CN109712706A - A kind of observation method and device based on deep learning - Google Patents
A kind of observation method and device based on deep learning Download PDFInfo
- Publication number
- CN109712706A CN109712706A CN201811556398.1A CN201811556398A CN109712706A CN 109712706 A CN109712706 A CN 109712706A CN 201811556398 A CN201811556398 A CN 201811556398A CN 109712706 A CN109712706 A CN 109712706A
- Authority
- CN
- China
- Prior art keywords
- image data
- data set
- observation
- deep learning
- tongue
- 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
- Image Analysis (AREA)
Abstract
The observation method based on deep learning that the invention discloses a kind of, comprising: obtain the image data set for being used for observation；More wheel marks are carried out to described image data set according to the feature of described image data set, and annotation results are corresponded in different characteristic data set units；It is trained using the image data of different characteristic concentrations and corresponding diagnostic result as observation model of the training dataset to deep learning；It is confirmed whether to be labeled training dataset again according to training result；The image data of observation to be diagnosed is inputted in the observation model, the observation model exports corresponding diagnostic result, improves accuracy, the speed of model algorithm of the image data mark for observation.
Description
Technical field
This application involves automated diagnostic fields, and in particular to a kind of observation method based on deep learning is related to simultaneously
A kind of observation device based on deep learning.
Background technique
Chinese medicine is thousands of years of crystallizations of wisdom of the Chinese nation, and tcm diagnosis is the important component of theory of traditional Chinese medical science, " hope,
Hear, ask, cut " it is most common, is also the tcm diagnosis method to show unique characteristics, usually it is also regarded as the contracting of tcm treatment according to syndrome differentiation
Shadow, or even the synonym being known as to a certain extent for Chinese medicine.With the development of rapid development, the international exchange of modern medicine,
Health care's theory obtains unprecedented attention, and Chinese medicine has also gradually moved towards internationalization, during more and more common people recognize
The deficiency of doctor trained in Western medicine can be made up at many aspects by curing, and start to receive theory of traditional Chinese medical science, learning Chinese medicine knowledge.
With the fast development of information technology, Chinese medicine also gradually combines with information technology, in distance medical diagnosis mould
In formula, doctor does not need to diagnose face-to-face with patient, provides convenience for patient.However remote traditional Chinese medical diagnostic techniques
Development still have many limitations, such as the accuracy of the mark of the decent eigen of tongue picture and face, the speed of model algorithm, with
And it is promoted in clinic, universal and application.
Summary of the invention
The application provides a kind of observation method based on deep learning, improves the standard of the image data mark for observation
True property, the speed of model algorithm.
A kind of observation method based on deep learning of the application characterized by comprising
Obtain the image data set for being used for observation；
More wheel marks are carried out to described image data set according to the feature of described image data set, and annotation results are corresponding
Into different characteristic data set units；
Using the image data of different characteristic concentrations and corresponding diagnostic result as training dataset to depth
The observation model of habit is trained；It is confirmed whether to be labeled training dataset again according to training result；
The image data of observation to be diagnosed is inputted in the observation model, the observation model exports corresponding diagnosis
As a result.
Preferably, described to obtain the image data set for being used for observation, include at least a kind of following image data set:
Whole observation, local observation, the tongue inspection, inspection of excreta, the image data set for hoping infantile finger veinlet.
Preferably, the feature according to described image data set carries out more wheel marks to described image data set, comprising:
To each image data set according to tagsort, and filter out the image data of error message；
Using the feature locations of described image data set as the position of target area, it is labeled using figure or text；
Annotation results are checked, the sample for not meeting tagsort is filtered out.
Preferably, it is described to each image data set according to tagsort, specifically, including:
If image data set is tongue color, image data set can be divided into pink tongue, pale tongue, deep red tongue, pale purple tongue, tongue
Red and tongue is dark red.
It is preferably, described to correspond to annotation results in different characteristic data set units, further includes:
The image data of error message and the sample for not meeting classification are not belonged in any characteristic data set unit.
Preferably, the mask method, for artificial mark.
It is preferably, described to be confirmed whether to be labeled training dataset again according to training result, comprising:
It is detected using observation model of the detection data collection to trained deep learning；
If the accuracy rate of diagnosis of the observation model of trained deep learning is lower than the threshold value of setting, again to training
Data set is labeled, and filters out the sample for not meeting tagsort.
The application provides a kind of observation device based on deep learning simultaneously characterized by comprising
Acquiring unit, for obtaining the image data set for being used for observation；
Unit is marked, for carrying out more wheel marks to described image data set according to the feature of described image data set, and
Annotation results are corresponded in different characteristic data set units；
Training unit, for using the image data of different characteristic concentrations and corresponding diagnostic result as training number
The observation model of deep learning is trained according to collection；It is confirmed whether to mark training dataset again according to training result
Note；
Output unit, the image data for observation that will be to be diagnosed input in the observation model, the observation model
Export corresponding diagnostic result.
Preferably, the mark unit, comprising:
Classification subelement, is used for each image data set according to tagsort, and filter out the image of error message
Data；
Subelement is marked, for using figure using the feature locations of described image data set as the position of target area
Or text is labeled；
Check that subelement filters out the sample for not meeting tagsort for checking annotation results.
Preferably, the training unit, further includes:
Detection sub-unit, for being detected using observation model of the detection data collection to trained deep learning；
Subelement is screened, if the accuracy rate of diagnosis of the observation model for trained deep learning is lower than the threshold of setting
Value, then be again labeled training dataset, filter out the sample for not meeting tagsort.
A kind of observation method based on deep learning provided by the present application is more by being carried out to image data set according to feature
Wheel mark, then corresponds to annotation results in different characteristic data set units, the image that different characteristics is concentrated
Data are trained as observation model of the training dataset to deep learning, and finally by the observation model, output is corresponded to
Diagnostic result, improve accuracy, the speed of model algorithm of the image data mark for observation, while can be in clinic
It is middle to promote, popularize and apply.
Detailed description of the invention
Fig. 1 is a kind of observation method flow schematic diagram based on deep learning provided by the embodiments of the present application；
Fig. 2 be the invention relates to remove interference after lingual diagnosis data set；
Fig. 3 be the invention relates to the training data image data and correspondence of concentrating different characteristics to concentrate
Diagnostic result relation schematic diagram；
Fig. 4 be the invention relates to be made into label file using training dataset；
Fig. 5 is a kind of observation schematic device based on deep learning provided by the embodiments of the present application.
Specific embodiment
Many details are explained in the following description in order to fully understand the application.But the application can be with
Much it is different from other way described herein to implement, those skilled in the art can be without prejudice to the application intension the case where
Under do similar popularization, therefore the application is not limited by following public specific implementation.
Fig. 1 is please referred to, Fig. 1 is a kind of observation method flow signal based on deep learning provided by the embodiments of the present application
Figure, is described in detail the application providing method below with reference to Fig. 1.
Step S101 obtains the image data set for being used for observation.
Tcm inspection is doctor with vision to the progress such as all visible signs of Whole Body and part and effluent
It purposefully observes, to understand health or morbid state, referred to as observation.The content of observation specifically include that the mind of observer, color,
It is situations such as shape, state, tongue picture, channels, skin, face nine orifices and excreta, secretion, the shape of secretion, color, quality etc., existing
Observation is divided into whole observation, local observation, the tongue inspection, inspection of excreta, hopes five Xiang Xushu such as infantile finger veinlet.Lingual diagnosis and the colour inspection of face
Though belonging to head face, because tongue picture, complexion reflection viscera are more accurate.Practical value is higher.Thus form complexion examine,
The unique traditional diagnostic method of two Chinese medicines of lingual diagnosis.It is first exactly to obtain the image data set for being used for observation, in this application, just with face
Based on the mass data of colour inspection and lingual diagnosis, the present processes are described in detail.
Step S102 carries out more wheel marks to described image data set according to the feature of described image data set, and will mark
Note result corresponds in different characteristic data set units.
Tcm inspection is segmented into complexion and examines and lingual diagnosis two major classes.Facial diagnosis is exactly to know zang-fu diseases through facial echo area
With the diagnostic method of health status, thus rapid healing.Facial diagnosis mainly has three big features: complexion, facial radiance and lip color；Chinese medicine
Tongue picture feature is broadly divided into two major classes, first is that tongue inspection matter, second is that tongue inspection tongue fur.Tongue inspection matter specifically includes that tongue color and ligulate etc..Tongue color
Be broadly divided into six kinds: pale tongue is white, and pink tongue, tongue is red, and pale tongue is purple, and tongue is dark red, and tongue is deep red；Ligulate specifically includes that fat or thin, pricking method, tooth
Trace and crackle etc..Tongue inspection tongue fur is broadly divided into: coating nature and coating colour.Coating nature is divided into thickness tongue fur and rotten greasy tongue fur；Coating colour is broadly divided into six kinds:
The white phase of whitish tongue, yellow tongue fur, yellow tongue fur and, tongue fur is greyish black, tongue fur is few and tongue fur without.Specifically it is classified as follows shown in table 1.
1 face of table as with tongue picture feature
More wheel marks are carried out to described image data set then according to the feature of described image data set, before mark,
Some disturbing factors of rejection image data are also needed, for example, lingual diagnosis data set, will exclude the other factors other than tongue body
Interference, so, it needs to be split tongue body based on image processing techniques, the tongue region that can be detected using SSD, use is ellipse
Circle region is partitioned into tongue in proportion, effectively eliminates the interference such as neck clothes.Lingual diagnosis data set such as Fig. 2 institute after removing interference
Show.When being labeled, firstly, to each image data set, if lingual diagnosis data set and facial diagnosis data set are according to the feature of table 1
Classification, when being labeled the first round, filters out the image data of error message, in this application, mask method is
Artificial mark, marks the first round, can select two professional mark personnel distinguish self-responsibilities each tongue picture and
Face is labeled respectively as feature, can have a broad classification as feature to tongue picture face, and filter out the image of error message
Data, and annotation results are corresponded into corresponding file, that is, in different characteristic data set units.
In order to avoid the one-sidedness of individual's mark, second wheel two mark personnel of mark are together to the annotation results of the first round
It carries out unified inspection and marks again, this time need to pick out uncertain sample, individually put in one file, it will
The image data of error message and the sample for not meeting classification, do not belong in any characteristic data set unit.
Step S103, the image data and the corresponding feature that different characteristics is concentrated are as training dataset
The observation model of deep learning is trained；It is confirmed whether to be labeled training dataset again according to training result.
After the mark of preceding two-wheeled, the image for the error message in image data that different characteristics is concentrated
Filtering has been carried out in data and the sample for not meeting classification, it is already possible to as training dataset to the observation mould of deep learning
Type is trained, and before training, can enhance transform method using data to increase the amount of input data, specifically, can be with
Increase transform method using following several data: 1. rotation transformations: Random-Rotation image certain angle changes the court of picture material
To；2. turning-over changed: along horizontal or vertical direction flipped image；3. scale transformation: amplifying according to a certain percentage or contract
Small image；4. translation transformation: being translated on the image plane to image with certain direction；5. using random or artificially defined
Mode specify range of translation and translating step, translated along the horizontal or vertical direction, change the position of picture material；6. ruler
Degree transformation: it to image according to specified scale factor, zooms in and out, changes the size or fog-level of picture material.So
With regard to being trained using observation model of the training dataset to deep learning.
The relationship such as Fig. 3 institute for the image data and corresponding diagnostic result that training data concentrates different characteristics to concentrate
Show, by taking lingual diagnosis as an example, then, the image data for just different characteristics being used to concentrate and corresponding diagnostic result are as training
Data set is trained the observation model of deep learning, can include multiple features due to the image data that characteristic is concentrated,
Or by taking lingual diagnosis as an example, the feature of lingual diagnosis may include the red and yellow tongue fur of tongue, then just obtaining corresponding examine according to this two features
Disconnected result.
When being trained using observation model of the training dataset to deep learning, it is necessary first to do training dataset
At label file, (train.txt) as shown in Figure 4, and the txt file of production is put into the i.e. trained picture of corresponding file
It is placed in a file with trained .txt.Different characteristic data set unit in file i.e. step S102.Right
After the completion of the observation model training of deep learning, it is confirmed whether to be labeled training dataset again according to training result, instructs
Practice the result is that being detected by using observation model of the detection data collection to trained deep learning, if trained
The accuracy rate of diagnosis of the observation model of deep learning lower than setting threshold value, that is, deep learning observation model accuracy rate
It is not obviously improved, then being labeled again to training dataset, filters out the sample for not meeting tagsort, specifically
, the mark of the second wheel is manually marked again；Third round is manually labeled on the basis of original two mark personnel
Increase a more professional mark personnel, while the carry out sample mark of another subsystem, handles second emphatically and take turns and select
Uncertain sample out.Then detection data collection is reused to examine the observation model of trained deep learning
It surveys, if the accuracy rate of diagnosis of the observation model of trained deep learning is higher than the threshold value of setting, that is, deep learning
The accuracy rate of observation model, which has, to be obviously improved, then the training of the observation model of deep learning is completed,
Step S104 inputs the image data of observation to be diagnosed in the observation model, the observation model output
Corresponding diagnostic result.
After the completion of training of the previous step to the observation model of deep learning, model can be used, and only needing will be wait diagnose
The image data of observation input in the observation model, the observation model exports corresponding diagnostic result.
Meanwhile in order to enrich training sample set, the accuracy rate of the observation model of deep learning is improved, it can be continuous perfect
Training sample set carries out model it is then possible to which the step of passing through front again carries out more wheel marks to the image data set of observation
Training detection etc., the accuracy rate of the observation model of deep learning provided by the present application is continuously improved.
The application provides a kind of observation device 500 based on deep learning simultaneously, please refer to Fig. 5 it is characterised by comprising:
Acquiring unit 510, for obtaining the image data set for being used for observation；
Unit 520 is marked, for carrying out more wheel marks to described image data set according to the feature of described image data set,
And annotation results are corresponded in different characteristic data set units；
Training unit 530, for using the image data of different characteristic concentrations and corresponding diagnostic result as instruction
Practice data set to be trained the observation model of deep learning；It is confirmed whether to carry out training dataset again according to training result
Mark；
Output unit 540, the image data for observation that will be to be diagnosed input in the observation model, the observation mould
Type exports corresponding diagnostic result.
Preferably, the mark unit, comprising:
Classification subelement, is used for each image data set according to tagsort, and filter out the image of error message
Data；
Subelement is marked, for using figure using the feature locations of described image data set as the position of target area
Or text is labeled；
Check that subelement filters out the sample for not meeting tagsort for checking annotation results.
Preferably, the training unit, further includes:
Detection sub-unit, for being detected using observation model of the detection data collection to trained deep learning；
Subelement is screened, if the accuracy rate of diagnosis of the observation model for trained deep learning is lower than the threshold of setting
Value, then be again labeled training dataset, filter out the sample for not meeting tagsort.
A kind of observation method based on deep learning provided by the present application is more by being carried out to image data set according to feature
Wheel mark, then corresponds to annotation results in different characteristic data set units, the image that different characteristics is concentrated
Data are trained as observation model of the training dataset to deep learning, and finally by the observation model, output is corresponded to
Diagnostic result, improve accuracy, the speed of model algorithm of the image data mark for observation, while can be in clinic
It is middle to promote, popularize and apply.
The above embodiments are merely illustrative of the technical scheme of the present invention and are not intended to be limiting thereof, although referring to above-described embodiment pair
The present invention is described in detail, those of ordinary skill in the art still can to a specific embodiment of the invention into
Row modifies perhaps equivalent replacement and these exist without departing from any modification of spirit and scope of the invention or equivalent replacement
Apply within pending claims of the invention.
Claims (10)
1. a kind of observation method based on deep learning characterized by comprising
Obtain the image data set for being used for observation；
More wheel marks are carried out to described image data set according to the feature of described image data set, and annotation results are corresponded to not
In same characteristic data set unit；
Using the image data of different characteristic concentrations and corresponding diagnostic result as training dataset to deep learning
Observation model is trained；It is confirmed whether to be labeled training dataset again according to training result；
The image data of observation to be diagnosed is inputted in the observation model, the observation model exports corresponding diagnosis knot
Fruit.
2. at least being wrapped the method according to claim 1, wherein described obtain the image data set for being used for observation
Include a kind of following image data set:
Whole observation, local observation, the tongue inspection, inspection of excreta, the image data set for hoping infantile finger veinlet.
3. method according to claim 1 or 2, which is characterized in that the feature according to described image data set is to institute
It states image data set and carries out more wheel marks, comprising:
To each image data set according to tagsort, and filter out the image data of error message；
Using the feature locations of described image data set as the position of target area, it is labeled using figure or text；
Annotation results are checked, the sample for not meeting tagsort is filtered out.
4. according to the method described in claim 3, it is characterized in that, it is described to each image data set according to tagsort,
Specifically, including:
If image data set is tongue color, it is red that image data set can be divided into pink tongue, pale tongue, deep red tongue, pale purple tongue, tongue, with
And tongue is dark red.
5. the method according to claim 1, wherein described correspond to annotation results different characteristic data sets
In unit, further includes:
The image data of error message and the sample for not meeting classification are not belonged in any characteristic data set unit.
6. the method according to claim 1, wherein the mask method, marks to be artificial.
7. the method according to claim 1, wherein described be confirmed whether according to training result to training dataset
It is labeled again, comprising:
It is detected using observation model of the detection data collection to trained deep learning；
If the accuracy rate of diagnosis of the observation model of trained deep learning is lower than the threshold value of setting, again to training data
Collection is labeled, and filters out the sample for not meeting tagsort.
8. a kind of observation device based on deep learning characterized by comprising
Acquiring unit, for obtaining the image data set for being used for observation；
Unit is marked, for carrying out more wheel marks to described image data set according to the feature of described image data set, and will mark
Note result corresponds in different characteristic data set units；
Training unit, for using the image data of different characteristic concentrations and corresponding diagnostic result as training dataset
The observation model of deep learning is trained；It is confirmed whether to be labeled training dataset again according to training result；
Output unit, the image data for observation that will be to be diagnosed input in the observation model, the observation model output
Corresponding diagnostic result.
9. device according to claim 8, which is characterized in that the mark unit, comprising:
Classification subelement, is used for each image data set according to tagsort, and filter out the image data of error message；
Subelement is marked, for using figure or text using the feature locations of described image data set as the position of target area
Word is labeled；
Check that subelement filters out the sample for not meeting tagsort for checking annotation results.
10. device according to claim 8, which is characterized in that the training unit, further includes:
Detection sub-unit, for being detected using observation model of the detection data collection to trained deep learning；
Subelement is screened, if the accuracy rate of diagnosis of the observation model for trained deep learning is lower than the threshold value of setting,
Then training dataset is labeled again, filters out the sample for not meeting tagsort.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811556398.1A CN109712706A (en) | 2018-12-19 | 2018-12-19 | A kind of observation method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811556398.1A CN109712706A (en) | 2018-12-19 | 2018-12-19 | A kind of observation method and device based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109712706A true CN109712706A (en) | 2019-05-03 |
Family
ID=66256942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811556398.1A Pending CN109712706A (en) | 2018-12-19 | 2018-12-19 | A kind of observation method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109712706A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110251084A (en) * | 2019-06-21 | 2019-09-20 | 福州数据技术研究院有限公司 | A kind of detection of tongue picture and recognition methods based on artificial intelligence |
| CN111105397A (en) * | 2019-12-19 | 2020-05-05 | 天津慧医谷科技有限公司 | Traditional Chinese medicine tongue picture data rapid labeling method and system based on deep learning |
| CN112652404A (en) * | 2020-12-30 | 2021-04-13 | 华南师范大学 | Traditional Chinese medicine self-service dialectical monitoring method based on big data artificial intelligence and robot |
| CN117315357A (en) * | 2023-09-27 | 2023-12-29 | 广东省新黄埔中医药联合创新研究院 | Image recognition method and related device based on traditional Chinese medicine deficiency-excess syndrome differentiation classification |
Citations (5)
| 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 |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| CN106683087A (en) * | 2016-12-26 | 2017-05-17 | 华南理工大学 | Coated tongue constitution distinguishing method based on depth neural network |
| CN106951925A (en) * | 2017-03-27 | 2017-07-14 | 成都小多科技有限公司 | Data processing method, device, server and system |
| CN107330889A (en) * | 2017-07-11 | 2017-11-07 | 北京工业大学 | A kind of traditional Chinese medical science tongue color coating colour automatic analysis method based on convolutional neural networks |
-
2018
- 2018-12-19 CN CN201811556398.1A patent/CN109712706A/en active Pending
Patent Citations (5)
| 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 |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| CN106683087A (en) * | 2016-12-26 | 2017-05-17 | 华南理工大学 | Coated tongue constitution distinguishing method based on depth neural network |
| CN106951925A (en) * | 2017-03-27 | 2017-07-14 | 成都小多科技有限公司 | Data processing method, device, server and system |
| CN107330889A (en) * | 2017-07-11 | 2017-11-07 | 北京工业大学 | A kind of traditional Chinese medical science tongue color coating colour automatic analysis method based on convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| 谢铮桂等: "基于免疫聚类的RBF神经网络在中医舌诊诊断中的应用", 《计算机应用与软件》, vol. 26, no. 04, 15 April 2009 (2009-04-15), pages 42 - 54 * |
| 郁磊等编著: "《短语识别与信息抽取的隐马尔可夫模型的方法研究》", 北京航空航天大学出版社, pages: 159 - 160 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110251084A (en) * | 2019-06-21 | 2019-09-20 | 福州数据技术研究院有限公司 | A kind of detection of tongue picture and recognition methods based on artificial intelligence |
| CN111105397A (en) * | 2019-12-19 | 2020-05-05 | 天津慧医谷科技有限公司 | Traditional Chinese medicine tongue picture data rapid labeling method and system based on deep learning |
| CN112652404A (en) * | 2020-12-30 | 2021-04-13 | 华南师范大学 | Traditional Chinese medicine self-service dialectical monitoring method based on big data artificial intelligence and robot |
| CN117315357A (en) * | 2023-09-27 | 2023-12-29 | 广东省新黄埔中医药联合创新研究院 | Image recognition method and related device based on traditional Chinese medicine deficiency-excess syndrome differentiation classification |
| CN117315357B (en) * | 2023-09-27 | 2024-04-30 | 广东省新黄埔中医药联合创新研究院 | Image recognition method and related device based on traditional Chinese medicine deficiency-excess syndrome differentiation classification |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109712706A (en) | A kind of observation method and device based on deep learning | |
| Raman et al. | Fundus photograph-based deep learning algorithms in detecting diabetic retinopathy | |
| CN108615051B (en) | Diabetic retina image classification method and system based on deep learning | |
| CN108056786B (en) | Bone age detection method and device based on deep learning | |
| WO2019180742A1 (en) | System and method for retinal fundus image semantic segmentation | |
| CN108665454A (en) | A kind of endoscopic image intelligent classification and irregular lesion region detection method | |
| CN110532424A (en) | A kind of lungs sound tagsort system and method based on deep learning and cloud platform | |
| CN108922599A (en) | A kind of accurate mask method of medical image lesion point based on MIL | |
| CN115862819B (en) | Medical image management method based on image processing | |
| Chen et al. | Missing teeth and restoration detection using dental panoramic radiography based on transfer learning with CNNs | |
| CN107423576A (en) | A kind of lung cancer identifying system based on deep neural network | |
| CN113139971A (en) | Tongue picture identification method and system based on artificial intelligence | |
| Wan et al. | An artificial intelligent risk classification method of high myopia based on fundus images | |
| CN109215021A (en) | A kind of cholelithiasis CT medical image method for quickly identifying based on deep learning | |
| CN109241963A (en) | Blutpunkte intelligent identification Method in capsule gastroscope image based on Adaboost machine learning | |
| Bragança et al. | Detection of glaucoma on fundus images using deep learning on a new image set obtained with a smartphone and handheld ophthalmoscope | |
| CN112863699B (en) | ESD preoperative discussion system based on mobile terminal | |
| CN109636864A (en) | A kind of tongue dividing method and system based on color correction Yu depth convolutional neural networks | |
| CN111967539B (en) | Recognition method and device for maxillofacial fracture based on CBCT database and terminal equipment | |
| Mao et al. | Deep Learning for Dental Diagnosis: A Novel Approach to Furcation Involvement Detection on Periapical Radiographs | |
| Chen et al. | Machine learning-guided prediction of central anterior chamber depth using slit lamp images from a portable smartphone device | |
| Hu et al. | Nurvid: A large expert-level video database for nursing procedure activity understanding | |
| CN113539476A (en) | Stomach endoscopic biopsy Raman image auxiliary diagnosis method and system based on artificial intelligence | |
| Hu et al. | Automatic diagnosis of infectious keratitis based on slit lamp images analysis | |
| Tobias et al. | Android Application for Chest X-ray Health Classification From a CNN Deep Learning TensorFlow Model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |