the following steps: acquiring an image segmentation result of a segmented image and a preset objective index, and determining an objective index value of the image segmentation result on the objective index; acquiring an objective index threshold value of an objective index, wherein the objective index threshold value is a threshold value determined according to a sample index numerical value and sample segmentation quality, the sample index numerical value is an index numerical value of a sample segmentation result of a sample segmentation image with the same image type as that of the segmented image on the objective index, and the sample segmentation quality is segmentation quality subjectively determined aiming at the sample segmentation image; and determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value. According to the technical scheme of the embodiment of the invention, the image segmentation quality is efficiently determined by the objective index threshold value which correlates the segmentation quality of subjective feeling with the objectively determined index value.

Description

Method, system, device and medium for determining image segmentation quality

Technical Field

The embodiment of the invention relates to the technical field of image processing, in particular to a method, a system, equipment and a medium for determining image segmentation quality.

Background

The image segmentation is a process of segmenting an image to be segmented into a plurality of specific regions with unique properties and extracting a region of interest from each region, and is a key step from image processing to image analysis, so that the image segmentation capable of providing better image segmentation quality is an important premise for image analysis.

In the current practice, the image segmentation quality is directly determined mainly by objective indexes set in advance. However, some detail features in the image to be segmented are difficult to directly express through objective indexes after all, and the detail features need to be subjectively identified through human eyes to obtain the image segmentation quality. Obviously, the existing scheme has the problem of low efficiency in determining the image segmentation quality.

Disclosure of Invention

The embodiment of the invention provides a method, a system, equipment and a medium for determining image segmentation quality, so as to realize the effect of efficiently determining the image segmentation quality.

In a first aspect, an embodiment of the present invention provides a method for determining image segmentation quality, which may include:

acquiring an image segmentation result of a segmented image and a preset objective index, and determining an objective index value of the image segmentation result on the objective index;

obtaining an objective index threshold value of an objective index, wherein the objective index threshold value is a threshold value determined according to a sample index value and sample segmentation quality, the sample index value is an index value of a sample segmentation result of a sample segmentation image with the same image category as that of the segmented image on the objective index, and the sample segmentation quality is segmentation quality determined subjectively aiming at the sample segmentation image;

and determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value.

In a second aspect, an embodiment of the present invention further provides a system for determining image segmentation quality, where the system may include:

the objective index numerical value determining module is used for acquiring an image segmentation result of the segmented image and preset objective indexes and determining objective index numerical values of the image segmentation result on the objective indexes;

an objective index threshold value obtaining module, configured to obtain an objective index threshold value of an objective index, where the objective index threshold value is a threshold value determined according to a sample index value and sample segmentation quality, the sample index value is an index value of a sample segmentation result of a sample segmentation image, which is the same as an image category of a segmented image, on an objective index, and the sample segmentation quality is segmentation quality subjectively determined for the sample segmentation image;

and the image segmentation quality determining module is used for determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value.

In a third aspect, an embodiment of the present invention further provides an apparatus for determining image segmentation quality, where the apparatus may include:

one or more processors;

a memory for storing one or more programs;

when executed by one or more processors, cause the one or more processors to implement a method for determining image segmentation quality as provided by any of the embodiments of the present invention.

In a fourth aspect, the embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method for determining image segmentation quality provided in any embodiment of the present invention.

According to the technical scheme of the embodiment of the invention, the image segmentation result of the segmented image and the preset objective index are obtained, and the objective index value of the image segmentation result on the objective index is determined, wherein the objective index value can show the image segmentation quality of the segmented image determined from an objective angle; acquiring an objective index threshold value of an objective index determined according to a sample index numerical value and sample segmentation quality, wherein the sample index numerical value is an index numerical value of a sample segmentation result of a sample segmentation image with the same image type as a segmented image on the objective index, the sample segmentation quality is segmentation quality determined subjectively aiming at the sample segmentation image, and the objective index threshold value is a threshold value capable of correlating the segmentation quality sensed subjectively and the objectively determined index numerical value; further, the image segmentation quality of the segmented image is determined based on a numerical relationship between the objective index value and the objective index threshold value. According to the technical scheme, the image segmentation quality is determined through the predetermined objective index threshold value capable of correlating the subjectively-sensed segmentation quality with the objectively-determined index value, so that the image segmentation quality capable of reflecting the subjective feeling is obtained under the condition that no artificial reference is needed, the determination accuracy of the image segmentation quality is guaranteed, the determination efficiency of the image segmentation quality is guaranteed, and the image segmentation quality determination scheme is a determination scheme of the subjectively-objectively combined image segmentation quality.

Drawings

Fig. 1 is a flowchart of a method for determining image segmentation quality in an embodiment of the present invention;

FIG. 2 is a schematic diagram illustrating calculation of objective index values of three-dimensional dice coefficients in a method for determining image segmentation quality according to an embodiment of the present invention;

FIG. 3 is a flowchart of model version selection during model iteration in a method for determining image segmentation quality according to an embodiment of the present invention;

FIG. 4 is a flow chart of another method for determining image segmentation quality in an embodiment of the present invention;

fig. 5 is a flowchart for determining image segmentation quality of each anatomical structure within a segmented image in another method for determining image segmentation quality in an embodiment of the present invention;

FIG. 6 is a flow chart of a further method for determining image segmentation quality in an embodiment of the present invention;

fig. 7 is a flowchart for determining consistency between model versions in still another method for determining image segmentation quality according to an embodiment of the present invention;

FIG. 8 is a flow chart of yet another method for determining image segmentation quality in an embodiment of the present invention;

FIG. 9 is a schematic diagram of a lung segment and a pulmonary artery and vein in another method for determining image segmentation quality according to the embodiment of the present invention;

FIG. 10 is a block diagram of an image segmentation quality determination system in an embodiment of the present invention;

fig. 11 is a schematic structural diagram of an apparatus for determining image segmentation quality in an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.

Fig. 1 is a flowchart of a method for determining image segmentation quality provided in an embodiment of the present invention. The present embodiment is applicable to a case where the image segmentation quality is determined quickly, and particularly, to a case where the image segmentation quality that can reflect the subjective feeling is determined quickly. The method can be executed by the image segmentation quality determination system provided by the embodiment of the invention, the system can be realized by software and/or hardware, and the system can be integrated on an image segmentation quality determination device, and the device can be various user terminals or servers.

Referring to fig. 1, the method of the embodiment of the present invention specifically includes the following steps:

s110, obtaining an image segmentation result of the segmented image and a preset objective index, and determining an objective index value of the image segmentation result on the objective index.

The image segmentation result may be a result obtained by performing image segmentation on a segmented image, and the segmented image may be a two-dimensional and/or three-dimensional natural image, a medical image, or the like, on which the image segmentation has been performed, and is not particularly limited herein. The objective index may be a preset index capable of objectively representing the image segmentation quality of the segmented image, such as an Average Surface Distance (ASD), a Hausdorff distance (Hausdorff distance), a three-dimensional dice (Volumetric dice) coefficient, and a centerline-based three-dimensional dice (cldie) coefficient, which are not specifically limited herein. Since rice can be understood as an aggregate similarity metric function, the above cldic can also be referred to as a three-dimensional centerline-based target coefficient, where the target coefficient can be a coefficient constructed based on the similarity metric function. In practical applications, optionally, when the segmented image is an image with a tubular structure subjected to image segmentation, such as a medical image of a blood vessel, a trachea, and the like, the objective index may include a three-dimensional dice coefficient based on a centerline, which may be used to measure whether the centerline (i.e., a skeleton) of a three-dimensional body is in the three-dimensional body, and is insensitive to the thickness of the tube diameter, and is more suitable for determining the image segmentation quality of the image with the tubular structure, thereby indirectly improving the segmentation accuracy of the tubular structure. The median line may be a line drawn from the midpoint of the cross-section of the tube in the tubular configuration.

The objective index value may include a value obtained after processing the image segmentation result based on the objective index. Illustratively, when the objective index is the average surface distance, it may represent the average of the distances of all surface points in the image segmentation result and the second gold standard segmentation result, and the smaller the value of the objective index, the smaller the contour difference between the two segmentation results. Specifically, the objective index value is calculated by the following formula:

a is a set of surface points a in the image segmentation result, and B is a set of surface points B in the second gold standard segmentation result.

Illustratively, when the objective index is a Hausdorff distance, the objective index value may be calculated by: for a point set A { a0, a1,. Eta } in the image segmentation result and a point set B { B0, B1, B2,. Eta } in the second golden standard segmentation result, taking a point a0 in A, calculating the distance from the point a0 to each point in B, and keeping the shortest distance d0; traversing each point in A, assuming that A comprises a0 and a1, and calculating d0 and d1; comparing all the distances { d0, d1}, selecting the longest distance, marking the longest distance as h, which is the one-way Hausdorff distance of A → B and is marked as h (A, B), and determining that the point in B must exist in a circle which takes a as the center of the circle and takes h as the radius for any point a in A; further, the character of A and B is exchanged, the one-way Hausdorff distance h (B, A) of B → A is calculated, and then the longest distance of h (A, B) and h (B, A) is selected as the two-way Hausdorff distance of A and B, and in this case, the smaller the objective index value, the smaller the contour difference between the image segmentation result and the second golden standard segmentation result.

Illustratively, when the objective index is a three-dimensional dice coefficient, referring to fig. 2, the correct segmentation result is the portion where the image segmentation result and the second golden standard segmentation result coincide, i.e., the image segmentation result and the second golden standard segmentation result are in alignmentWhen the image is segmented, the part is correctly segmented, and the objective index value can be calculated by the following formula:

wherein true positive represents the result of correct segmentation, false positive represents the part of the image segmentation result except the result of correct segmentation, and false negative represents the part of the second golden standard segmentation result except the result of correct segmentation.

For another example, when the objective index is a three-dimensional dice coefficient based on a centerline, the numerical value of the objective index may be calculated by the following formula:

wherein V _L Is the second golden standard segmentation result, V _P Is the result of image segmentation, S _L Is based on V _L The resulting center line, S _P Is based on V _P The resulting median line.

And S120, obtaining an objective index threshold value of the objective index, wherein the objective index threshold value is a threshold value determined according to a sample index numerical value and sample segmentation quality, the sample index numerical value is an index numerical value of a sample segmentation result of the sample segmentation image, which is the same as the image category of the segmented image, on the objective index, and the sample segmentation quality is the segmentation quality which is subjectively determined aiming at the sample segmentation image.

The objective index threshold may be a threshold predetermined for the objective index, and the determination process is as follows: acquiring a sample segmentation result of a sample segmentation image with the same image category as that of a segmented image, and determining a sample index value of the sample segmentation result on an objective index, wherein the sample segmentation image and the segmented image have the same target type, such as a pulmonary blood vessel image (the target is a pulmonary blood vessel at this time), a chest blood vessel image (the target is a chest blood vessel at this time), and the like; the determination process of the sample index value and the objective index value is similar, and is not described herein again. And acquiring the subjectively determined sample segmentation quality aiming at the sample segmentation image, namely the sample segmentation quality comprises the segmentation quality of the sample segmentation image which is artificially and subjectively determined according to the sample segmentation result. Further, an objective index threshold value that correlates the subjective segmentation quality with the objectively determined index value is determined based on the sample index value and the sample segmentation quality. It should be noted that, on the basis of not updating the objective index threshold, the above determination process of the objective index threshold needs to be performed only once, that is, the step requiring human operation needs to be performed only once. On this basis, the updating process of the objective index threshold may be performed when the subjective requirement of the user changes (for example, the standard good for image segmentation changes), the fineness of the automatic segmentation exceeds the gold standard (which may cause the objective index value to be poor, but actually the image segmentation effect is good), and the like, and is not specifically limited herein. In addition, the threshold number of the objective index threshold value can be one, two or more, which is related to the total number of the sample segmentation qualities, for example, when there are 2 sample segmentation qualities, the threshold number can be 1; when there are 3 sample segmentation qualities, the threshold number may be 2; and so on.

And S130, determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value.

The objective index threshold value can be a threshold value which can correlate the segmentation quality sensed subjectively and the index value determined objectively, so that the image segmentation quality of the segmented image determined according to the numerical relationship between the objective index value and the objective index threshold value can reflect the segmentation quality sensed subjectively, and the determination process does not need human participation, namely the image segmentation quality which can reflect the subjective feeling can be obtained through the objective index value under the condition of not needing human participation, so that the determination efficiency of the image segmentation quality is ensured while the determination precision of the image segmentation quality is ensured. In practical applications, optionally, the image segmentation quality may be represented by a numerical relationship, such as an objective index value greater than, less than, or equal to an objective index threshold value; the image segmentation quality can be represented by subjective feeling reflected by a numerical relationship, and if the image segmentation quality of the subjective feeling is qualified when the numerical value of the objective index is greater than or equal to the threshold value of the objective index, or else, the image segmentation quality of the subjective feeling is unqualified, the automatically determined image segmentation quality can be represented by qualification or unqualified; etc., and are not specifically limited herein. Optionally, the image segmentation result with poor image segmentation quality may be output, and then the output image segmentation result is evaluated manually, that is, each image segmentation result does not need to be evaluated manually, and only the image segmentation result with poor image segmentation quality that is output after automatic determination needs to be evaluated, so that the determination efficiency of the image segmentation quality is improved.

According to the technical scheme of the embodiment of the invention, the image segmentation result of the segmented image and the preset objective index are obtained, and the objective index value of the image segmentation result on the objective index is determined, wherein the objective index value can show the image segmentation quality of the segmented image determined from an objective angle; acquiring an objective index threshold value of an objective index determined according to a sample index numerical value and sample segmentation quality, wherein the sample index numerical value is an index numerical value of a sample segmentation result of a sample segmentation image with the same image type as a segmented image on the objective index, the sample segmentation quality is segmentation quality determined subjectively aiming at the sample segmentation image, and the objective index threshold value is a threshold value capable of correlating the segmentation quality sensed subjectively and the objectively determined index numerical value; further, the image segmentation quality of the segmented image is determined based on a numerical relationship between the objective index value and the objective index threshold value. According to the technical scheme, the image segmentation quality is determined through the predetermined objective index threshold value which can correlate the subjectively sensed segmentation quality with the objectively determined index value, so that the image segmentation quality which can reflect the subjective feeling is obtained under the condition of no need of artificial parameters, the determination accuracy of the image segmentation quality is guaranteed, the determination efficiency of the image segmentation quality is guaranteed, and the method is a determination scheme of the subjectively combined image segmentation quality.

On this basis, an optional technical solution is that obtaining an image segmentation result of a segmented image may include: acquiring a second golden standard segmentation result of the segmented image and an image segmentation result obtained by segmenting the segmented image in different image segmentation modes; determining the objective index value of the image segmentation result on the objective index can include: aiming at each image segmentation result, determining an objective index value of the image segmentation result on an objective index according to a second gold standard segmentation result; the method for determining the image segmentation quality may further include: and determining a target division mode from the image division modes according to the objective index value of each image division result. The second gold standard segmentation result can be a gold standard segmentation result manually marked on the segmented image; the image segmentation method may be a method adopted when the segmented image is segmented, such as an image segmentation algorithm based on a threshold, an image segmentation algorithm based on a region, an image segmentation algorithm based on an edge, an image segmentation algorithm based on a preset theory, an image segmentation model based on deep learning, and the like, and it should be noted that the different image segmentation methods may be different types of image segmentation algorithms, different versions of image segmentation models, and the like, and are not specifically limited herein. In other words, the number of results of the second gold standard segmentation result may be one, the number of results of the image segmentation result may be at least two, and the obtaining manner of each image segmentation result may differ. On the basis, the objective index value of each image segmentation result on the objective index can be respectively determined according to the second golden standard segmentation result, and then the target segmentation mode capable of obtaining a better objective index value is determined from each image segmentation mode according to the objective index value of each image segmentation result, so that the method is suitable for application scenes of selecting a better image segmentation algorithm from each image segmentation algorithm, selecting a better model version in a model iteration process and the like. For example, referring to fig. 3, each image segmentation result may be a result output by an image segmentation model of different versions, and each image segmentation result and the second golden standard segmentation result are input as a group into an objective evaluation system, which may be a system for determining an objective index value, and a corresponding objective index value is obtained according to the output result of the objective evaluation system, so that a more optimal model version may be selected according to each objective index value. The sample segmentation quality may also be referred to as a subjective evaluation result, and the sample index value and/or the objective index value may also be referred to as an objective evaluation result.

Fig. 4 is a flowchart of another method for determining image segmentation quality provided in the embodiment of the present invention. The present embodiment is optimized based on the above technical solutions. In this embodiment, optionally, the objective index threshold is predetermined by the following steps: acquiring a sample segmentation result and a first gold standard segmentation result of a sample segmentation image, and determining a sample index value of the sample segmentation result according to the first gold standard segmentation result; determining samples by taking the sample index numerical value and the acquired sample segmentation quality as a group of first threshold values, and determining objective index threshold values of the samples based on a plurality of groups of first threshold values; accordingly, determining the objective index value of the image segmentation result on the objective index may include: and obtaining a second gold standard segmentation result of the segmented image, and determining an objective index value of the image segmentation result on an objective index according to the second gold standard segmentation result. The same or corresponding terms as those in the above embodiments are not explained in detail herein.

Referring to fig. 4, the method of this embodiment may specifically include the following steps:

s210, obtaining a sample segmentation result and a first gold standard segmentation result of the sample segmentation image and a preset objective index, and determining a sample index value of the sample segmentation result on the objective index according to the first gold standard segmentation result.

The first gold standard segmentation result can be a gold standard segmentation result obtained by manually segmenting and labeling the sample segmentation image; the determination process of the sample index value is similar to that of the objective index value, and is obtained by processing the segmentation result and the first gold standard segmentation result based on the objective index, and is not described herein again.

S220, subjectively determined sample segmentation quality of the sample segmentation image is obtained, the sample index numerical value and the sample segmentation quality are used as a group of first threshold values to determine a sample, and an objective index threshold value of an objective index is determined by the sample based on the multiple groups of first threshold values.

For example, as the segmentation quality of some samples in the multiple sets of first threshold determination samples is the same, and the segmentation quality of some samples is different, for each set of first threshold determination samples having the same segmentation quality of the samples, the sample index values in the first threshold determination samples having the same segmentation quality of the samples may be counted to obtain a first value statistical result, where a specific statistical manner may be a mean value, a median value, a mode, a minimum value, a maximum value, and the like, and is not specifically limited herein; the objective index threshold is determined according to at least one first numerical statistical result, that is, the objective index threshold may be determined according to one, two or more first numerical statistical results, where each first numerical statistical result corresponds to a first threshold determination sample having the same sample segmentation quality. Of course, the objective index threshold may also be determined based on multiple sets of first threshold determination samples in other ways, which are not specifically limited herein.

And S230, acquiring an image segmentation result of the segmented image with the same image type as the sample segmented image and a second gold standard segmentation result, and determining an objective index value of the image segmentation result on an objective index according to the second gold standard segmentation result.

S240, determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value.

According to the technical scheme of the embodiment of the invention, a sample segmentation result and a first gold standard segmentation result of a sample segmentation image are obtained, and then a sample index value of the sample segmentation result is determined according to the first gold standard segmentation result; furthermore, the sample index value and the sample segmentation quality are used as a group of first threshold values to determine a sample, and the sample is determined based on a plurality of groups of first threshold values to determine an objective index threshold value, so that the image segmentation quality meeting the actual image segmentation requirement of the user can be obtained based on the objective index threshold value subsequently, because the second golden standard segmentation result involved in determining the objective index value possibly has a difference with the actual image segmentation requirement of the user.

In order to better understand the specific implementation process of the above technical solution, in consideration of application scenarios that may be involved in the embodiments of the present invention, the following describes the above technical solution exemplarily with reference to specific examples. For example, as shown in fig. 5, a sample segmentation image is a segmented medical image of a lung of a certain object, the image segmentation result includes at least two image segmentation regions segmented according to the anatomical structure of the lung in the medical image (i.e., image segmentation regions of blood vessels in a lung segment), and the objective index is cldie, for example, a sample segmentation region of a blood vessel in each lung segment of each object is obtained; for each sample segmentation area, inputting the sample segmentation area and the corresponding first gold standard segmentation area into an objective evaluation system to obtain a sample index value of CLDicice; acquiring the sample segmentation quality (such as qualified, unqualified and the like) of the sample segmentation area, wherein the sample segmentation quality only needs to be acquired once; and taking the sample index numerical value and the sample segmentation quality as a group of first threshold determination samples, matching the sample index numerical value and the sample segmentation quality in each group of first threshold determination samples, and performing statistical analysis to obtain an objective index threshold value, so that the subjective feeling corresponding to the sample index numerical value which is greater than or equal to the objective index threshold value is qualified, and otherwise, the subjective feeling is unqualified, thereby obtaining the objective evaluation system combined with subjective prior. Therefore, after the image segmentation region and the second gold standard segmentation region are subsequently input into the objective evaluation system combined with subjective prior, the image segmentation quality capable of reflecting subjective feeling can be obtained. Furthermore, a doctor only needs to subjectively evaluate the blood vessels in the pulmonary segment of the object under the objective index threshold value, and does not need to subjectively evaluate the blood vessels in each pulmonary segment of each object, so that the subjective evaluation efficiency is greatly improved. In practical applications, the objective index threshold corresponding to the blood vessel in each lung segment may be the same or different.

Fig. 6 is a flowchart of still another method for determining image segmentation quality provided in the embodiment of the present invention. The present embodiment is optimized based on the above technical solutions. In this embodiment, optionally, the sample segmentation result includes a first sample segmentation result and a second sample segmentation result, the sample segmentation quality includes a first sample segmentation quality subjectively determined for the first sample segmentation result and a second sample segmentation quality subjectively determined for the second sample segmentation result, and the objective index threshold may be predetermined by the following steps: obtaining a first sample segmentation result and a second sample segmentation result, and determining a sample index value according to the first sample segmentation result and the second sample segmentation result; acquiring a first sample segmentation quality and a second sample segmentation quality, determining samples by taking the first sample segmentation quality, the second sample segmentation quality and sample index values as a group of second threshold values, and determining objective index threshold values of the samples based on the plurality of groups of second threshold values; accordingly, the image segmentation result includes a first image segmentation result and a second image segmentation result, and determining an objective index value of the image segmentation result on an objective index may include: determining an objective index value of the image segmentation result on an objective index according to the first image segmentation result and the second image segmentation result; accordingly, determining the image segmentation quality of the segmented image may include: it is determined whether an image segmentation quality of a first image segmentation result of the segmented image and an image segmentation quality of a second image segmentation result of the segmented image are consistent. The same or corresponding terms as those in the above embodiments are not explained in detail herein.

Referring to fig. 6, the method of this embodiment may specifically include the following steps:

s310, obtaining a sample segmentation result and a sample segmentation quality of a sample segmentation image, and a preset objective index, wherein the sample segmentation result comprises a first sample segmentation result and a second sample segmentation result, and the sample segmentation quality comprises a first sample segmentation quality subjectively determined for the first sample segmentation result and a second sample segmentation quality subjectively determined for the second sample segmentation result.

Wherein the first sample segmentation result and the second sample segmentation result may be sample segmentation results obtained based on different image segmentation modes, the first sample segmentation quality may be a sample segmentation quality for the first sample segmentation result, and the second sample segmentation quality may be a sample segmentation quality for the second sample segmentation result.

And S320, determining a sample index value of the sample segmentation result on the objective index according to the first sample segmentation result and the second sample segmentation result.

The step can be understood as that the second sample segmentation result is used as a first gold standard segmentation result, and a sample index value of the first sample segmentation result on the objective index is determined; it may also be understood as determining a sample index value of the second sample segmentation result on the objective index using the first sample segmentation result as the first gold standard segmentation result. In other words, in order to reduce the workload of manual labeling, the first gold standard segmentation result is no longer involved in determining the sample index value, and the step of determining the objective index threshold in S330 is combined, so that the objective index threshold for determining whether the subjectively perceived segmentation quality changes can be obtained.

S330, determining samples by taking the first sample segmentation quality, the second sample segmentation quality and the sample index value as a group of second threshold values, and determining objective index threshold values of the samples based on the plurality of groups of second threshold values.

For example, since some of the first sample division quality and the second sample division quality in the multiple sets of second threshold determination samples are the same (if both are qualified or unqualified), and some are different (if one is qualified and the other is unqualified), determining the sample by using a third threshold in each set of second threshold determination samples, wherein the first sample division quality and the second sample division quality in the sample are the same, and counting the sample index values in each third threshold determination sample to obtain a second numerical statistic result, and then determining the objective index threshold according to at least one second numerical statistic result; and/or determining samples by aiming at fourth threshold values with different first sample segmentation quality and second sample segmentation quality in each group of second threshold values, counting sample index values in each fourth threshold value determination sample to obtain third value statistical results, and determining objective index threshold values according to at least one third value statistical result. The statistical methods of the first, second and third numerical statistical results may be the same or different. In addition, the objective index threshold may also be determined based on the plurality of sets of second threshold determination samples in other ways, which are not specifically limited herein.

S340, obtaining an image segmentation result of the segmented image with the same image category as the sample segmented image, wherein the image segmentation result comprises a first image segmentation result and a second image segmentation result.

The image segmentation modes adopted in the obtaining process of the first image segmentation result and the first sample segmentation result can be the same or different, and are not specifically limited herein; the second image segmentation result and the second sample segmentation result are similar, and are not described herein again.

And S350, determining an objective index value of the image segmentation result on the objective index according to the first image segmentation result and the second image segmentation result.

The determination process of the objective index value and the sample index value is similar, and is not described herein again.

S360, determining whether the image segmentation quality of the first image segmentation result of the segmented image is consistent with the image segmentation quality of the second image segmentation result according to the numerical relation between the objective index numerical value and the objective index threshold value.

The objective index threshold is a threshold obtained according to the sample index value, the first sample segmentation quality and the second sample segmentation quality, so that the numerical relationship between the objective index threshold and the sample index value can reflect whether the first sample segmentation quality and the second sample segmentation quality are consistent, that is, whether the segmentation quality of subjective feeling is changed, which means that whether the image segmentation quality of the first image segmentation result is consistent with the image segmentation quality of the second image segmentation result can be determined according to the numerical relationship between the objective index value and the objective index threshold, which is suitable for an application scenario for quickly determining the superiority of two image segmentation modes, that is, for an application scenario for quickly determining the conformity of two image segmentation models, that is, for quickly determining whether the image segmentation model of the current version is superior to or inferior to the image segmentation performance of the image segmentation model of the previous version (where the image segmentation quality of the two image segmentation results is inconsistent or inferior to the image segmentation quality).

According to the technical scheme of the embodiment of the invention, the sample index value is determined according to the first sample segmentation result and the second sample segmentation result, then the first sample segmentation quality, the second sample segmentation quality and the sample index value are used as a group of second threshold values to determine the sample, and the sample is determined based on a plurality of groups of second threshold values to determine the objective index threshold value which can reflect whether the segmentation quality sensed subjectively changes; in this way, after the objective index value is determined according to the first image segmentation result and the second image segmentation result, whether the image segmentation qualities of the two image segmentation results are consistent or not can be determined according to the numerical relationship between the objective index value and the objective index threshold value, so that the effect of quickly determining whether the image segmentation qualities of the image segmentation results obtained based on different image segmentation modes are changed or not is achieved.

In order to better understand a specific implementation process of the foregoing technical solution, in consideration of an application scenario that may be related to an embodiment of the present invention, an exemplary description is given below with reference to a specific example. For example, taking the segmented image including the segmented medical image including the target portion, and the image segmentation result including at least two image segmentation regions segmented according to the anatomical structure of the target portion in the medical image as an example, in the model iteration process, in a normal case, the change of the image segmentation model of the current version relative to the image segmentation model of the previous version may be small, which reflects that only a small part of the image segmentation regions on the image segmentation result may have changed, so it is not necessary to evaluate the image segmentation quality of the image segmentation model of the current version through the second golden standard segmentation result; meanwhile, the image segmentation model of the current version may be optimized only for some image segmentation areas, so that there is a need for evaluation of these image segmentation areas, and thus the technical solution described above may be adopted to determine the consistency of the two image segmentation models.

Specifically, referring to fig. 7, taking an example that a sample segmentation image is a pulmonary blood vessel image of a certain object, a sample segmentation region is a segmentation result of a certain blood vessel in a lung segment, and an objective index is cldic, a first sample segmentation region and a second sample segmentation region of a blood vessel in each lung segment of each object output by two versions of image segmentation models, and a first sample segmentation quality and a second sample segmentation quality (such as pass, fail, etc.) for the two sample segmentation regions are respectively obtained, and the two sample segmentation qualities only need to be obtained once; inputting a first sample segmentation area and a second sample segmentation area of blood vessels in the same lung segment belonging to the same object into an objective evaluation system to obtain a sample index value of CLDice; the sample index value, the first sample segmentation quality and the second sample segmentation quality are used as a group of second threshold value determination samples, the sample index values in each group of second threshold value determination samples are matched with the two sample segmentation qualities, objective index threshold values are obtained through statistical analysis, so that subjective feelings corresponding to the sample index values which are larger than or equal to the objective index threshold values are qualified or unqualified (namely, the segmentation quality sensed subjectively does not change), correspondingly, the subjective feelings corresponding to the sample index values which are smaller than the objective index threshold values are qualified, and unqualified (namely, the segmentation quality sensed subjectively changes), and accordingly, an objective evaluation system combined with subjective priors can be obtained. In this way, after two image segmentation regions of a blood vessel in a certain lung segment of a certain subject are subsequently input into the objective evaluation system combined with subjective priors, it can be determined whether the subjectively perceived image segmentation quality for the two image segmentation regions changes. Therefore, relative to the previous version, related personnel only need to subjectively evaluate whether the image segmentation quality of the image segmentation model of the current version is better or worse for the image segmentation region with large change (namely, blood vessels in the lung segment under the objective index threshold), so that the effect of rapidly inspecting the expression of the image segmentation models of the two versions relative to the changed image segmentation region is realized, and further the subjective evaluation efficiency is greatly improved.

Fig. 8 is a flowchart of still another method for determining image segmentation quality according to an embodiment of the present invention. The present embodiment is optimized based on the above technical solutions. In this embodiment, optionally, the segmented image is a segmented medical image including the target region, and the image segmentation result includes at least two image segmentation regions segmented according to the anatomical structure of the target region in the medical image; determining the objective index value of the image segmentation result on the objective index may include: determining an objective index value of each image segmentation area on an objective index; determining image segmentation quality of the segmented image may include: image segmentation quality of an image segmentation region in a segmented image is determined. The same or corresponding terms as those in the above embodiments are not explained in detail herein.

Referring to fig. 8, the method of this embodiment may specifically include the following steps:

s410, obtaining an image segmentation result of a segmented image and a preset objective index, wherein the segmented image is a segmented medical image containing a target part, and the image segmentation result comprises at least two image segmentation areas segmented according to the anatomical structure of the target part in the medical image.

The target region may be a region in a human body, the region may be a tissue or an organ, such as a lung, a chest, a heart, etc., and the segmented image may be a medical image that includes the target region and has been segmented. Since the target region can be divided into at least two parts according to the anatomical structure of the target region, at least two image segmentation regions can be obtained after segmenting the segmented image, and each image segmentation region can correspond to a respective anatomical structure.

S420, determining an objective index value of the image segmentation area on the objective index for each image segmentation area.

And S430, obtaining an objective index threshold value of the objective index, wherein the objective index threshold value is a threshold value determined according to a sample index numerical value and sample segmentation quality, the sample index numerical value is an index numerical value of a sample segmentation result of the sample segmentation image with the same image category as that of the segmented image on the objective index, and the sample segmentation quality is segmentation quality subjectively determined aiming at the sample segmentation image.

The objective index threshold value in this step is a threshold value suitable for each anatomical structure. In practical applications, optionally, the objective index threshold corresponding to each anatomical structure may be determined separately, and specifically, for each anatomical structure, the objective index threshold may be determined according to a sample index value of a sample segmentation region corresponding to the anatomical structure on the objective index and a sample segmentation quality for the sample segmentation region.

S440, determining image segmentation quality of image segmentation areas in the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value.

Wherein, the step is to determine the image segmentation quality based on objective index threshold values suitable for each anatomical structure. In practical applications, optionally, the image segmentation quality may also be determined based on an objective index threshold corresponding to the anatomical structure, and specifically, for an image segmentation region of which the image segmentation quality is to be determined, the image segmentation quality of the image segmentation region may be determined according to an objective index value and an objective index threshold matching the anatomical structure in the image segmentation region.

According to the technical scheme of the embodiment of the invention, the image segmentation quality of each image segmentation area in the segmented image is respectively evaluated, so that the problems that the image segmentation quality of a certain image segmentation area is covered due to the good overall image segmentation quality and the image segmentation quality of which image segmentation area cannot be directly given is poor when the image segmentation quality of the whole segmented image is directly evaluated are solved, the image segmentation area with poor image segmentation quality can be rapidly positioned, and the convenience of subsequent targeted subjective evaluation is improved.

In order to better understand the specific implementation process of the above technical solution, in consideration of application scenarios that may be involved in the embodiments of the present invention, the following describes the above technical solution exemplarily with reference to specific examples. For example, the segmented image is a pulmonary vessel image, which may involve three concepts of a lung segment, a pulmonary vessel, and a pulmonary trachea. In particular, the lung tissue in which each lung segment including the bronchus and its branches is located is called a bronchus lung segment (bronchus single segments), which is conical, with the tip facing the lung portal and the bottom located on the lung surface, and the connective tissue and pulmonary vein branches are separated between adjacent lung segments. Within the lung segment, branches of the pulmonary artery accompany the pulmonary segment bronchial tubes, but branches of the pulmonary veins are distributed between the lung segments. The left and right lungs typically have 10 bronchopulmonary segments, sometimes with common bronchi in the left lung, and in this case the left lung can be divided into 8 segments, as an example, see fig. 9. The pulmonary blood vessels are divided into pulmonary arteries and pulmonary veins, which are structured to deliver deoxygenated blood flow to the surrounding lungs and oxygenated blood flow to systemic circulation, respectively. These highly complex branched structures support the lung with blood in close association with it to deliver fresh gas to the terminal air sacs (alveoli) through the process of respiration. Segmentation of the pulmonary arterial and pulmonary venous vascular tree plays a very important role in the clinic, as it enables identification of the vascular tree of pulmonary embolism (local obstruction), detection of pulmonary signs of hypertension, and differentiation between the vasculature and local turbidity (for detection of lung cancer and other local lesions); the vessel tree can also be used as a roadmap to track lung tissue across lung volumes to continuously monitor the time variation of the lungs or across time. The pulmonary airways are the gas transport channels between the pulmonary capillaries and the air, and are pathologically involved in various pulmonary diseases such as Chronic Obstructive Pulmonary Disease (COPD), lung cancer and inflammatory lung disease. The segmentation of airway images from computed tomography plays an important role in analysis, and accurate three-dimensional lung airway segmentation will greatly benefit for diagnosis and preoperative assessment of airway diseases.

When the objective index value is determined in units of image segmentation areas, numerical statistics can be performed from different dimensions. In an exemplary manner, the first and second electrodes are,

(1) For a review of the image segmentation quality based on anatomical partitioning of an object dimension (i.e. an object), see table 1. Specifically, the three-dimensional image segmentation result of the pulmonary artery and vein blood vessel is segmented into 18 image segmentation areas, and the image segmentation quality of the image segmentation area of the artery and vein blood vessel in each lung segment range is respectively determined, so that objective index values of four objective indexes of the artery and vein of each lung segment of each object in a test set can be obtained and cannot be submerged by the overall objective index values, wherein the test set comprises a plurality of objects. Meanwhile, when the objective index value of the blood vessel under the individual lung segment of a certain object is reduced, the specific position with poor image segmentation quality can be rapidly and definitely positioned, and targeted subjective examination is facilitated.

TABLE 1 examination of image segmentation quality effects based on anatomical partitioning of object dimensions

(2) Examination of the effect of image segmentation quality based on anatomical partitioning of the test set dimensions is shown in table 2. Specifically, each objective index value in table 2 is an average value of objective index values of each object in the test set on the corresponding objective index, which is calculated by taking a certain lung segment as a unit, and is set to determine which lung segment has poor image segmentation quality in the currently adopted image segmentation method.

TABLE 2 examination of set dimensions for image segmentation quality effects based on anatomical partitioning

(3) See table 3 for a review of the image segmentation quality effect for object dimensions. Specifically, each objective index value in table 3 is an average value of objective index values of lung segments of a certain object on the corresponding objective index, which is calculated by taking the certain object as a unit, and is suitable for selecting a model version in a model iteration process.

TABLE 3 examination of image segmentation quality effects for object dimensions

(4) Examination of the image segmentation quality effect of the test set dimension is shown in table 4. Specifically, table 4 shows the overall image segmentation quality effect from the test set, and the artery and vein statistics are performed separately, which is suitable for selecting the model version in the model iteration process.

TABLE 4 examination of image segmentation quality effects for test set dimensions

It should be noted that the first two dimensions are relatively fine dimensions, and the second two dimensions are relatively coarse (starting from the whole). The technical solutions set forth in the embodiments of the present invention can be matched with the several dimensions to achieve corresponding technical effects, and the specific matching manner is not specifically limited herein.

Fig. 10 is a block diagram of a system for determining image segmentation quality according to an embodiment of the present invention, which is configured to execute a method for determining image segmentation quality according to any of the embodiments described above. The system and the method for determining the image segmentation quality of the embodiments belong to the same inventive concept, and details which are not described in detail in the embodiments of the system for determining the image segmentation quality can refer to the embodiments of the method for determining the image segmentation quality. Referring to fig. 10, the system may specifically include: an objective index value determination module 510, an objective index threshold acquisition module 520, and an image segmentation quality determination module 530.

The objective index value determining module 510 is configured to obtain an image segmentation result of the segmented image and a preset objective index, and determine an objective index value of the image segmentation result on the objective index;

an objective index threshold obtaining module 520, configured to obtain an objective index threshold of an objective index, where the objective index threshold is a threshold determined according to a sample index value and sample segmentation quality, the sample index value is an index value of a sample segmentation result of a sample segmentation image of the same image category as that of the segmented image on an objective index, and the sample segmentation quality is a segmentation quality subjectively determined for the sample segmentation image;

an image segmentation quality determination module 530, configured to determine an image segmentation quality of the segmented image according to a numerical relationship between the objective index value and the objective index threshold value.

Optionally, the objective index threshold is predetermined by the following modules:

the first sample index value determining module is used for obtaining a sample segmentation result and a first gold standard segmentation result of the sample segmentation image, and determining a sample index value of the sample segmentation result according to the first gold standard segmentation result;

the first objective index threshold value determining module is used for determining samples by taking the sample index values and the acquired sample segmentation quality as a group of first threshold values and determining the objective index threshold values of the samples on the basis of a plurality of groups of first threshold values;

the objective indicator value determination module 510 may include:

and the first objective index numerical value determining unit is used for acquiring a second gold standard segmentation result of the segmented image and determining an objective index numerical value of the image segmentation result on the objective index according to the second gold standard segmentation result.

On this basis, optionally, the first objective indicator threshold determining module may include:

a first numerical value statistical result obtaining unit, configured to determine, for each group of first threshold values, each first threshold value determination sample having the same sample segmentation quality in the samples, and perform statistics on sample index numerical values in each first threshold value determination sample having the same sample segmentation quality to obtain a first numerical value statistical result;

a first objective index threshold value determination unit for determining an objective index threshold value according to at least one first numerical statistical result.

Optionally, the sample segmentation result includes a first sample segmentation result and a second sample segmentation result, and the sample segmentation quality includes a first sample segmentation quality subjectively determined for the first sample segmentation result and a second sample segmentation quality subjectively determined for the second sample segmentation result;

the objective index threshold is predetermined by:

the second sample index value determining module is used for obtaining a first sample segmentation result and a second sample segmentation result and determining a sample index value according to the first sample segmentation result and the second sample segmentation result;

the second objective index threshold value determining module is used for acquiring the first sample segmentation quality and the second sample segmentation quality, determining samples by taking the first sample segmentation quality, the second sample segmentation quality and the sample index numerical value as a group of second threshold values, and determining the objective index threshold value of the samples based on a plurality of groups of second threshold values;

the image segmentation result includes a first image segmentation result and a second image segmentation result, and the objective index value determination module 510 may include:

a second objective index value determining unit, configured to determine an objective index value of the image segmentation result on the objective index according to the first image segmentation result and the second image segmentation result;

the image segmentation quality determination module 530 may include:

a first image segmentation quality determination unit for determining whether an image segmentation quality of a first image segmentation result of the segmented image and an image segmentation quality of a second image segmentation result are consistent.

On this basis, optionally, the second objective index threshold determining module may include:

a second numerical value statistical result obtaining unit, configured to determine, for each group of second threshold values, third threshold value determination samples in which the first sample segmentation quality and the second sample segmentation quality are the same, and perform statistics on sample index values in each third threshold value determination sample to obtain a second numerical value statistical result;

a second objective index threshold value determination unit for determining an objective index threshold value based on at least one second numerical statistic result; and/or the presence of a gas in the gas,

a third numerical value statistical result obtaining unit, configured to determine, for each group of second threshold values, fourth threshold determination samples with different first sample segmentation qualities and second sample segmentation qualities, and perform statistics on sample index values in each fourth threshold determination sample to obtain a third numerical value statistical result;

and the third objective index threshold value determining unit is used for determining the objective index threshold value according to at least one third numerical value statistical result.

Optionally, the segmented image is a segmented medical image including the target region, and the image segmentation result includes at least two image segmentation regions segmented according to the anatomical structure of the target region in the medical image; the objective indicator value determination module 510 may include:

a third objective index value determination unit, configured to determine, for each image segmentation region, an objective index value of the image segmentation region on an objective index;

the image segmentation quality determination module 530 may include:

and a second image segmentation quality determination unit for determining the image segmentation quality of the image segmentation area in the segmented image.

Optionally, the objective index value determining module 510 may include:

the segmentation result acquisition unit is used for acquiring a second gold standard segmentation result of the segmented image and an image segmentation result obtained by segmenting the segmented image in different image segmentation modes;

the objective indicator value determination module 510 may include:

a fourth objective index numerical value determining unit, configured to determine, for each image segmentation result, an objective index numerical value of the image segmentation result on the objective index according to the second gold standard segmentation result;

the system for determining the image segmentation quality may further include:

and the target segmentation mode determining module is used for determining a target segmentation mode from the image segmentation modes according to the objective index value of each image segmentation result.

Alternatively, when the segmented image includes an image having a tubular structure subjected to image segmentation, the objective index may include a three-dimensional centerline-based target coefficient, and the target coefficient may include a coefficient constructed based on a similarity metric function.

The image segmentation quality determining system provided by the embodiment of the invention obtains the image segmentation result of the segmented image and the preset objective index through the objective index value determining module, and determines the objective index value of the image segmentation result on the objective index, wherein the objective index value can show the image segmentation quality of the segmented image determined from an objective angle; obtaining an objective index threshold value of an objective index according to a sample index numerical value and sample segmentation quality, wherein the sample index numerical value is an index numerical value of a sample segmentation result of a sample segmentation image with the same image type as a segmented image on the objective index, the sample segmentation quality is segmentation quality which is subjectively determined aiming at the sample segmentation image, and the objective index threshold value is a threshold value which can correlate the segmentation quality which is subjectively sensed with the objectively determined index numerical value; and further, determining the image segmentation quality of the segmented image according to the numerical relation between the objective index numerical value and the objective index threshold value through an image segmentation quality determination module. The system determines the image segmentation quality through the predetermined objective index threshold value which can correlate the segmentation quality sensed subjectively and the objectively determined index value, so that the image segmentation quality which can reflect the subjective feeling is obtained under the condition of no need of artificial reference, the determination efficiency of the image segmentation quality is ensured while the determination precision of the image segmentation quality is ensured, and the method is a determination scheme of the image segmentation quality which combines subjectivity and objectivity.

The image segmentation quality determining system provided by the embodiment of the invention can execute the image segmentation quality determining method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the executing method.

It should be noted that, in the embodiment of the system for determining image segmentation quality, the included units and modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.

Fig. 11 is a schematic structural diagram of an apparatus for determining image segmentation quality according to an embodiment of the present invention, as shown in fig. 11, the apparatus includes a memory 610, a processor 620, an input system 630, and an output system 640. The number of processors 620 in the device may be one or more, and one processor 620 is taken as an example in fig. 11; the memory 610, processor 620, input system 630, and output system 640 in the device may be connected by a bus or other means, such as by bus 650 in fig. 11.

The memory 610 serves as a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the image segmentation quality determination method in the embodiment of the present invention (for example, the objective index value determination module 510, the objective index threshold value acquisition module 520, and the image segmentation quality determination module 530 in the image segmentation quality determination system). The processor 620 executes various functional applications of the device and data processing by executing software programs, instructions, and modules stored in the memory 610, that is, implements the above-described method of determining the image segmentation quality.

The memory 610 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function; the storage data area may store data created according to use of the device, and the like. Further, the memory 610 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, memory 610 may further include memory located remotely from processor 620, which may be connected to devices through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The input system 630 may be used to receive entered numeric or character information and to generate key signal inputs relating to user settings and function controls of the system. The output system 640 may include a display device such as a display screen.

An embodiment of the present invention provides a storage medium containing computer-executable instructions, which when executed by a computer processor, are configured to perform a method for determining image segmentation quality, the method comprising:

acquiring an objective index threshold value of an objective index, wherein the objective index threshold value is a threshold value determined according to a sample index numerical value and sample segmentation quality, the sample index numerical value is an index numerical value of a sample segmentation result of a sample segmentation image with the same image type as that of the segmented image on the objective index, and the sample segmentation quality is segmentation quality subjectively determined aiming at the sample segmentation image;

Of course, the storage medium containing the computer-executable instructions provided by the embodiments of the present invention is not limited to the method operations described above, and may also perform related operations in the method for determining image segmentation quality provided by any embodiment of the present invention.

From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly, can also be implemented by hardware, but the former is a better embodiment in many cases. With this understanding, the technical solutions of the present invention may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a FLASH Memory (FLASH), a hard disk or an optical disk of a computer, and includes instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for determining image segmentation quality is characterized by comprising the following steps:

obtaining an objective index threshold value of the objective index, wherein the objective index threshold value is a threshold value determined according to a sample index value and a sample segmentation quality, the sample index value is an index value of a sample segmentation result of a sample segmentation image of the same image category as the segmented image on the objective index, and the sample segmentation quality is a segmentation quality subjectively determined for the sample segmentation image;

2. The method according to claim 1, wherein the objective index threshold is predetermined by:

obtaining the sample segmentation result and a first gold standard segmentation result of the sample segmentation image, and determining the sample index value of the sample segmentation result according to the first gold standard segmentation result;

determining samples by using the sample index values and the acquired sample segmentation quality as a group of first threshold values, and determining the objective index threshold value based on a plurality of groups of the first threshold values;

the determining the objective index value of the image segmentation result on the objective index comprises:

and acquiring a second gold standard segmentation result of the segmented image, and determining an objective index value of the image segmentation result on the objective index according to the second gold standard segmentation result.

3. The method according to claim 2, wherein determining the objective index threshold based on the plurality of sets of the first threshold determination samples comprises:

for each first threshold determination sample with the same sample segmentation quality in each group of first threshold determination samples, counting the sample index values in each first threshold determination sample with the same sample segmentation quality to obtain a first value statistical result;

determining the objective index threshold according to at least one of the first numerical statistics.

4. The method according to claim 1, wherein the sample segmentation result comprises a first sample segmentation result and a second sample segmentation result, the sample segmentation quality comprises a first sample segmentation quality subjectively determined for the first sample segmentation result and a second sample segmentation quality subjectively determined for the second sample segmentation result, and the objective index threshold is predetermined by:

obtaining the first sample segmentation result and the second sample segmentation result, and determining the sample index value according to the first sample segmentation result and the second sample segmentation result;

acquiring the first sample segmentation quality and the second sample segmentation quality, determining samples by taking the first sample segmentation quality, the second sample segmentation quality and the sample index value as a group of second threshold values, and determining the objective index threshold value based on a plurality of groups of second threshold values;

wherein the determining the sample index value according to the first sample segmentation result and the second sample segmentation result comprises: taking the second sample segmentation result as a first gold standard segmentation result, and determining the sample index value of the first sample segmentation result on an objective index; or, taking the first sample segmentation result as a first gold standard segmentation result, and determining the sample index value of the second sample segmentation result on an objective index;

the image segmentation result comprises a first image segmentation result and a second image segmentation result, and the determining the objective index value of the image segmentation result on the objective index comprises:

determining an objective index value of the image segmentation result on the objective index according to the first image segmentation result and the second image segmentation result;

the determining the image segmentation quality of the segmented image comprises:

determining whether an image segmentation quality of the first image segmentation result of the segmented image and an image segmentation quality of the second image segmentation result of the segmented image are consistent.

5. The method according to claim 4, wherein determining the objective indicator threshold based on the plurality of sets of the second threshold determination samples comprises:

determining third threshold determination samples with the same first sample segmentation quality and second sample segmentation quality in each group of second threshold determination samples, and counting the sample index values in each third threshold determination sample to obtain a second numerical value counting result;

determining the objective index threshold value according to at least one second numerical statistic result; and/or the presence of a gas in the gas,

determining fourth threshold determination samples with different first sample segmentation quality and second sample segmentation quality in each group of second threshold determination samples, and counting the sample index values in each fourth threshold determination sample to obtain a third value statistical result;

and determining the objective index threshold value according to at least one third numerical statistical result.

6. The method according to claim 1, wherein the segmented image is a segmented medical image containing a target region, and the image segmentation result includes at least two image segmentation regions segmented according to an anatomical structure of the target region in the medical image;

for each image segmentation region, determining an objective index value of the image segmentation region on the objective index;

determining an image segmentation quality of the image segmentation region in the segmented image.

7. The method of claim 1, wherein obtaining image segmentation results for segmented images comprises: acquiring a second golden standard segmentation result of the segmented image and an image segmentation result obtained by segmenting the segmented image in different image segmentation modes;

for each image segmentation result, determining an objective index value of the image segmentation result on the objective index according to the second gold standard segmentation result;

the method further comprises the following steps:

and determining a target segmentation mode from the image segmentation modes according to the objective index value of each image segmentation result.

8. The method according to claim 1, wherein when the segmented image comprises an image having a tubular structure that has been image segmented, the objective index comprises a midline-based three-dimensional target coefficient including a coefficient constructed based on a similarity metric function.

9. A system for determining image segmentation quality, comprising:

the objective index value determining module is used for acquiring an image segmentation result of the segmented image and a preset objective index and determining an objective index value of the image segmentation result on the objective index;

an objective index threshold obtaining module, configured to obtain an objective index threshold of the objective index, where the objective index threshold is a threshold determined according to a sample index value and sample segmentation quality, the sample index value is an index value of a sample segmentation result of a sample segmentation image of the same image category as the segmented image on the objective index, and the sample segmentation quality is segmentation quality determined subjectively for the sample segmentation image;

10. An apparatus for determining image segmentation quality, comprising:

one or more processors;

a memory for storing one or more programs;

when executed by the one or more processors, cause the one or more processors to implement the method of determining image segmentation quality as claimed in any one of claims 1 to 8.

11. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method of determining the quality of image segmentation according to any one of claims 1 to 8.