CN103914841A - Bacterium division and classification method based on superpixels and in-depth learning and application thereof - Google Patents
Bacterium division and classification method based on superpixels and in-depth learning and application thereof Download PDFInfo
- Publication number
- CN103914841A CN103914841A CN201410133805.3A CN201410133805A CN103914841A CN 103914841 A CN103914841 A CN 103914841A CN 201410133805 A CN201410133805 A CN 201410133805A CN 103914841 A CN103914841 A CN 103914841A
- Authority
- CN
- China
- Prior art keywords
- bacterium
- degree
- super pixel
- cut apart
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The invention provides a bacterium division and classification method based on superpixels and in-depth learning. The method comprises the steps that an in-depth learning classifier is established through a training image set, a tested image is compared with the in-depth learning classifier, and a classification result is output; the comparison comprises the steps that the tested image passes through all GRB channels to be subjected to median filtering, and noise is removed; the superpixels are calculated; the color, shape and dimension characteristics of each superpixel region are calculated; primary filtering is conducted on each superpixel region through priori knowledge, and each superpixel region is divided, so that candidate bacterium regions are determined; feature extraction is conducted on the candidate bacterium regions; extracted features are compared with classification standards in the classifier, and classification is completed according to similarity. The invention further provides application of the method in diagnosis of vagina bacteria. According to the bacterium division and classification method, division is conducted through calculation of the superpixels, classification is conducted through in-depth learning, and therefore the bacterium division and classification method has the advantages that the identification degree is high, cost is low, implementation is simple, and popularization is easy.
Description
Technical field
The present invention relates to a kind of image and cut apart and sorting technique, relate in particular to a kind of bacterium based on super pixel and degree of depth study and cut apart and sorting technique and the application in vaginal bacteria diagnosis thereof.
Background technology
Be modal disease among women by bacterial vaginal disease or infection (vaginal candida/trichomonad, gonococcus, Chlamydia), also cause very high mortality ratio.The classification of carrying out vaginal disease by bacteria types and counting has very important effect in clinical practice.At present, be widely used and prevent that the diagnostic method of vaginal disease from being cytology examination, this kind of method depends on clinician's experience to a great extent, obtains diagnostic result accurately with this.In order to address this problem, the automatic diagnostics of design and development prevents vaginal disease, and inflammation and cancer have attracted a lot of people's interest, and becomes a focus.
Previously there is a large amount of research work to cut apart nucleus, but only utilized the segmentation performance of nucleus information approach still can not be satisfactory.Nuclear dividing method is broadly divided into multiple nucleus segmentation thresholds to be chosen, Hough conversion and watershed method.But these methods are the nucleus based on normal condition, instead of there are disease or abnormal nucleus.Recently, also having to cut apart has pathology and normal cell, and the report of cutting apart of cell mixing.
As Yeoman, C.J., Thomas, S.M., Miller, M.E.B., Ulanov, A.V., Torralba, M., Lucas, S., Gillis, M., Cregger, M., Gomez, A., Ho, M., Leigh, S.R., Stumpf, R., Creedon, D.J., Smith, M.A., Weisbaum, J.S., Nelson, K.E., Wilson, B.A., White, B.A.:A Multi-Omic Systems-Based Approach Reveals Metabolic Markers of Bacterial Vaginosis and Insight into the Disease.PLoS ONE8 (2), e56111 (2013) [1].
The work that cell splits from image, current a lot of algorithms are all based on nucleus independently, and to various kinds of cell to cut apart research less.Cut apart and can be divided into parametric filtering based on shape and marginal information nucleus, movable contour model and difference maximize.In addition, cut apart work and also obtained because merging previous knowledge and local feature the performance showing.
Therefore, how to design a kind of bacterium based on super pixel and degree of depth study and cut apart and sorting technique and application thereof, become new R&D direction.
Summary of the invention
The present invention is directed to this problem, provide a kind of bacterium based on super pixel and degree of depth study to cut apart and sorting technique and application thereof.
Bacterium based on super pixel and degree of depth study provided by the invention is cut apart and sorting technique, comprises by training plan image set and sets up degree of deep learning classification device, and test pattern and degree of deep learning classification device are contrasted to the result of output category; The process of described contrast comprises: described test pattern is done to medium filtering through each GRB passage, and remove noise; Calculate super pixel; Calculate color, shape, the size characteristic of each super pixel region; Utilize priori to carry out preliminary filtering to each super pixel region, and cut apart to determine candidate bacterium region; Feature extraction is carried out in described candidate bacterium region; Criteria for classification in the feature of extraction and described sorter is contrasted, complete classification according to similarity.
Preferably, the described step of setting up degree of deep learning classification device comprises: initialization convolutional neural networks; Receive the image pattern from described training plan image set; Actual output is calculated in forward transmission; Calculate the poor of actual output and target output; Utilize the backpropagation of minimization error approach to adjust weights; Judge whether the frequency of training that reaches predetermined; If do not reached, continue and receive above-mentioned image pattern, and adjust weights; If reached, output category standard, and deposit in described degree of deep learning classification device.
Preferably, the super pixel of described calculating is the algorithm that adopts linear iteration cluster.
Vaginal bacteria based on super pixel and degree of depth study provided by the invention is cut apart and categorizing system, comprising: training plan image set, for storing the sample image of vaginal bacteria; Degree of deep learning classification device, for storing the criteria for classification arriving based on described training plan image set gained; Pretreatment module, for test pattern is carried out to pre-service, calculate super pixel and calculate color, shape, the size characteristic of each super pixel region, utilize priori to carry out preliminary filtering to each super pixel region, and cut apart to determine candidate bacterium region; Characteristic extracting module, for carrying out feature extraction to described candidate bacterium region; Judge module, for the criteria for classification of the feature of extraction and described degree of deep learning classification device is contrasted, completes classification according to similarity.
Preferably, described degree of deep learning classification device also comprises: initialization module, for initialization convolutional neural networks; Receiver module, for receiving the image pattern from described training plan image set; Adjusting module, for calculating actual output to transmitting, calculating actual output with the poor of target output and utilize the backpropagation of minimization error approach to adjust weights; Counting module, for the frequency of training that judges whether to reach predetermined, and after reaching, notice adjusting module output category standard.
Preferably, the super pixel of described calculating is the algorithm that adopts linear iteration cluster.
Bacterium based on super pixel and degree of depth study of the present invention is cut apart and sorting technique and application thereof, is calculated and is cut apart and learn to classify by the degree of depth by super pixel, has advantages of high identification, low cost, the simple and easy popularization of enforcement.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of setting up the method for degree of deep learning classification device in the present invention.
Fig. 2 is that the bacterium that the present invention is based on the study of super pixel and the degree of depth is cut apart and the schematic flow sheet of sorting technique.
Fig. 3 is that the bacterium that the present invention is based on the study of super pixel and the degree of depth is cut apart and the structural representation of categorizing system.
Fig. 4 is the schematic diagram of medial vagina common bacteria of the present invention.
Fig. 5 is calculating and the effect schematic diagram of cutting apart of carrying out and feature extraction based on super pixel in the present invention.
Fig. 6 is typical vagina image and the result to vaginal bacteria segmentation and classification in the present invention.
Fig. 7 is the exemplary plot of segmentation effect in the present invention.
Fig. 8 is the exemplary plot of cutting apart comparative result in the present invention.
Fig. 9 is the classification results comparative example figure of different sorters in the present invention.
Embodiment
The invention provides a kind of bacterium based on super pixel and degree of depth study cuts apart and sorting technique, can be applicable to identification and the classification of each bacterioid, there is the advantages such as quick, accurate, with low cost, idiographic flow refers to embodiment 1, the application of said method in the diagnosis of vaginal bacteria is also provided, and concrete system architecture refers to embodiment 2.
Embodiment 1
Incorporated by reference to Fig. 1 and Fig. 2, shown in, for the bacterium based on super pixel and degree of depth study in the present invention is cut apart and sorting technique.Mainly comprise the following steps:
One, set up degree of deep learning classification device by training plan image set;
Two, test pattern and degree of deep learning classification device are contrasted to the result of output category.
Refer to Fig. 1, be depicted as the step of setting up degree of deep learning classification device, specifically comprise:
In step S101, initialization convolutional neural networks.
In step S102, receive the image pattern from described training plan image set.
In step S103, actual output is calculated in forward transmission.
In step S104, calculate the poor of actual output and target output.
In step S105, utilize the backpropagation of minimization error approach to adjust weights.
In step S106, judge whether the frequency of training that reaches predetermined.
If do not reached, return to step S102 and continue and receive above-mentioned image pattern, and adjust weights.
If reached, in step S107, output category standard and priori, and deposit in described degree of deep learning classification device.
Refer to Fig. 2, be depicted as bacterium based on the study of super pixel and the degree of depth and cut apart and the process flow diagram of sorting technique, what this figure more stressed is the contrast step of test pattern, comprising:
In step S201, described test pattern is done to medium filtering through each GRB passage, and remove noise.
In step S202, calculate super pixel.
Because super pixel algorithm is very compact and feature flexibly, and the atomic region pixel with perception meaning can be flocked together, therefore, be applied in computer aided calculation and can reduce costs and labour.
In step S203, calculate color, shape, the size characteristic of each super pixel region.
In step S204, utilize priori to carry out preliminary filtering to each super pixel region, and cut apart to determine candidate bacterium region.
Wherein, priori, comprise: the numerical value upper and lower limit of the each RGB passage in bacterium region, the upper and lower limit of bacterium length and width, the upper and lower limit of bacterium field color average, the upper and lower limit of bacterium area, girth, a rough estimates value of namely by degree of deep learning classification table, enough samples being done.
In step S205, feature extraction is carried out in described candidate bacterium region.
In step S206, the criteria for classification in the feature of extraction and described sorter is contrasted.
In step S207, complete classification according to similarity.
Similarity, the namely score of the feature of candidate region after the criteria for classification of sorter contrasts, the numerical value of the test mark of gained and the label of affiliated bacteria types is more close, is considered to similarity higher, just this test zone is classified as to the bacterium that changes type.
In other embodiments, when similarity is during lower than certain predetermined threshold value, according to the needs of the state of an illness or pathology, the too low candidate region of similarity is abandoned or paid close attention to.
Bacterium based on super pixel and degree of depth study of the present invention is cut apart and sorting technique, is calculated and is cut apart and learn to classify by the degree of depth by super pixel, has advantages of high identification, low cost, the simple and easy popularization of enforcement.
Embodiment 2
Refer to Fig. 3, be depicted as a kind of vaginal bacteria based on super pixel and degree of depth study in the present invention and cut apart and categorizing system.Described system comprises training plan image set 10, degree of deep learning classification device 20, pretreatment module 30, characteristic extracting module 40 and judge module 50.
Training plan image set 10, for storing the sample image of vaginal bacteria.
Degree of deep learning classification device 20, for storing the criteria for classification arriving based on described training plan image set gained.
In the present embodiment, described degree of deep learning classification device 20, also comprises: initialization module 21, for initialization convolutional neural networks; Receiver module 22, for receiving the image pattern from described training plan image set 10; Adjusting module 23, calculates actual output, calculates actual output with the poor of target output and utilize the backpropagation of minimization error approach to adjust weights for forward transmission; Counting module 24, for the frequency of training that judges whether to reach predetermined, and after reaching, notice adjusting module 23 output category standards.
Refer to Fig. 4, be depicted as the description of the criteria for classification of vaginal bacteria.Learn from practice, vaginal bacteria is divided into Four types: comprise positive bacillus, negative bacillus, positive coccus and negative cocci.
From color differentiating: positive be generally blue, negative general pinkiness;
Distinguish from shape: bacillus is generally shaft-like, it is spherical that coccus is generally, but in overlapping situation, have different forms;
Distinguish from area: positive bacillus is substantially all greater than 250, negative and positive bacillus is substantially all greater than 150 but be slightly smaller than positively, and the area of positive coccus and negative cocci is generally all less than 200.
In addition, degree of deep learning classification device 20, also calculates and stores priori for being in course of adjustment, comprise: the numerical value upper and lower limit of the each RGB passage in bacterium region, the upper and lower limit of bacterium length and width, the upper and lower limit of bacterium field color average, the upper and lower limit of bacterium area, girth.
Pretreatment module 30, for test pattern is carried out to pre-service, calculate super pixel and calculate color, shape, the size characteristic of each super pixel region, utilize priori to carry out preliminary filtering to each super pixel region, and cut apart to determine candidate bacterium region.
Wherein, calculate super pixel, can utilize the algorithm of the super pixel of simple linear iteration cluster (SLIC), for bacterium can provide good border.The feature that SLIC algorithm has simply, remembers fast and efficiently, can reduce complicacy and raising speed by a pre-treatment step.And there is cluster advantage and superperformance, be applicable to very much cutting apart bacterium.Characteristic extracting module 40, for carrying out feature extraction to described candidate bacterium region.
Refer to Fig. 5, be depicted as that pretreatment module and characteristic extracting module are calculated based on super pixel and the design sketch of cutting apart of carrying out and feature extraction.Eigenwert resolving power in Fig. 5 is higher, can isolate fast coccus and bacillus.The border of each super pixel has real image boundary and is closely related.And because the border of image is retained, the method for super pixel is very effective to cutting apart, and ensure the accuracy of its follow-up extraction.
Judge module 50, for the criteria for classification of the feature of extraction and described degree of deep learning classification device is contrasted, completes classification according to similarity.
Similarity, the namely score of the feature of candidate region after the criteria for classification of sorter contrasts, the numerical value of the test mark of gained and the label of affiliated bacteria types is more close, is considered to similarity higher, just this test zone is classified as to the bacterium that changes type.
In other embodiments, when similarity is during lower than certain predetermined threshold value, according to the needs of the state of an illness or pathology, the too low candidate region of similarity is abandoned or paid close attention to.
Experimental result
One, material and data
From South Mountain, Shenzhen hospital, to the women of 18 to 50 years old, 105 slides are collected altogether for the collection of vaginal cell data.Then use Olympus BX43 microscope, it is (oily mirror under the condition of 100 times at objective lens magnification, its numerical aperture is 1.25), each slide is at least gathered to 105 reasonable visuals field as deal with data, the rgb image of the jpg that the picture obtaining is 1360*1024.Obtain after raw data, under doctor's guidance, the picture often running at work taking doctor is standard, has chosen altogether the test picture of 319 pictures as algorithm.
Two, segmentation result
The popular approach of the assessment to system performance is all used as accuracy rate, recall rate, degree of accuracy, sensitivity, specificity, F1 estimate with Zijdenbos index of similarity (ZSI) etc.In 6 kinds of typical cell images, identification different bacterium manually and automatic segmentation result more as shown in Figure 6.As can be seen from Figure 6, based on the manual operations of doctors experience, the effect of auto Segmentation is also very good.Quantitative test for segmentation result shows at Fig. 7.Segmentation performance is very good as can be seen from Figure 7.Fig. 8 has listed the comparison on the distinct methods in segmentation result.We can see, the segmentation result proposing is better than partitioning algorithm used in background technology [1].
Three, classification results
Utilize different sorters as degree of depth study (CNN), reverse transmittance nerve network (BNN), probabilistic neural network (PNN), support vector machine (SVM) and learn vector quantization (LVQ) vaginal cell is classified, and carrying out Performance Ratio.BNN is the sorting technique based on supervision, and the method is adjusted the weight of each layer by using gradient to minimize objective function.
Fig. 9 has summed up the performance of the sorter different based on representational feature.Can find out from the result that Fig. 9 is quantitative, the sorting technique of CNN is better than other sorting techniques, and the method that has simultaneously shown degree of depth study is the method for very effective vaginal bacteria classification.But due to its limited network layer, the result of BNN is quite poor.Obviously, enough high to the classification performance of vaginal bacteria.As can be seen from the results, the classification of vaginal cell meets the requirement using in real life.
Four, conclusion
Herein to super pixel and the application of dark learning art and vaginal cell cut apart and classification is studied, utilize these two kinds of technology that vaginal bacteria is divided into four classes.Experimental result shows, the algorithm of super pixel segmentation is most suitable to cutting apart of vaginal bacteria, has obtained the effect of highly significant.Based on the super pixel tag of degree of deep learning training, and vaginal bacteria is identified, also obtained extraordinary Classification and Identification result.The algorithm proposing has obtained the effect of significantly cutting apart and classify and has shown, vaginal bacteria automatic classification and identification contribute to doctor's diagnosis, and accuracy rate is high.
The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (8)
1. the bacterium based on super pixel and degree of depth study is cut apart and a sorting technique, comprises by training plan image set and sets up degree of deep learning classification device, and test pattern and described degree of deep learning classification device are contrasted to the result of output category; It is characterized in that: the process of described contrast comprises:
Described test pattern is done to medium filtering through each GRB passage, and remove noise;
Calculate super pixel;
Calculate color, shape, the size characteristic of each super pixel region;
Utilize priori to carry out preliminary filtering to described each super pixel region, and cut apart to determine candidate bacterium region;
Feature extraction is carried out in described candidate bacterium region;
Criteria for classification in the feature of extraction and described degree of deep learning classification device is contrasted;
Export the result of described classification according to similarity.
2. bacterium as claimed in claim 1 is cut apart and sorting technique, it is characterized in that, the described step of setting up degree of deep learning classification device comprises:
Initialization convolutional neural networks;
Receive the image pattern from described training plan image set;
Actual output is calculated in forward transmission;
Calculate the poor of actual output and target output;
Utilize the backpropagation of minimization error approach to adjust weights;
Judge whether the frequency of training that reaches predetermined;
If do not reached, continue and receive above-mentioned image pattern, and adjust weights;
If reached, export described criteria for classification, and deposit in described degree of deep learning classification device.
3. bacterium as claimed in claim 1 is cut apart and sorting technique, it is characterized in that, the super pixel of described calculating is the algorithm that adopts linear iteration cluster.
4. bacterium as claimed in claim 1 is cut apart and sorting technique, it is characterized in that, when described similarity is during lower than predetermined threshold value, abandons or pay close attention to its corresponding candidate bacterium region.
5. the vaginal bacteria based on super pixel and degree of depth study is cut apart and a categorizing system, it is characterized in that, comprising:
Training plan image set, for storing the sample image of vaginal bacteria;
Degree of deep learning classification device, for storing the criteria for classification arriving based on described training plan image set gained;
Pretreatment module, for test pattern is carried out to pre-service, calculate super pixel and calculate color, shape, the size characteristic of each super pixel region, utilize priori to carry out preliminary filtering to each super pixel region, and cut apart to determine candidate bacterium region;
Characteristic extracting module, for carrying out feature extraction to described candidate bacterium region;
Judge module, for the criteria for classification of the feature of extraction and described degree of deep learning classification device is contrasted, completes classification according to similarity.
6. vaginal bacteria as claimed in claim 5 is cut apart and categorizing system, it is characterized in that, described degree of deep learning classification device also comprises:
Initialization module, for initialization convolutional neural networks;
Receiver module, for receiving the image pattern from described training plan image set;
Adjusting module, calculates actual output, calculates actual output with the poor of target output and utilize the backpropagation of minimization error approach to adjust weights for forward transmission;
Counting module, for the frequency of training that judges whether to reach predetermined, and after reaching, notice adjusting module output category standard.
7. vaginal bacteria as claimed in claim 5 is cut apart and categorizing system, it is characterized in that, the super pixel of described calculating is the algorithm that adopts linear iteration cluster.
8. vaginal bacteria as claimed in claim 5 is cut apart and sorting technique, it is characterized in that, when described similarity is during lower than predetermined threshold value, abandons or pay close attention to its corresponding candidate bacterium region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410133805.3A CN103914841B (en) | 2014-04-03 | 2014-04-03 | Based on the segmentation of the vaginal bacteria of super-pixel and deep learning and categorizing system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410133805.3A CN103914841B (en) | 2014-04-03 | 2014-04-03 | Based on the segmentation of the vaginal bacteria of super-pixel and deep learning and categorizing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103914841A true CN103914841A (en) | 2014-07-09 |
| CN103914841B CN103914841B (en) | 2018-03-09 |
Family
ID=51040497
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410133805.3A Active CN103914841B (en) | 2014-04-03 | 2014-04-03 | Based on the segmentation of the vaginal bacteria of super-pixel and deep learning and categorizing system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103914841B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117611A (en) * | 2015-09-23 | 2015-12-02 | 北京科技大学 | Determining method and system for traditional Chinese medicine tongue diagnosis model based on convolution neural networks |
| WO2016192612A1 (en) * | 2015-06-02 | 2016-12-08 | 陈宽 | Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof |
| CN106651886A (en) * | 2017-01-03 | 2017-05-10 | 北京工业大学 | Cloud image segmentation method based on superpixel clustering optimization CNN |
| CN107945144A (en) * | 2017-11-20 | 2018-04-20 | 宁波江丰生物信息技术有限公司 | Tubercle bacillus fluorescence Sputum smears recognition methods based on deep neural network |
| CN108231190A (en) * | 2017-12-12 | 2018-06-29 | 北京市商汤科技开发有限公司 | Handle the method for image and nerve network system, equipment, medium, program |
| CN108447047A (en) * | 2018-02-11 | 2018-08-24 | 深圳市恒扬数据股份有限公司 | Acid-fast bacilli detection method and device |
| CN108629768A (en) * | 2018-04-29 | 2018-10-09 | 山东省计算中心（国家超级计算济南中心） | The dividing method of epithelial tissue in a kind of oesophagus pathological image |
| CN108961301A (en) * | 2018-07-12 | 2018-12-07 | 中国海洋大学 | It is a kind of based on the unsupervised Chaetoceros image partition method classified pixel-by-pixel |
| WO2019073312A1 (en) * | 2017-10-13 | 2019-04-18 | Sigtuple Technologies Private Limited | Method and device for integrating image channels in a deep learning model for classification |
| CN111062346A (en) * | 2019-12-21 | 2020-04-24 | 电子科技大学 | Automatic leukocyte positioning detection and classification recognition system and method |
| CN111429361A (en) * | 2020-02-17 | 2020-07-17 | 马健锦 | Bacterial colony counting method based on machine vision |
| WO2020158107A1 (en) * | 2019-01-28 | 2020-08-06 | 日本たばこ産業株式会社 | Sample quality determining method using fluorescent image, program, and device |
| CN112907456A (en) * | 2019-12-04 | 2021-06-04 | 四川大学 | Deep neural network image denoising method based on global smooth constraint prior model |
| CN112949378A (en) * | 2020-12-30 | 2021-06-11 | 至微生物智能科技(厦门)有限公司 | Bacterial microscopic image segmentation method based on deep learning network |
| CN112966645A (en) * | 2021-03-24 | 2021-06-15 | 山东仕达思生物产业有限公司 | Intelligent detection and classification counting method for multiple types of bacilli in gynecological microecology |
| CN116602712A (en) * | 2023-05-25 | 2023-08-18 | 广州斐博生物科技有限公司 | Female vaginal secretion sample based evaluation system |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101900737A (en) * | 2010-06-10 | 2010-12-01 | 上海理工大学 | Automatic identification system for urinary sediment visible components based on support vector machine |
| CN102750544A (en) * | 2012-06-01 | 2012-10-24 | 浙江捷尚视觉科技有限公司 | Detection system and detection method of rule-breaking driving that safety belt is not fastened and based on plate number recognition |
| WO2013037357A1 (en) * | 2011-09-14 | 2013-03-21 | Eads Deutschland Gmbh | Automatic learning method for the automatic learning of forms of appearance of objects in images |
| CN103093470A (en) * | 2013-01-23 | 2013-05-08 | 天津大学 | Rapid multi-modal image synergy segmentation method with unrelated scale feature |
| CN103164858A (en) * | 2013-03-20 | 2013-06-19 | 浙江大学 | Adhered crowd segmenting and tracking methods based on superpixel and graph model |
| US8526723B2 (en) * | 2009-06-23 | 2013-09-03 | Los Alamos National Security, Llc | System and method for the detection of anomalies in an image |
| CN103353938A (en) * | 2013-06-14 | 2013-10-16 | 山东大学 | Cell membrane segmentation method based on hierarchy-level characteristic |
| CN103400368A (en) * | 2013-06-25 | 2013-11-20 | 西安电子科技大学 | Parallel rapid SAR image segmentation method based on graph theory and superpixel |
| CN103544697A (en) * | 2013-09-30 | 2014-01-29 | 南京信息工程大学 | Hypergraph spectrum analysis based image segmentation method |
| CN103679719A (en) * | 2013-12-06 | 2014-03-26 | 河海大学 | Image segmentation method |
-
2014
- 2014-04-03 CN CN201410133805.3A patent/CN103914841B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8526723B2 (en) * | 2009-06-23 | 2013-09-03 | Los Alamos National Security, Llc | System and method for the detection of anomalies in an image |
| CN101900737A (en) * | 2010-06-10 | 2010-12-01 | 上海理工大学 | Automatic identification system for urinary sediment visible components based on support vector machine |
| WO2013037357A1 (en) * | 2011-09-14 | 2013-03-21 | Eads Deutschland Gmbh | Automatic learning method for the automatic learning of forms of appearance of objects in images |
| CN102750544A (en) * | 2012-06-01 | 2012-10-24 | 浙江捷尚视觉科技有限公司 | Detection system and detection method of rule-breaking driving that safety belt is not fastened and based on plate number recognition |
| CN103093470A (en) * | 2013-01-23 | 2013-05-08 | 天津大学 | Rapid multi-modal image synergy segmentation method with unrelated scale feature |
| CN103164858A (en) * | 2013-03-20 | 2013-06-19 | 浙江大学 | Adhered crowd segmenting and tracking methods based on superpixel and graph model |
| CN103353938A (en) * | 2013-06-14 | 2013-10-16 | 山东大学 | Cell membrane segmentation method based on hierarchy-level characteristic |
| CN103400368A (en) * | 2013-06-25 | 2013-11-20 | 西安电子科技大学 | Parallel rapid SAR image segmentation method based on graph theory and superpixel |
| CN103544697A (en) * | 2013-09-30 | 2014-01-29 | 南京信息工程大学 | Hypergraph spectrum analysis based image segmentation method |
| CN103679719A (en) * | 2013-12-06 | 2014-03-26 | 河海大学 | Image segmentation method |
Non-Patent Citations (2)
| Title |
|---|
| 李葆青: "基于卷积神经网络的模式分类器", 《大连大学学报》 * |
| 赵明珠: "细胞病理图像的特征分析与分类识别", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016192612A1 (en) * | 2015-06-02 | 2016-12-08 | 陈宽 | Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof |
| IL255856A (en) * | 2015-06-02 | 2018-01-31 | Chen Kuan | Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof |
| JP2018529134A (en) * | 2015-06-02 | 2018-10-04 | 寛 陳 | Medical data analysis method based on deep learning and its intelligent analyzer |
| US11200982B2 (en) | 2015-06-02 | 2021-12-14 | Infervision Medical Technology Co., Ltd. | Method for analysing medical treatment data based on deep learning and intelligence analyser thereof |
| CN105117611B (en) * | 2015-09-23 | 2018-06-12 | 北京科技大学 | Based on the determining method and system of the TCM tongue diagnosis model of convolutional Neural metanetwork |
| CN105117611A (en) * | 2015-09-23 | 2015-12-02 | 北京科技大学 | Determining method and system for traditional Chinese medicine tongue diagnosis model based on convolution neural networks |
| CN106651886A (en) * | 2017-01-03 | 2017-05-10 | 北京工业大学 | Cloud image segmentation method based on superpixel clustering optimization CNN |
| WO2019073312A1 (en) * | 2017-10-13 | 2019-04-18 | Sigtuple Technologies Private Limited | Method and device for integrating image channels in a deep learning model for classification |
| CN107945144A (en) * | 2017-11-20 | 2018-04-20 | 宁波江丰生物信息技术有限公司 | Tubercle bacillus fluorescence Sputum smears recognition methods based on deep neural network |
| CN107945144B (en) * | 2017-11-20 | 2022-02-25 | 宁波江丰生物信息技术有限公司 | Mycobacterium tuberculosis fluorescence sputum smear identification method based on deep neural network |
| CN108231190B (en) * | 2017-12-12 | 2020-10-30 | 北京市商汤科技开发有限公司 | Method of processing image, neural network system, device, and medium |
| CN108231190A (en) * | 2017-12-12 | 2018-06-29 | 北京市商汤科技开发有限公司 | Handle the method for image and nerve network system, equipment, medium, program |
| CN108447047A (en) * | 2018-02-11 | 2018-08-24 | 深圳市恒扬数据股份有限公司 | Acid-fast bacilli detection method and device |
| CN108629768A (en) * | 2018-04-29 | 2018-10-09 | 山东省计算中心（国家超级计算济南中心） | The dividing method of epithelial tissue in a kind of oesophagus pathological image |
| CN108629768B (en) * | 2018-04-29 | 2022-01-21 | 山东省计算中心（国家超级计算济南中心） | Method for segmenting epithelial tissue in esophageal pathology image |
| CN108961301B (en) * | 2018-07-12 | 2021-11-02 | 中国海洋大学 | Chaetoceros image segmentation method based on unsupervised pixel-by-pixel classification |
| CN108961301A (en) * | 2018-07-12 | 2018-12-07 | 中国海洋大学 | It is a kind of based on the unsupervised Chaetoceros image partition method classified pixel-by-pixel |
| JPWO2020158107A1 (en) * | 2019-01-28 | 2021-09-30 | 日本たばこ産業株式会社 | Sample quality determination method, program, and equipment using fluorescent images |
| WO2020158107A1 (en) * | 2019-01-28 | 2020-08-06 | 日本たばこ産業株式会社 | Sample quality determining method using fluorescent image, program, and device |
| JP7134421B2 (en) | 2019-01-28 | 2022-09-12 | 日本たばこ産業株式会社 | Sample quality determination method, program, and device using fluorescence image |
| CN112907456A (en) * | 2019-12-04 | 2021-06-04 | 四川大学 | Deep neural network image denoising method based on global smooth constraint prior model |
| CN112907456B (en) * | 2019-12-04 | 2022-06-10 | 四川大学 | Deep neural network image denoising method based on global smooth constraint prior model |
| CN111062346A (en) * | 2019-12-21 | 2020-04-24 | 电子科技大学 | Automatic leukocyte positioning detection and classification recognition system and method |
| CN111429361A (en) * | 2020-02-17 | 2020-07-17 | 马健锦 | Bacterial colony counting method based on machine vision |
| CN112949378A (en) * | 2020-12-30 | 2021-06-11 | 至微生物智能科技(厦门)有限公司 | Bacterial microscopic image segmentation method based on deep learning network |
| CN112966645A (en) * | 2021-03-24 | 2021-06-15 | 山东仕达思生物产业有限公司 | Intelligent detection and classification counting method for multiple types of bacilli in gynecological microecology |
| CN112966645B (en) * | 2021-03-24 | 2022-04-08 | 山东仕达思生物产业有限公司 | Intelligent detection and classification counting method for multiple types of bacilli in gynecological microecology |
| CN116602712A (en) * | 2023-05-25 | 2023-08-18 | 广州斐博生物科技有限公司 | Female vaginal secretion sample based evaluation system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103914841B (en) | 2018-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103914841A (en) | Bacterium division and classification method based on superpixels and in-depth learning and application thereof | |
| Xiang et al. | A novel automation-assisted cervical cancer reading method based on convolutional neural network | |
| CN110428432B (en) | Deep neural network algorithm for automatically segmenting colon gland image | |
| CN101900737A (en) | Automatic identification system for urinary sediment visible components based on support vector machine | |
| CN105122308A (en) | Systems and methods for multiplexed biomarker quantitation using single cell segmentation on sequentially stained tissue | |
| WO2010003041A2 (en) | Mitotic figure detector and counter system and method for detecting and counting mitotic figures | |
| WO2021227708A1 (en) | Method and device for identifying abnormal cell in sample to be tested, and storage medium | |
| US10769432B2 (en) | Automated parameterization image pattern recognition method | |
| Mao et al. | Iteratively training classifiers for circulating tumor cell detection | |
| Zhang et al. | Segmentation of overlapping cells in cervical smears based on spatial relationship and overlapping translucency light transmission model | |
| CN103679184A (en) | Method for leukocyte automatic identification based on relevant vector machine | |
| CN105184808A (en) | Automatic segmentation method for foreground and background of optical field image | |
| Sornapudi et al. | Comparing deep learning models for multi-cell classification in liquid-based cervical cytology image | |
| Rulaningtyas et al. | Multi patch approach in K-means clustering method for color image segmentation in pulmonary tuberculosis identification | |
| CN112784767A (en) | Cell example segmentation algorithm based on leukocyte microscopic image | |
| WO2020119624A1 (en) | Class-sensitive edge detection method based on deep learning | |
| CN115206495A (en) | Renal cancer pathological image analysis method and system based on CoAtNet deep learning and intelligent microscopic device | |
| Sheela Shiney et al. | Deep auto encoder based extreme learning system for automatic segmentation of cervical cells | |
| CN114387596A (en) | Automatic interpretation system for cytopathology smear | |
| Jiang et al. | A systematic review of deep learning-based cervical cytology screening: from cell identification to whole slide image analysis | |
| WO2021164320A1 (en) | Computer vision based catheter feature acquisition method and apparatus and intelligent microscope | |
| Vale et al. | Automatic segmentation and classification of blood components in microscopic images using a fuzzy approach | |
| Arya et al. | Three segmentation techniques to predict the dysplasia in cervical cells in the presence of debris | |
| Karunakar et al. | An unparagoned application for red blood cell counting using marker controlled watershed algorithm for android mobile | |
| Liu et al. | Artificial classification of cervical squamous lesions in ThinPrep cytologic tests using a deep convolutional neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |