CN111403027A - Rare disease picture searching method based on rare mining - Google Patents

Rare disease picture searching method based on rare mining Download PDF

Info

Publication number
CN111403027A
CN111403027A CN202010185084.6A CN202010185084A CN111403027A CN 111403027 A CN111403027 A CN 111403027A CN 202010185084 A CN202010185084 A CN 202010185084A CN 111403027 A CN111403027 A CN 111403027A
Authority
CN
China
Prior art keywords
rare
picture
pictures
user
class
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010185084.6A
Other languages
Chinese (zh)
Other versions
CN111403027B (en
Inventor
刘振广
杨家旭
钱鹏
杨文武
纪首领
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang Gongshang University
Original Assignee
Zhejiang Gongshang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang Gongshang University filed Critical Zhejiang Gongshang University
Priority to CN202010185084.6A priority Critical patent/CN111403027B/en
Publication of CN111403027A publication Critical patent/CN111403027A/en
Application granted granted Critical
Publication of CN111403027B publication Critical patent/CN111403027B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Biomedical Technology (AREA)
  • Library & Information Science (AREA)
  • Theoretical Computer Science (AREA)
  • Pathology (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The invention discloses a rare disease picture searching method based on rare type mining, which is completed by two steps of rare type detection and rare type development, realizes accurate and efficient rare disease picture searching, and specifically comprises the following steps: preprocessing a large data set off line; online query of rare class detection; interactively updating the interest condition of the data by the user; and developing rare classes by using positive and negative samples. In order to solve the speed problem of rare class detection, the time complexity of single query is reduced to a degree sufficient for real-time feedback through the global setting and offline processing of variables, meanwhile, the rare class development result is ensured to be in line with the reality by utilizing human-computer interaction, the waste of human resources due to the obtainment of a large number of irrelevant pictures is avoided, the defect of the current method based on the problem is filled, and the method not only can help people to better obtain economic benefits from medical pictures, but also has reference value in relevant research fields.

Description

Rare disease picture searching method based on rare mining
Technical Field
The invention belongs to the technical field of data mining, and particularly relates to a rare disease picture searching method based on rare mining.
Background
In recent years, with the progress of medical means, more and more instrument detection results can be presented and stored in the form of pictures, and various medical pictures cover real affected parts photos from microscopic bacterial virus structures to macroscopic testees, so that the medical instrument has the characteristics of huge data volume and high potential utilization value. Generally, the medical pictures are analyzed by using a data mining technology, and the classification corresponding to the examination results of some common diseases and healthy people can be obtained from the view point of the pictures, so that the medical pictures are used for further teaching and scientific research and promote the progress of the medical technology.
The pictures corresponding to different classifications with the highest research value are the pictures corresponding to the focuses of the rare diseases; taking polycystic kidney disease as an example, the prevalence rate of the disease is about 1-2%, the disease condition is not obvious, the examination is often easily confused with simple renal cyst, and the treatment is delayed, but one obvious characteristic of the disease is that the size of the kidney is increased, medical pictures obtained by ultrasound or CT are different from common renal cyst to a certain extent and belong to different types of pictures, and the pictures belong to rare types relative to detection pictures corresponding to healthy people and common diseases. The rare disease pictures in the rare category have certain similarity as the expression of the same disease, but the number of the rare disease pictures is very rare due to the characteristics of low disease incidence and few picture-form inspection results; therefore, in the conventional data analysis and data mining technology, due to the influence of main classification, pictures corresponding to rare diseases are difficult to distinguish.
Rare mining technology is needed for detecting the rare disease pictures belonging to rare classes, the rare mining can detect the pictures of the rare diseases belonging to rare classes from a large number of medical picture data sets, and other pictures in a large data set corresponding to the same disease can be found out through analysis. Rare disease pictures belonging to rare classes detected from a large number of medical picture data sets belong to rare class detection, however, a detection process of an existing rare class detection method is usually square time complexity, and if rare disease pictures concerned by researchers do not exist in a query result, a long time is needed for adjusting an input parameter once to obtain a result again, and real-time feedback cannot be obtained. In practical situations, if one is eager to analyze and study some serious rare diseases, a long query time may delay the medical study and delay the treatment of the patient.
On the basis of rare class detection, all picture information corresponding to rare diseases concerned by researchers possibly existing in a large data set is found out to belong to rare class development. In general, rare class development work is to gradually search all other similar pictures by taking a few rare disease pictures detected by rare classes as a core through the characteristic that data in the same class have higher similarity. In the existing method, only the characteristic of high similarity of data under the same rare class is considered in the process, the actual situation is not considered, the influence of small number of core pictures is caused, the obtained result possibly comprises a large number of pictures which are similar in shape but not corresponding to rare diseases, and manpower and material resources are wasted.
In summary, there is a need for a method for searching rare disease pictures by using rare mining, which can realize human-computer interaction, and the following problems exist in the conventional rare mining method: (1) only the statistical characteristics of the rare data are considered, and the practical significance of the data is ignored; (2) the computation of the pictures corresponding to the rare classes detected by the user at a time is complex, and the requirement of real-time interaction cannot be met. Therefore, the design and implementation of an interactive rare type mining rare disease detection method considering practical research significance inevitably brings huge economic value and practical value.
Disclosure of Invention
In order to solve the problems that the query speed is low and the actual situation cannot be considered in the rare type mining process of rare disease pictures, the invention provides a rare disease picture searching method based on rare type mining, which optimizes the query speed in the traditional rare type detection (RCD) process and reduces the query time complexity of rare type detection by combining offline preprocessing and online query; on the basis of a small number of detected rare disease pictures, the invention combines the actual situation through human-computer interaction, provides a new rare class development (RCE) method, interactively updates the searched rare disease picture range, ensures that the human resources are maximally utilized, and finds all rare disease pictures under a large data set.
A rare disease picture searching method based on rare type mining comprises two parts, namely rare type detection and rare type development; wherein the rare class detection process is as follows:
A1. acquiring search characteristic parameters input by a user;
A2. calculating the rare index of each picture in the picture library;
A3. searching pictures of the rare class index in the upper and lower limit intervals and feeding back the pictures to the user, stopping detection if at least one rare disease picture which the user is interested in exists in a feedback result, otherwise, finely adjusting the search characteristic parameters by the user and then returning to execute the step A1;
the process of the rare class development is as follows:
B1. forming a positive sample set by the rare disease pictures interested by the user from the pictures fed back to the user by the rare detection, and forming a negative sample set by the other pictures;
B2. determining adjacent sample sets of each picture in the positive sample set and taking the union of the adjacent sample sets as phi;
B3. for any picture in the set phi, calculating the positive sample distance r between the picture and the positive sample set+And the negative sample distance r between the negative sample and the negative sample set-；
B4. Extracting the picture with the maximum r value from the set phi and providing the picture to a user, wherein r is r--r+If the user is interested, adding the picture into the positive sample set, and if the user is not interested, adding the picture into the negative sample set;
B5. circularly executing the steps B2-B4 until phi- ∩+-∩-For space-time termination, the positive sample set at this time is taken as the final output result, wherein ∩+As the intersection of the set of positive samples with the set Φ, ∩-Is the intersection of the set of negative samples and the set phi.
Further, the search feature parameters in the step a1 are triples<k,sup,slow>Wherein s isupAnd slowRespectively setting upper and lower limit values of rare class indexes of the rare disease pictures for users, k being a given natural number used for representing the size of the adjacent sample set of the pictures and k ∈ [ k ]min,kmax]，kminAnd kmaxRespectively, given interval upper and lower limit values.
Further, in the step a2, the rare class index of each picture is calculated by the following formula;
Figure BDA0002413898880000031
wherein:
Figure BDA0002413898880000032
the method comprises the steps that a rare class index of an ith picture in a picture library is given, i is a natural number larger than 0, a set of adjacent samples of the ith picture is determined through a KNN (K-Nearest Neighbor) algorithm, the number of the pictures in the set is K, d1～dkAnd corresponding to Euclidean distances between k pictures in the set and the ith picture in the picture library, wherein the Euclidean distances are in the order from small to large, and avg { } is an averaging function.
Further, in the step a2, the interval [ k ] is pairedmin,kmax]Discretization is carried out, and the rare index of all pictures in the picture library under each K value in the traversal interval is calculated and stored, so that the rare index of the pictures can be directly called without calculation when the search characteristic parameters are subsequently adjusted, and the time consumption and waste of repeatedly calculating the K neighbor relation can be avoided.
Furthermore, a hierarchical statistical method is adopted for the rare class indexes of the pictures without the k value, and interval division is carried out according to the size of the rare class indexes through statistical processing, so that the single online rare class detection process of the user can be completed more quickly.
Further, in the step B3, the positive sample distance r is calculated by the following formula+；
Figure BDA0002413898880000041
Wherein: x is the characteristic vector of any picture in the set phi, P is the characteristic matrix formed by the characteristic vectors of all pictures in the positive sample set, the characteristic vector is formed by all pixel values of the pictures, and omega1Is a weight vector consisting of n weight values, and the n weight values cumulatively equal to 1, n being the number of pictures in the positive sample set,
Figure BDA0002413898880000042
is a weight coefficient, ω1And
Figure BDA0002413898880000043
by aligning the sample distance r+Minimization determination, | | | purple2Is a two-norm.
Further, in the step B3, the negative sample distance r is calculated by the following formula-；
Figure BDA0002413898880000044
Wherein: x is the characteristic vector of any picture in the set phi, G is the characteristic matrix formed by the characteristic vectors of all pictures in the negative sample set, the characteristic vector is formed by all pixel values of the pictures, and omega is2Is a weight vector consisting of m weight values, and the m weight values are cumulatively equal to 1, m is the number of pictures in the negative sample set,
Figure BDA0002413898880000045
is a weight coefficient, ω2And
Figure BDA0002413898880000046
by a distance r to the negative sample-Minimization determination, | | | purple2Is a two-norm.
The invention provides a rare disease picture searching method based on rare type mining by combining rare type detection RCD and rare type development RCE, which greatly improves the detection speed of rare disease pictures, has stronger pertinence than the existing rare type mining method based on the traditional method, fills the vacancy of the rare disease picture searching method in real-time human-computer interaction, has high practical value and strong reference significance, and has the following specific beneficial technical effects and innovativeness:
1. the invention utilizes the global setting of the parameters to carry out off-line processing on the pictures in the big data set, thereby reducing the time complexity of on-line query from a square level to a logarithmic level.
2. According to the invention, the positive and negative samples are established, and after each possible rare disease picture is developed, the user is interactively inquired whether the positive and negative samples are accurate and updated, so that the result is ensured to be in line with the reality.
3. The invention explains the realization of man-machine interaction in rare mining work, so that the rare mining work under a large data set can be better applied in wider fields.
Drawings
FIG. 1 is a diagram showing the arrangement of rare class indices in the k-interval range in the rare class detection of the present invention.
FIG. 2 is a schematic diagram of positive and negative sample sets in rare class development for user interest.
Detailed Description
In order to describe the present invention more specifically, the following detailed description of the embodiments of the present invention is provided with reference to the accompanying drawings.
The rare disease picture searching method based on rare type mining is realized by sequentially detecting RCD through rare type and developing RCE through rare type, and the problems that the query speed is low and the user interest cannot be met are solved, so that man-machine interaction is realized.
The rare detection RCD preprocesses a big data set picture to be processed through an offline module, so that the time for a user to adjust parameters online to obtain a required rare disease picture is shortened, and the method specifically comprises the following steps:
(1) and (4) performing off-line preprocessing on all medical picture data in the large data set, and calculating and storing the rare index of each picture.
The rare index s is an index for determining whether data belongs to a rare class, specifically, rare class indexes corresponding to rare disease pictures are distributed in a specific interval, the rare disease pictures in the rare class can be effectively detected through a rare class index threshold, and the rare class index has the following calculation formula:
Figure BDA0002413898880000051
wherein: i denotes a picture x numbered i in the big data setiAnd d represents a picture x after the picture is expanded into a vector by pixeliThe Euclidean distance of the picture corresponding to the index of d in the feature space of the vector, k represents the picture xiExpanding the K neighbor relation in the feature space after the vector, wherein the subscript of d represents the picture xiThe K neighbor relation of the image x is in accordance with the image xiThe pictures corresponding to the Euclidean distance are sorted from small to large, 1 represents the picture closest to the Euclidean distance, 2 represents the picture closest to the second, and so on, and avg represents the average value of the distance d in the brackets.
The off-line module preprocessing is to pre-calculate and store the rare index of all pictures in the big data set, that is, to globally set the parameters, the process is based on the characteristic that after a parameter k in the rare index is given, the rare index corresponding to all picture data is limited, and in combination with the characteristic that the parameter k is an integer and the possible values thereof are also the countable number, as shown in fig. 1, all possible parameters k (the upper limit is k) of all picture data in the big data set are calculatedmaxLower limit of kmin) And taking the corresponding rare index s.
The K neighbor relation and the grading curve are the keys for effectively performing global setting and statistical planning on parameters; the K nearest neighbor relation is that Euclidean distances in a feature space after each picture in the big data set is expanded into vectors are calculated, the distances are sorted, and x is selected for each selected pictureiThe k pictures k before the Euclidean distance are xiThe K neighbor relation pictures are calculated to obtain the front K corresponding to all the picturesmaxAfter each image is close to the image, when the rare index of each image under different conditions is calculated in the overall setting of the parameters, the K neighbor relation images under all possible parameter K values of each image can be directly found, and the waste of repeatedly calculating the K neighbor relation is avoided.
The grading curve is a grading statistical method for the rare class indexes s of all picture data under different k values in the global setting of parameters, and the single online rare class detection process of a user can be completed more quickly by performing interval division according to the size of the rare class indexes through statistical processing.
(2) And accepting the characteristic parameters input by the user online.
The feature parameters input online by user rare detection refer to feature triples<k，sl Yao w，sup>Wherein k is a parameter for calculating the rare class index, and corresponds to a predicted value of the number of rare disease pictures possibly existing in a large data set by a user in practical application, sl Yao wAnd supThe image is used for screening the image of the rare class index in the large data set within the threshold range under the k value input by the user, and the image is regarded as the image belonging to the rare class and used as the detection result.
(3) And searching all the pictures meeting the requirements in the corresponding range of the user input parameters on line in the preprocessing result.
The on-line search of rare class detection refers to positioning a corresponding rare class index sequence in off-line processing through k in parameters input by a user and positioning s by utilizing binary searchl Yao wAnd supIn the interval, the rare class index is satisfied
Figure BDA0002413898880000061
The picture corresponding to the index of (1) is output as a picture belonging to the rare class.
(4) And (3) feeding the result back to the user, judging whether the picture corresponding to the concerned rare disease exists or not by the user, if not, adjusting the input parameters and repeating the steps (2), (3) and (4), and if so, ending the rare class detection.
The rare class development RCE interactively completes the extension of rare classes by taking a picture corresponding to a rare disease as a center in a rare class detection result, and specifically comprises the following steps:
(1) and (3) taking the rare disease picture obtained from the rare detection RCD as a positive sample P and the non-rare disease picture as a negative sample G, and searching a picture set adjacent to the positive sample.
The picture set adjacent to the positive sample means that k is preset for each positive sample picture in the positive sample0And calculating the K neighbor relation to obtain a union set of all K neighbor relation picture sets.
(2) As shown in FIG. 2, each picture x in the picture set that does not belong to the positive and negative samples is calculatediPositive sample distance to positive sample
Figure BDA0002413898880000071
And negative example distance to negative example
Figure BDA0002413898880000072
Positive sample distance
Figure BDA0002413898880000073
There is the following calculation formula, namely L2 standard regularization distance formula:
Figure BDA0002413898880000074
wherein: p represents all pictures in the positive sample, that is, each picture is expanded into a vector and then is juxtaposed to form a matrix. x is the number ofiRepresenting the ith picture of the non-positive and negative sample pictures in the picture set adjacent to the positive sample, wherein the numerical value of the ith picture is a vector of the picture after the picture is expanded in the feature space; omega1And
Figure BDA0002413898880000075
is a weight, ω1Is a vector
Figure BDA0002413898880000076
Is a scalar and subject to constraints
Figure BDA0002413898880000077
Limit, ω1And
Figure BDA0002413898880000078
is such that the calculated positive sample distance
Figure BDA0002413898880000079
And taking the minimum corresponding value. Similarly, the negative sample distance
Figure BDA00024138988800000710
There is the formula:
Figure BDA00024138988800000711
wherein: g denotes all pictures in the negative example, xiSame as the picture in the positive sample distance calculation, ω2And
Figure BDA00024138988800000712
to correspond to
Figure BDA00024138988800000713
Under the condition of making negative sample distance
Figure BDA00024138988800000714
And taking the minimum corresponding weight.
(3) Taking distances from the set
Figure BDA00024138988800000715
And (4) the largest picture is given to the user for judgment, if the user is interested, the picture is taken as a positive sample, and if the user is not interested, the picture is taken as a negative sample.
(4) And (3) updating the pictures in the positive sample and negative sample sets, searching a picture set adjacent to the new positive sample, and repeating the steps (2), (3) and (4) until no pictures with non-positive and negative samples exist in the set.
The embodiments described above are presented to enable a person having ordinary skill in the art to make and use the invention. It will be readily apparent to those skilled in the art that various modifications to the above-described embodiments may be made, and the generic principles defined herein may be applied to other embodiments without the use of inventive faculty. Therefore, the present invention is not limited to the above embodiments, and those skilled in the art should make improvements and modifications to the present invention based on the disclosure of the present invention within the protection scope of the present invention.

Claims (8)

1. A rare disease picture searching method based on rare type mining comprises two parts, namely rare type detection and rare type development; the rare class detection process is characterized by comprising the following steps:
A1. acquiring search characteristic parameters input by a user;
A2. calculating the rare index of each picture in the picture library;
A3. searching pictures of the rare class index in the upper and lower limit intervals and feeding back the pictures to the user, stopping detection if at least one rare disease picture which the user is interested in exists in a feedback result, otherwise, finely adjusting the search characteristic parameters by the user and then returning to execute the step A1;
the process of the rare class development is as follows:
B1. forming a positive sample set by the rare disease pictures interested by the user from the pictures fed back to the user by the rare detection, and forming a negative sample set by the other pictures;
B2. determining adjacent sample sets of each picture in the positive sample set and taking the union of the adjacent sample sets as phi;
B3. for any picture in the set phi, calculating the positive sample distance r between the picture and the positive sample set+And the negative sample distance r between the negative sample and the negative sample set-；
B4. Extracting the picture with the maximum r value from the set phi and providing the picture to a user, wherein r is r--r+If the user is interested, the user will beAdding the picture into a positive sample set, and adding the picture into a negative sample set if the user is not interested;
B5. circularly executing the steps B2-B4 until phi- ∩+-∩-For space-time termination, the positive sample set at this time is taken as the final output result, wherein ∩+As the intersection of the set of positive samples with the set Φ, ∩-Is the intersection of the set of negative samples and the set phi.
2. The rare disease picture searching method of claim 1, wherein: the search feature parameters in step a1 are triplets<k,sup,slow>Wherein s isupAnd slowRespectively setting upper and lower limit values of rare class indexes of the rare disease pictures for users, k being a given natural number used for representing the size of the adjacent sample set of the pictures and k ∈ [ k ]min,kmax]，kminAnd kmaxRespectively, given interval upper and lower limit values.
3. The rare disease picture searching method of claim 2, wherein: in the step A2, calculating the rare class index of each picture by the following formula;
Figure FDA0002413898870000011
wherein:
Figure FDA0002413898870000012
the method comprises the steps that a rare class index of the ith picture in a picture library is represented, i is a natural number larger than 0, a set of adjacent samples of the ith picture is determined through a KNN algorithm, the number of the pictures in the set is k, d1～dkAnd corresponding to Euclidean distances between k pictures in the set and the ith picture in the picture library, wherein the Euclidean distances are in the order from small to large, and avg { } is an averaging function.
4. The rare disease picture searching method of claim 2, wherein: what is needed isFor the interval [ k ] in the step A2min,kmax]Discretizing, calculating and storing the rare index of all pictures in the picture library under each k value in the traversal interval, so that the rare index of the pictures can be directly called without calculation when the search characteristic parameters are subsequently adjusted.
5. The rare disease picture searching method of claim 4, wherein: the rare class indexes of the pictures without the k value are divided into intervals according to the size of the rare class indexes through statistical processing by adopting a grading statistical method, so that the single online rare class detection process of a user can be completed more quickly.
6. The rare disease picture searching method of claim 1, wherein: in the step B3, the positive sample distance r is calculated by the following formula+；
Figure FDA0002413898870000021
Wherein: x is the characteristic vector of any picture in the set phi, P is the characteristic matrix formed by the characteristic vectors of all pictures in the positive sample set, the characteristic vector is formed by all pixel values of the pictures, and omega1Is a weight vector consisting of n weight values, and the n weight values cumulatively equal to 1, n being the number of pictures in the positive sample set,
Figure FDA0002413898870000026
is a weight coefficient, ω1And
Figure FDA0002413898870000025
by aligning the sample distance r+Minimization determination, | | | purple2Is a two-norm.
7. The rare disease picture searching method of claim 1, wherein: in the step B3, the negative sample distance r is calculated by the following formula-；
Figure FDA0002413898870000022
Wherein: x is the characteristic vector of any picture in the set phi, G is the characteristic matrix formed by the characteristic vectors of all pictures in the negative sample set, the characteristic vector is formed by all pixel values of the pictures, and omega is2Is a weight vector consisting of m weight values, and the m weight values are cumulatively equal to 1, m is the number of pictures in the negative sample set,
Figure FDA0002413898870000023
is a weight coefficient, ω2And
Figure FDA0002413898870000024
by a distance r to the negative sample-Minimization determination, | | | purple2Is a two-norm.
8. The rare disease picture searching method of claim 1, wherein: the method reduces the time complexity of single query to the degree of real-time feedback through the global setting and offline processing of variables, ensures that rare development results are in line with reality by utilizing human-computer interaction, avoids the waste of human resources due to the obtainment of a large number of irrelevant pictures, fills the defects of the current method based on the problem, can help people to better obtain economic benefits from medical pictures, and has reference value in relevant research fields.
CN202010185084.6A 2020-03-17 2020-03-17 Rare disease picture searching method based on rare class mining Active CN111403027B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010185084.6A CN111403027B (en) 2020-03-17 2020-03-17 Rare disease picture searching method based on rare class mining

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010185084.6A CN111403027B (en) 2020-03-17 2020-03-17 Rare disease picture searching method based on rare class mining

Publications (2)

Publication Number Publication Date
CN111403027A true CN111403027A (en) 2020-07-10
CN111403027B CN111403027B (en) 2023-06-27

Family

ID=71428932

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010185084.6A Active CN111403027B (en) 2020-03-17 2020-03-17 Rare disease picture searching method based on rare class mining

Country Status (1)

Country Link
CN (1) CN111403027B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140348419A1 (en) * 2013-05-23 2014-11-27 International Business Machines Corporation Auto-Maintained Document Classification
US20150363551A1 (en) * 2013-01-31 2015-12-17 Renaud CEZAR Process for identifying rare events
CN107066515A (en) * 2017-01-23 2017-08-18 武汉万般上品信息技术有限公司 A kind of quick search technology towards rare class data in big data
CN107145901A (en) * 2017-04-24 2017-09-08 武汉大学 A kind of method for quickly querying towards rare class data in big data
CN109948705A (en) * 2019-03-20 2019-06-28 武汉大学 A kind of rare class detection method and device based on k neighbour's figure

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150363551A1 (en) * 2013-01-31 2015-12-17 Renaud CEZAR Process for identifying rare events
US20140348419A1 (en) * 2013-05-23 2014-11-27 International Business Machines Corporation Auto-Maintained Document Classification
CN107066515A (en) * 2017-01-23 2017-08-18 武汉万般上品信息技术有限公司 A kind of quick search technology towards rare class data in big data
CN107145901A (en) * 2017-04-24 2017-09-08 武汉大学 A kind of method for quickly querying towards rare class data in big data
CN109948705A (en) * 2019-03-20 2019-06-28 武汉大学 A kind of rare class detection method and device based on k neighbour's figure

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
HANFEI LIN ET AL.: "RCLens: Interactive Rare Category Exploration and Identification" *
JINGRUI HE ET AL.: "Nearest-neighbor-based active learning for rare category detection" *
JINGRUI HE ET AL: "Rare Category Characterization" *
SONG WANG ET AL.: "Fast Rare Category Detection Using Nearest Centriod Neighborhood" *
王淞;黄浩;余果;梁楠;王黎维;孙月明;: "一种基于k近邻图的稀有类检测算法" *
黄浩 等: "基于加权边界度的稀有类检测算法" *

Also Published As

Publication number Publication date
CN111403027B (en) 2023-06-27

Similar Documents

Publication Publication Date Title
CN109993102B (en) Similar face retrieval method, device and storage medium
CN110827398B (en) Automatic semantic segmentation method for indoor three-dimensional point cloud based on deep neural network
CN111259940B (en) Target detection method based on space attention map
CN111160407B (en) Deep learning target detection method and system
CN109685765B (en) X-ray film pneumonia result prediction device based on convolutional neural network
CN111429407A (en) Chest X-ray disease detection device and method based on two-channel separation network
CN111524140B (en) Medical image semantic segmentation method based on CNN and random forest method
CN111798424B (en) Medical image-based nodule detection method and device and electronic equipment
CN113222149A (en) Model training method, device, equipment and storage medium
CN115880720A (en) Non-labeling scene self-adaptive human body posture and shape estimation method based on confidence degree sharing
CN114782948A (en) Global interpretation method and system for cervical liquid-based cytology smear
Meng et al. Clinical applications of graph neural networks in computational histopathology: A review
CN112084913B (en) End-to-end human body detection and attribute identification method
CN117371511A (en) Training method, device, equipment and storage medium for image classification model
Zhang et al. A method of deep belief network image classification based on probability measure rough set theory
WO2023226217A1 (en) Microsatellite instability prediction system and construction method therefor, terminal device, and medium
CN116188428A (en) Bridging multi-source domain self-adaptive cross-domain histopathological image recognition method
CN111403027A (en) Rare disease picture searching method based on rare mining
CN109460768A (en) A kind of text detection and minimizing technology for histopathology micro-image
Hassanien et al. Enhanced rough sets rule reduction algorithm for classification digital mammography
Mao et al. W-net: A two-stage convolutional network for nucleus detection in histopathology image
Gonzalez Obando et al. Icytomine: A user-friendly tool for integrating workflows on whole slide images
CN112529081A (en) Real-time semantic segmentation method based on efficient attention calibration
AlTalli et al. Chest pathology detection in x-ray scans using social spider optimization algorithm with generalization deep learning
CN117454957B (en) Reasoning training system for image processing neural network model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant