CN109830300A - Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing - Google Patents
Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing Download PDFInfo
- Publication number
- CN109830300A CN109830300A CN201910129439.7A CN201910129439A CN109830300A CN 109830300 A CN109830300 A CN 109830300A CN 201910129439 A CN201910129439 A CN 201910129439A CN 109830300 A CN109830300 A CN 109830300A
- Authority
- CN
- China
- Prior art keywords
- thyroid nodule
- thyroid
- analysis
- training
- analyzing
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention is suitable for computer field, provides a kind of thyroid nodule analysis method, which comprises the thyroid nodule for receiving user's input analyzes parameter, and the thyroid nodule analysis parameter includes user basic information and thyroid gland essential information;Parameter is analyzed according to the thyroid nodule and trains the thyroid nodule analysis model generated to determine the pernicious risk of user's thyroid nodule and exports, and the trained thyroid nodule analysis model generated is based on random forests algorithm or to be based on neural network algorithm training generation.Thyroid nodule analysis method provided in an embodiment of the present invention, diagnostic result can directly directly be exported according to the relevant information that user inputs, diagnosis process is convenient and efficient, and the accuracy rate of diagnostic result is effectively ensured based on random forests algorithm or the thyroid nodule analysis model generated based on neural network algorithm training used in diagnosis process.
Description
Technical field
The present invention relates to computer fields, more particularly to a kind of thyroid nodule analysis method, device, computer equipment
And readable storage medium storing program for executing.
Background technique
Thyroid gland is the endocrine gland of neck, is made of two blades by isthmus connection, thyroid gland can secrete first shape
Glandular hormone, main to influence metabolic rate and protein synthesis, thyroid gland can also generate calcitonin hormone, rise emphatically in calcium homeostasis
The effect wanted.
And in the prior art to the diagnosis of thyroid disease, due to lacking the thyroid gland detection device of perfect in shape and function, cause
Diagnosis process is very time-consuming and laborious, it usually needs veteran doctor goes to judge, accuracy is influenced by machinery equipment and experience
It is larger.
As it can be seen that diagnostic result is not quasi- enough in the prior art to the diagnostic method of thyroid disease there is also time-consuming and laborious
True technical problem.
Summary of the invention
The embodiment of the present invention provides a kind of thyroid nodule analysis method, it is intended to which solve to mention in above-mentioned background technique shows
Have technology in the diagnosis of thyroid disease there is also process is time-consuming and laborious, technical problem that result is not accurate enough.
The embodiment of the present invention provides a kind of thyroid nodule analysis method, which comprises
The thyroid nodule for receiving user's input analyzes parameter, and the thyroid nodule analysis parameter includes that user believes substantially
Breath and thyroid gland essential information, the basic information of thyroid gland includes Thyroid echogenicity, internal component, aspect ratio, calcification, reality
When tissue elasticity imaging and cervical lymph node state etc.;
User's first is determined according to the thyroid nodule analysis model that the thyroid nodule analyzes parameter and training generation
The pernicious risk of shape gland tubercle simultaneously exports, the thyroid nodule analysis model that the training generates be based on random forests algorithm or
It is generated based on neural network algorithm training.
The embodiment of the present invention also provides a kind of thyroid nodule analytical equipment, comprising:
Thyroid nodule analyzes parameter receiving unit, and the thyroid nodule for receiving user's input analyzes parameter, described
It includes user basic information and thyroid gland essential information that thyroid nodule, which analyzes parameter, and the basic information of thyroid gland includes first
Shape gland echo, internal component, aspect ratio, calcification, Real time Organization elastogram and cervical lymph node state etc.;
The pernicious risk output unit of thyroid nodule, for analyzing parameter and training generation according to the thyroid nodule
Thyroid nodule analysis model determine the pernicious risk of user's thyroid nodule and export, it is described training generate thyroid nodule
Analysis model is generated based on random forests algorithm or based on neural network algorithm training.
The embodiment of the present invention also provides a kind of computer equipment, and the computer equipment includes memory and processor, institute
It states and is stored with computer program in memory, when the computer program is executed by the processor, so that the processor is held
The step of row as described above thyroid nodule analysis method.
The embodiment of the present invention also provides a kind of computer readable storage medium, stores on the computer readable storage medium
There is computer program, when the computer program is executed by processor, so that the processor executes first shape as described above
The step of gland tubercle analysis method.
Thyroid nodule analysis method provided in an embodiment of the present invention, the thyroid nodule by receiving user's input are analyzed
Parameter, thyroid nodule analysis parameter include user basic information and thyroid gland essential information, and first based on the received
The thyroid nodule analysis model that shape gland tubercle analysis parameter and training generate directly determines the pernicious wind of user's thyroid nodule
Danger simultaneously exports.Thyroid nodule analysis method provided in an embodiment of the present invention, end user inputs corresponding information can be direct
Diagnostic result is exported, diagnosis process is very convenient quick, and thyroid nodule analysis model used in diagnosis process is base
It is generated in random forests algorithm or based on neural network algorithm training, output is effectively guaranteed in stability with higher
Thyroid nodule risk accuracy rate.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of thyroid nodule analysis method provided in an embodiment of the present invention;
Fig. 1 (a) is that a kind of thyroid nodule provided in an embodiment of the present invention analyzes parameter input interface;
Fig. 2 is a kind of trained thyroid nodule analysis model of the generation based on random forest provided in an embodiment of the present invention
The step flow chart of method;
Fig. 3 is that the training that another embodiment of the present invention provides generates the thyroid nodule analysis model based on random forest
Method step flow chart;
Fig. 4 is the step flow chart provided in an embodiment of the present invention for generating thyroid nodule analysis decision tree;
Fig. 5 is that a kind of trained thyroid nodule of the generation based on neural network algorithm provided in an embodiment of the present invention analyzes mould
The step flow chart of the method for type;
Fig. 6 is a kind of structural schematic diagram of thyroid nodule analytical equipment provided in an embodiment of the present invention;
Fig. 7 is the structural schematic diagram for the thyroid nodule analytical equipment that another embodiment of the present invention provides;
Fig. 8 is the structural schematic diagram for the random forests algorithm training unit that another embodiment of the present invention provides.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Thyroid nodule analysis method provided in an embodiment of the present invention is led to based on big data and artificial intelligence learning algorithm
2064 thyroid gland case experimental datas from the first affiliated hospital, Ji'nan University of acquisition are crossed as sample, are based on artificial intelligence
Learning algorithm training generates the thyroid nodule analysis model that stability is strong, accuracy rate is high, so that subsequent carry out thyroid gland knot again
When section diagnosis, according to the thyroid nodule analysis parameter that user inputs and the thyroid nodule analysis model generated can be trained
The pernicious risk of thyroid nodule for directly determining and inputting user, effectively improves the efficiency and standard of Diagnosis of Thyroid
True rate.
Fig. 1 is a kind of step flow chart of thyroid nodule analysis method provided in an embodiment of the present invention, and details are as follows.
Step S101, the thyroid nodule for receiving user's input analyze parameter.
In embodiments of the present invention, the thyroid nodule analysis parameter includes that user basic information and thyroid gland are basic
Information, the user basic information include user's gender and age, and the basic information of thyroid gland includes Thyroid echogenicity, interior
Portion's composition, aspect ratio, calcification, Real time Organization elasticity imaging and cervical lymph node information, can in conjunction with the related content of following Fig. 4
Know, the empirical entropy of above-mentioned listed thyroid gland essential information is relatively large.
As a preferred embodiment of the present invention, thyroid gland disease is diagnosed to further increase thyroid nodule analysis method
The accuracy rate of disease, the thyroid gland essential information further include thyroid nodule size, boundary, shape, Color Doppler Flow letter
Number, thyroid gland coating information, sound swoon information and the past thyroid disease information, in conjunction with following Fig. 4 related content it is found that on
The empirical entropy for stating listed thyroid gland essential information is relatively small.
In embodiments of the present invention, the user can directly input the thyroid nodule analysis by the corresponding page
Parameter, for thyroid gland essential information options in the page presentation, for example, a kind of feasible thyroid nodule analysis
Parameter input interface please refers to Fig. 1 (a).
Step S102, the thyroid nodule analysis model for analyzing parameter and training generation according to the thyroid nodule are true
Determine the pernicious risk of user's thyroid nodule and exports.
In embodiments of the present invention, the thyroid nodule analysis model that the training generates be based on random forests algorithm or
Person is based on neural network algorithm training and generates.
In embodiments of the present invention, disease accurately can be diagnosed rapidly based on artificial intelligence study.
In embodiments of the present invention, neural network algorithm training is still either based on based on random forests algorithm and generates first
Shape gland tubercle analysis model requires to acquire a large amount of sample experimental data, and sample experimental data is abundanter, the first that training generates
Shape gland tubercle analysis model stability is stronger, accuracy rate is also higher.Wherein, thyroid nodule analysis provided in an embodiment of the present invention
In model, it is to be come from by 2064 of acquisition and south is big that training, which generates sample data used in thyroid nodule analysis model,
Learn the thyroid gland case analysis of experimental data acquisition of the first affiliated hospital.
In embodiments of the present invention, the thyroid nodule analysis method is close to the accuracy rate of diagnosis of thyroid illness
90%.
In embodiments of the present invention, training generates the flow diagram of the thyroid nodule analysis model based on random forest
Please refer to Fig. 2~Fig. 4 and its explanation.
Thyroid nodule analysis method provided in an embodiment of the present invention, the thyroid nodule by receiving user's input are analyzed
Parameter, thyroid nodule analysis parameter include user basic information and thyroid gland essential information, and first based on the received
The thyroid nodule analysis model that shape gland tubercle analysis parameter and training generate directly determines the pernicious wind of user's thyroid nodule
Danger simultaneously exports.Thyroid nodule analysis method provided in an embodiment of the present invention, end user inputs corresponding information can be direct
Diagnostic result is exported, diagnosis process is very convenient quick, and thyroid nodule analysis model used in diagnosis process is base
It is generated in random forests algorithm or based on neural network algorithm training, output is effectively guaranteed in stability with higher
The pernicious risk of thyroid nodule accuracy rate.
Fig. 1 (a) is that a kind of thyroid nodule provided in an embodiment of the present invention analyzes parameter input interface.
In embodiments of the present invention, parameter is analyzed for each thyroid nodule, interface has shown that corresponding options,
User directly clicks corresponding options according to demand.
Fig. 2 is a kind of trained thyroid nodule analysis model of the generation based on random forest provided in an embodiment of the present invention
The step flow chart of method, details are as follows.
Random forest refers to a kind of algorithm for being trained and predicting to sample using more decision trees, that is to say, that random
Forest algorithm is the algorithm comprising multiple decision trees.Each decision tree can once predict result,
And random forest is exactly based on multiple decision tree outputs, and votes the result of output, to efficiently solve single
The problem of decision tree over-fitting that may be present.
Step S201, obtain multiple thyroid nodule analyzing and training samples and with the thyroid nodule analyzing and training sample
This corresponding pernicious risk of thyroid nodule.
In embodiments of the present invention, multiple thyroid nodule analyzing and training samples of the acquisition are abovementioned steps S102
Thyroid gland case experimental data of 2064 by acquisition mentioned in explanation from the first affiliated hospital, Ji'nan University
What analysis obtained, further, general extraction 80% is used as training sample, and remaining 20% is used as test sample, the test
The use of sample generates the thyroid nodule analysis model based on random forest referring specifically to the another kind training shown in Fig. 3
Method flow diagram and its explanation.
It in embodiments of the present invention, include specific thyroid gland in the thyroid nodule analyzing and training sample of the acquisition
Tubercle analysis parameter further also obtains the pernicious risk of thyroid nodule corresponding with each training sample.
Step S202 generates multiple first according to the multiple thyroid nodule analyzing and training sample and according to default rule
Shape gland tubercle analyzing and training sample set.
As a possible embodiments of the invention, according to the first shape of the rule extraction preset quantity of preset random sampling
Gland tubercle analyzing and training sample repeats the above steps repeatedly, can be generated more as thyroid nodule analyzing and training sample set
A thyroid nodule analyzing and training sample set.
As a preferred embodiment of the present invention, found after adjusting model using different parameter value many experiments,
500 thyroid nodule analyzing and training samples are extracted by random sampling every time and generate a thyroid nodule analyzing and training sample
Set, repeats the above steps 3 times, generates 3 thyroid nodule analyzing and training sample sets and trains the thyroid nodule point come
It is best to analyse modelling effect.
Step S203, according to the thyroid nodule analyzing and training in the first thyroid nodule analyzing and training sample set
Sample training generates the first thyroid nodule analysis decision tree.
In embodiments of the present invention, one, which is determined, to be determined for each thyroid nodule analyzing and training sample set
Plan tree.
In embodiments of the present invention, thyroid nodule analysis decision is determined according to thyroid nodule analyzing and training sample set
The step flow chart of tree please refers to Fig. 4 and its explanation.
Step S204, the thyroid gland knot by multiple thyroid nodule analysis decision tree combination producings based on random forests algorithm
Save analysis model.
In embodiments of the present invention, it as a kind of most succinct model, directly combines the more decision trees and can give birth to
At the thyroid nodule analysis model based on random forests algorithm, i.e., when inputting one group of output valve, for each decision tree
There is an output as a result, using the most output result of frequency of occurrence as final output.
The embodiment of the invention provides a kind of trained methods for generating the thyroid nodule analysis model based on random forest
Flow chart by collecting sample data, and generates multiple thyroid nodule analyzing and training sample sets, for each thyroid gland
Tubercle analyzing and training sample set can determine unique decision tree based on decision Tree algorithms, by more decision tree groups of generation
It closes and produces the thyroid nodule analysis model based on random forests algorithm.
Fig. 3 is that the training that another embodiment of the present invention provides generates the thyroid nodule analysis model based on random forest
Method step flow chart, details are as follows.
In embodiments of the present invention, it is contemplated that the accuracy rate of each decision tree is different in random forest, therefore can pass through
Thyroid nodule analysis test sample calculates separately the predictablity rate of each decision tree, using accuracy rate as each decision tree
Weight, decision tree higher for predictablity rate, weight is higher, decision tree lower for predictablity rate, weight compared with
Low, the result for the thyroid nodule analysis model output that the training method training after optimization obtains is more accurate.
In embodiments of the present invention, with Fig. 2 shows it is a kind of it is trained generation based on random forest thyroid nodule analysis
The step flow chart of the method for model the difference is that, analyzed in the step S203 according to first thyroid nodule
After thyroid nodule analyzing and training sample training in training sample set generates the first thyroid nodule analysis decision tree, also wrap
It includes:
Step S301 obtains multiple thyroid nodule analysis test samples and analyzes test specimens with the thyroid nodule
This corresponding pernicious risk of thyroid nodule.
In embodiments of the present invention, the explanation for please referring to abovementioned steps S201, in 2064 case samples of acquisition
In, wherein 80% sample is used for training pattern, the sample of residue 20% is then used for test herein.
Step S302 analyzes test sample and the pernicious wind of corresponding thyroid nodule according to the multiple thyroid nodule
Danger calculates the accuracy rate of the first thyroid nodule analysis decision tree.
In embodiments of the present invention, the multiple thyroid nodule analysis test sample is output to the first thyroid nodule
Available thyroid nodule corresponding with the multiple thyroid nodule analysis test sample responds risk in analysis decision tree,
According to the multiple thyroid nodule corresponding response risk of analysis test sample and actually obtaining with the multiple first
The corresponding pernicious risk of thyroid nodule of shape gland tubercle analysis test sample, which compares, can obtain accuracy rate, i.e. response risk
It is consistent with actual conditions, then shows that the thyroid nodule analysis test sample is predicted successfully, if response risk and actual conditions
It is not consistent, then shows the thyroid nodule analysis test sample prediction of failure, analyzed according to the successful thyroid nodule of prediction
The thyroid nodule of test sample quantity and prediction of failure analysis test sample quantity can calculate accuracy rate.
Step S303 determines the first thyroid nodule point according to the accuracy rate of the first thyroid nodule analysis decision tree
Analyse the weight of decision tree.
In the embodiment of the present invention, thyroid nodule analysis decision tree higher for accuracy rate, determining weight is relatively
Height, thyroid nodule analysis decision tree lower for accuracy rate, determining weight are relatively low.
In embodiments of the present invention, the weight of determining thyroid nodule analysis decision tree is divided with corresponding thyroid nodule
It analyses decision tree cooperation and generates thyroid nodule analysis model.
As a possible embodiments of the invention, directly using the accuracy rate of the first thyroid nodule analysis decision tree as
The weight of first thyroid nodule analysis decision tree.
As another possible embodiments of the invention, according to the first thyroid nodule analysis decision tree and preset difference
Relationship between threshold value determines the weight of the first thyroid nodule analysis decision tree, such as accuracy rate is higher than to 90% first shape
The weight of gland tubercle analysis decision tree is denoted as 3.
The step S204 is specific by multiple thyroid nodule analysis decision tree combination producing thyroid nodule analysis models
Include:
Step S304, according to multiple thyroid nodule analysis decision trees and with the thyroid nodule analysis decision tree phase
The weight combination producing thyroid nodule analysis model answered.
In embodiments of the present invention, when inputting one group of output valve, for each decision tree have one output as a result,
The weight of each decision tree is considered simultaneously, is calculated the ratio of corresponding output result, will be exported the highest determination of success ratio
For final output.
Another training provided in an embodiment of the present invention generates the side of the thyroid nodule analysis model based on random forest
The step flow chart of method is assessed using each decision tree of the test sample to generation, calculates the accuracy rate of decision tree, and
The weight of the decision book is determined based on accuracy rate, decision tree higher for accuracy rate, weight is bigger, effectively further mentions
The stability of the high thyroid nodule analysis model based on random forest ultimately generated and the accuracy rate of output.
Fig. 4 is the step flow chart provided in an embodiment of the present invention for generating thyroid nodule analysis decision tree, and details are as follows.
In embodiments of the present invention, the main step that thyroid nodule analysis decision tree is generated based on decision Tree algorithms is provided
Rapid flow chart specifically asks embodiment listed in each step explanation.
Step S401 calculates the warp of each thyroid nodule analysis parameter in the first thyroid nodule analyzing and training sample set
Test entropy.
It in embodiments of the present invention, include multiple characteristic nodes in the thyroid nodule analysis decision tree, wherein each
A characteristic node is all that the thyroid nodule in aforementioned analyzes parameter.
In embodiments of the present invention, it is contemplated that different thyroid nodule analysis parameters are to thyroid nodule analytic process
" contribution " is different, i.e., influence factor is different, we utilize empirical entropy for indicating thyroid nodule analysis parameter to first here
The impact factor of shape gland tubercle analytic process, empirical entropy is bigger, then influences bigger, is preferentially considered as choosing.
In embodiments of the present invention, the calculation formula of the empirical entropy is as follows:
Wherein, pi、xi、yiIt can be regarded as probability, with 15 sample D1~D15For, output result includes two kinds of M and N,
Wherein 9 outputs are M, and 6 outputs are N, then for one of characteristic parameter A, it is assumed that characteristic parameter A value has three
Kind, including A1、A2And A3, and be A in value1X1In the output of a sample, there is M1A output is M, N1A output is N, is taken
Value is A2X2In a sample output, M2A output is M, N2A output is N, in the X that value is A33In a sample, M3A output
For M, N3A output is N.
The then empirical entropy of sample D:
For feature A, empirical entropy relative to sample D:
Step S402 will choose multiple thyroid nodules according to the empirical entropy and analyze parameter.
Step S403 analyzes parameter according to multiple thyroid nodules of the selection and determines that the analysis of the first thyroid nodule is determined
Plan tree.
As a kind of possible embodiments of the invention, after choosing first thyroid nodule analysis parameter, according to described
Thyroid nodule analyzes parameter and the first thyroid nodule analyzing and training sample set is divided into multiple thyroid nodules point
It analyses training sample subset to close, and calculates remaining training sample to each thyroid nodule analyzing and training sample set after segmentation is total
The empirical entropy of this set, repeats the above steps, until the empirical entropy of remaining thyroid nodule analysis parameter is less than preset threshold value.
At this point, the thyroid nodule analysis parameter combination of selection can determine thyroid nodule analysis decision tree.
As a kind of possible embodiments of the invention, with X sample and A, B, C, for five characteristic parameters of D, E, point
Not Ji Suan A, B, C, X sample is then divided into set X according to B wherein assuming that B is maximum by the empirical entropy of D, EB1And XB2, and after
Continuous to calculate A, C, D, tetra- characteristic parameters of E are in set XB1And XB2Empirical entropy, and according to the maximum feature of the empirical entropy of selection
Parameter continues to set XB1And XB2Continue to divide, until the empirical entropy of residue character parameter does not surpass in some set
When crossing preset threshold value, then the set, which divides, completes, and when all set, which divide, to be completed, can determine corresponding decision at this time
Tree.
Fig. 5 is that a kind of trained thyroid nodule of the generation based on neural network algorithm provided in an embodiment of the present invention analyzes mould
The step flow chart of the method for type, details are as follows.
In embodiments of the present invention, different from random forests algorithm, neural network algorithm be a kind of learning ability relatively more
Add outstanding artificial intelligence learning method, the effect of the thyroid nodule analysis model generated is trained more by neural network algorithm
Add excellent, but process time-consuming trained accordingly is longer.
Step S501, obtain multiple thyroid nodule analyzing and training samples and with the thyroid nodule analyzing and training sample
This corresponding thyroid nodule target risk.
In embodiments of the present invention, the thyroid nodule analyzing and training sample of the acquisition is identical as abovementioned steps S201.
Step S502 contains the initialization thyroid nodule analysis model of variable element based on Establishment of Neural Model.
In embodiments of the present invention, the formula of the thyroid nodule analysis model is as follows:
R1=σ (T*W1+b1)
In embodiments of the present invention, the T is each thyroid nodule analysis in multiple thyroid nodule analyzing and training samples
The matrix-vector that parameter is constituted, W1For the first weight matrix coefficient, b1For the first bias vector, wherein W1With b1For variable element
As a preferred embodiment of the present invention, the thyroid nodule analysis model includes multilayer, that is, be will acquire
R1T as next layer is inputted.
In embodiments of the present invention, the σ indicates Sigmoid function (S sigmoid growth curve), specific function expression are as follows:
Step S503, according to the first thyroid nodule analyzing and training sample and the thyroid nodule analysis model
Determine that thyroid nodule corresponding with the first thyroid nodule analyzing and training sample responds risk.
In embodiments of the present invention, ginseng is analyzed according to each thyroid nodule in the first thyroid nodule analyzing and training sample
Number generates corresponding matrix-vector, and inputs above-mentioned shown thyroid nodule analysis model for the matrix-vector as T
In, can obtain final output matrix vector R, the R is thyroid nodule response risk Metrics vector, wherein every in R
The accordingly result of the corresponding thyroid nodule analyzing and training sample of one numerical value.
Step S504 responds risk according to thyroid nodule corresponding with the first thyroid nodule analyzing and training sample
And thyroid nodule target risk determines first-loss value.
In embodiments of the present invention, it is assumed that thyroid nodule corresponding with the first thyroid nodule analyzing and training sample
The matrix-vector of target risk is R ', and the matrix-vector that thyroid nodule responds risk is R, wherein R, corresponding in R ' matrix-vector
The numerical value of position be respectively y andThe then calculation formula of first-loss value:
Step S505, judges whether the penalty values of multiple thyroid nodule analyzing and training samples meet preset condition.
In embodiments of the present invention, by penalty values can training of judgement generate thyroid nodule analysis model whether
It meets the requirements.
It is determined as a possible embodiments of the invention by judging the size relation of penalty values Yu preset threshold value
Whether the thyroid nodule analysis model meets the requirements, and when penalty values are less than preset threshold value, shows thyroid nodule point
Analysing model, propaedeutics is completed, when penalty values are greater than preset threshold value, shows that thyroid nodule analysis model is not yet trained
It completes.
Step S506 adjusts the variable element in the thyroid nodule analysis model based on back-propagation algorithm, and returns
It is back to the step S503.
In embodiments of the present invention, when the penalty values are unsatisfactory for preset condition, show the thyroid nodule point
It is imperfect in analysis model, it can be adjusted according to penalty values by back-propagation algorithm variable in thyroid nodule analysis model
Parameter, including weight matrix coefficient and bias vector.
Current thyroid nodule analysis model is determined as described generating based on neural network algorithm training by step S507
Thyroid nodule analysis model.
In embodiments of the present invention, when the penalty values meet preset condition, show thyroid nodule analysis at this time
It is perfect in model, input value can be made and accordingly export corresponding result.
The thyroid nodule analysis model based on neural network algorithm that training provided in an embodiment of the present invention generates, is compared
In Fig. 2 shows the thyroid nodule analysis model based on random forest, can effectively improve thyroid nodule analysis method
Accuracy rate.
Fig. 6 is a kind of structural schematic diagram of thyroid nodule analytical equipment provided in an embodiment of the present invention, for the ease of saying
It is bright, part related to the embodiment of the present invention is only shown.
In embodiments of the present invention, the thyroid nodule analytical equipment includes thyroid nodule analysis parameter receiving unit
601 and the pernicious risk output unit 602 of thyroid nodule.
The thyroid nodule analyzes parameter receiving unit 601, for receiving the thyroid nodule analysis ginseng of user's input
Number.
In embodiments of the present invention, the thyroid nodule analysis parameter includes that user basic information and thyroid gland are basic
Information, the user basic information include user's gender and age, and the basic information of thyroid gland includes Thyroid echogenicity, interior
Portion's composition, aspect ratio, calcification, Real time Organization elasticity imaging and cervical lymph node information, can in conjunction with the related content of following Fig. 4
Know, the empirical entropy of above-mentioned listed thyroid gland essential information is relatively large.
As a preferred embodiment of the present invention, thyroid gland disease is diagnosed to further increase thyroid nodule analysis method
The accuracy rate of disease, the thyroid gland essential information further include thyroid nodule size, boundary, shape, Color Doppler Flow letter
Number, thyroid gland coating information, sound swoon information and the past thyroid disease information, in conjunction with following Fig. 4 related content it is found that on
The empirical entropy for stating listed thyroid gland essential information is relatively small.
The pernicious risk output unit 602 of thyroid nodule, for according to the thyroid nodule analyze parameter and
The thyroid nodule analysis model that training generates determines the pernicious risk of user's thyroid nodule and exports.
In embodiments of the present invention, the thyroid nodule analysis model that the training generates be based on random forests algorithm or
Person is based on neural network algorithm training and generates.
In embodiments of the present invention, disease accurately can be diagnosed rapidly based on artificial intelligence study.
In embodiments of the present invention, neural network algorithm training is still either based on based on random forests algorithm and generates first
Shape gland tubercle analysis model requires to acquire a large amount of sample experimental data, and sample experimental data is abundanter, the first that training generates
Shape gland tubercle analysis model stability is stronger, accuracy rate is also higher.Wherein, thyroid nodule analysis provided in an embodiment of the present invention
In model, it is to be come from by 2064 of acquisition and south is big that training, which generates sample data used in thyroid nodule analysis model,
Learn the thyroid gland case analysis of experimental data acquisition of the first affiliated hospital.
In embodiments of the present invention, the thyroid nodule analysis method is close to the accuracy rate of diagnosis of thyroid illness
90%.
Thyroid nodule analytical equipment provided in an embodiment of the present invention, the thyroid nodule by receiving user's input are analyzed
Parameter, thyroid nodule analysis parameter include user basic information and thyroid gland essential information, and first based on the received
The thyroid nodule analysis model that shape gland tubercle analysis parameter and training generate directly determines the pernicious wind of user's thyroid nodule
Danger simultaneously exports.Thyroid nodule analysis method provided in an embodiment of the present invention, end user inputs corresponding information can be direct
Diagnostic result is exported, diagnosis process is very convenient quick, and thyroid nodule analysis model used in diagnosis process is base
It is generated in random forests algorithm or based on neural network algorithm training, output is effectively guaranteed in stability with higher
The pernicious risk of thyroid nodule accuracy rate.
Fig. 7 is the structural schematic diagram for the thyroid nodule analytical equipment that another embodiment of the present invention provides, for the ease of
Illustrate, part related to the embodiment of the present invention is only shown.
In embodiments of the present invention, with a kind of thyroid nodule analytical equipment shown in Fig. 6 the difference is that, it is described
Thyroid nodule analytical equipment further includes random forests algorithm training unit 701, generates for training and is based on random forests algorithm
Thyroid nodule analysis model, wherein the random forests algorithm training unit 701 include training sample obtain module 701a,
Training sample set closes generation module 701b, thyroid nodule analysis decision tree generation module 701c and thyroid nodule and analyzes mould
Type generation module 701d.
The training sample obtains module 701a, for obtain multiple thyroid nodule analyzing and training samples and with it is described
The pernicious risk of the corresponding thyroid nodule of thyroid nodule analyzing and training sample.
In embodiments of the present invention, what multiple thyroid nodule analyzing and training samples of the acquisition were as previously mentioned is logical
2064 thyroid gland case analysis of experimental data acquisitions from the first affiliated hospital, Ji'nan University of acquisition are crossed, further
, general extraction 80% is used as training sample, and remaining 20% is used as test sample, and the use of the test sample please specifically join
Read method flow diagram and its explanation that the another kind training shown in Fig. 3 generates the thyroid nodule analysis model based on random forest
Explanation.
It in embodiments of the present invention, include specific thyroid gland in the thyroid nodule analyzing and training sample of the acquisition
Tubercle analysis parameter further also obtains the pernicious risk of thyroid nodule corresponding with each training sample.
The training sample set closes generation module 701b, is used for according to the multiple thyroid nodule analyzing and training sample simultaneously
Multiple thyroid nodule analyzing and training sample sets are generated according to default rule.
As a possible embodiments of the invention, according to the first shape of the rule extraction preset quantity of preset random sampling
Gland tubercle analyzing and training sample repeats the above steps repeatedly, can be generated more as thyroid nodule analyzing and training sample set
A thyroid nodule analyzing and training sample set.
As a preferred embodiment of the present invention, found after adjusting model using different parameter value many experiments,
500 thyroid nodule analyzing and training samples are extracted by random sampling every time and generate a thyroid nodule analyzing and training sample
Set, repeats the above steps 3 times, generates 3 thyroid nodule analyzing and training sample sets and trains the thyroid nodule point come
It is best to analyse modelling effect.
The thyroid nodule analysis decision tree generation module 701c is instructed for being analyzed according to first thyroid nodule
The thyroid nodule analyzing and training sample training practiced in sample set generates the first thyroid nodule analysis decision tree.
In embodiments of the present invention, one, which is determined, to be determined for each thyroid nodule analyzing and training sample set
Plan tree.
In embodiments of the present invention, thyroid nodule analysis decision is determined according to thyroid nodule analyzing and training sample set
The step flow chart of tree please refers to Fig. 4 and its explanation.
The thyroid nodule analysis model generation module 701d, by multiple thyroid nodule analysis decision tree combination producings
Thyroid nodule analysis model based on random forests algorithm.
In embodiments of the present invention, it as a kind of most succinct model, directly combines the more decision trees and can give birth to
At the thyroid nodule analysis model based on random forests algorithm, i.e., when inputting one group of output valve, for each decision tree
There is an output as a result, using the most output result of frequency of occurrence as final output.
Fig. 8 is the structural schematic diagram for the random forests algorithm training unit that another embodiment of the present invention provides, in order to just
In explanation, part related to the embodiment of the present invention is only shown.
In embodiments of the present invention, it is contemplated that the accuracy rate of each decision tree is different in random forest, therefore can pass through
Thyroid nodule analysis test sample calculates separately the predictablity rate of each decision tree, using accuracy rate as each decision tree
Weight, decision tree higher for predictablity rate, weight is higher, decision tree lower for predictablity rate, weight compared with
Low, the result for the thyroid nodule analysis model output that the random forests algorithm training unit training after optimization obtains is more quasi-
Really.
In embodiments of the present invention, it is different from the structural schematic diagram of the random forests algorithm training unit shown in Fig. 7 it
It is in further including that test sample obtains module 801, accuracy rate determining module 802 and weight determination module 803, the first
Shape gland tubercle analysis model generation module 701d is specially thyroid nodule analysis model generation module 804.
The test sample obtains module 801, for obtain multiple thyroid nodules analysis test samples and with it is described
Thyroid nodule analyzes the corresponding pernicious risk of thyroid nodule of test sample.
In embodiments of the present invention, the explanation for please referring to abovementioned steps S201, in 2064 case samples of acquisition
In, wherein 80% sample is used for training pattern, the sample of residue 20% is then used for test herein.
The accuracy rate determining module 802, for analyzing test sample and correspondence according to the multiple thyroid nodule
Pernicious the first thyroid nodule of the Risk Calculation analysis decision tree of thyroid nodule accuracy rate.
In embodiments of the present invention, the multiple thyroid nodule analysis test sample is output to the first thyroid nodule
Available thyroid nodule corresponding with the multiple thyroid nodule analysis test sample responds risk in analysis decision tree,
According to the multiple thyroid nodule corresponding response risk of analysis test sample and actually obtaining with the multiple first
The corresponding pernicious risk of thyroid nodule of shape gland tubercle analysis test sample, which compares, can obtain accuracy rate, i.e. response risk
It is consistent with actual conditions, then shows that the thyroid nodule analysis test sample is predicted successfully, if response risk and actual conditions
It is not consistent, then shows the thyroid nodule analysis test sample prediction of failure, analyzed according to the successful thyroid nodule of prediction
The thyroid nodule of test sample quantity and prediction of failure analysis test sample quantity can calculate accuracy rate.
The weight determination module 803, for being determined according to the accuracy rate of the first thyroid nodule analysis decision tree
The weight of first thyroid nodule analysis decision tree.
In the embodiment of the present invention, thyroid nodule analysis decision tree higher for accuracy rate, determining weight is relatively
Height, thyroid nodule analysis decision tree lower for accuracy rate, determining weight are relatively low.
In embodiments of the present invention, the weight of determining thyroid nodule analysis decision tree is divided with corresponding thyroid nodule
It analyses decision tree cooperation and generates thyroid nodule analysis model.
As a possible embodiments of the invention, directly using the accuracy rate of the first thyroid nodule analysis decision tree as
The weight of first thyroid nodule analysis decision tree.
As another possible embodiments of the invention, according to the first thyroid nodule analysis decision tree and preset difference
Relationship between threshold value determines the weight of the first thyroid nodule analysis decision tree, such as accuracy rate is higher than to 90% first shape
The weight of gland tubercle analysis decision tree is denoted as 3.
The thyroid nodule analysis model generation module 804, for according to multiple thyroid nodule analysis decision trees with
And weight combination producing thyroid nodule analysis model corresponding with the thyroid nodule analysis decision tree.
In embodiments of the present invention, when inputting one group of output valve, for each decision tree have one output as a result,
The weight of each decision tree is considered simultaneously, is calculated the ratio of corresponding output result, will be exported the highest determination of success ratio
For final output.
The structure of another kind random forests algorithm training unit provided in an embodiment of the present invention, by utilizing test sample pair
The each decision tree generated is assessed, and calculates the accuracy rate of decision tree, and the power of the decision book is determined based on accuracy rate
Weight, decision tree higher for accuracy rate, weight is bigger, effectively further improve ultimately generate based on random forest
The stability of thyroid nodule analysis model and the accuracy rate of output.
The embodiment of the present invention provides a kind of computer equipment, which includes memory and processor, described to deposit
Computer program is stored in reservoir, when the computer program is executed by the processor, so that the processor executes such as
Figure 1 above in 5 it is any it is diagrammatically shown go out thyroid nodule analysis method the step of.
Illustratively, computer program can be divided into one or more modules, one or more module is stored
In memory, and by processor it executes, to complete the present invention.One or more modules, which can be, can complete specific function
Series of computation machine program instruction section, the instruction segment is for describing implementation procedure of the computer program in computer installation.Example
Such as, computer program can be divided into the step of thyroid nodule analysis method that above-mentioned each embodiment of the method provides.
It will be understood by those skilled in the art that the description of above-mentioned computer installation is only example, do not constitute to calculating
The restriction of machine device may include component more more or fewer than foregoing description, perhaps combine certain components or different portions
Part, such as may include input-output equipment, network access equipment, bus etc..
Alleged processor can be central processing unit (Central Processing Unit, CPU), can also be it
His general processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng the processor is the control centre of the computer installation, utilizes various interfaces and the entire computer installation of connection
Various pieces.
The memory can be used for storing the computer program and/or module, and the processor is by operation or executes
Computer program in the memory and/or module are stored, and calls the data being stored in memory, described in realization
The various functions of computer installation.The memory can mainly include storing program area and storage data area, wherein storage program
It area can application program (such as sound-playing function, image player function etc.) needed for storage program area, at least one function
Deng;Storage data area, which can be stored, uses created data (such as audio data, phone directory etc.) etc. according to mobile phone.In addition,
Memory may include high-speed random access memory, can also include nonvolatile memory, such as hard disk, memory, grafting
Formula hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card
(Flash Card), at least one disk memory, flush memory device or other volatile solid-state parts.
If the integrated module/unit of the computer installation is realized in the form of SFU software functional unit and as independent
Product when selling or using, can store in a computer readable storage medium.Based on this understanding, the present invention is real
All or part of the process in existing above-described embodiment method, can also instruct relevant hardware come complete by computer program
At the computer program can be stored in a computer readable storage medium, which is being executed by processor
When, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program code, described
Computer program code can be source code form, object identification code form, executable file or certain intermediate forms etc..The meter
Calculation machine readable medium may include: can carry the computer program code any entity or device, recording medium, USB flash disk,
Mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), random access memory
Device (RAM, Random Access Memory), electric carrier signal, electric signal and software distribution medium etc..
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (10)
1. a kind of thyroid nodule analysis method, which is characterized in that the described method includes:
The thyroid nodule for receiving user's input analyzes parameter, thyroid nodule analysis parameter include user basic information with
And thyroid gland essential information, the basic information of thyroid gland include Thyroid echogenicity, internal component, aspect ratio, calcification, real-time bullet
Property imaging and cervical lymph node state etc.;
User's thyroid gland is determined according to the thyroid nodule analysis model that the thyroid nodule analyzes parameter and training generation
The pernicious risk of tubercle simultaneously exports, and the thyroid nodule analysis model that the training generates is based on random forests algorithm or to be based on
Neural network algorithm training generates.
2. thyroid nodule analysis method according to claim 1, which is characterized in that the thyroid nodule analysis model
It is to be generated based on random forests algorithm training, trains the step of generating the thyroid nodule analysis model, specifically include:
Obtain multiple thyroid nodule analyzing and training samples and first shape corresponding with the thyroid nodule analyzing and training sample
The pernicious risk of gland tubercle;
Multiple thyroid nodules analyses are generated according to the multiple thyroid nodule analyzing and training sample and according to default rule
Training sample set includes that the thyroid nodule of preset quantity analyzes instruction in the thyroid nodule analyzing and training sample set
Practice sample;
It is generated according to the thyroid nodule analyzing and training sample training in the first thyroid nodule analyzing and training sample set
First thyroid nodule analysis decision tree;
Thyroid nodule analysis model by multiple thyroid nodule analysis decision tree combination producings based on random forests algorithm.
3. thyroid nodule analysis method according to claim 2, which is characterized in that described according to the first first shape
Thyroid nodule analyzing and training sample training in gland tubercle analyzing and training sample set generates the analysis of the first thyroid nodule and determines
After the step of plan tree, further includes:
Obtain multiple thyroid nodule analysis test samples and first shape corresponding with thyroid nodule analysis test sample
The pernicious risk of gland tubercle;
Test sample and pernicious the first first of Risk Calculation of corresponding thyroid nodule are analyzed according to the multiple thyroid nodule
The accuracy rate of shape gland tubercle analysis decision tree;
The power of the first thyroid nodule analysis decision tree is determined according to the accuracy rate of the first thyroid nodule analysis decision tree
Weight;
It is described by multiple thyroid nodule analysis decision tree combination producing thyroid nodule analysis models the step of specifically:
It is combined according to multiple thyroid nodule analysis decision trees and with the corresponding weight of thyroid nodule analysis decision tree
Generate thyroid nodule analysis model.
4. thyroid nodule analysis method according to claim 2, which is characterized in that described according to first thyroid gland
Thyroid nodule analyzing and training sample training in tubercle analyzing and training sample set generates the first thyroid nodule analysis decision
The step of tree, specifically includes:
Calculate the empirical entropy of each thyroid nodule analysis parameter in the first thyroid nodule analyzing and training sample set, the experience
Entropy is used to indicate thyroid nodule analysis parameter to the impact factor of thyroid nodule analytic process;
Multiple thyroid nodule analysis parameters will be chosen according to the empirical entropy;
Parameter, which is analyzed, according to multiple thyroid nodules of the selection determines the first thyroid nodule analysis decision tree.
5. thyroid nodule analysis method according to claim 1, which is characterized in that the thyroid nodule analysis model
It is to be generated based on neural network algorithm training, trains the step of generating the thyroid nodule analysis model, specifically include:
Obtain multiple thyroid nodule analyzing and training samples and first shape corresponding with the thyroid nodule analyzing and training sample
The pernicious risk of gland tubercle;
Contain the initialization thyroid nodule analysis model of variable element based on Establishment of Neural Model;
It is determining with described the according to the first thyroid nodule analyzing and training sample and the thyroid nodule analysis model
The corresponding thyroid nodule of one thyroid nodule analyzing and training sample responds risk;
Risk and thyroid gland knot are responded according to thyroid nodule corresponding with the first thyroid nodule analyzing and training sample
It saves pernicious risk and determines first-loss value;
Judge whether the penalty values of multiple thyroid nodule analyzing and training samples meet preset condition;
When judging that the penalty values are unsatisfactory for preset condition, the thyroid nodule analysis is adjusted based on back-propagation algorithm
Variable element in model, and be back to described according to the first thyroid nodule analyzing and training sample and the thyroid gland
Tubercle analysis model determines the step of thyroid nodule response risk corresponding with the first thyroid nodule analyzing and training sample
Suddenly;
When judging that the penalty values meet preset condition, current thyroid nodule analysis model is determined as described based on mind
The thyroid nodule analysis model generated through network algorithm training.
6. a kind of thyroid nodule analytical equipment characterized by comprising
Thyroid nodule analyzes parameter receiving unit, and the thyroid nodule for receiving user's input analyzes parameter, the first shape
It includes user basic information and thyroid gland essential information that gland tubercle, which analyzes parameter, and the basic information of thyroid gland includes thyroid gland
Echo, internal component, aspect ratio, calcification, Real time Organization elastogram and cervical lymph node state etc.;
The pernicious risk output unit of thyroid nodule, the first for analyzing parameter according to the thyroid nodule and training generates
Shape gland tubercle analysis model determines the pernicious risk of user's thyroid nodule and exports that the thyroid nodule that the training generates is analyzed
Model is generated based on random forests algorithm or based on neural network algorithm training.
7. thyroid nodule analytical equipment according to claim 5, which is characterized in that further include random forests algorithm training
Unit generates the thyroid nodule analysis model based on random forests algorithm for training, and the random forests algorithm training is single
Member includes:
Training sample obtains module, for obtaining multiple thyroid nodule analyzing and training samples and dividing with the thyroid nodule
Analyse the corresponding pernicious risk of thyroid nodule of training sample;
Training sample set closes generation module, for according to the multiple thyroid nodule analyzing and training sample and according to preset rule
Multiple thyroid nodule analyzing and training sample sets are then generated, include default in the thyroid nodule analyzing and training sample set
Quantity thyroid nodule analyzing and training sample;
Thyroid nodule analysis decision tree generation module, for according in the first thyroid nodule analyzing and training sample set
Thyroid nodule analyzing and training sample training generate the first thyroid nodule analysis decision tree;
Multiple thyroid nodule analysis decision tree combination producings are based on random forest by thyroid nodule analysis model generation module
The thyroid nodule analysis model of algorithm.
8. device according to claim 7, which is characterized in that the random forests algorithm training unit further include:
Test sample obtains module, for obtaining multiple thyroid nodule analysis test samples and dividing with the thyroid nodule
Analyse the corresponding pernicious risk of thyroid nodule of test sample;
Accuracy rate determining module, for analyzing test sample and corresponding thyroid nodule according to the multiple thyroid nodule
The accuracy rate of pernicious the first thyroid nodule of Risk Calculation analysis decision tree;
Weight determination module, for determining the first thyroid gland knot according to the accuracy rate of the first thyroid nodule analysis decision tree
Save the weight of analysis decision tree;
The thyroid nodule analysis model generation module, for according to multiple thyroid nodule analysis decision trees and with it is described
The corresponding weight combination producing thyroid nodule analysis model of thyroid nodule analysis decision tree.
9. a kind of computer equipment, which is characterized in that including memory and processor, computer journey is stored in the memory
Sequence, when the computer program is executed by the processor, so that the processor perform claim requires any one of 1 to 5 power
Benefit requires the step of thyroid nodule analysis method.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program, when the computer program is executed by processor, so that the processor perform claim requires any one of 1 to 5 right
It is required that the step of described thyroid nodule analysis method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129439.7A CN109830300A (en) | 2019-02-21 | 2019-02-21 | Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129439.7A CN109830300A (en) | 2019-02-21 | 2019-02-21 | Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109830300A true CN109830300A (en) | 2019-05-31 |
Family
ID=66864090
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910129439.7A Pending CN109830300A (en) | 2019-02-21 | 2019-02-21 | Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109830300A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113053495A (en) * | 2019-12-27 | 2021-06-29 | 天津幸福生命科技有限公司 | Evaluation method and device for image information |
CN113450325A (en) * | 2021-06-28 | 2021-09-28 | 什维新智医疗科技(上海)有限公司 | Thyroid nodule benign and malignant recognition device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105550583A (en) * | 2015-12-22 | 2016-05-04 | 电子科技大学 | Random forest classification method based detection method for malicious application in Android platform |
CN109087703A (en) * | 2018-08-24 | 2018-12-25 | 南京大学 | Abdominal cavity CT image peritonaeum metastatic marker method based on depth convolutional neural networks |
-
2019
- 2019-02-21 CN CN201910129439.7A patent/CN109830300A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105550583A (en) * | 2015-12-22 | 2016-05-04 | 电子科技大学 | Random forest classification method based detection method for malicious application in Android platform |
CN109087703A (en) * | 2018-08-24 | 2018-12-25 | 南京大学 | Abdominal cavity CT image peritonaeum metastatic marker method based on depth convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
郑冰: "基于生物信息学的复杂疾病诊断模型研究", 《万方数据》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113053495A (en) * | 2019-12-27 | 2021-06-29 | 天津幸福生命科技有限公司 | Evaluation method and device for image information |
CN113450325A (en) * | 2021-06-28 | 2021-09-28 | 什维新智医疗科技(上海)有限公司 | Thyroid nodule benign and malignant recognition device |
CN113450325B (en) * | 2021-06-28 | 2022-09-09 | 什维新智医疗科技(上海)有限公司 | Thyroid nodule benign and malignant recognition device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11023689B1 (en) | Applied artificial intelligence technology for narrative generation using an invocable analysis service with analysis libraries | |
CN110335259B (en) | Medical image identification method and device and storage medium | |
Shi et al. | Represent, compare, and learn: A similarity-aware framework for class-agnostic counting | |
Le et al. | Interactive boundary prediction for object selection | |
CN105701120B (en) | The method and apparatus for determining semantic matching degree | |
CN109886388A (en) | A kind of training sample data extending method and device based on variation self-encoding encoder | |
CN109543701A (en) | Vision significance method for detecting area and device | |
CN106503863A (en) | Based on the Forecasting Methodology of the age characteristicss of decision-tree model, system and terminal | |
Luo et al. | SFA: small faces attention face detector | |
CN110503155A (en) | A kind of method and relevant apparatus, server of information classification | |
KR20210113336A (en) | Method for predicting target object properties based on machine learning and related devices | |
CN111476192A (en) | Intercepted image synthesis method based on intelligent traffic and big data cloud server | |
CN112215696A (en) | Personal credit evaluation and interpretation method, device, equipment and storage medium based on time sequence attribution analysis | |
CN109830300A (en) | Thyroid nodule analysis method, device, computer equipment and readable storage medium storing program for executing | |
KR20220059287A (en) | Attention-based stacking method for time series forecasting | |
CN113554657A (en) | Super-pixel segmentation method and system based on attention mechanism and convolutional neural network | |
Guo et al. | Leafmask: Towards greater accuracy on leaf segmentation | |
CN109919296A (en) | A kind of deep neural network training method, device and computer equipment | |
Kalash et al. | Relative saliency and ranking: Models, metrics, data and benchmarks | |
CN110490058A (en) | Training method, device, system and the computer-readable medium of pedestrian detection model | |
CN113158084B (en) | Method, device, computer equipment and storage medium for processing movement track data | |
CN109711315A (en) | A kind of method and device of Lung neoplasm analysis | |
CN113486925A (en) | Model training method, fundus image generation method, model evaluation method and device | |
CN110276802B (en) | Method, device and equipment for positioning pathological tissue in medical image | |
CN113096080A (en) | Image analysis method and system |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190531 |