CN112164059A

CN112164059A - Focus detection method, device and related product

Info

Publication number: CN112164059A
Application number: CN202011125994.1A
Authority: CN
Inventors: 彭成宝; 邱文旭; 孟庆余
Original assignee: Shenyang Neusoft Intelligent Medical Technology Research Institute Co Ltd
Current assignee: Shenyang Neusoft Intelligent Medical Technology Research Institute Co Ltd
Priority date: 2020-10-20
Filing date: 2020-10-20
Publication date: 2021-01-01
Anticipated expiration: 2040-10-20
Also published as: CN112164059B

Abstract

The application discloses a focus detection method, a focus detection device and a related product. In the method, an analytic hierarchy process is used for determining a focus detection algorithm from a plurality of focus detection algorithms as a target algorithm; and detecting the focus in the target medical image by using the target algorithm. According to the technical scheme, a target algorithm is determined from multiple focus detection algorithms to serve as a target layer, multiple efficacy indexes are used as a criterion layer, multiple focus detection algorithms are used as scheme layers, and a three-layer structure hierarchical model of the target layer, the criterion layer and the scheme layer is built in the mode. By applying the technical scheme of the application, an algorithm which meets the requirement of an efficiency index and is suitable for detecting the focus in the target medical image can be finally determined from a plurality of focus detection algorithms by applying an analytic hierarchy process on the basis of the hierarchical model. According to the technical scheme, the method and the device can assist the user in selecting the focus detection algorithm, reduce the difficulty of the user in selecting the algorithm and improve the focus detection efficiency.

Description

Focus detection method, device and related product

Technical Field

The present application relates to the field of image processing technologies, and in particular, to a method and an apparatus for detecting a lesion, and a related product.

Background

With the development of electronic computer technology and medical imaging technology, rich information contained in medical images can play an important role in clinical diagnosis, medical teaching, scientific research and other aspects. The type of medical image is related to the imaging mode, and the types mainly include Computed Tomography (CT) image, Magnetic Resonance (MR) image, and the like. Due to the complexity, diversity, and interdisciplinary nature of medical image types and disease types, existing lesion detection and marking technologies mainly focus on providing lesion detection and marking methods for specific lesions and medical image types, resulting in increasing kinds and numbers of lesion detection algorithms or tools. Even the same lesion and the same medical image type have produced numerous lesion detection algorithms.

In many cases, the user (e.g., medical staff) does not know the characteristics of various lesion detection algorithms completely and perfectly, and the number of lesion detection algorithms is large, which makes it difficult for the user to select, compare and decide an appropriate lesion detection algorithm among the lesion detection algorithms. Thus, the lesion detection efficiency is also affected.

Disclosure of Invention

Based on the problems, the application provides a method and a device for detecting the focus and a related product, so as to assist a user in selecting a focus detection algorithm, reduce the difficulty in selecting the algorithm and improve the efficiency of focus detection.

The embodiment of the application discloses the following technical scheme:

in a first aspect, the present application provides a lesion detection method, comprising:

determining a focus detection algorithm from a plurality of focus detection algorithms by using an analytic hierarchy process as a target algorithm;

detecting a lesion in the target medical image using the target algorithm;

the analytic hierarchy process relates to a hierarchical model comprising: a target layer, a criterion layer and a scheme layer; wherein the target layer is a target algorithm determined from the multiple lesion detection algorithms; the criterion layer includes: a plurality of performance indicators; the scheme layer comprises: the plurality of lesion detection algorithms.

Optionally, before the determining, by using an analytic hierarchy process, one lesion detection algorithm from a plurality of lesion detection algorithms as a target algorithm, the method further comprises:

obtaining the multiple focus detection algorithms from a focus detection algorithm library by indexing according to the type of the focus in the target medical image and the type of the target medical image; the focus detection algorithm library is constructed in advance, and the focus detection algorithm library takes focus types as main indexes and medical image types as sub-indexes.

Optionally, the determining, by using an analytic hierarchy process, one lesion detection algorithm from among multiple lesion detection algorithms as a target algorithm specifically includes:

obtaining a weight of the criterion layer relative to the target layer, and obtaining a weight of the scheme layer relative to the criterion layer;

obtaining a comprehensive efficacy index of each lesion detection algorithm in the plurality of lesion detection algorithms relative to the target layer according to the weight of the criterion layer relative to the target layer and the weight of the scheme layer relative to the criterion layer;

and taking a lesion detection algorithm with the highest comprehensive efficacy index relative to the target layer as the target algorithm.

Optionally, the obtaining the weight of the criterion layer relative to the target layer specifically includes:

constructing a judgment matrix of the plurality of performance indexes relative to the target layer according to the performance index importance index; the performance index importance index includes: an importance index of each performance indicator of the plurality of performance indicators relative to other performance indicators;

obtaining a feature vector according to the judgment matrix, and taking elements in the feature vector as the weight of the criterion layer relative to the target layer; the number of elements in the feature vector is the same as the number of the performance indicators and corresponds to the performance indicators one by one.

Optionally, the obtaining a feature vector according to the determination matrix, and using an element in the feature vector as a weight of a corresponding performance index in the criterion layer with respect to the target layer specifically includes:

normalizing the judgment matrix along each column to obtain a normalized judgment matrix;

summing each row of the normalized judgment matrix to obtain a characteristic vector;

normalizing the preliminary feature vector to obtain a normalized feature vector; and taking elements in the normalized feature vector as the weight of the corresponding performance index in the criterion layer relative to the target layer.

Optionally, after the obtaining the normalized feature vector, the method further includes:

obtaining the maximum characteristic root of the judgment matrix according to the judgment matrix, the quantity of the performance indexes in the criterion layer and the normalized characteristic vector;

obtaining a consistency index of the judgment matrix according to the maximum characteristic root and the number of the performance indexes in the criterion layer;

obtaining the consistency ratio of the judgment matrix according to the average random consistency index and the consistency index of the judgment matrix;

comparing the consistency ratio with a preset threshold value, and determining that the judgment matrix does not need to be adjusted when the consistency ratio is smaller than the preset threshold value; otherwise, determining that the judgment matrix needs to be adjusted.

Optionally, the obtaining the weight of the scheme layer relative to the criterion layer specifically includes:

obtaining a performance index for each of the plurality of lesion detection algorithms for each of the plurality of performance metrics;

the weight of the protocol layer relative to the criteria layer is derived from the efficacy index for each lesion detection algorithm.

Optionally, the obtaining the weight of the plan layer relative to the criterion layer according to the efficacy index of each lesion detection algorithm on each efficacy index specifically includes:

taking the ratio of the efficacy index of a lesion detection algorithm on one efficacy index to the sum of the efficacy indexes of the various lesion detection algorithms on the efficacy index as the normalization weight of the lesion detection algorithm on the efficacy index;

constructing a performance index vector corresponding to each performance index; the number of elements contained in the performance index vector is the same as the number of types of the lesion detection algorithms in the multiple lesion detection algorithms and corresponds to the number of types of the lesion detection algorithms one by one, and the elements contained in the performance index vector are the normalized weights of the corresponding lesion detection algorithms on the performance indexes;

constructing a performance index matrix by using performance index vectors corresponding to the performance indexes respectively, and taking elements in the performance index matrix as weights of the scheme layer relative to the criterion layer; the elements in the same row in the performance index matrix correspond to the same lesion detection algorithm, and the elements in the same column correspond to the same performance index.

Optionally, the obtaining a comprehensive efficacy index of each lesion detection algorithm in the plurality of lesion detection algorithms with respect to the target layer according to the weight of the criterion layer with respect to the target layer and the weight of the scheme layer with respect to the criterion layer specifically includes:

multiplying the performance index matrix and the normalized feature vector to obtain a comprehensive performance index vector; the number of elements in the comprehensive efficacy index vector is the same as the number of the types of the multiple lesion detection algorithms and corresponds to the types of the multiple lesion detection algorithms one by one; elements contained in the comprehensive efficacy index vector are comprehensive efficacy indexes of the corresponding lesion detection algorithm relative to the target layer;

the taking the lesion detection algorithm with the highest comprehensive efficacy index relative to the target layer as the target algorithm specifically comprises the following steps:

and sequencing the values of all elements in the comprehensive performance index vector, and determining a focus detection algorithm corresponding to the element with the largest value as the target algorithm.

Optionally, the plurality of performance indicators includes at least the following six performance indicators:

accuracy, coverage, output quality, number of parameters, detection speed and computational power requirements.

After the detecting a lesion in a target medical image using the target algorithm, the method further comprises:

outputting the focus detection result to a specified position in the form of an XML file; the lesion detection result includes at least one type of information:

the image area covered by the focus in the target medical image is a set of pixel coordinates, focus edge information, focus center position coordinates, or focus size information.

Optionally, after the detecting a lesion in a target medical image using the target algorithm, the method further comprises:

marking the detected focus in the target medical image in any one of the following ways:

marking the entire lesion, marking the edge of the lesion, marking the center of the lesion, or marking the lesion with a rectangular box.

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

the target algorithm determining module is used for determining a focus detection algorithm from a plurality of focus detection algorithms as a target algorithm by utilizing an analytic hierarchy process;

the focus detection module is used for detecting a focus in the target medical image by using the target algorithm;

the analytic hierarchy process involves the following layers: a target layer, a criterion layer and a scheme layer; wherein the target layer is a target algorithm determined from the multiple lesion detection algorithms; the criterion layer includes: a plurality of performance indicators; the scheme layer comprises: the plurality of lesion detection algorithms.

Optionally, the lesion detection device further comprises:

and the focus detection algorithm library construction module is used for constructing a focus detection algorithm library, and the focus detection algorithm library takes the focus type as a main index and takes the medical image type as a sub-index.

And the algorithm indexing module is used for indexing the multiple focus detection algorithms from a focus detection algorithm library according to the types of the focuses in the target medical images and the types of the target medical images.

Optionally, the target algorithm determining module specifically includes:

a first weight obtaining unit, configured to obtain a weight of the criterion layer relative to the target layer;

a second weight obtaining unit, configured to obtain a weight of the scheme layer relative to the criterion layer;

a comprehensive efficacy index obtaining unit, configured to obtain a comprehensive efficacy index of each lesion detection algorithm in the plurality of lesion detection algorithms with respect to the target layer according to the weight of the criterion layer with respect to the target layer and the weight of the solution layer with respect to the criterion layer;

and the target algorithm determining unit is used for taking a lesion detection algorithm with the highest comprehensive efficacy index relative to the target layer as the target algorithm.

Optionally, the first weight obtaining unit specifically includes:

a judgment matrix construction subunit, configured to construct a judgment matrix of the multiple performance indicators with respect to the target layer according to the performance indicator importance index; the performance index importance index includes: an importance index of each performance indicator of the plurality of performance indicators relative to other performance indicators;

the characteristic vector obtaining subunit is configured to obtain a characteristic vector according to the determination matrix, and use an element in the characteristic vector as a weight of the criterion layer relative to the target layer; the number of elements in the feature vector is the same as the number of the performance indicators and corresponds to the performance indicators one by one.

Optionally, the feature vector obtaining subunit specifically includes: a first subunit, a second subunit, and a third subunit;

the first subunit is used for normalizing the judgment matrix along each column to obtain a normalized judgment matrix; summing each row of the normalized judgment matrix to obtain a characteristic vector;

the second subunit is used for normalizing the preliminary feature vector to obtain a normalized feature vector;

a third subunit, configured to use elements in the normalized feature vector as weights of corresponding performance indicators in the criterion layer with respect to the target layer.

Optionally, the first weight obtaining unit of the lesion detection device further includes:

the consistency check subunit is used for obtaining the maximum characteristic root of the judgment matrix according to the judgment matrix, the quantity of the performance indexes in the criterion layer and the normalized characteristic vector; obtaining a consistency index of the judgment matrix according to the maximum characteristic root and the number of the performance indexes in the criterion layer; obtaining the consistency ratio of the judgment matrix according to the average random consistency index and the consistency index of the judgment matrix; comparing the consistency ratio with a preset threshold value, and determining that the judgment matrix does not need to be adjusted when the consistency ratio is smaller than the preset threshold value; otherwise, determining that the judgment matrix needs to be adjusted.

And determining whether the consistency degree of the judgment matrix meets the requirement or not by judging the consistency of the judgment matrix. Under the condition that the requirements are not met, the numerical values of elements in the characteristic vectors are more reliable and accurate by adjusting the judgment matrix, the error probability is reduced, and the selected target algorithm is more accurate.

Optionally, the second weight obtaining unit specifically includes:

a performance index obtaining subunit, configured to obtain a performance index of each of the plurality of lesion detection algorithms on each of the plurality of performance indices;

and the weight acquisition subunit is used for obtaining the weight of the scheme layer relative to the criterion layer according to the efficacy index of each lesion detection algorithm on each efficacy index.

Optionally, the weight obtaining subunit specifically includes: a fourth subunit, a fifth subunit and a sixth subunit;

the fourth subunit is configured to use a ratio of the performance index of one lesion detection algorithm on one performance index to a sum of the performance indexes of the plurality of lesion detection algorithms on the performance indexes, as a normalization weight of the lesion detection algorithm on the performance index;

a fifth subunit, configured to construct a performance index vector corresponding to each performance index; the number of elements contained in the performance index vector is the same as the number of types of the lesion detection algorithms in the multiple lesion detection algorithms and corresponds to the number of types of the lesion detection algorithms one by one, and the elements contained in the performance index vector are the normalized weights of the corresponding lesion detection algorithms on the performance indexes;

a sixth subunit, configured to construct a performance index matrix by using the performance index vectors corresponding to the performance indexes, and use elements in the performance index matrix as weights of the scheme layer relative to the criterion layer; the elements in the same row in the performance index matrix correspond to the same lesion detection algorithm, and the elements in the same column correspond to the same performance index.

Optionally, the comprehensive performance index obtaining unit is specifically configured to multiply the performance index matrix and the normalized feature vector to obtain a comprehensive performance index vector; the number of elements in the comprehensive efficacy index vector is the same as the number of the types of the multiple lesion detection algorithms and corresponds to the types of the multiple lesion detection algorithms one by one; elements contained in the comprehensive efficacy index vector are comprehensive efficacy indexes of the corresponding lesion detection algorithm relative to the target layer;

and the target algorithm determining unit is specifically used for sorting the numerical values of the elements in the comprehensive performance index vector and determining the focus detection algorithm corresponding to the element with the largest numerical value as the target algorithm.

Optionally, the apparatus further comprises a lesion detection result output module, configured to output a lesion detection result to a specified location in the form of an XML file; the lesion detection result includes at least one type of information:

Optionally, the apparatus further comprises: a lesion marking module for marking the detected lesion in the target medical image in any one of the following ways:

In a third aspect, the present application provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a lesion detection method as provided in the first aspect.

In a fourth aspect, the present application provides a processor for executing a computer program, which when executed performs the lesion detection method as provided in the first aspect.

Compared with the prior art, the method has the following beneficial effects:

in the focus detection method provided by the application, one focus detection algorithm is determined from multiple focus detection algorithms by using an analytic hierarchy process and is used as a target algorithm; and detecting the focus in the target medical image by using the target algorithm. According to the technical scheme, a target algorithm is determined from multiple focus detection algorithms to serve as a target layer, multiple efficacy indexes are used as a criterion layer, multiple focus detection algorithms are used as scheme layers, and a three-layer structure hierarchical model of the target layer, the criterion layer and the scheme layer is built in the mode. By applying the technical scheme of the application, an algorithm which meets the requirement of an efficiency index and is suitable for detecting the focus in the target medical image can be finally determined from a plurality of focus detection algorithms by applying an analytic hierarchy process on the basis of the hierarchical model. According to the technical scheme, the method and the device can assist the user in selecting the focus detection algorithm, reduce the difficulty of the user in selecting the algorithm and improve the focus detection efficiency.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.

Fig. 1 is a flowchart of a lesion detection method according to an embodiment of the present disclosure;

fig. 2 is a schematic diagram of a three-layer structure hierarchical model of a target layer-a criterion layer-a structure layer according to an embodiment of the present application;

fig. 3 is a schematic diagram of an index of a lesion detection algorithm library according to an embodiment of the present disclosure;

fig. 4 is a flowchart of another lesion detection method provided in the embodiments of the present application;

FIG. 5 is a schematic flow chart of obtaining weights of a criterion layer relative to a target layer and obtaining weights of a scheme layer relative to the criterion layer according to an embodiment of the present application;

fig. 6 is a schematic flowchart of a process for obtaining eigenvectors according to a determination matrix according to an embodiment of the present application;

FIG. 7 is a schematic diagram of a process for obtaining weights of a scenario layer relative to a criteria layer according to a performance index according to an embodiment of the present application;

fig. 8 is a schematic flowchart illustrating a process of determining consistency of a determination matrix and adjusting the determination matrix according to an embodiment of the present disclosure;

fig. 9 is a schematic structural diagram of a lesion detection apparatus according to an embodiment of the present application;

fig. 10 is a schematic structural diagram of another lesion detection apparatus according to an embodiment of the present application;

fig. 11 is a schematic structural diagram of a target algorithm determining module according to an embodiment of the present disclosure;

fig. 12 is a schematic structural diagram of another lesion detection apparatus according to an embodiment of the present application;

fig. 13 is a schematic structural diagram of another lesion detection apparatus according to an embodiment of the present application;

fig. 14 is a hardware configuration diagram of a lesion detection apparatus according to an embodiment of the present application.

Detailed Description

As described above, there are numerous existing lesion detection algorithms, in contrast, some algorithms may be advantageous in some respects and disadvantageous in others; other algorithms may be reversed. On one hand, the user can hardly clearly and accurately memorize the advantages and disadvantages of various focus detection algorithms; on the other hand, the selection of the lesion detection algorithm also puts high requirements on the professional knowledge literacy of the user. This presents a challenge to lesion detection: when a user needs to detect a focus in a medical image, the user can hardly select the focus from the focus, and meanwhile, the user has high dependence on professional knowledge. It will be appreciated that the difficulty in selecting an algorithm among many lesion detection algorithms also affects the efficiency of lesion detection.

Based on the above problems, the inventors have studied and provided a lesion detection method, device and related products. In The technical scheme of The application, firstly, an Analytic Hierarchy Process (AHP) is applied to a three-layer structure hierarchical model of a target layer, a criterion layer and a scheme layer, and a focus detection algorithm is determined from a plurality of focus detection algorithms to serve as a target algorithm. And then, detecting the focus in the target medical image by using the target algorithm. According to the technical scheme, the dependence degree on the professional knowledge of the user is reduced, the user does not need to memorize the advantages and disadvantages of various focus detection algorithms in a large quantity, the user is assisted to select the focus detection algorithm, the difficulty in selecting the algorithm is reduced, and meanwhile, the focus detection efficiency is correspondingly improved.

In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Method embodiment

Referring to fig. 1, a flowchart of a lesion detection method according to an embodiment of the present application is shown. As shown in fig. 1, the method for detecting a lesion provided in the embodiment of the present application includes:

step 101: and determining a focus detection algorithm from a plurality of focus detection algorithms by using an analytic hierarchy process as a target algorithm.

Fig. 2 is a schematic diagram of a three-layer structure hierarchical model of a target layer-a criterion layer-a structure layer according to an embodiment of the present application. In this step, a hierarchical analysis is performed based on the hierarchical model of the three-layer structure shown in fig. 2, and a lesion detection algorithm is finally determined from a plurality of lesion detection algorithms.

In an embodiment of the present application, the target layer is a layer that identifies a target algorithm from the plurality of lesion detection algorithms. The target layer may be used as the final destination for the hierarchical analysis. The criterion layer comprises a plurality of performance indexes, and the performance indexes are specifically used as the criterion for screening in the screening algorithm. The protocol layer includes a variety of lesion detection algorithms. Various focus detection algorithms in the scheme layer all belong to the algorithms to be selected.

The performance index refers to the performance of the focus detection algorithm in detecting the focus. For example: accuracy, coverage, output quality, number of parameters, speed of detection, and computational requirements, examples of which are different performance indicators. In practical applications, the number and specific types of performance indicators in the criterion layer are not limited, and may be added or deleted based on the six performance indicators provided in the above examples. The number and specific types of performance indicators of the criterion layer may be set by default or customized by the user.

It will be appreciated that certain lesion detection algorithms may be dedicated to detecting a particular type of lesion, or to detecting several particular types of lesions. Some lesion detection algorithms may be dedicated to detecting lesions of a specific medical image type, for example algorithm 1 is used to detect lesions in CT images and algorithm 2 is used to detect lesions in MR images. In one possible implementation, the multiple lesion detection algorithms included in the recipe layer may specifically be a lesion of a target type (a type of a lesion to be detected is definite) of a target medical image (a type of a medical image is definite) that a current user needs to detect.

An example is provided below:

the user wishes to detect the presence of lung nodules in the current MR image. The multiple lesion detection algorithms in the protocol layer are all algorithms suitable for detecting lung nodules in MR images.

In order to determine a target algorithm from a plurality of lesion detection algorithms in a solution layer, when the step 101 is implemented specifically, the following two aspects may be implemented: in a first aspect, the efficacy of each lesion detection algorithm in a solution layer on each efficacy index in a criteria layer is obtained; in a second aspect, the importance of each performance indicator in the criteria layer to the target layer is obtained. The first aspect can be assessed based on historical test data for a variety of lesion detection algorithms, e.g., algorithm 1 with higher accuracy and higher computational effort requirements, algorithm 2 with lower accuracy and faster detection speed. The second aspect may refer to the actual requirement of the user to detect the lesion, for example, the user has high requirements on the detection speed, accuracy and coverage rate of the lesion detection algorithm. After the results are obtained in the two aspects, the results in the two aspects are combined to select the optimal detection algorithm, namely the target algorithm, which meets the actual requirements of the user from a plurality of focus detection algorithms in a scheme layer.

In one possible implementation, a library of lesion detection algorithms is pre-established prior to execution of step 101. The focus detection algorithm library comprises detection algorithms which are respectively suitable for various focus types and various medical image types. As an example, in the lesion detection algorithm library, the lesion type is used as a main index, and the medical image type is used as a sub-index. In other examples, the medical image type may also be used as a main index and the lesion type may be used as a sub-index in the lesion detection algorithm library.

Fig. 3 is a schematic diagram of an index of a lesion detection algorithm library according to an embodiment of the present disclosure. As can be seen from fig. 3, a plurality of detection algorithms for a specific lesion type can be searched according to the main index; and then, according to the sub-indexes, a detection algorithm for detecting the specific focus type in the specific medical image type in a plurality of detection algorithms can be inquired. For example, by indexing through a library of lesion detection algorithms, a total of n kinds of lesion detection algorithms are known to be suitable for detecting lung nodules in MR images. In practical applications, the n-th focus detection algorithm may be listed in a scheme layer.

The lesion detection algorithm may be a locally stored algorithm or an algorithm that exists in a foreign location (e.g., a server or other terminal device). The focus detection algorithm library stores the calling interface of each focus detection algorithm. After the target algorithm is screened out in the step 101, the target algorithm can be called through a calling interface of the target algorithm to execute the subsequent step 102.

Step 102: and detecting a focus in the target medical image by using the target algorithm.

On the basis that the target algorithm is determined, relevant programs of the target algorithm are operated, and then focus detection can be carried out on the target medical image. Since the detection process of each detection algorithm may be different, the specific implementation process of this step is not limited herein.

The above is the focus detection method provided in the embodiments of the present application. In the method, firstly, an analytic hierarchy process is utilized to determine a focus detection algorithm from a plurality of focus detection algorithms as a target algorithm; and detecting the focus in the target medical image by using the target algorithm. According to the technical scheme, a target algorithm is determined from multiple focus detection algorithms to serve as a target layer, multiple efficacy indexes are used as a criterion layer, multiple focus detection algorithms are used as scheme layers, and a three-layer structure hierarchical model of the target layer, the criterion layer and the scheme layer is built in the mode. By applying the focus detection method provided by the embodiment of the application, an algorithm which meets the requirement of an efficiency index and is suitable for detecting the focus in a target medical image can be finally determined from a plurality of focus detection algorithms by applying an analytic hierarchy process on the basis of a hierarchical model. The focus detection method provided by the embodiment can assist a user in selecting the most suitable focus detection algorithm from a plurality of focus detection algorithms, so that the selection difficulty of the algorithms is reduced, and the focus detection efficiency is correspondingly improved.

The embodiment of the application also provides another focus detection method. A specific implementation of the method is described below in conjunction with fig. 4.

Referring to fig. 4, a flowchart of another lesion detection method according to an embodiment of the present application is shown. As shown in fig. 4, the method includes:

step 401: weights of the criterion layer relative to the target layer are obtained, and weights of the scheme layer relative to the criterion layer are obtained.

In the following description, an implementation process of obtaining the weight of the criterion layer relative to the target layer is described in conjunction with S4011 to S4012, and an implementation process of obtaining the weight of the scheme layer relative to the criterion layer is described in conjunction with S4013 to S4014. It should be noted that the serial numbers S4011 to S4014 are only used to distinguish different operation steps, and are not used as a limitation of the execution sequence. For example, S4011-S4012 can be executed first, and then S4013-S4014 can be executed; or S4013-S4014 can be executed first, and then S4011-S4012 can be executed; S4011-S4012 and S4013-S4014 can also be performed in parallel. See FIG. 5 for S4011-S4014.

S4011: constructing a judgment matrix of a plurality of performance indexes relative to the target layer according to the performance index importance index; the performance index importance index includes: an importance index of each performance indicator of the plurality of performance indicators relative to other performance indicators.

The performance index importance index may be specifically provided by a user. The user may specify the performance index importance index in the form of the following table.

TABLE 1 performance index importance index schematic table

In Table 1, a_ijThe performance index importance index represents the importance of the ith performance index relative to the jth performance index in the criterion layer. As an exampleThe performance index importance index covers 1-9 levels. a is_ijAnd a_jiIs an inverse relation, i.e. a_ji＝1/a_ijSo that the user only needs to specify a_ij(i<j) The importance index of (a). For the case of i ═ j, a_ij1. For example, the user gives a₁₂Given a value of 5, it can be understood that the accuracy is 5 times more important than the coverage, and a₂₁The automatic setting is 1/5, and the importance degree of the coverage rate is 1/5 of the importance degree of the accuracy rate.

With reference to table 1, a determination matrix of a plurality of performance indicators in the criterion layer with respect to the target layer can be obtained:

please refer to table 1 for determining the meaning of each element in the matrix a, which is not described herein.

S4012: obtaining a feature vector according to the judgment matrix, and taking elements in the feature vector as the weight of the criterion layer relative to the target layer; the number of elements in the feature vector is the same as the number of the plurality of performance indicators and corresponds to one another.

An exemplary implementation of this step is described below in conjunction with S40121-S40123, please refer specifically to fig. 6.

The performance index importance index set by the user may have a problem of large number difference, for example, the sum of the first row elements of the matrix a is determined to be 5, and the sum of the second row elements of the matrix a is determined to be 50. This easily causes the difference of the significance indexes of the performance indexes to be too large, thereby reducing the reference value of some elements in the judgment matrix. For this purpose, the decision matrix may be first normalized, see S40121.

S40121: and normalizing the judgment matrix along each column to obtain the normalized judgment matrix.

Normalizing the column vectors on the basis of the judgment matrix A to obtain a normalized judgment matrix as follows:

in the normalized decision matrix B, the sum of the elements of each column is 1. The element meanings of the normalized judgment matrix B and the normalized judgment matrix A at the same position are unchanged.

S40122: and summing each row of the normalized judgment matrix to obtain a feature vector.

The feature vectors are as follows:

wherein, W₁＝b₁₁+b₁₂+b₁₃+b₁₄+b₁₅+b₁₆W, the rest₂～W₆The calculation of (c) is analogized in the same way. In the feature vector W, each element corresponds to a feature index: w₁～W₆And respectively corresponding to the requirements of accuracy, coverage rate, output quality, parameter quantity, detection speed and computational power.

In the conversion process from the judgment matrix a to the judgment matrix B, only the column direction is normalized. Considering that each element in the feature vector W is the sum of the elements in the same row of the judgment matrix B, adding the result W₁～W₆There may also be problems with large differences in the quantities, e.g. W₁＝0.84，W₂＝0.57，W₃0.09. Thus, the application value of the elements in the feature vector is influenced. In order to avoid the above problem, in the embodiment of the present application, the feature vector W may be further normalized to ensure an application value of the performance index importance index set by the user, see S40123.

S40123: and normalizing the preliminary feature vector to obtain a normalized feature vector.

Normalizing the feature vector obtained in S40122 yields the following vector w:

in the normalized feature vector w, the sum of the elements is 1. The meaning of the elements of the normalized feature vector W and the feature vector W at the same position is unchanged. Element w in normalized feature vector w₁～w₆May be used as a weight of the corresponding performance metric in the criterion layer relative to the target layer. For example, w₁Representing the weight of accuracy with respect to the target layer, w₂Representing the weight of the coverage relative to the target layer. If w is₁＜w₂I.e. representing the target algorithm required for the user, the user pays more attention to the coverage when detecting the lesion than to the accuracy.

Through S4011-S4012, the implementation process of obtaining the weight of each performance index in the criterion layer relative to the target layer is obtained, and the implementation process of obtaining the weight of the scheme layer relative to the criterion layer is introduced in combination with S4013-S4014.

S4013: a performance index for each of a plurality of performance indicators for each of a plurality of lesion detection algorithms is obtained.

In practical applications, the performance index of each lesion detection algorithm on each performance index may be obtained by analyzing in advance according to historical test data of a plurality of lesion detection algorithms (e.g., lesion detection results, lesion detection time, etc. in the same medical image using the different lesion detection algorithms). For example, algorithm 1 requires performance indices on 6 different performance metrics for accuracy, coverage, output quality, number of parameters, speed of measurement and power when performing lesion detection.

In the embodiment of the present application, the performance index obtained according to the historical test data may be stored, so as to be directly retrieved when S4013 is executed. It should be noted that, when the historical test data is updated, the performance index of each lesion detection algorithm on each performance index can be updated accordingly through the analysis result.

S4014: the weight of the solution layer relative to the criterion layer is obtained according to the efficacy index of each lesion detection algorithm on each efficacy index.

An exemplary implementation of this step is described below in conjunction with S40141-S40143, please refer specifically to fig. 7.

S40141: the ratio of the performance index of a lesion detection algorithm on one performance index to the sum of the performance indexes of a plurality of lesion detection algorithms on the performance index respectively is used as the normalization weight of the lesion detection algorithm on the performance index.

The step is executed to obtain a normalized weight of a lesion detection algorithm on a performance index, and the specific implementation manner is to obtain a performance index of each of all lesion detection algorithms in a scheme layer on the performance index, wherein a ratio of the performance index of the lesion detection algorithm on the performance index to a sum of the performance indexes of all the lesion detection algorithms in the scheme layer on the performance index can be used as the normalized weight of the lesion detection algorithm on the performance index. The following is a specific calculation formula:

in this formula, S_ijNormalized weight of the ith lesion detection algorithm in the solution layer over the jth efficacy index in the criterion layer. P_ijCalculating the performance index of the algorithm i on the performance index for the performance index of the ith focus detection algorithm on the jth performance index in the criterion layer in the scheme layer, wherein i and k are values from 1 to n, n is the total number of focus detection algorithms in the scheme layer, and n is an integer greater than 1. j is a positive integer, and j is not more than the total number of the performance indicators in the criterion layer. It is understood that, on the same performance index, the higher the normalized weight of a certain lesion detection algorithm, the better the performance of the algorithm on the performance index.

By way of example, assume that the criteria layer includes accuracy, coverage, output quality, number of parameters, reckoning speed, and effort require 6 different performance metrics. S_i ^{Accurate and accurate}，S_i ^Covering，S_i ^{Output of}，S_i ^Parameter(s)，S_i ^{Speed of rotation}，S_i ^{Computing power}Sequentially corresponds to the above 6 performance indexesThe normalized weight of the i-th lesion algorithm on each performance index is expressed, which is equivalent to that j is 1, 2, 3, 4, 5, and 6 in the above formula.

S40142: constructing a performance index vector corresponding to each performance index; the number of elements contained in the performance index vector is the same as the number of types of lesion detection algorithms in various lesion detection algorithms and corresponds to each other one by one, and the elements contained in the performance index vector are normalization weights of the corresponding lesion detection algorithms on the performance indexes.

Continuing with the example above, S^{Accurate and accurate}，S^Covering，S^{Output of}，S^Parameter(s)，S^{Speed of rotation}，S^{Computing power}The performance index vectors corresponding to 6 performance indexes, which are required by accuracy, coverage, output quality, parameter quantity, measurement speed and calculation power, are respectively expressed as follows:

s40143: constructing a performance index matrix by using performance index vectors corresponding to the performance indexes respectively, and taking elements in the performance index matrix as weights of the scheme layer relative to the criterion layer; the elements in the same row in the performance index matrix correspond to the same lesion detection algorithm, and the elements in the same column correspond to the same performance index.

The performance index matrix S is represented as follows:

step 402: and obtaining the comprehensive efficacy index of each focus detection algorithm in the multiple focus detection algorithms relative to the target layer according to the weight of the criterion layer relative to the target layer and the weight of the scheme layer relative to the criterion layer.

See S4011-S4012 of fig. 5 for a way of obtaining weights of the criterion layer relative to the target layer, and see S4013-S4014 of fig. 5 for a way of obtaining weights of the scheme layer relative to the criterion layer. In conjunction with the above description, an exemplary implementation of this step 402 will be described below with elements in the feature vector W or the normalized feature vector W as weights of the criterion layer relative to the target layer and elements in the performance index matrix S as weights of the scheme layer relative to the criterion layer.

And multiplying the performance index matrix S by the normalized feature vector w to obtain a comprehensive performance index vector E, which is shown in the following calculation formula. Element E in the composite Performance index vector E₁，E₂，…，E_nThe number of the focus detection algorithms is the same as the number n of the types of the focus detection algorithms and corresponds to one. Element E contained in the composite Performance index vector E₁，E₂，…，E_nIs the comprehensive efficacy index of the corresponding lesion detection algorithm relative to the target layer. For example, the 1 st lesion detection algorithm has a combined efficacy index relative to the target layer of E₁The comprehensive efficacy index of the nth focus detection algorithm relative to the target layer is E_n。

Step 403: and taking a lesion detection algorithm with the highest comprehensive efficacy index relative to the target layer as the target algorithm.

In the embodiment of the application, the elements in the performance index matrix S represent the objective performance of various focus detection algorithms in the scheme layer on each performance index in the criterion layer; the elements in the normalized feature vector w represent the expectation of the user on each performance index of the finally selected target algorithm in the criterion layer, and the larger the numerical value of the elements, the higher the expectation of the user on the performance index of the selected target algorithm is. Based on the comprehensive performance index vector E obtained in step 402, the numerical values of the elements in the comprehensive performance index vector E may be sorted in this step, and the lesion detection algorithm corresponding to the element with the largest numerical value is determined to be used as the detection algorithmIs a target algorithm. For example, the largest element in the overall performance indicator vector E is E after sorting₃Then the 3 rd lesion detection algorithm in the solution layer is selected as the target algorithm.

Step 404: and detecting a focus in the target medical image by using the target algorithm.

The implementation manner of this step is basically the same as that of step 102 in the foregoing embodiment. The process of detecting lesions depends on the target algorithm itself. Therefore, the detailed implementation of step 404 is not described herein.

The above is the focus detection method provided in the embodiments of the present application. In the method, the weight of a criterion layer relative to a target layer is obtained based on customer requirements, and the weight of a scheme layer relative to the criterion layer is obtained based on objective and actual performance (historical test data) of each focus detection algorithm; and obtaining the comprehensive efficacy index of each focus detection algorithm in the multiple focus detection algorithms relative to the target layer according to the weight of the criterion layer relative to the target layer and the weight of the scheme layer relative to the criterion layer. And finally, sequencing the comprehensive performance indexes of the focus detection algorithms, determining a target algorithm, and realizing focus detection by the target algorithm. When the target algorithm determined in the above way is used for detecting the focus of a target medical image, the requirements of a user on various performance indexes are better met, and the advantages of the target algorithm are exerted. The technical scheme can effectively assist the user to select the focus detection algorithm which best meets the requirements of the user among various focus detection algorithms, reduce the difficulty of algorithm selection and further improve the focus detection efficiency.

In an example implementation manner, after the step 102 or the step 404 of the foregoing embodiment is executed, the lesion detection result may be further output. For example, see step 405 of FIG. 4: outputting the lesion detection result to a designated position in the form of an XML (Extensible Markup Language) file; the lesion detection result includes at least one type of information: a set of pixel coordinates in an image region covered by the lesion in the target medical image, information of an edge of the lesion, coordinates of a center position of the lesion, or information of a size of the lesion.

The specified location may be a folder that specifies a path.

In an example implementation, after step 102 or step 404 of the foregoing embodiment is executed, the lesion in the target medical image may be further marked to assist the user in visually viewing the lesion in the medical image. For example, see step 406 of FIG. 4: marking the detected focus in the target medical image in any one of the following ways: marking the entire lesion, marking the edge of the lesion, marking the center of the lesion, or marking the lesion with a rectangular box.

It should be noted that, in the embodiment of the present application, the marking manner of the lesion may depend on the actual requirements of the user and the type of information included in the lesion output result. For example, if the lesion detection result includes the coordinates of the center of the lesion, the lesion may be marked in a manner that marks the center of the lesion.

In the lesion detection method provided in the foregoing embodiment, it is mentioned that the feature vector W may be normalized to obtain a normalized feature vector W. Because the feature vector W is obtained according to the judgment matrix A, each element in the judgment matrix A is provided by a user, the possibility of poor consistency of the judgment matrix A exists, and the selection of a target algorithm is easily influenced. Based on this problem, after obtaining the normalized feature vector w, i.e. after S40123 is executed, the method for detecting a lesion may further include the following steps S40124-S40127, and refer to fig. 8 in particular:

s40124: and obtaining the maximum characteristic root of the judgment matrix according to the judgment matrix, the quantity of the performance indexes in the criterion layer and the normalized characteristic vector.

The maximum characteristic root lambda calculation formula of the judgment matrix A is as follows:

in the above formula, a represents a judgment matrix, q represents the total number of performance indexes in the criterion layer, q > 1, and j is any integer between 1 and q. w is the normalized eigenvector, wj represents the jth element in the normalized eigenvector w, and (Aw) j is the jth element in the vector obtained by multiplying the matrix A and the vector w. The number of columns of the matrix a and the total number of elements of the vector w are both q, and the columns of the matrix a correspond to the performance indexes one to one. The assumption criterion layer includes accuracy, coverage, output quality, parameter number, measurement speed and calculation force, which require 6 different performance indexes, i.e. q is 6.

S40125: and obtaining the consistency index of the judgment matrix according to the maximum characteristic root and the quantity of the performance indexes in the criterion layer.

The consistency index CI is calculated as follows:

in the above formula, λ represents the maximum characteristic root of the judgment matrix a, q represents the total number of performance indexes in the criterion layer, and q > 1.

S40126: and obtaining the consistency ratio of the judgment matrix according to the average random consistency index and the consistency index of the judgment matrix.

The calculation formula of the consistency ratio CR of the judgment matrix a is as follows:

in the above formula, CI represents a consistency index of the determination matrix a, and RI represents an average random consistency index. RI is typically an empirically selected standard value, and in the AHP method RI is related to the order of the decision matrix a. Since the order of the determination matrix a is consistent with the total number of the performance indexes, the order of the determination matrix a is also q.

The following table provides the average random consistency index RI for several matrices with different values of the a order q.

TABLE 2 average random consistency index RI for different orders of decision matrix

Order q	2	3	4	5	6	7	8	9	10
										RI	0	0.58	0.90	1.12	1.24	1.32	1.41	1.45	1.49

As shown in table 2 above, when q is 6, RI is 1.24.

S40127: comparing the consistency ratio with a preset threshold value, judging whether the consistency ratio is smaller than the preset threshold value, if so, entering S4013; if not, S0 is entered.

The consistency ratio of the judgment matrix a is obtained by S40126. Here, the consistency ratio is compared with a preset threshold, that is, whether the consistency of the judgment matrix a meets the requirement of the preset threshold is determined. And when the consistency ratio is smaller than the preset threshold value, the consistency of the judgment matrix A meets the requirement, and the judgment matrix A does not need to be adjusted. And when the consistency ratio is larger than or equal to the preset threshold value, the consistency of the judgment matrix A is poor, the requirement is not met, and the judgment matrix A needs to be readjusted.

If the requirement for judging the consistency of the matrix is high, a smaller preset threshold value can be set. As an example, the preset threshold is 0.1.

S0: and determining that the judgment matrix needs to be adjusted, and automatically adjusting the judgment matrix or displaying a prompt message that a user needs to adjust the judgment matrix.

After S0 is executed, in the case of automatically adjusting the determination matrix, the process may directly proceed to S40121 again with the adjusted determination matrix; for the case that the user autonomously adjusts the decision matrix (i.e. re-provides the performance index importance index, see table 1), after S0 is executed, S4011 is entered to construct a new decision matrix according to the performance index importance index re-provided by the user.

Through the steps, the consistency of the judgment matrix A can be judged to determine whether the consistency degree meets the requirement or not. Under the condition that the requirement is not met, the numerical values of elements in the characteristic vector W are more reliable and accurate by adjusting the judgment matrix, the error probability is reduced, and the selected target algorithm is more accurate.

Based on the lesion detection method provided by the foregoing embodiment, correspondingly, the present application further provides a lesion detection apparatus. Specific implementations of the apparatus are described below in conjunction with the embodiments and the figures.

Device embodiment

Referring to fig. 9, a schematic structural diagram of a lesion detection apparatus according to an embodiment of the present disclosure is shown. As shown in fig. 9, the lesion detection apparatus 90 includes:

a target algorithm determining module 901, configured to determine a lesion detection algorithm from multiple lesion detection algorithms by using an analytic hierarchy process as a target algorithm;

a lesion detection module 902 for detecting a lesion in the target medical image using the target algorithm;

The above embodiments are directed to a lesion detection apparatus provided in the embodiments of the present application. According to the technical scheme, a target algorithm is determined from multiple focus detection algorithms to serve as a target layer, multiple efficacy indexes are used as a criterion layer, multiple focus detection algorithms are used as scheme layers, and a three-layer structure hierarchical model of the target layer, the criterion layer and the scheme layer is built in the mode. By applying the technical scheme of the application, an algorithm which meets the requirement of an efficiency index and is suitable for detecting the focus in the target medical image can be finally determined from a plurality of focus detection algorithms by applying an analytic hierarchy process on the basis of the hierarchical model. According to the technical scheme, the method and the device can assist the user in selecting the focus detection algorithm, reduce the difficulty of the user in selecting the algorithm and improve the focus detection efficiency.

Referring to fig. 10, the structure of another lesion detection apparatus is illustrated. As shown in fig. 10, the lesion detection apparatus 100 further includes:

a lesion detection algorithm library construction module 1001 configured to construct a lesion detection algorithm library, where the lesion detection algorithm library uses a lesion type as a main index and a medical image type as a sub-index.

An algorithm indexing module 1002, configured to index the multiple lesion detection algorithms from a lesion detection algorithm library according to the type of a lesion in the target medical image and the type of the target medical image.

Referring optionally to FIG. 11, a structure of a target algorithm determination module 901 is illustrated. As shown in fig. 11, the target algorithm determining module 901 specifically includes:

a first weight obtaining unit 9011, configured to obtain a weight of the criterion layer relative to the target layer;

a second weight obtaining unit 9012, configured to obtain a weight of the scheme layer relative to the criterion layer;

a comprehensive performance index obtaining unit 9013, configured to obtain, according to the weight of the criterion layer relative to the target layer and the weight of the scheme layer relative to the criterion layer, a comprehensive performance index of each of the multiple lesion detection algorithms relative to the target layer;

a target algorithm determination unit 9014, configured to use, as the target algorithm, a lesion detection algorithm with a highest comprehensive efficacy index with respect to the target layer.

When the target algorithm determined in the above way is used for detecting the focus of a target medical image, the requirements of a user on various performance indexes are better met, and the advantages of the target algorithm are exerted. The technical scheme can effectively assist the user to select the focus detection algorithm which best meets the requirements of the user among various focus detection algorithms, reduce the difficulty of algorithm selection and further improve the focus detection efficiency.

Optionally, the first weight obtaining unit 9011 specifically includes:

Optionally, the first weight obtaining unit 9011 of the lesion detection apparatus further includes:

Optionally, the second weight obtaining unit 9012 specifically includes:

Optionally, the comprehensive performance index obtaining unit 9013 is specifically configured to multiply the performance index matrix and the normalized feature vector to obtain a comprehensive performance index vector; the number of elements in the comprehensive efficacy index vector is the same as the number of the types of the multiple lesion detection algorithms and corresponds to the types of the multiple lesion detection algorithms one by one; elements contained in the comprehensive efficacy index vector are comprehensive efficacy indexes of the corresponding lesion detection algorithm relative to the target layer;

a target algorithm determining unit 9014, configured to perform numerical sorting on each element in the comprehensive performance index vector, and determine a lesion detection algorithm corresponding to an element with a largest numerical value as the target algorithm.

Referring to fig. 12, the structure of yet another lesion detection apparatus is illustrated. As shown in fig. 12, the apparatus 120 includes: the target algorithm determining module 901, the focus detecting module 902, further include a focus detecting result outputting module 1201, for outputting the focus detecting result to the designated position in the form of XML file; the lesion detection result includes at least one type of information:

The lesion detection apparatus 120 shown in fig. 12 may facilitate a user to extract a lesion detection result by outputting the lesion detection result, thereby implementing further analysis.

Referring to fig. 13, the structure of yet another lesion detection apparatus is illustrated. As shown in fig. 13, the apparatus 130 includes: the target algorithm determining module 901 and the lesion detecting module 902 further include: a lesion marking module 1301, configured to mark the detected lesion in the target medical image in any one of the following ways:

The lesion detection apparatus 130 shown in fig. 13 may assist a user in visually viewing a lesion in a medical image by marking the lesion.

Based on the lesion detection method and device provided by the foregoing embodiments, the embodiments of the present application further provide a computer-readable storage medium. The storage medium has a program stored thereon, and the program, when executed by a processor, implements some or all of the steps of the lesion detection method as claimed in the aforementioned method embodiments of the present application.

The storage medium may be a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and various media capable of storing program codes.

Based on the lesion detection method, the lesion detection device and the storage medium provided by the foregoing embodiments, the present application provides a processor. The processor is configured to execute a program, where the program executes to perform some or all of the steps of the lesion detection method provided in the foregoing method embodiments.

Based on the storage medium and the processor provided by the foregoing embodiments, the present application further provides a lesion detection apparatus. Referring to fig. 14, a hardware configuration diagram of a lesion detection apparatus according to the present embodiment is shown.

As shown in fig. 14, the lesion detection apparatus includes: a memory 1401, a processor 1402, a communication bus 1403, and a communication interface 1404.

The memory 1401 stores a program that can be executed on the processor, and when the program is executed, part or all of the steps in the lesion detection method provided in the foregoing method embodiments of the present application are implemented. The memory 1401 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 volatile solid-state storage device.

In the lesion detection device, the processor 1402 and the memory 1401 communicate signaling, logic instructions, etc. via a communication bus 1403. The device is capable of communicative interaction with other devices via a communication interface 1404.

In the embodiment of the present application, the lesion detection device may be implemented by a device (e.g., a CT machine or an MR device) that locally generates a medical image, or may be implemented by other devices, such as a terminal (e.g., a laptop, a desktop, a mobile phone, etc.) that communicates with the device that generates a medical image, or may be implemented by a physical server. In addition, the focus detection method and device provided by the embodiment of the application can also be realized on a cloud server.

It should be noted that, in the present specification, all the embodiments are described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, they are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described embodiments of the apparatus and system are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts suggested as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

The above description is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method of lesion detection, comprising:

detecting a lesion in the target medical image using the target algorithm;

2. The lesion detection method of claim 1, wherein before the determining, using an analytic hierarchy process, one lesion detection algorithm from among a plurality of lesion detection algorithms as a target algorithm, the method further comprises:

3. The method according to claim 1, wherein the determining a lesion detection algorithm from a plurality of lesion detection algorithms as a target algorithm by using an analytic hierarchy process comprises:

4. The lesion detection method according to claim 3, wherein the obtaining the weight of the criterion layer relative to the target layer specifically comprises:

5. The lesion detection method according to claim 4, wherein the obtaining of the feature vector according to the determination matrix and the weighting of the corresponding performance index in the criterion layer with respect to the target layer using the elements in the feature vector comprise:

6. The lesion detection method of claim 5, wherein after the deriving the normalized feature vector, the method further comprises:

7. The lesion detection method according to claim 5 or 6, wherein the obtaining the weight of the plan layer relative to the criterion layer specifically comprises:

8. A lesion detection apparatus, comprising:

9. A computer-readable storage medium, in which a computer program is stored which, when executed by a processor, implements a lesion detection method according to any one of claims 1 to 7.

10. A processor configured to run a computer program, the program when executed performing the lesion detection method of any one of claims 1 to 7.