CN106780466A - A kind of cervical cell image-recognizing method based on convolutional neural networks - Google Patents

A kind of cervical cell image-recognizing method based on convolutional neural networks Download PDF

Info

Publication number
CN106780466A
CN106780466A CN201611189662.3A CN201611189662A CN106780466A CN 106780466 A CN106780466 A CN 106780466A CN 201611189662 A CN201611189662 A CN 201611189662A CN 106780466 A CN106780466 A CN 106780466A
Authority
CN
China
Prior art keywords
convolutional neural
neural networks
layer
cervical cell
feature
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
Application number
CN201611189662.3A
Other languages
Chinese (zh)
Inventor
郭磊
罗晓曙
何富运
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangxi Normal University
Original Assignee
Guangxi Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangxi Normal University filed Critical Guangxi Normal University
Priority to CN201611189662.3A priority Critical patent/CN106780466A/en
Publication of CN106780466A publication Critical patent/CN106780466A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G06V20/695—Preprocessing, e.g. image segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of cervical cell image-recognizing method based on convolutional neural networks, it is characterized in that, comprise the following steps:1）Prepare training sample；2）Build convolutional neural networks layer；（3）Build two graders；（4）It is identified result：Cervical cell picture input to be tested is improved into convolutional neural networks, convolutional neural networks is improved and is identified automatically, sorts out.This method high degree of automation, adaptive ability are strong, can not only improve the efficiency of cervical cell image recognition, and can improve the accuracy rate of cervical cell image recognition.

Description

A kind of cervical cell image-recognizing method based on convolutional neural networks
Technical field
The present invention relates to cell image processing technology field, specifically a kind of cervical cell figure based on convolutional neural networks As recognition methods.
Background technology
In China, cervical cell the image-recognizing method traditional at present main still artificial diagosis technology of Pasteur, Ba Shiren Work diagosis technology will lean on people in a large amount of cell images of Microscopic observation, and working strength is big, and easily make one to feel fatigue, identification Accuracy rate and recognition efficiency are low.
The content of the invention
The purpose of the present invention is directed to the deficiencies in the prior art, and provides a kind of cervical cell based on convolutional neural networks Image-recognizing method.
This method high degree of automation, adaptive ability are strong, can not only improve the efficiency of cervical cell image recognition, and And can also improve the accuracy rate of cervical cell image recognition.
Realizing the technical scheme of the object of the invention is：
A kind of cervical cell image-recognizing method based on convolutional neural networks, comprises the following steps:
1) training sample is prepared：
(1-1) reads in cervical cell image in existing picture library as training sample and classifies：The all of palace that will be read in Neck cell image is divided into normal cervix cell training sample and lesion cervical cell training sample；
(1-2) gray processing：It is gray level image block by cervical cell image preprocessing, and by the colour in cervical cell image Picture is converted into gray level image, then the gray level image size for obtaining is normalized to the gray level image block of 32*32；
2) convolutional neural networks layer is built：Build one have self-adapting estimation classification feature including adding BN algorithms Convolutional neural networks are improved, it is a neutral net for multilayer to improve convolutional neural networks, is used as by trainable convolution kernel Wave filter, is successively filtered to image, and each layer of filter result is carried out into Automatic Combined, is finally automatically extracted out to classification Best feature, has extracted after feature, according to class categories difference from all characteristic parameters, carries out parametric classification, it Group is carried out to the characteristic parameter between different classes of afterwards and is closed training and is recognized, and according to the difference of recognition result, adjusting training Characteristic vector, when according to this characteristic parameter combination obtain recognition result be less than before recognition result when, then according to existing Characteristic vector, add or delete corresponding characteristic parameter, discrimination higher is obtained during to again identifying that；
The feature of cervical cell is extracted including form, colourity, optical density, textural characteristics etc., wherein morphological feature includes thin Born of the same parents (core) area, girth, height, width, circularity, rectangular degree, elongation etc., chromaticity include nucleus (matter) in RGB Average, variance and colourity variation coefficient on color component etc., optical density feature include the integral optical density of nucleus (matter), put down Equal optical density, optical density coefficient etc., textural characteristics use two kinds of features of the features of Haralick two and Tamura, totally four kinds of textures Feature；
Convolutional neural networks are improved to be produced through after network processes by view data directly as network inputs variable Classification number as a result, realization is processed end to end, and the complicated mistake such as image preprocessing and manual features extraction is reduced or avoided Journey；
(3) two graders are built：By step 1) the normal cervix cell training sample that obtains and the training of lesion cervical cell Sample input improves convolutional neural networks, improves convolutional neural networks and is trained to be capable of identify that normal cervix cell and lesion palace Two graders of neck cell；
(4) it is identified result：Cervical cell picture input to be tested is improved into convolutional neural networks, convolution god is improved It is identified automatically through network, is sorted out.
The improvement convolutional neural networks of the addition BN algorithms are to add to criticize in classical convolutional neural networks structure Normalization algorithm, i.e., connect one BN layers after each convolutional layer and down-sampling layer, and 3 are connected after down-sampling S2 Full articulamentum, equally each accessed after full articulamentum it is corresponding BN layer, the full articulamentum of last layer using sigmoid activation primitives, Direct output image.
Described batch of normalization algorithm, i.e. BN algorithms are：
BN layers and convolutional layer, down-sampling layer are also the Rotating fields in network as full articulamentum, are a normalization Process, be standardized dimension：
The output of last layer is changed into standardized normal distribution by formula (1), and mini-batch boarding steps are used in training process Degree declines, E [x(k)] refer to per a collection of training data neuron x(k)Desired value,It is x(k)The standard deviation of activation value, Simultaneously to prevent from influenceing each layer of feature for learning, conversion reconstruct is introduced, addition can learning parameter γ and β：
Formula (2) to a certain extent reduces back the feature for being changed into normal distribution, allows it to keep its initial distribution Trend, reducing degree is got by convolutional neural networks oneself study, is set Original activation value can be recovered, that is, recover original certain layer feature acquired, the introducing of γ and β can allow convolutional neural networks Habit recovers the primitive network feature distribution to be learnt, and Batch Normalization forward conduction formula are：
Formula (3) is the input-mean of each sample in a collection of mini-batch in all samples on same neuron, Formula (4) is the input variance of each sample in a collection of mini-batch in all samples on same neuron, and formula (5) is handle The knot that the input of each sample in this crowd of mini-batch in all samples on same neuron is obtained after being standardized Really, formula (6) is isomorphic convert.
This method high degree of automation, adaptive ability are strong, can not only improve the efficiency of cervical cell image recognition, and And can also improve the accuracy rate of cervical cell image recognition.
Brief description of the drawings
Fig. 1 is C3 layers of characteristic pattern combination；
Fig. 2 is that embodiment improves convolutional neural networks structural representation；
Fig. 3 is the method flow schematic diagram of embodiment.
Specific embodiment
Present invention is further elaborated with reference to the accompanying drawings and examples, but is not limitation of the invention.
Embodiment：
A kind of reference picture 3, cervical cell image-recognizing method based on convolutional neural networks, comprises the following steps:
1) training sample is prepared：
(1-1) reads in cervical cell image in existing picture library as training sample and classifies：The all of palace that will be read in Neck cell image is divided into normal cervix cell training sample and lesion cervical cell training sample；
(1-2) gray processing：It is gray level image block by cervical cell image preprocessing, and by the colour in cervical cell image Picture is converted into gray level image, then the gray level image size for obtaining is normalized to the gray level image block of 32*32；
2) convolutional neural networks layer is built：Build one have self-adapting estimation classification feature including adding BN algorithms Convolutional neural networks are improved, it is a neutral net for multilayer to improve convolutional neural networks, is used as by trainable convolution kernel Wave filter, is successively filtered to image, and each layer of filter result is carried out into Automatic Combined, is finally automatically extracted out to classification Best feature, has extracted after feature, according to class categories difference from all characteristic parameters, carries out parametric classification, it Group is carried out to the characteristic parameter between different classes of afterwards and is closed training and is recognized, and according to the difference of recognition result, adjusting training Characteristic vector, when according to this characteristic parameter combination obtain recognition result be less than before recognition result when, then according to existing Characteristic vector, add or delete corresponding characteristic parameter, discrimination higher is obtained during to again identifying that；
As shown in Fig. 2 the improvement convolutional neural networks used in the present embodiment are a neutral nets for multilayer, improve and add Enter BN layers, one BN layers is connected after each convolutional layer and down-sampling layer, 3 full connections are connected after down-sampling S2 Layer, equally respectively accesses corresponding BN layers after full articulamentum.Convolutional neural networks are the topological structures for aiming at two dimensional image and designing, most Important the characteristics of be feature extraction with pattern classification while carrying out, the pattern classification better than shallow-layer machine learning algorithm will be carried additionally Take characteristics of image.Additionally, the weights of convolutional neural networks share the training parameter for reducing network, along with its multiple feature is carried Take, make it have very strong robustness；
11 layers of convolutional neural networks as shown in Figure 2, including two convolutional layers, i.e. feature extraction layer, two down-sampling layers, That is Feature Mapping layer, four BN layer and three connect layer entirely, and preceding convolutional layer twice is all followed by one BN layers, Ran Houjie after C layers One layer is used for asking local weighted i.e. S layers average of down-sampling layer as Further Feature Extraction, this distinctive feature extraction twice The structure being combined makes the ability that network has certain tolerance noise in pattern classification to input picture, that is, show as network Robustness；
The input layer of convolutional neural networks is the gray-scale map of 32*32, and convolutional layer C1 has 6 characteristic patterns, and down-sampling layer S2 has 6 Characteristic pattern is opened, convolutional layer C3 is combined after convolution by S2 layers of 6 characteristic patterns and obtained 16 characteristic patterns, combination such as Fig. 1 institutes Show, down-sampling layer S4 there are 16 characteristic patterns, full articulamentum C5 sets 84 nodes, and full articulamentum F6 sets 120 nodes, output Layer sets two nodes；
C1 layers by after the convolution mask convolution of 5*5, the size of 6 characteristic patterns is 28*28, in characteristic pattern each nerve Unit is connected with the convolution mask of 5*5 in input, each wave filter 5*5 totally 25 unit's parameters and an offset parameter, and totally 6 filter Device, (5*5+1) * 6=156 are individual altogether can training parameter, altogether 156* (28*28)=122304 connection；S2 layers through down-sampling 6 characteristic patterns of 14*14 are obtained afterwards, each unit in characteristic pattern is connected with the 2*2 fields of character pair figure in C1, S2 layers every 4 of individual unit inputs are added, be multiplied by one can training parameter, then add one and can train biasing, the acting as of down-sampling obscures Image, does not overlap during using 2*2 sample template, because the size of each characteristic pattern is 14*14 in S2, (2*2+1) * altogether (14*14)=5880 connect；C3 layers has the 5*5*60+16=1516 can training parameter；
The feature of cervical cell is extracted including form, colourity, optical density, textural characteristics etc., wherein morphological feature includes thin Born of the same parents (core) area, girth, height, width, circularity, rectangular degree, elongation etc., chromaticity include nucleus (matter) in RGB Average, variance and colourity variation coefficient on color component etc., optical density feature include the integral optical density of nucleus (matter), put down Equal optical density, optical density coefficient etc., textural characteristics use two kinds of features of the features of Haralick two and Tamura, totally four kinds of textures Feature；
Convolutional neural networks are improved to be produced through after network processes by view data directly as network inputs variable Classification number as a result, realization is processed end to end, and the complicated mistake such as image preprocessing and manual features extraction is reduced or avoided Journey；
(3) two graders are built：By step 1) the normal cervix cell training sample that obtains and the training of lesion cervical cell Sample input improves convolutional neural networks, improves convolutional neural networks and is trained to be capable of identify that normal cervix cell and lesion palace Two graders of neck cell, specially：Training parameter is treated in initializing network with some different small random numbers, to such as Fig. 2 Improve convolutional neural networks to be input into N1 training sample to train improvement convolutional neural networks, each sample includes input vector, Preferable output vector, input vector is sent to output layer by converting layer by layer, obtains reality output vector；
Using cross entropy loss function, convolutional neural networks parameter is improved with reference to the adjustment of backpropagation BP algorithm, using equal The backpropagation of square error, completes training；
(4) it is identified result：Cervical cell picture input to be tested is improved into convolutional neural networks, convolution god is improved It is identified automatically through network, is sorted out.
The improvement convolutional neural networks of the addition BN algorithms are to add to criticize in classical convolutional neural networks structure Normalization algorithm, i.e., connect one BN layers after each convolutional layer and down-sampling layer, and 3 are connected after down-sampling S2 Full articulamentum, equally each accessed after full articulamentum it is corresponding BN layer, the full articulamentum of last layer using sigmoid activation primitives, Direct output image.
Described batch of normalization algorithm, i.e. BN algorithms are：
BN layers and convolutional layer, down-sampling layer are also the Rotating fields in network as full articulamentum, are a normalization Process, be standardized dimension：
The output of last layer is changed into standardized normal distribution by formula (1), and mini-batch boarding steps are used in training process Degree declines, E [x(k)] refer to per a collection of training data neuron x(k)Desired valueIt is x(k)The standard deviation of activation value, Simultaneously to prevent from influenceing each layer of feature for learning, conversion reconstruct is introduced, addition can learning parameter γ and β：
Formula (2) to a certain extent reduces back the feature for being changed into normal distribution, allows it to keep its initial distribution Trend, reducing degree is got by convolutional neural networks oneself study, is set Original activation value can be recovered, that is, recover original certain layer feature acquired, the introducing of γ and β can allow convolutional neural networks Habit recovers the primitive network feature distribution to be learnt, and Batch Normalization forward conduction formula are：
Formula (3) is the input-mean of each sample in a collection of mini-batch in all samples on same neuron, Formula (4) is the input variance of each sample in a collection of mini-batch in all samples on same neuron, and formula (5) is handle The knot that the input of each sample in this crowd of mini-batch in all samples on same neuron is obtained after being standardized Really, formula (6) is isomorphic convert.

Claims (3)

1. a kind of cervical cell image-recognizing method based on convolutional neural networks, it is characterized in that, comprise the following steps:
1) training sample is prepared：
(1-1) reads in cervical cell image in existing picture library as training sample and classifies：The all of uterine neck that will be read in is thin Born of the same parents' image is divided into normal cervix cell training sample and lesion cervical cell training sample；
(1-2) gray processing：It is gray level image block by cervical cell image preprocessing, and by the colour picture in cervical cell image Gray level image is converted into, then the gray level image size for obtaining is normalized to the gray level image block of 32*32；
2) convolutional neural networks layer is built：Build an improvement including adding BN algorithms with self-adapting estimation classification feature Convolutional neural networks, it is a neutral net for multilayer to improve convolutional neural networks, and filtering is used as by trainable convolution kernel Device, is successively filtered to image, and each layer of filter result is carried out into Automatic Combined, and finally automatically extract out most has to classification The feature of profit, has extracted after feature, according to class categories difference from all characteristic parameters, carries out parametric classification, right afterwards Characteristic parameter between different classes of carries out group and closes training and recognize, and according to the difference of recognition result, adjusting training feature Vector, when according to this characteristic parameter combination obtain recognition result be less than before recognition result when, then according to existing spy Vector being levied, corresponding characteristic parameter is added or delete, discrimination higher is obtained during to again identifying that；
(3) two graders are built：By step 1) the normal cervix cell training sample and lesion cervical cell training sample that obtain Input improves convolutional neural networks, improves convolutional neural networks and is trained to be capable of identify that normal cervix cell and lesion uterine neck are thin Two graders of born of the same parents；
(4) it is identified result：Cervical cell picture input to be tested is improved into convolutional neural networks, convolutional Neural net is improved Network is identified, sorts out automatically.
2. the cervical cell image-recognizing method based on convolutional neural networks according to claim 1, it is characterized in that, it is described The improvement convolutional neural networks for adding BN algorithms are that batch normalization algorithm is added in classical convolutional neural networks structure, i.e., One BN layers is connected after each convolutional layer and down-sampling layer, 3 full articulamentums, Quan Lian are connected after down-sampling S2 Connect and equally each access after layer corresponding BN layers, the full articulamentum of last layer uses sigmoid activation primitives, direct output image.
3. the cervical cell image-recognizing method based on convolutional neural networks according to claim 2, it is characterized in that, it is described Normalization algorithm is criticized, i.e. BN algorithms are：
BN layers and convolutional layer, down-sampling layer are also the Rotating fields in network as full articulamentum, are a normalized mistakes Journey, is standardized dimension：
x ^ ( k ) = x ( k ) - E [ x ( k ) ] V a r [ x ( k ) ] - - - ( 1 )
The output of last layer is changed into standardized normal distribution by formula (1), using under mini-batch stochastic gradients in training process Drop, E [x(k)] refer to per a collection of training data neuron x(k)Desired value,It is x(k)The standard deviation of activation value, while To prevent influenceing each layer of feature for learning, conversion reconstruct is introduced, addition can learning parameter γ and β:
y ( k ) = γ ( k ) x ^ ( k ) + β ( k ) - - - ( 2 )
Formula (2) to a certain extent reduces back the feature for being changed into normal distribution, allows it to keep its initial distribution trend, Reducing degree is got by convolutional neural networks oneself study, is setCan be with Recover original activation value, that is, recover original certain layer feature acquired, the introducing of γ and β can allow convolutional neural networks study extensive Primitive network of the appearing again feature distribution to be learnt, Batch Normalization forward conduction formula are：
μ B → 1 m Σ i = 1 m x i - - - ( 3 )
σ B 2 ← 1 m Σ i = 1 m ( x i - μ B ) 2 - - - ( 4 )
x i ^ ← x i - μ B σ B 2 + ϵ - - - ( 5 )
y i ← γ x i ^ + β ≡ BN γ , β ( x i ) - - - ( 6 )
Formula (3) is the input-mean of each sample in a collection of mini-batch in all samples on same neuron, formula (4) It is input variance of each sample in all samples in a collection of mini-batch on same neuron, formula (5) is this batch The result that the input of each sample in mini-batch in all samples on same neuron is obtained after being standardized, formula (6) it is isomorphic convert.
CN201611189662.3A 2016-12-21 2016-12-21 A kind of cervical cell image-recognizing method based on convolutional neural networks Pending CN106780466A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611189662.3A CN106780466A (en) 2016-12-21 2016-12-21 A kind of cervical cell image-recognizing method based on convolutional neural networks

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611189662.3A CN106780466A (en) 2016-12-21 2016-12-21 A kind of cervical cell image-recognizing method based on convolutional neural networks

Publications (1)

Publication Number Publication Date
CN106780466A true CN106780466A (en) 2017-05-31

Family

ID=58896738

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611189662.3A Pending CN106780466A (en) 2016-12-21 2016-12-21 A kind of cervical cell image-recognizing method based on convolutional neural networks

Country Status (1)

Country Link
CN (1) CN106780466A (en)

Cited By (41)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107609503A (en) * 2017-09-05 2018-01-19 刘宇红 Intelligent cancerous tumor cell identifying system and method, cloud platform, server, computer
CN107680082A (en) * 2017-09-11 2018-02-09 宁夏医科大学 Lung tumor identification method based on depth convolutional neural networks and global characteristics
CN108109102A (en) * 2017-12-28 2018-06-01 珠海市君天电子科技有限公司 A kind of data processing method, device, electronic equipment and storage medium
CN108154164A (en) * 2017-11-15 2018-06-12 上海微波技术研究所（中国电子科技集团公司第五十研究所） Signal of communication modulation classification system and method based on deep learning
CN108197606A (en) * 2018-01-31 2018-06-22 浙江大学 The recognition methods of abnormal cell in a kind of pathological section based on multiple dimensioned expansion convolution
CN108281183A (en) * 2018-01-30 2018-07-13 重庆大学 Cervical smear image diagnostic system based on convolutional neural networks and transfer learning
CN108334909A (en) * 2018-03-09 2018-07-27 南京天数信息科技有限公司 Cervical carcinoma TCT digital slices data analysing methods based on ResNet
CN108345871A (en) * 2018-03-20 2018-07-31 宁波江丰生物信息技术有限公司 A kind of cervical carcinoma slice recognition methods
CN108395986A (en) * 2018-03-23 2018-08-14 余晖 Human papilloma virus automatic parting direction detection device based on deep learning
CN108416379A (en) * 2018-03-01 2018-08-17 北京羽医甘蓝信息技术有限公司 Method and apparatus for handling cervical cell image
CN108491765A (en) * 2018-03-05 2018-09-04 中国农业大学 A kind of classifying identification method and system of vegetables image
CN108510482A (en) * 2018-03-22 2018-09-07 姚书忠 Cervical carcinoma detection method, device, equipment and medium based on gynecatoptron image
CN108520206A (en) * 2018-03-22 2018-09-11 南京大学 A kind of fungi microscopic image identification method based on full convolutional neural networks
CN108665463A (en) * 2018-03-30 2018-10-16 哈尔滨理工大学 A kind of cervical cell image partition method generating network based on confrontation type
CN108765374A (en) * 2018-04-27 2018-11-06 华南理工大学 A kind of method of abnormal core region screening in cervical smear image
CN109154989A (en) * 2018-03-02 2019-01-04 香港应用科技研究院有限公司 The classification performance that convolutional neural networks apply cancer cell screening is improved using mask
CN109190441A (en) * 2018-06-21 2019-01-11 丁彦青 Female genital tract cells pathology intelligent method for classifying, diagnostic equipment and storage medium
CN109255340A (en) * 2018-10-29 2019-01-22 东北大学 It is a kind of to merge a variety of face identification methods for improving VGG network
CN109325586A (en) * 2018-12-05 2019-02-12 北京航空航天大学合肥创新研究院 Deep neural network system based on composite object function
CN109460720A (en) * 2018-10-25 2019-03-12 中科院成都信息技术股份有限公司 Ballot paper recognition methods based on convolutional neural networks
CN109816623A (en) * 2017-11-17 2019-05-28 国立癌症研究中心 Method for analyzing image, device and program and study are over Deep Learning algorithm and its manufacturing method
CN109948429A (en) * 2019-01-28 2019-06-28 上海依智医疗技术有限公司 Image analysis method, device, electronic equipment and computer-readable medium
CN109993122A (en) * 2019-04-02 2019-07-09 中国石油大学(华东) A kind of pedestrian based on depth convolutional neural networks multiplies staircase anomaly detection method
CN110009050A (en) * 2019-04-10 2019-07-12 杭州智团信息技术有限公司 A kind of classification method and device of cell
CN110020714A (en) * 2018-01-10 2019-07-16 阿里巴巴集团控股有限公司 Model training and data analysing method, device, equipment and storage medium
US10354122B1 (en) 2018-03-02 2019-07-16 Hong Kong Applied Science and Technology Research Institute Company Limited Using masks to improve classification performance of convolutional neural networks with applications to cancer-cell screening
CN110070175A (en) * 2019-04-12 2019-07-30 北京市商汤科技开发有限公司 Image processing method, model training method and device, electronic equipment
CN110352461A (en) * 2017-11-16 2019-10-18 韩国宝之铂株式会社 For determining the method and apparatus that cervical carcinoma whether occurs in subject
WO2019238072A1 (en) * 2018-06-13 2019-12-19 深圳市商汤科技有限公司 Normalization method, apparatus and device for deep neural network, and storage medium
US10586129B2 (en) 2018-02-21 2020-03-10 International Business Machines Corporation Generating artificial images for use in neural networks
CN110946552A (en) * 2019-10-30 2020-04-03 南京航空航天大学 Cervical cancer pre-lesion screening method combining spectrum and image
CN111052177A (en) * 2017-09-14 2020-04-21 三菱电机株式会社 Abnormality detection device
CN111274903A (en) * 2020-01-15 2020-06-12 合肥工业大学 Cervical cell image classification method based on graph convolution neural network
CN111524132A (en) * 2020-05-09 2020-08-11 腾讯科技（深圳）有限公司 Method, device and storage medium for identifying abnormal cells in sample to be detected
CN111783571A (en) * 2020-06-17 2020-10-16 陕西中医药大学 Cervical cell automatic classification model establishment and cervical cell automatic classification method
CN112204674A (en) * 2018-03-07 2021-01-08 维迪控股私人有限公司 Method for identifying biological material by microscopy
CN112613505A (en) * 2020-12-18 2021-04-06 安徽丹姆斯生物科技有限公司 Cell micronucleus identification, positioning and counting method based on deep learning
CN113222044A (en) * 2021-05-25 2021-08-06 合肥工业大学 Cervical fluid-based cell classification method based on ternary attention and scale correlation fusion
CN114266717A (en) * 2020-09-25 2022-04-01 天津科技大学 Parallel capsule network cervical cancer cell detection method based on Inception module
CN114708589A (en) * 2022-04-02 2022-07-05 大连海事大学 Cervical cell classification method based on deep learning
CN115359483A (en) * 2022-08-09 2022-11-18 杭州迪英加科技有限公司 Training method for cervical fluid-based cell classification

Cited By (53)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107609503A (en) * 2017-09-05 2018-01-19 刘宇红 Intelligent cancerous tumor cell identifying system and method, cloud platform, server, computer
CN107680082A (en) * 2017-09-11 2018-02-09 宁夏医科大学 Lung tumor identification method based on depth convolutional neural networks and global characteristics
CN111052177A (en) * 2017-09-14 2020-04-21 三菱电机株式会社 Abnormality detection device
CN108154164A (en) * 2017-11-15 2018-06-12 上海微波技术研究所（中国电子科技集团公司第五十研究所） Signal of communication modulation classification system and method based on deep learning
CN110352461A (en) * 2017-11-16 2019-10-18 韩国宝之铂株式会社 For determining the method and apparatus that cervical carcinoma whether occurs in subject
CN109816623A (en) * 2017-11-17 2019-05-28 国立癌症研究中心 Method for analyzing image, device and program and study are over Deep Learning algorithm and its manufacturing method
CN108109102B (en) * 2017-12-28 2021-11-23 珠海市君天电子科技有限公司 Data processing method and device, electronic equipment and storage medium
CN108109102A (en) * 2017-12-28 2018-06-01 珠海市君天电子科技有限公司 A kind of data processing method, device, electronic equipment and storage medium
CN110020714B (en) * 2018-01-10 2023-05-30 斑马智行网络(香港)有限公司 Model training and data analysis method, device, equipment and storage medium
CN110020714A (en) * 2018-01-10 2019-07-16 阿里巴巴集团控股有限公司 Model training and data analysing method, device, equipment and storage medium
CN108281183A (en) * 2018-01-30 2018-07-13 重庆大学 Cervical smear image diagnostic system based on convolutional neural networks and transfer learning
CN108197606A (en) * 2018-01-31 2018-06-22 浙江大学 The recognition methods of abnormal cell in a kind of pathological section based on multiple dimensioned expansion convolution
US10586129B2 (en) 2018-02-21 2020-03-10 International Business Machines Corporation Generating artificial images for use in neural networks
CN108416379A (en) * 2018-03-01 2018-08-17 北京羽医甘蓝信息技术有限公司 Method and apparatus for handling cervical cell image
CN109154989A (en) * 2018-03-02 2019-01-04 香港应用科技研究院有限公司 The classification performance that convolutional neural networks apply cancer cell screening is improved using mask
US10354122B1 (en) 2018-03-02 2019-07-16 Hong Kong Applied Science and Technology Research Institute Company Limited Using masks to improve classification performance of convolutional neural networks with applications to cancer-cell screening
CN109154989B (en) * 2018-03-02 2021-07-06 香港应用科技研究院有限公司 Using masks to improve classification performance of convolutional neural networks for cancer cell screening applications
WO2019165649A1 (en) * 2018-03-02 2019-09-06 Hong Kong Applied Science and Technology Research Institute Company Limited Using masks to improve classification performance of convolutional neural networks with applications to cancer-cell screening
CN108491765A (en) * 2018-03-05 2018-09-04 中国农业大学 A kind of classifying identification method and system of vegetables image
CN112204674A (en) * 2018-03-07 2021-01-08 维迪控股私人有限公司 Method for identifying biological material by microscopy
CN108334909A (en) * 2018-03-09 2018-07-27 南京天数信息科技有限公司 Cervical carcinoma TCT digital slices data analysing methods based on ResNet
CN108334909B (en) * 2018-03-09 2020-06-16 上海天数智芯半导体有限公司 Cervical cancer TCT digital slice data analysis system based on ResNet
CN108345871A (en) * 2018-03-20 2018-07-31 宁波江丰生物信息技术有限公司 A kind of cervical carcinoma slice recognition methods
CN108520206A (en) * 2018-03-22 2018-09-11 南京大学 A kind of fungi microscopic image identification method based on full convolutional neural networks
CN108510482A (en) * 2018-03-22 2018-09-07 姚书忠 Cervical carcinoma detection method, device, equipment and medium based on gynecatoptron image
CN108395986A (en) * 2018-03-23 2018-08-14 余晖 Human papilloma virus automatic parting direction detection device based on deep learning
CN108665463A (en) * 2018-03-30 2018-10-16 哈尔滨理工大学 A kind of cervical cell image partition method generating network based on confrontation type
CN108765374B (en) * 2018-04-27 2022-03-29 华南理工大学 Method for screening abnormal nuclear area in cervical smear image
CN108765374A (en) * 2018-04-27 2018-11-06 华南理工大学 A kind of method of abnormal core region screening in cervical smear image
WO2019238072A1 (en) * 2018-06-13 2019-12-19 深圳市商汤科技有限公司 Normalization method, apparatus and device for deep neural network, and storage medium
CN109190441A (en) * 2018-06-21 2019-01-11 丁彦青 Female genital tract cells pathology intelligent method for classifying, diagnostic equipment and storage medium
CN109190441B (en) * 2018-06-21 2022-11-08 丁彦青 Intelligent classification method, diagnostic instrument and storage medium for female genital tract cell pathology
CN109460720A (en) * 2018-10-25 2019-03-12 中科院成都信息技术股份有限公司 Ballot paper recognition methods based on convolutional neural networks
CN109255340A (en) * 2018-10-29 2019-01-22 东北大学 It is a kind of to merge a variety of face identification methods for improving VGG network
CN109325586A (en) * 2018-12-05 2019-02-12 北京航空航天大学合肥创新研究院 Deep neural network system based on composite object function
CN109325586B (en) * 2018-12-05 2022-03-04 北京航空航天大学合肥创新研究院 System for denoising electroencephalogram signal
CN109948429A (en) * 2019-01-28 2019-06-28 上海依智医疗技术有限公司 Image analysis method, device, electronic equipment and computer-readable medium
CN109993122A (en) * 2019-04-02 2019-07-09 中国石油大学(华东) A kind of pedestrian based on depth convolutional neural networks multiplies staircase anomaly detection method
CN110009050A (en) * 2019-04-10 2019-07-12 杭州智团信息技术有限公司 A kind of classification method and device of cell
CN110070175A (en) * 2019-04-12 2019-07-30 北京市商汤科技开发有限公司 Image processing method, model training method and device, electronic equipment
CN110070175B (en) * 2019-04-12 2021-07-02 北京市商汤科技开发有限公司 Image processing method, model training method and device and electronic equipment
CN110946552A (en) * 2019-10-30 2020-04-03 南京航空航天大学 Cervical cancer pre-lesion screening method combining spectrum and image
CN111274903A (en) * 2020-01-15 2020-06-12 合肥工业大学 Cervical cell image classification method based on graph convolution neural network
CN111274903B (en) * 2020-01-15 2022-12-06 合肥工业大学 Cervical cell image classification method based on graph convolution neural network
CN111524132A (en) * 2020-05-09 2020-08-11 腾讯科技（深圳）有限公司 Method, device and storage medium for identifying abnormal cells in sample to be detected
CN111783571A (en) * 2020-06-17 2020-10-16 陕西中医药大学 Cervical cell automatic classification model establishment and cervical cell automatic classification method
CN114266717A (en) * 2020-09-25 2022-04-01 天津科技大学 Parallel capsule network cervical cancer cell detection method based on Inception module
CN114266717B (en) * 2020-09-25 2024-08-27 天津科技大学 Parallel capsule network cervical cancer cell detection method based on Inception modules
CN112613505A (en) * 2020-12-18 2021-04-06 安徽丹姆斯生物科技有限公司 Cell micronucleus identification, positioning and counting method based on deep learning
CN112613505B (en) * 2020-12-18 2024-08-09 合肥码鑫生物科技有限公司 Cell micronucleus identification, positioning and counting method based on deep learning
CN113222044A (en) * 2021-05-25 2021-08-06 合肥工业大学 Cervical fluid-based cell classification method based on ternary attention and scale correlation fusion
CN114708589A (en) * 2022-04-02 2022-07-05 大连海事大学 Cervical cell classification method based on deep learning
CN115359483A (en) * 2022-08-09 2022-11-18 杭州迪英加科技有限公司 Training method for cervical fluid-based cell classification

Similar Documents

Publication Publication Date Title
CN106780466A (en) A kind of cervical cell image-recognizing method based on convolutional neural networks
CN111369563B (en) Semantic segmentation method based on pyramid void convolutional network
CN106203395B (en) Face attribute recognition method based on multitask deep learning
CN106169081B (en) A kind of image classification and processing method based on different illumination
CN107016406A (en) The pest and disease damage image generating method of network is resisted based on production
CN106650786A (en) Image recognition method based on multi-column convolutional neural network fuzzy evaluation
CN106446942A (en) Crop disease identification method based on incremental learning
CN106156793A (en) Extract in conjunction with further feature and the classification method of medical image of shallow-layer feature extraction
CN108615010A (en) Facial expression recognizing method based on the fusion of parallel convolutional neural networks characteristic pattern
CN109118485A (en) Digestive endoscope image classification based on multitask neural network cancer detection system early
CN107506740A (en) A kind of Human bodys' response method based on Three dimensional convolution neutral net and transfer learning model
CN106682704A (en) Method of disease image identification based on hybrid convolutional neural network fused with context information
CN109711426A (en) A kind of pathological picture sorter and method based on GAN and transfer learning
CN107563999A (en) A kind of chip defect recognition methods based on convolutional neural networks
CN107133616A (en) A kind of non-division character locating and recognition methods based on deep learning
CN108109160A (en) It is a kind of that interactive GrabCut tongue bodies dividing method is exempted from based on deep learning
CN107239514A (en) A kind of plants identification method and system based on convolutional neural networks
CN107066934A (en) Tumor stomach cell image recognition decision maker, method and tumor stomach section identification decision equipment
CN109117897A (en) Image processing method, device and readable storage medium storing program for executing based on convolutional neural networks
CN107133651A (en) The functional magnetic resonance imaging data classification method of subgraph is differentiated based on super-network
CN110555836A (en) Automatic identification method and system for standard fetal section in ultrasonic image
CN108053398A (en) A kind of melanoma automatic testing method of semi-supervised feature learning
CN111178121B (en) Pest image positioning and identifying method based on spatial feature and depth feature enhancement technology
CN107424145A (en) The dividing method of nuclear magnetic resonance image based on three-dimensional full convolutional neural networks
CN108960299A (en) A kind of recognition methods of multiclass Mental imagery EEG signals

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: 20170531