CN118315080B - Automatic similar case recommending method, equipment, cluster and medium - Google Patents

Automatic similar case recommending method, equipment, cluster and medium Download PDF

Info

Publication number
CN118315080B
CN118315080B CN202410726143.4A CN202410726143A CN118315080B CN 118315080 B CN118315080 B CN 118315080B CN 202410726143 A CN202410726143 A CN 202410726143A CN 118315080 B CN118315080 B CN 118315080B
Authority
CN
China
Prior art keywords
hash
bucket
cases
images
algorithm
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.)
Active
Application number
CN202410726143.4A
Other languages
Chinese (zh)
Other versions
CN118315080A (en
Inventor
王欢
张文亮
庄良婷
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.)
First Affiliated Hospital of Guangzhou Medical University
Original Assignee
First Affiliated Hospital of Guangzhou Medical 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 First Affiliated Hospital of Guangzhou Medical University filed Critical First Affiliated Hospital of Guangzhou Medical University
Priority to CN202410726143.4A priority Critical patent/CN118315080B/en
Publication of CN118315080A publication Critical patent/CN118315080A/en
Application granted granted Critical
Publication of CN118315080B publication Critical patent/CN118315080B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Analysis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The application provides an automatic similar case recommending method, equipment, a cluster and a medium. The method comprises the following steps: extracting a first feature vector from the first image; calculating a first local sensitive hash algorithm on the first feature vector to obtain a first hash value, and calculating a second local sensitive hash algorithm on the first feature vector to obtain a second hash value; and recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket under the condition that the first hash value is located in the first hash bucket of the first hash table and the second hash value is located in the second hash bucket of the second hash table, wherein the hash value obtained by calculating the feature vector extracted from the images in the first hash bucket through a second local sensitive hash algorithm belongs to the second hash bucket, and the hash value obtained by calculating the feature vector extracted from the images in the second hash bucket through the first local sensitive hash algorithm belongs to the first hash bucket.

Description

Automatic similar case recommending method, equipment, cluster and medium
Technical Field
The invention relates to the field of image processing, in particular to an automatic similar case recommending method, equipment, a cluster and a medium.
Background
In medical diagnosis, the former cases often have important reference values. However, the number of cases is very large, the manual search for similar cases is a huge project, and the time is very long, but the search result cannot meet the requirements of users yet.
Disclosure of Invention
The application provides an automatic similar case recommending method, equipment, a cluster and a medium, which can automatically and quickly find similar cases and recommend the similar cases.
In a first aspect, there is provided a method for automatic recommendation of similar cases, including:
extracting a first feature vector from the first image;
Calculating the first feature vector by adopting a first local sensitive hash algorithm to obtain a first hash value, and calculating the first feature vector by adopting a second local sensitive hash algorithm to obtain a second hash value, wherein the first local sensitive hash algorithm is different from the second local sensitive hash algorithm;
And recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket under the condition that the first hash value is located in a first hash bucket of the first hash table and the second hash value is located in a second hash bucket of the second hash table, wherein the first hash table corresponds to the first local sensitive hash algorithm, the second hash table corresponds to the second local sensitive hash algorithm, the first hash table comprises a plurality of hash buckets including the first hash bucket, the second hash table comprises a plurality of hash buckets including the second hash bucket, the hash value obtained by calculating the feature vector extracted from the images in the first hash bucket by adopting the second local sensitive hash algorithm belongs to the second hash bucket, and the hash value obtained by calculating the feature vector extracted from the images in the second hash bucket by adopting the first local sensitive hash algorithm belongs to the first hash bucket.
In some possible designs, storing the cases corresponding to the first image in the second hash bucket if the number of cases in the first hash bucket minus the number of cases in the second hash bucket is greater than or equal to a number threshold;
And storing the cases corresponding to the first image into the first hash bucket when the number of cases in the first hash bucket minus the number of cases in the second hash bucket is less than a number threshold.
In some possible designs, before storing the case corresponding to the first image in the second hash bucket or storing the case corresponding to the first image in the first hash bucket, the method further comprises:
and storing the case corresponding to the first image into the second hash bucket or storing the case corresponding to the first image into the first hash bucket under the condition that a confirmation instruction input by a user is received.
In some possible designs, recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket includes:
And recommending the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket according to the similarity, wherein the similarity of the cases corresponding to the first images is higher when the hash value calculated by the images of the cases in the first hash bucket through a first local sensitive hash algorithm is closer to the first hash value, and the similarity of the cases corresponding to the first images is higher when the hash value calculated by the images of the cases in the second hash bucket through a second local sensitive hash algorithm is closer to the second hash value.
In some possible designs, recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket includes:
And recommending the case corresponding to the image in the first hash bucket and the case corresponding to the image in the second hash bucket according to the level of the doctor.
In some possible designs, the higher the similarity of the plurality of input values of the first locality-sensitive hashing algorithm and the second locality-sensitive hashing algorithm, the higher the similarity of the corresponding output values of the plurality of input values.
In some possible designs, the first locality sensitive hashing algorithm comprises any one of a minimum value hashing algorithm, a similar hashing algorithm, a random projection hashing algorithm, a spectral hashing algorithm; the second locality sensitive hashing algorithm comprises another one of a minimum value hashing algorithm, a similar hashing algorithm, a random projection hashing algorithm, and a spectral hashing algorithm.
In a second aspect, an automatic similar case recommendation device is provided. The device comprises an extraction unit, a calculation unit and a recommendation unit,
The extraction unit is used for extracting a first feature vector from the first image;
The computing unit is used for computing the first feature vector by adopting a first local sensitive hash algorithm to obtain a first hash value, and computing the first feature vector by adopting a second local sensitive hash algorithm to obtain a second hash value, wherein the first local sensitive hash algorithm is different from the second local sensitive hash algorithm;
The recommending unit is configured to recommend a case corresponding to an image in the first hash bucket and a case corresponding to an image in the second hash bucket when the first hash value is located in a first hash bucket of the first hash table and the second hash value is located in a second hash bucket of the second hash table, where a hash value obtained by calculating a feature vector extracted from the image in the first hash bucket by using a second local sensitive hash algorithm belongs to the second hash bucket, and a hash value obtained by calculating a feature vector extracted from the image in the second hash bucket by using the first local sensitive hash algorithm belongs to the first hash bucket.
In a third aspect, a computing device is provided, comprising: a processor and a memory, wherein the memory is for storing instructions, the processor is for executing the instructions in the memory to perform the method of any of the first aspects.
In a fourth aspect, a computing cluster is provided, comprising a plurality of computing devices, wherein each computing device comprises a processor and a memory, the memory for storing instructions, the processor for executing the instructions in the memory to perform the method of any of the first aspects.
In a fifth aspect, there is provided a computer readable storage medium comprising instructions which, when executed by a computing device, perform the method of any of the first aspects.
In the above scheme, the first feature vector is obtained by extracting features of the first image, and the first feature vector is calculated by adopting two different local sensitive hash algorithms, if the two different algorithms fall into hash buckets corresponding to two hash tables corresponding to the same class of cases, the first image is considered to belong to the class of cases, and the cases in the two hash buckets are recommended to the user. The first hash table is indexed according to the hash value calculated by the first local sensitive hash algorithm, and the second hash table is indexed according to the hash value calculated by the second local sensitive hash algorithm, so that the retrieval speed is very high. Because a plurality of different local sensitive hash algorithms are adopted at the same time, under the condition that the plurality of hash algorithms determine that the cases corresponding to the first image are the same case, the recommendation of the case is carried out, the accuracy of the recommendation can be effectively improved, and the cases of the same type can be respectively placed in a plurality of different hash tables, so that the storage capacity of the case can be increased under the condition that the retrieval speed is not reduced.
Drawings
In order to more clearly describe the embodiments of the present invention or the technical solutions in the background art, the following description will describe the drawings that are required to be used in the embodiments of the present invention or the background art.
FIG. 1 is a schematic flow chart of an automatic similar case recommending method provided by the application;
FIG. 2 is a schematic diagram of a convolutional neural network according to the present application;
FIG. 3 is a schematic diagram of another convolutional neural network provided by the present application;
fig. 4 is a schematic structural diagram of a computing device provided by the present application.
Detailed Description
Embodiments of the present invention will be described below with reference to the accompanying drawings in the embodiments of the present invention. The terminology used in the description of the embodiments of the invention herein is for the purpose of describing particular embodiments of the invention only and is not intended to be limiting of the invention.
Referring to fig. 1, fig. 1 is a schematic flow chart of an automatic similar case recommending method provided by the application. As shown in fig. 1, the automatic similar case recommending method of the present application includes:
s101: a first feature vector is extracted from the first image.
In some possible embodiments, the first image may be one or more of an X-ray image, a computed tomography (Computed Tomography, CT) image, a nuclear magnetic resonance (Magnetic Resonance Imaging, MRI) image, an ultrasound examination (B-mode ultrasonography) image, a positron emission tomography (Positron Emission Tomography, PET), or the like. The first image may be a whole body image of the human body or an image of a local organ.
In some possible embodiments, the first image may be obtained after one or more of image processing, such as image segmentation, image extraction, image filtering, image denoising, image stitching, image enhancement, and the like, of the original image.
In some possible embodiments, the first feature vector may be extracted from the first image by a feature extraction algorithm. Wherein the feature extraction algorithm may comprise a combination of one or more of the following: principal component analysis (PRINCIPAL COMPONENT ANALYSIS, PCA), independent component analysis (INDEPENDENT COMPONENT ANALYSIS, ICA), linear discriminant analysis (LINEAR DISCRIMINANT ANALYSIS, LDA), local binary patterns (Local Binary Patterns, LBP), scale-invariant feature transforms (Scale-INVARIANT FEATURE TRANSFORM, SIFT), directional gradient histograms (Histogram of Oriented Gradients, HOG), convolutional neural networks (Convolutional Neural Networks, CNN).
In a specific embodiment, a convolutional neural network is illustrated as an example of a feature extraction algorithm, and as shown in fig. 2, the convolutional neural network may include a plurality of feature extractors.
The feature extractor includes a convolution layer and a pooling layer. The convolution layer is used for performing convolution processing, and the convolution processing process can be regarded as that a trainable first filter is used for convolving an input image or a convolution feature plane (feature map) to obtain a feature image, and the convolution layer refers to a neuron layer in a convolution neural network for convolving the input image and the filter. In the convolutional layer of the convolutional neural network, one neuron may be connected with only a part of adjacent layer neurons. A convolutional layer typically contains a number of feature planes, each of which may be composed of a number of neural elements arranged in a rectangular pattern. Neural elements of the same feature plane share weights, where the shared weights are convolution kernels. A convolution kernel can be understood as a position-independent way of extracting image information. So we can use the same learned image information for all positions on the input image. In the same convolution layer, a plurality of convolution kernels may be used to extract different image information, and in general, the greater the number of convolution kernels, the more abundant the image information reflected by the convolution operation. The pooling layer is used for performing pooling processing, and the pooling processing can be regarded as processing of sampling the input image, and the input image can be subjected to dimension reduction through the pooling processing, that is, the pooling layer can reduce the dimension of the input image with a larger image size into a characteristic image with a smaller image size.
In a more specific embodiment, taking convolutional neural network as an example, as shown in fig. 3, the feature extraction algorithm includes a feature extractor 1 (including a convolutional layer 1 and a pooling layer 1), a feature extractor 2 (including a convolutional layer 2 and a pooling layer 2), and a convolutional layer 3.
In the feature extractor 1:
Will be of the size of Is input to the feature extractor 1 to obtain a plurality of sizesIs a feature image of (1). Wherein, Is greater than a value ofIs a numerical value of (2). In practice, the dimensions may be first of allIs a first image of (1)Input convolution layer 1, thereby obtainingWith a size ofConvolved imageAnd then willWith a size ofConvolved imageInputting the pooling layer 1 to obtainWith a size ofIs a pooled image of (1). Here the number of the elements is the number,With a size ofIs a pooled image of (1)Namely, isWith a size ofIs a feature image of (1). The specific calculation process in the convolution layer 1 and the pooling layer 1 is as follows:
in convolution layer 1:
Will be of the size of Is a first image of (1)As input to the convolutional layer 1, throughEach convolution kernel=1,2,…,) Convolution generation of (a)With a size ofIs a convolution image of (1)=1,2,…,) Wherein each convolution imageThe generation process of (a) is specifically as follows:
wherein, Represented as using convolution kernelsFor the first imageThe convolution operation is performed in a manner that the same is denoted as padding,Represented as a value of the offset,Represented as the result of the convolution calculation,Represented as an activation function, the present invention employs relu functions.
In pooling layer 1:
Output convolutional layer 1 With a size ofIs a convolution image of (1)As an input of the pooling layer, generating after pooling through a pooling windowWith a size ofIs a pooled image of (1)=1,2,…,) Wherein each pooled imageThe generation process of (a) is specifically as follows:
Wherein maxpool is denoted maximum pooling. It should be understood that, by taking maximum pooling as an example, in practical application, mean pooling may also be used, and the like, which is not specifically limited herein.
In the feature extractor 2:
Will be With a size ofIs a pooled image of (1)=1,2,…,) Input to the feature extractor 2, thereby obtaining a plurality of dimensionsIs a feature image of (1). Wherein, Is greater than a value ofIs a numerical value of (2). In practice, it is possible to first applyWith a size ofIs a pooled image of (1)=1,2,…,) Input convolution layer 2, thereby obtainingWith a size ofConvolved image=1,2,…,) And then willWith a size ofConvolved image=1,2,…,) Input into pooling layer 2 to obtainWith a size ofIs a pooled image of (1)=1,2,…,). Here the number of the elements is the number,With a size ofIs a pooled image of (1)=1,2,…,) Namely, isWith a size ofIs a feature image of (1). The specific calculation process in the convolution layer 2 and the pooling layer 2 is as follows:
In convolution layer 2:
Will be of the size of Is a pooled image of (1)=1,2,…,) As input to the convolutional layer 2, throughEach convolution kernel=1,2,…,) Convolution generation of (a)With a size ofIs a convolution image of (1)=1,2,…,) Wherein each convolution imageThe generation process of (a) is specifically as follows:
wherein, Represented as using convolution kernelsTo pooling imagesThe convolution operation is performed in a manner that the same is denoted as padding,Represented as a value of the offset,Represented as the result of the convolution calculation,Represented as an activation function, the present invention employs relu functions.
In pooling layer 2:
Output convolutional layer 2 With a size ofIs a convolution image of (1)As an input of the pooling layer, generating after pooling through a pooling windowWith a size ofIs a pooled image of (1)=1,2,…,) Wherein each pooled imageThe generation process of (a) is specifically as follows:
Wherein maxpool is denoted maximum pooling. It should be understood that, by taking maximum pooling as an example, in practical application, mean pooling may also be used, and the like, which is not specifically limited herein.
In convolution layer 3:
Will be With a size ofIs a pooled image of (1)=1,2,…,) As input to the convolutional layer 3, throughEach convolution kernel=1,2,…,) Convolution generation of (a)With a size ofIs a convolution image of (1)=1,2,…,) Wherein each convolution imageThe generation process of (a) is specifically as follows:
wherein, Represented as using convolution kernelsTo pooling imagesThe convolution operation is performed in a manner that the same is denoted as padding,Represented as a value of the offset,Represented as the result of the convolution calculation,Represented as an activation function, the present invention employs relu functions.
The first feature vector can be obtained by stitching the plurality of convolution images.
S102: and calculating the first eigenvector by adopting a first local sensitive hash algorithm to obtain a first hash value, and calculating the first eigenvector by adopting a second local sensitive hash algorithm to obtain a second hash value.
In some possible embodiments, the first locality sensitive hashing algorithm and the second locality sensitive hashing algorithm are both hashing algorithms with higher similarity for a plurality of input values, and higher similarity for output values corresponding to the plurality of input values. That is, the first locality sensitive hashing algorithm and the second locality sensitive hashing algorithm are both algorithms that map similar input values to neighboring hash values. The first locality sensitive hashing algorithm comprises any one of a minimum value hashing (minhash) algorithm, a similar hashing (simhash) algorithm, a random projection hashing algorithm (Random Projection), a spectral hashing algorithm (SPECTRAL HASHING). The second locality sensitive hashing algorithm comprises another of a minimum value hashing algorithm, a similar hashing algorithm, a random projection hashing algorithm, and a spectral hashing algorithm. Here, the first locality sensitive hashing algorithm must be different from the second locality sensitive hashing algorithm.
In some possible embodiments, the first locality sensitive hashing algorithm corresponds to a first hash table, and the second locality sensitive hashing algorithm corresponds to a second hash table. That is, the first hash table is indexed by the hash value calculated by the first locality sensitive hashing algorithm, and the second hash table is indexed by the hash value calculated by the second locality sensitive hashing algorithm. The first hash table includes a plurality of hash buckets, and multiple similar hash values in the first hash table may share the same hash bucket or each hash value corresponds to one hash bucket. The second hash table includes a plurality of hash buckets, and multiple similar hash values in the second hash table may share the same hash bucket or each hash value corresponds to one hash bucket.
Each hash bucket may include a plurality of hash structures, with the depth of the hash bucket being the number of hash structures that the hash bucket includes. In a particular embodiment, each hash structure includes one or more of a visit information, a complaint, a current medical history, a past history, a family history, a physical examination, a laboratory examination, a diagnosis, a treatment regimen, a follow-up condition, and the like. The treatment information comprises treatment date, treatment department, treatment doctor and the like. Complaints include patient self-describing principal symptoms or discomfort. The present history includes a current description of the patient's condition, including symptoms, duration, frequency of episodes, etc. The past history includes past disease history, operation history, drug allergy history, etc. Family history includes whether the patient family members have a similar disease or genetic history. Physical examinations include physical examinations of a patient including height, weight, blood pressure, heart rate, and the like. Laboratory tests include blood tests, urine tests, imaging tests, and the like. Diagnosis includes preliminary or final diagnosis of a disease in a patient. Treatment regimens include drug therapy, surgical therapy, rehabilitation programs, and the like. Follow-up conditions include patient treatment progress, efficacy assessment, recurrent conditions, and the like. The above-described hash structure is merely a specific example, and in practical applications, the hash structure may further include more or fewer fields, and the length of each field may be set as needed.
The hash buckets in the first hash table and the hash buckets in the second hash table have a one-to-one correspondence. Two cases of the same class are in the same hash bucket of the first hash table with two hash values calculated by the first locality sensitive hashing algorithm, and in the same hash bucket of the second hash table with two hash values calculated by the second locality sensitive hashing algorithm. Because two cases are assigned to the same bucket by two different locality sensitive hashing algorithms, if they are assigned to the same bucket twice, the likelihood of similarity of the two cases is greater than if they were assigned to the same bucket by calculation using one locality sensitive hashing algorithm, because the influence of chance is eliminated. Here, the first hash bucket of the first hash table and the second hash bucket of the second hash table are buckets of the same class of cases respectively corresponding in two different hash tables. That is, the hash value calculated by the feature vector extracted from the image in the first hash bucket using the first locality sensitive hashing algorithm belongs to the first hash bucket, and the hash value calculated by the feature vector extracted from the image in the first hash bucket using the second locality sensitive hashing algorithm belongs to the second hash bucket. The hash value obtained by calculating the feature vector extracted from the image in the second hash bucket by adopting the first local sensitive hash algorithm belongs to the first hash bucket, and the hash value obtained by calculating by adopting the second local sensitive hash algorithm belongs to the second hash bucket. Thus, the cases in the first hash bucket and the cases in the second hash bucket are the same type of cases.
S103: it is determined whether the first hash value is located in a first hash bucket of the first hash table and the second hash value is located in a second hash bucket of the second hash table. If the first hash value is located in the first hash bucket of the first hash table and the second hash value is located in the second hash bucket of the second hash table, the process proceeds to step S104, and if not, the flow ends.
S104: and recommending the case corresponding to the image in the first hash bucket and the case corresponding to the image in the second hash bucket.
In some possible embodiments, since the case corresponding to the image in the first hash bucket and the case corresponding to the image in the second hash bucket are both the same type of case, the case corresponding to the image in the first hash bucket and the case corresponding to the image in the second hash bucket may be recommended. In making the recommendation, the recommendation may be made in the following manner:
(1) And recommending the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket according to the similarity. Assuming that one of the cases in the first hash bucket is the second image, the similarity of the first image and the second image may be calculated according to the following manner:
And calculating the similarity between the first hash value and the second hash value according to a first hash value calculated by a first local sensitive hash algorithm according to a first feature vector extracted from the first image and a second hash value calculated by a first local sensitive hash algorithm according to a second feature vector extracted from the second image. The calculation method of the similarity comprises the following steps: euclidean distance (Euclidean Distance), manhattan distance (MANHATTAN DISTANCE), cosine similarity (Cosine Similarity), jaccard similarity (Jaccard Similarity), pearson correlation coefficient (Pearson Correlation Coefficient), hamming distance (HAMMING DISTANCE), and so forth. Therefore, the closer the hash value calculated by the first local sensitive hash algorithm is to the first hash value, the higher the similarity of the case corresponding to the first image, and the closer the hash value calculated by the second local sensitive hash algorithm is to the second hash value, the higher the similarity of the case corresponding to the first image.
(2) And recommending the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket according to the level of the doctor.
Because the higher the level of the treatment plan of the main doctor and the reference value of the diagnosis are often higher, the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket can be recommended in accordance with the level of the main doctor. The levels of the primary doctor include the primary doctor, the secondary primary doctor, the inpatients, the practitioners, and the like.
S105: the number of cases in the first hash bucket and the number of cases in the second hash bucket are obtained, and whether the number of cases in the first hash bucket minus the number of cases in the second hash bucket is greater than or equal to a number threshold is calculated. In the case where the number of cases in the first hash bucket minus the number of cases in the second hash bucket is greater than or equal to the number threshold, the flow proceeds to step S106, otherwise the flow proceeds to step S107.
In some possible embodiments, the number of cases in the first hash bucket of the first hash table and the number of cases in the second hash bucket of the second hash table are obtained, the number of cases in the first hash bucket and the number of cases in the second hash bucket are compared to obtain a comparison result, and then, according to the comparison result, it is determined whether the cases corresponding to the first image are stored in the first hash bucket of the first hash table or whether the cases corresponding to the first image are stored in the second hash bucket of the second hash table. For example, in a case where the number of cases in the first hash bucket minus the number of cases in the second hash bucket is greater than or equal to a number threshold, the cases corresponding to the first image may be stored in the second hash bucket of the second hash table. And storing the cases corresponding to the first image into the first hash bucket of the first hash table under the condition that the number of cases in the first hash bucket minus the number of cases in the second hash bucket is smaller than a number threshold. Wherein the number threshold may be set as desired or empirically, and the number threshold may be set to 1,2,3 or more, the smaller the value of the number threshold is, the smaller the depth difference between the first hash bucket of the first hash table and the second hash bucket of the second hash table is, and the larger the value of the number threshold is, the larger the depth difference between the first hash bucket of the first hash table and the second hash bucket of the second hash table is. The number threshold can be adjusted at any time as needed. For example, the hash buckets in the first hash table and the second hash table have 4 hash structures, respectively, denoted as a first hash structure, a second hash structure, a third hash structure, and a fourth hash structure. The case corresponding to the first image is preferably placed in the first hash structure of the first hash table, the second hash structure, the first hash structure of the second hash table, the second hash structure, the third hash structure of the first hash table, the third hash structure of the second hash table, the fourth hash structure of the first hash table, and the fourth hash structure of the second hash table. For another example, the buckets in the first hash table and the second hash table have 4 hash structures, respectively, denoted as a first hash structure, a second hash structure, a third hash structure, and a fourth hash structure. The case corresponding to the first image is preferably placed in the first hash structure of the first hash table, then in the first hash structure of the second hash table, then in the second hash structure and the third hash structure of the first hash table, then in the second hash structure and the third hash structure of the second hash table, then in the fourth hash structure of the first hash table, and finally in the fourth hash structure of the second hash table. In addition to the above examples, the case may be placed in a first hash table, then placed in a second hash table, and so on, and the case is placed in an even manner, and in practical applications, the case may be placed in a first hash table, then placed in a second hash table, and so on, without limitation.
S106: and storing the case corresponding to the first image into a second hash bucket.
S107: and storing the case corresponding to the first image into a first hash bucket.
In the above scheme, the first feature vector is obtained by extracting features of the first image, and the first feature vector is calculated by adopting two different local sensitive hash algorithms, if the two different algorithms fall into hash buckets corresponding to two hash tables corresponding to the same class of cases, the first image is considered to belong to the class of cases, and the cases in the two hash buckets are recommended to the user. The first hash table is indexed according to the hash value calculated by the first local sensitive hash algorithm, and the second hash table is indexed according to the hash value calculated by the second local sensitive hash algorithm, so that the retrieval speed is very high. Because a plurality of different local sensitive hash algorithms are adopted at the same time, under the condition that the plurality of hash algorithms determine that the cases corresponding to the first image are the same case, the recommendation of the case is carried out, the accuracy of the recommendation can be effectively improved, and the cases of the same type can be respectively placed in a plurality of different hash tables, so that the storage capacity of the case can be increased under the condition that the retrieval speed is not reduced.
Referring to fig. 4, fig. 4 is a schematic structural diagram of a computing device according to the present application. As shown in fig. 4, the computing device 400 includes: one or more processing units 410, a communication interface 420, and a memory 430.
The processing unit 410, the communication interface 420, and the memory 430 are interconnected by a bus 440. Optionally, the computing device 400 may further include an input/output interface 450, where the input/output interface 450 is connected to an input/output device for receiving parameters set by a user, etc. The computing device 400 can be used to implement some or all of the functionality of the device embodiments or system embodiments of the present application described above; the processing unit 410 can also be used to implement some or all of the operating steps of the method embodiments in the embodiments of the application described above. For example, specific implementations of the computing device 400 performing various operations may refer to specific details in the above-described embodiments, such as the processing unit 410 being configured to perform some or all of the steps of the above-described method embodiments or some or all of the operations in the above-described method embodiments. For another example, in an embodiment of the present application, the computing device 400 may be used to implement some or all of the functionality of one or more components of the apparatus embodiments described above, and the communication interface 420 may be used in particular for communication functions and the like necessary to implement the functionality of those apparatuses, components, and the processing unit 410 may be used in particular for processing functions and the like necessary to implement the functionality of those apparatuses, components.
It should be appreciated that the computing device 400 of fig. 4 may include one or more processing units 410, and that the plurality of processing units 410 may cooperatively provide processing power in a parallelized connection, a serialized connection, a serial-parallel connection, or any connection, or the plurality of processing units 410 may constitute a processor sequence or processor array, or the plurality of processing units 410 may be separated into a main processor and an auxiliary processor, or the plurality of processing units 410 may have different architectures such as employing heterogeneous computing architectures. In addition, the computing device 400 shown in FIG. 4, the associated structural and functional descriptions are exemplary and not limiting. In some example embodiments, computing device 400 may include more or fewer components than shown in fig. 4, or combine certain components, or split certain components, or have a different arrangement of components.
The processing unit 410 may have various specific implementations, for example, the processing unit 410 may include one or more of a central processing unit (central processingunit, CPU), a graphics processor (graphic processing unit, GPU), a neural network processor (neural-networkprocessing unit, NPU), a tensor processor (tensor processing unit, TPU), or a data processor (data processing unit, DPU), and the embodiment of the present application is not limited in particular. The processing unit 410 may also be a single-core processor or a multi-core processor. The processing unit 410 may be formed by a combination of a CPU and a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (programmable logicdevice, PLD), or a combination thereof. The PLD may be a complex programmable logic device (complexprogrammable logic device, CPLD), a field-programmable gate array (FPGA) GATE ARRAY, generic array logic (GENERIC ARRAY logic, GAL), or any combination thereof. The processing unit 410 may also be implemented solely with logic devices incorporating processing logic, such as an FPGA or Digital Signal Processor (DSP), etc. The communication interface 420 may be a wired interface, which may be an ethernet interface, a local area network (local interconnect network, LIN), etc., or a wireless interface, which may be a cellular network interface, or use a wireless lan interface, etc., for communicating with other modules or devices.
The memory 430 may be a nonvolatile memory such as read-only memory (ROM), programmable ROM (PROM), erasable programmable ROM (erasable PROM, EPROM), electrically erasable programmable ROM (electricallyEPROM, EEPROM), or flash memory. Memory 430 may also be volatile memory, which may be random access memory (randomaccess memory, RAM) used as external cache memory. By way of example, and not limitation, many forms of RAM are available, such as static random access memory (STATIC RAM, SRAM), dynamic random access memory (DYNAMIC RAM, DRAM), synchronous Dynamic Random Access Memory (SDRAM), double data rate synchronous dynamic random access memory (double DATARATE SDRAM, DDRSDRAM), enhanced synchronous dynamic random access memory (ENHANCED SDRAM, ESDRAM), synchronous link dynamic random access memory (SYNCHLINK DRAM, SLDRAM), and direct memory bus random access memory (direct rambus RAM, DR RAM). Memory 430 may also be used to store program code and data such that processing unit 410 invokes the program code stored in memory 430 to perform some or all of the operational steps of the method embodiments described above, or to perform corresponding functions of the apparatus embodiments described above. Moreover, computing device 400 may contain more or fewer components than shown in FIG. 4, or may have a different configuration of components.
Bus 440 may be a peripheral component interconnect express (PERIPHERAL COMPONENT INTERCONNECT EXPRESS, PCIe) bus, or an extended industry standard architecture (extended industry standard architecture, EISA) bus, a unified bus (unified bus, ubus or UB), a computer quick link (compute express link, CXL), a cache coherent interconnect protocol (cache coherentinterconnect for accelerators, CCIX), or the like. The bus 440 may be divided into an address bus, a data bus, a control bus, and the like. The bus 440 may include a power bus, a control bus, a status signal bus, and the like in addition to a data bus. But is shown with only one bold line in fig. 4 for clarity of illustration, but does not represent only one bus or one type of bus.
Embodiments of the present application also provide a system that includes a plurality of computing devices, where each computing device may be structured as described above. The functions or operations that may be implemented by the system may refer to specific implementation steps in the above method embodiments and/or specific functions described in the above apparatus embodiments, which are not described herein.
Embodiments of the present application also provide a computer-readable storage medium having stored therein computer instructions which, when executed on a computer device (e.g., one or more processors), implement the method steps of the method embodiments described above. The specific implementation of the processor of the computer readable storage medium in executing the above method steps may refer to specific operations described in the above method embodiments and/or specific functions described in the above apparatus embodiments, which are not described herein again.
Embodiments of the present application also provide a computer program product comprising instructions stored on a computer-readable storage medium, which when run on a computer device, cause the computer device to perform the method steps in the method embodiments described above.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions which, when loaded and executed on a computer, produce, in whole or in part, a process or function in accordance with embodiments of the present invention. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one network site, computer, server, or data center to another network site, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer and may also be a data storage device, such as a server, data center, etc., that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape, etc.), an optical medium (e.g., DVD, etc.), or a semiconductor medium (e.g., solid state disk), etc.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to related descriptions of other embodiments.

Claims (9)

1. An automatic similar case recommendation method, comprising:
extracting a first feature vector from the first image;
Calculating the first feature vector by adopting a first local sensitive hash algorithm to obtain a first hash value, and calculating the first feature vector by adopting a second local sensitive hash algorithm to obtain a second hash value, wherein the first local sensitive hash algorithm is different from the second local sensitive hash algorithm;
recommending cases corresponding to images in a first hash table and cases corresponding to images in a second hash table under the condition that a first hash value is located in a first hash table and a second hash value is located in a second hash table of the first hash table, wherein the first hash table corresponds to the first local sensitive hash algorithm, the second hash table corresponds to the second local sensitive hash algorithm, the first hash table comprises a plurality of hash tables including the first hash table, the second hash table comprises a plurality of hash tables including the second hash table, a hash value obtained by calculating a feature vector extracted from images in the first hash table by adopting the second local sensitive hash algorithm belongs to the second hash table, and a hash value obtained by calculating a feature vector extracted from images in the second hash table by adopting the first local sensitive hash algorithm belongs to the first hash table;
the method further comprises the steps of:
Acquiring a number of cases in the first hash bucket and a number of cases in the second hash bucket;
storing the cases corresponding to the first image into the second hash bucket if the number of cases in the first hash bucket minus the number of cases in the second hash bucket is greater than or equal to a number threshold;
And storing the cases corresponding to the first image into the first hash bucket when the number of cases in the first hash bucket minus the number of cases in the second hash bucket is less than a number threshold.
2. The method of claim 1, wherein prior to storing the case corresponding to the first image in the second hash bucket or storing the case corresponding to the first image in the first hash bucket, the method further comprises:
and storing the case corresponding to the first image into the second hash bucket or storing the case corresponding to the first image into the first hash bucket under the condition that a confirmation instruction input by a user is received.
3. The method of claim 1 or 2, wherein recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket comprises:
Recommending the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket according to the similarity, wherein the similarity between the case images in the first hash bucket and the cases corresponding to the first images is higher when the hash values calculated by the first local sensitive hash algorithm are closer to the first hash values, and the similarity between the case images in the second hash bucket and the cases corresponding to the first images is higher when the hash values calculated by the second local sensitive hash algorithm are closer to the second hash values.
4. The method of claim 1 or 2, wherein recommending cases corresponding to the images in the first hash bucket and cases corresponding to the images in the second hash bucket comprises:
And recommending the cases corresponding to the images in the first hash bucket and the cases corresponding to the images in the second hash bucket according to the level of the doctor.
5. The method according to claim 1 or 2, wherein the higher the similarity of the plurality of input values of the first locality-sensitive hashing algorithm and the second locality-sensitive hashing algorithm, the higher the similarity of the corresponding output values of the plurality of input values.
6. The method of claim 5, wherein the first locality sensitive hashing algorithm comprises any one of a minimum hashing algorithm, a similar hashing algorithm, a random projection hashing algorithm, a spectral hashing algorithm; the second locality sensitive hashing algorithm comprises another one of a minimum value hashing algorithm, a similar hashing algorithm, a random projection hashing algorithm, and a spectral hashing algorithm.
7. A computing device, comprising: a processor and a memory, wherein the memory is for storing instructions, the processor is for executing the instructions in the memory to perform the method of any of claims 1-6.
8. A computing cluster comprising a plurality of computing devices, wherein each computing device comprises a processor and a memory, the memory to store instructions, the processor to execute the instructions in the memory to perform the method of any of claims 1-6.
9. A computer readable storage medium comprising instructions which, when executed by a computing device, perform the method of any of claims 1-6.
CN202410726143.4A 2024-06-06 2024-06-06 Automatic similar case recommending method, equipment, cluster and medium Active CN118315080B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410726143.4A CN118315080B (en) 2024-06-06 2024-06-06 Automatic similar case recommending method, equipment, cluster and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410726143.4A CN118315080B (en) 2024-06-06 2024-06-06 Automatic similar case recommending method, equipment, cluster and medium

Publications (2)

Publication Number Publication Date
CN118315080A CN118315080A (en) 2024-07-09
CN118315080B true CN118315080B (en) 2024-09-24

Family

ID=91730170

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410726143.4A Active CN118315080B (en) 2024-06-06 2024-06-06 Automatic similar case recommending method, equipment, cluster and medium

Country Status (1)

Country Link
CN (1) CN118315080B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106570141A (en) * 2016-11-04 2017-04-19 中国科学院自动化研究所 Method for detecting approximately repeated image

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11263541B2 (en) * 2017-09-27 2022-03-01 Oracle International Corporation Ensembled decision systems using feature hashing models
CN113190551A (en) * 2021-04-20 2021-07-30 北京百度网讯科技有限公司 Feature retrieval system construction method, feature retrieval method, device and equipment
CN114064948A (en) * 2021-10-15 2022-02-18 西安深信科创信息技术有限公司 Hash image retrieval method and device based on generalized average pooling strategy

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106570141A (en) * 2016-11-04 2017-04-19 中国科学院自动化研究所 Method for detecting approximately repeated image

Also Published As

Publication number Publication date
CN118315080A (en) 2024-07-09

Similar Documents

Publication Publication Date Title
Jasti et al. Computational technique based on machine learning and image processing for medical image analysis of breast cancer diagnosis
KR101857624B1 (en) Medical diagnosis method applied clinical information and apparatus using the same
Öztürk Stacked auto-encoder based tagging with deep features for content-based medical image retrieval
CN110008971B (en) Image processing method, image processing device, computer-readable storage medium and computer equipment
CN111709485B (en) Medical image processing method, device and computer equipment
Marghalani et al. Automatic classification of brain tumor and Alzheimer’s disease in MRI
US12027267B2 (en) Information processing apparatus, information processing system, information processing method, and non-transitory computer-readable storage medium for computer-aided diagnosis
CN109242863B (en) Ischemic stroke image region segmentation method and device
Khan et al. Hybrid‐NET: A fusion of DenseNet169 and advanced machine learning classifiers for enhanced brain tumor diagnosis
Sadad et al. Internet of medical things embedding deep learning with data augmentation for mammogram density classification
CN113469981B (en) Image processing method, device and storage medium
Bingol NCA‐based hybrid convolutional neural network model for classification of cervical cancer on gauss‐enhanced pap‐smear images
Badawy et al. A two-stage renal disease classification based on transfer learning with hyperparameters optimization
CN112530550A (en) Image report generation method and device, computer equipment and storage medium
Patil et al. Predictive modelling of brain tumor detection using deep learning
Venkatesan et al. NFU-Net: an automated framework for the detection of neurotrophic foot ulcer using deep convolutional neural network
Bouamrane et al. Enhancing lung cancer detection and classification using machine learning and deep learning techniques: A comparative study
Polat et al. Brain tumor classification by using a novel convolutional neural network structure
CN117036894B (en) Multi-mode data classification method and device based on deep learning and computer equipment
Irfan et al. Skin cancer prediction using deep learning techniques
Jadwaa X‐Ray Lung Image Classification Using a Canny Edge Detector
Shen et al. Multicontext multitask learning networks for mass detection in mammogram
CN118315080B (en) Automatic similar case recommending method, equipment, cluster and medium
Basak et al. F-unet: A modified u-net architecture for segmentation of stroke lesion
Türk et al. Machine learning of kidney tumors and diagnosis and classification by deep learning methods

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