CN110335230A - A kind of endoscopic image lesion real-time detection method and device - Google Patents
A kind of endoscopic image lesion real-time detection method and device Download PDFInfo
- Publication number
- CN110335230A CN110335230A CN201910254178.1A CN201910254178A CN110335230A CN 110335230 A CN110335230 A CN 110335230A CN 201910254178 A CN201910254178 A CN 201910254178A CN 110335230 A CN110335230 A CN 110335230A
- Authority
- CN
- China
- Prior art keywords
- image
- lesion
- detection
- buffer area
- endoscopic
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B1/00—Instruments for performing medical examinations of the interior of cavities or tubes of the body by visual or photographical inspection, e.g. endoscopes; Illuminating arrangements therefor
- A61B1/04—Instruments for performing medical examinations of the interior of cavities or tubes of the body by visual or photographical inspection, e.g. endoscopes; Illuminating arrangements therefor combined with photographic or television appliances
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/10024—Color 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
- G06T2207/10—Image acquisition modality
- G06T2207/10068—Endoscopic 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Surgery (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Optics & Photonics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Processing (AREA)
- Endoscopes (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to medical image processing technology field, specially a kind of endoscopic image lesion real-time detection method and device.The method of the present invention includes the present image of capture card acquisition endoscopic assistance output；Present image pretreatment；Call the lesion region in lesion detection model inspection present image；Testing result post-processing；Image before and after lesion detection is stored in buffer area；Buffer area is detected, reads image from buffer area；In interface display lesion detection result.Apparatus of the present invention include scope capture apparatus, capture card, lesion detection program module, buffer area, display equipment；Capture card, lesion detection program module, buffer area are placed in main frame.The present invention can obtain the video of scope in real time and detect, and export the video before and after lesion detection, observe convenient for doctor.The present invention is to different endoscopic assistance good compatibilities, and installation is simple, plug and play, and doctor can be assisted to improve clinical diagnosis efficiency, has wide application prospects in actual clinical diagnosis.
Description
Technical field
The invention belongs to medical image processing technology fields, and in particular to a kind of endoscopic image lesion real-time detection method and
Device.
Background technique
Medical endoscope inspection is the technological means that sight glass feeding body cavities are checked and treated.With endoscopic diagnosis
The progress of technology, scope are widely used in enterogastric diseases, celiaca, pancreas and disease of biliary tract, respiratory disease, uropoiesis
The inspection of tract disease plays an important role in clinical diagnosis and treatment.For example, oesophagoscope and Sigmoidoscope are considered as oesophagus
Cancer, the Main Diagnosis method of colorectal cancer can be effectively reduced death by the detection and excision of early stage tumour
Risk [1-3].It is previous studies have shown that increase by 1.0% adenoma recall rate (ADR) [4-6], the risk of colorectal cancer can be made to drop
Low 3.0%.However, the effect of endoscopy is influenced very greatly by the experience of clinical endoscopic doctor, and then influence the accuracy rate of diagnosis.
In recent years, with the rapid development of artificial intelligence technology, in answering for the computer aided detection field (CADe)
Be increasingly taken seriously [7,8], at the same time, also there is the correlative study [9- of artificial intelligence auxiliary detection in scope field
11].Karkanis et al. for the first time detects the endoscopic image of colorectal polyp, verification and measurement ratio > 90% [9]. Misawa
Et al. propose a kind of real-time polyp detection method based on deep learning, be able to detect 94% polyp [10].2018, four
The People's Hospital Chuan Sheng, Harvard Medical School, BIDMC hospital and scientific & technical corporation Wision A.I. use the method for deep learning and preceding
The data verification of looking forward or upwards property has reached 94.38% [11] to the susceptibility of polyp detection in colonoscopy picture.
Although the studies above has theoretically reached higher accuracy rate, there is no landings in actual Medical treatment
In.If designing a kind of real-time lesion detection method and device of scope, theoretical research is applied to clinical diagnosis, is had very big
Clinical meaning.
Summary of the invention
For overcome the deficiencies in the prior art, the purpose of the present invention is to provide a kind of endoscopic image lesion real-time detection sides
Theory is applied to clinical diagnosis by method and device, realizes that lesion automatically analyzes in endoscopic image, for diagnosis provides ginseng
It examines.
Endoscopic image lesion real-time detection method provided by the invention, specific steps are as follows:
(1) present image of capture card acquisition endoscopic assistance output；
(2) pretreatment of present image；
(3) lesion region in lesion detection model inspection present image is called；
(4) post-processing of testing result；
(5) image before and after lesion detection is stored in buffer area；
(6) image is read from buffer area in timing detection buffer area；
(7) in interface display lesion detection result.
In step (1), the capture card is mounted in the host of computer.The method of capture card acquisition scope present image
Are as follows: the application programming interface for calling capture card obtains current endoscopic image.
In the present invention, main frame can use higher hardware configuration, such as: the central processing unit of 2.4GHz or more,
The running memory of 8GB or more, stream handle number 3584, core frequency 1481MHz or more, the video card of 11GB memory.
In step (2), the pretreated method of present image are as follows: cut in present image and detection process is incoherent
Part, incoherent part such as black background, patient related information, shooting time etc..
In step (3), the lesion detection model is the irregular lesion region detection model based on deep learning, knot
Structure is full convolutional neural networks, using the image of arbitrary size as input, generates Pixel-level identical with input picture size
Prediction result；To there is the data set of Pixel-level mark as training set, the network after the completion of training can be used for predicting interior model
Each pixel is the probability of abnormal pixel in mirror image；Given thresholdT, prediction probability is greater than threshold valueTPixel classifications be disease
Transshaping element, obtains irregular lesion detection result.
In the present invention, the full convolutional neural networks can be conspicuousness detection network, semantic segmentation network etc., such as Hou
[12] et al. conspicuousness detects network and semantic segmentation network SegNet [13], FCN [14], DeepLab [15] etc..
The model threshold obtained by heterogeneous networks structure, training dataTSetting it is different, need to rule of thumb set.Such as use Hou
[12] et al. conspicuousness detects network, and region existing for polyp is as abnormal area using in colonoscopy, by the pixel of polyp locations
It is labeled as 1, normal region is labeled as 0, and training obtains polyposis intestinalis detection model, threshold valueTIt is set in 0.05-0.95；With digestion
The road region Zao Ai is as abnormal area and is labeled as 1, and normal region is labeled as 0, and training obtains alimentary canal morning cancer detection model, threshold
ValueTIt is set in 0.1-0.8.
In the present invention, model treatment speed is in the 25 frames/more than second.
In step (4), the method for the post-processing detection result are as follows: according to the exposure mask that detection model exports, modify scope
The color for the pixel that exposure mask is 1 in image restores black background, patient related information, shooting time etc., in order to intuitively open up
Show the region of lesion.
In step (5), the method by the image deposit buffer area before and after lesion detection are as follows: endoscopic image detection finishes
Afterwards, it attempts to obtain buffer area lock；After obtaining buffer area lock, image before and after treatment is put into buffer area, release lock.
In step (6), the method for reading image from buffer area are as follows: at regular intervals, whether detection buffer area has
Image；If there is image, it tries obtain buffer area lock, obtain the image of taking-up buffer area after lock, release lock；It will test result
Being sent into buffer area helps to alleviate the unmatched problem of speed between lesion detection speed and display frequency.
In step (7), the method in interface display lesion detection result are as follows: there are two window, a windows at interface
For showing that the image for having testing result, another window are used to show the original image of acquisition, in order to user's comparison
Detect the image of front and back, auxiliary diagnosis.
Further, the realization of the method is by two Process flowcharts: detection procedure and show process, wherein detection procedure
It is responsible for step (1)-(5), show process is responsible for step (6), (7).Detection procedure successively executes step (1)-(5) and completes a figure
After the detection of picture, then step (1) is jumped back to, the present frame for continuing to obtain endoscopic assistance output is handled；Show process timing
Check that buffer area is read image and shown if there is image in buffer area.
The real-time lesion detection device of endoscopic image provided by the invention, including scope capture apparatus, capture card, lesion detection
Program module, buffer area, display equipment；Wherein capture card, lesion detection program module, buffer area are placed in main frame；
The output interface of scope capture apparatus is connected to the input interface of capture card, and the output interface of computer is connected to display equipment
Input interface.
Wherein, the lesion detection program module includes: present image pretreatment submodule, lesion detection model submodule
Block, testing result post-process submodule；These three word modules successively execute in the operation of step (2), step (3) and step (4)
Hold.
The installation steps of apparatus of the present invention are simple, plug and play, can be compatible with different scope capture apparatus.Scope figure
As the flow direction of data in a device are as follows: scope capture apparatus, capture card, lesion detection program module, buffer area, display equipment.
The beneficial effects of the present invention are: the present invention answers the advanced computer aided processing method based on artificial intelligence
For practicing.It does not need manually to participate in actual use, reducing human factor influences, and real-time output test result is
Diagnosis provides reference, improves the quality and efficiency of diagnosis.In addition, the present invention can be compatible with different endoscopic assistances, side
Just assembling, easily expansion, plug and play can popularize in clinical diagnosis.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention.
Fig. 2 is the device of the invention structure chart.
Fig. 3 is the options interface that embodiment detects different lesions.
Fig. 4 is the display interface figure of embodiment.
Specific embodiment
Embodiment of the present invention is described in detail below, but protection scope of the present invention is not limited to the implementation
Example.
The detection that lesion in real-time endoscopic image is realized using the process that Fig. 1 is introduced, the structure shown using Fig. 2
Figure configuration device.
The configuration of device:
(1) capture card is mounted in the PCIe interface of host of computer；
(2) endoscopic assistance used is Olympus, and endoscopic assistance has a DVI interface, image is exported, by the DVI of endoscopic assistance
Output interface is connected to the DVI input of capture card；
(3) DVI that the DVI output of main frame is connected to display is inputted.
The realization of lesion detection in real-time endoscopic image:
(1) present image of capture card acquisition endoscopic assistance output.The application programming interface (API) provided by capture card
Carry out Image Acquisition.The API of capture card is encapsulated into a class XIStream with C++ programming language.Class XIStream has as follows
Method:
VideoCaptureCount: the capture card number on existing equipment is returned to, for judging whether current device has acquisition
Card；
OpenVideo: video interface is opened；
CloseVideo: video interface is closed；
Start: start to acquire；
Stop: terminate acquisition；
SetMaxHeight: the maximum height of setting acquisition image (width is according to equal proportion scaling)；
GetCurrentFrame: current collected frame is obtained；
In order to which by the Python module in the library Boost of C Plus Plus, XIStream class is exported to pyd text across language call
Part is called for Python.
(2) pre-process present image, cut present image in and the incoherent part of detection process.General endoscopic assistance is clapped
The content taken the photograph is red, and in three Color Channels of red, green, blue of image, the pixel value of red channel is higher；Without relevant portion
It is mostly black and a small amount of white font, red channel pixel value is lower, can be cut according to red channel pixel value uncorrelated
Part, and record and cut position.
(3) lesion region in lesion detection model inspection present image is called.Lesion detection model is based on depth
The irregular lesion region detection model practised, structure are the conspicuousness detection net that Hou [12] et al. are published in 2017 CVPR
Network is realized with Python and TensorFlow frame, by largely there is the training of the data of mark；Training data is Pixel-level
Mark, therefore detection model can predict the classification of each pixel, export the exposure mask of lesion region, i.e., irregular prediction knot
Fruit；Two endoscopic image lesion detection models, respectively polyposis intestinalis detection model, alimentary canal morning cancer detection model are trained altogether；
When training colonoscopy polyp detection model, the pixel of polyp locations is labeled as 1 in training data, and normal region is labeled as 0, intestines
Mirror Image Adjusting subtracts mean value to 224 × 224 to size 300 × 300, random cropping is unified.Model parameter is by pre-training
The initialization of VGG-16 [16] disaggregated model.If initial learning rate is 0.0001, attenuation rate 0.9, every two periodic attenuation is primary.
With the method for small lot stochastic gradient descent, loss function is minimized.That criticizes is sized to 12, and loss function is the coke of Pixel-level
Point loss [17].After the completion of training, image to be detected is inputtedI, model output andIThe identical exposure mask of size；It covers
Value in film is bigger, and corresponding position is that the probability of polyp is bigger；Rule of thumb, given thresholdTIt is 0.95, whenM(i,j)>TWhen,
Then thinkI(i,j)Lesion has occurred；Otherwise it is assumed thatI(i,j)Normally, whereini, jFor the coordinate of pixel.The lesion region of detection
ForM(i,j)>TCorresponding region；
The training of alimentary canal morning cancer detection model and detection process are similar with polyposis intestinalis detection model, only training data and threshold
Value setting is different: the pixel of alimentary canal morning cancer position is labeled as 1 in training data, and normal region is labeled as 0, given thresholdTFor
0.2。
(4) post-processing detection modifies the pixel that exposure mask is 1 in endoscopic image as a result, according to the exposure mask that detection model exports
Color be green, recorded according to the cutting of preprocessing process, restore black background, patient related information, shooting time etc..
(5) image before and after lesion detection is stored in buffer area.After endoscopic image is processed, attempt to obtain buffer area
Lock；After obtaining buffer area lock, image before and after treatment is put into buffer area, then release lock.
(6) image is read from buffer area in timing detection buffer area.At regular intervals, whether detection buffer area has image,
If there is image, it tries obtain buffer area lock, obtain the image of taking-up buffer area after lock, release lock.
(7) in interface display lesion detection result.Interface includes two display windows, before one of window display detection
Image, the image after the display detection of window, interface is realized using the library PyQt5 of Python.
For realization process there are two thread, a thread, which is responsible for data acquisition, lesion detection, will test result is stored in buffering
Qu Zhong；Another thread is responsible for reading image from buffer area, is shown in interface.
Fig. 3 is the options interface that different lesions are detected in embodiment, shares polyposis intestinalis detection, alimentary canal morning cancer detection two
A option.
Fig. 4 is the display interface of embodiment, and left side black surround shows that present image, right side black surround show figure after testing
Picture.
Bibliography
[1] Seeff, L.C. et al. How many endoscopies are performed for colorectal
cancer screening Results from CDC's survey of endoscopic capacity.
Gastroenterology 127, 1670-7 (2004).
[2] Zauber, A.G. et al. Colonoscopic polypectomy and long-term prevention
of colorectal -cancer deaths. New England Journal of Medicine 366, 687-696
(2012).
[3] Brenner, H., Chang-Claude, J., Seiler, C.M., Rickert, A. &
Hoffmeister, M. Protection from colorectal cancer after colonoscopy: a
population-based, case-control study. Annals of internal medicine 154, 22-30
(2011).
[4] Corley, D.A. et al. Adenoma detection rate and risk of colorectal
cancer and death. New england journal of medicine 370, 1298-1306 (2014).
[5] Jemal, A., Center, M.M., DeSantis, C. & Ward, E.M. Global patterns of
cancer incidence and mortality rates and trends. Cancer Epidemiology and
Prevention Biomarkers 19, 1893-1907 (2010).
[6] Wu, S., Zhu, W., Thompson, P. & Hannun, Y.A. Evaluating intrinsic and
non-intrinsic cancer risk factors. Nature communications 9, 3490 (2018).
[7] Esteva, A.,Kuprel, B., Novoa, R.A., et al. Dermatologist-level
classification of skin cancer with deep neural networks. Nature 542, 115-118
(2017).
[8] Gulshan, V., Peng, L., Coram, M., et al. Development and Validation
of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal
Fundus Photographs. JAMA 316, 2402-2410 (2016).
[9] Karkanis, S.A., Iakovidis, D.K., Maroulis, D.E., et al. Computer-
aided tumor detection in endoscopic video using color wavelet features. IEEE
Transactions on Information Technology in Biomedicine 7, 141-152 (2003).
[10] Misawa, M., Kudo, S., Mori, Y., et al. Artificial Intelligence-
Assisted Polyp Detection for Colonoscopy: Initial Experience.
Gastroenterology 154, 2027-2029 (2018).
[11] Wang, P., Xiao, X., Glissen Brown, J. R., et al. Development and
validation of a deep-learning algorithm for the detection of polyps during
colonoscopy. Nature Biomedical Engineering 2, 741-748 (2018).
[12] Hou, Q., Cheng, M., Hu, X., Borji, A., Tu, Z. & Torr, P. H. Deeply
supervised salient object detection with short connections. IEEE Conference
on Computer Vision and Pattern Recognition, 3203-3212 (2017).
[13] Badrinarayanan, V., Kendall, A., Cipolla, A. SegNet: A Deep
Convolutional Encoder-Decoder Architecture for Image Segmentation. IEEE
Transactions on Pattern Analysis & Machine Intelligence, 39, 2481-2495
(2017).
[14] Long, J., Shelhamer, E. & Darrell, T. Fully convolutional networks
for semantic segmentation. IEEE International Conference on Computer Vision,
3431-3440 (2015).
[15] Chen, L., Papandreou, G., Kokkinos, I., Murphy, K. & Yuille, A. L.
DeepLab: Semantic image segmentation with deep convolutional nets, atrous
convolution, and fully connected CRFs. IEEE Transactions on Pattern Analysis
and Machine Intelligence 40, 834-848 (2018).
[16] Simonyan, K. & Zisserman A. Very deep convolutional networks for
large-scale image recognition. International Conference on Representation
Learning, (2014).
[17] Lin, T., Goyal, P., Girshick, R., He, K. & Dollar, P. Focal loss for
dense object detection. IEEE International Conference on Computer Vision,
2999-3007 (2017).。
Claims (8)
1. method is surveyed in a kind of endoscopic image lesion inspection in real time, which is characterized in that specific steps are as follows:
(1) present image of capture card acquisition endoscopic assistance output；
(2) pretreatment of present image；
(3) lesion region in lesion detection model inspection present image is called；
(4) post-processing of testing result；
(5) image before and after lesion detection is stored in buffer area；
(6) image is read from buffer area in timing detection buffer area；
(7) in interface display lesion detection result；
In step (1), the capture card is mounted in the host of computer；The method of capture card acquisition scope present image are as follows:
The application programming interface for calling capture card, obtains current endoscopic image；
In step (2), the preprocess method of the present image are as follows: in the cutting present image and incoherent portion of detection process
Point, incoherent part includes black background, patient related information, shooting time；
In step (3), the lesion detection model is the irregular lesion region detection model based on deep learning；Its structure is
Full convolutional neural networks generate the prediction of Pixel-level identical with input picture size using the image of arbitrary size as input
As a result；For model to there is the data set of Pixel-level mark as training set, the network after the completion of training can be used for predicting scope figure
Each pixel is the probability of abnormal pixel as in；Given thresholdT, prediction probability is greater than threshold valueTPixel classifications be lesion picture
Element obtains irregular lesion detection result；
In step (4), the post-processing approach of the testing result are as follows: according to the exposure mask that detection model exports, modify endoscopic image
The color for the pixel that middle exposure mask is 1, restores incoherent part, in order to intuitively show the region of lesion；
In step (5), the method by the image deposit buffer area before and after lesion detection are as follows: after endoscopic image detects,
It attempts to obtain buffer area lock；After obtaining buffer area lock, image before and after treatment is put into buffer area, release lock；
In step (6), the method for reading image from buffer area are as follows: at regular intervals, whether detection buffer area has image；
If there is image, it tries obtain buffer area lock, obtain the image of taking-up buffer area after lock, release lock；It will test result feeding
Buffer area helps to alleviate the unmatched problem of speed between lesion detection speed and display frequency.
2. endoscopic image lesion real-time detection method according to claim 1, which is characterized in that in step (7), it is described
The method of interface display lesion detection result are as follows: there are two window, a windows to be used to show the figure with testing result at interface
Picture, another window is used to show the original image of acquisition, in order to the image before and after user's contrasting detection, auxiliary diagnosis.
3. endoscopic image lesion real-time detection method according to claim 2, which is characterized in that by two Process flowcharts:
Detection procedure and show process, wherein detection procedure is responsible for step (1)-(5), and show process is responsible for step (6), (7)；Detect into
After Cheng Yici executes the detection that an image is completed in step (1)-(5), then step (1) is jumped back to, it is defeated to continue acquisition endoscopic assistance
Present frame out is handled；It reads image if there is image in buffer area and shows in show process regular check buffer area.
4. endoscopic image lesion real-time detection method according to claim 3, which is characterized in that the full convolutional Neural net
Network is that conspicuousness detects network, semantic segmentation network；Model thresholdTAccording to heterogeneous networks structure, training data, rule of thumb set
It is fixed.
5. endoscopic image lesion real-time detection method according to claim 4, which is characterized in that detect net for conspicuousness
Network, the pixel of polyp locations is labeled as 1, normal region is labeled as abnormal area by region existing for polyp using in colonoscopy
0, training obtains polyposis intestinalis detection model, threshold valueTIt is set in 0.05-0.95；Using alimentary canal morning cancer region as abnormal area,
And it is labeled as 1, normal region is labeled as 0, and training obtains alimentary canal morning cancer detection model, threshold valueTIt is set in 0.1-0.8.
6. endoscopic image lesion real-time detection method described in one of -5 according to claim 1, which is characterized in that the computer
The hardware configuration of host are as follows: the central processing unit of 2.4GHz or more, the running memory of 8GB or more, stream handle number 3584, core
The video card of frequency of heart 1481MHz or more, 11GB memory.
7. endoscopic image lesion real-time detection method described in one of -5 according to claim 1, which is characterized in that the lesion inspection
Model treatment speed is surveyed in the 25 frames/more than second.
8. a kind of endoscopic image lesion real-time detection apparatus based on one of -7 the methods according to claim 1, feature exist
In, including scope capture apparatus, capture card, lesion detection program module, buffer area, display equipment；Wherein capture card, lesion inspection
Survey program module, buffer area is placed in main frame；The input that the output interface of scope capture apparatus is connected to capture card connects
Mouthful, the output interface of computer is connected to the input interface of display equipment；Wherein, the lesion detection program module includes: to work as
Preceding image preprocessing submodule, lesion detection model submodule, testing result post-process submodule；These three word modules are successively held
The operation content of row step (2), step (3) and step (4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910254178.1A CN110335230A (en) | 2019-03-30 | 2019-03-30 | A kind of endoscopic image lesion real-time detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910254178.1A CN110335230A (en) | 2019-03-30 | 2019-03-30 | A kind of endoscopic image lesion real-time detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110335230A true CN110335230A (en) | 2019-10-15 |
Family
ID=68139708
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910254178.1A Pending CN110335230A (en) | 2019-03-30 | 2019-03-30 | A kind of endoscopic image lesion real-time detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110335230A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110974306A (en) * | 2019-12-17 | 2020-04-10 | 山东大学齐鲁医院 | System for discernment and location pancreas neuroendocrine tumour under ultrasonic endoscope |
| CN111047582A (en) * | 2019-12-17 | 2020-04-21 | 山东大学齐鲁医院 | Crohn's disease auxiliary diagnosis system under enteroscope based on degree of depth learning |
| CN111091559A (en) * | 2019-12-17 | 2020-05-01 | 山东大学齐鲁医院 | Depth learning-based auxiliary diagnosis system for small intestine sub-scope lymphoma |
| CN111932507A (en) * | 2020-07-31 | 2020-11-13 | 苏州慧维智能医疗科技有限公司 | Method for identifying lesion in real time based on digestive endoscopy |
| CN112102256A (en) * | 2020-08-22 | 2020-12-18 | 复旦大学 | Narrow-band endoscopic image-oriented cancer focus detection and diagnosis system for early esophageal squamous carcinoma |
| CN112200773A (en) * | 2020-09-17 | 2021-01-08 | 苏州慧维智能医疗科技有限公司 | Large intestine polyp detection method based on encoder and decoder of cavity convolution |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101778226A (en) * | 2010-01-25 | 2010-07-14 | 深圳市同洲电子股份有限公司 | High-definition image sawtooth-prevention method, device and digital television receiving terminal |
| CN102497552A (en) * | 2011-12-07 | 2012-06-13 | 北京大学 | Management control method of decoding buffer zone of video bit stream containing constructed images |
| US20150078615A1 (en) * | 2013-09-18 | 2015-03-19 | Cerner Innovation, Inc. | Marking and tracking an area of interest during endoscopy |
| CN104732708A (en) * | 2015-03-13 | 2015-06-24 | 河南城建学院 | Hydropower station warning system and method |
| CN107333164A (en) * | 2016-04-29 | 2017-11-07 | 北京学而思教育科技有限公司 | A kind of image processing method and device |
| CN108665454A (en) * | 2018-05-11 | 2018-10-16 | 复旦大学 | A kind of endoscopic image intelligent classification and irregular lesion region detection method |
-
2019
- 2019-03-30 CN CN201910254178.1A patent/CN110335230A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101778226A (en) * | 2010-01-25 | 2010-07-14 | 深圳市同洲电子股份有限公司 | High-definition image sawtooth-prevention method, device and digital television receiving terminal |
| CN102497552A (en) * | 2011-12-07 | 2012-06-13 | 北京大学 | Management control method of decoding buffer zone of video bit stream containing constructed images |
| US20150078615A1 (en) * | 2013-09-18 | 2015-03-19 | Cerner Innovation, Inc. | Marking and tracking an area of interest during endoscopy |
| CN104732708A (en) * | 2015-03-13 | 2015-06-24 | 河南城建学院 | Hydropower station warning system and method |
| CN107333164A (en) * | 2016-04-29 | 2017-11-07 | 北京学而思教育科技有限公司 | A kind of image processing method and device |
| CN108665454A (en) * | 2018-05-11 | 2018-10-16 | 复旦大学 | A kind of endoscopic image intelligent classification and irregular lesion region detection method |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110974306A (en) * | 2019-12-17 | 2020-04-10 | 山东大学齐鲁医院 | System for discernment and location pancreas neuroendocrine tumour under ultrasonic endoscope |
| CN111047582A (en) * | 2019-12-17 | 2020-04-21 | 山东大学齐鲁医院 | Crohn's disease auxiliary diagnosis system under enteroscope based on degree of depth learning |
| CN111091559A (en) * | 2019-12-17 | 2020-05-01 | 山东大学齐鲁医院 | Depth learning-based auxiliary diagnosis system for small intestine sub-scope lymphoma |
| CN111932507A (en) * | 2020-07-31 | 2020-11-13 | 苏州慧维智能医疗科技有限公司 | Method for identifying lesion in real time based on digestive endoscopy |
| CN112102256A (en) * | 2020-08-22 | 2020-12-18 | 复旦大学 | Narrow-band endoscopic image-oriented cancer focus detection and diagnosis system for early esophageal squamous carcinoma |
| CN112102256B (en) * | 2020-08-22 | 2022-04-12 | 复旦大学 | Narrow-band endoscopic image-oriented cancer focus detection and diagnosis system for early esophageal squamous carcinoma |
| CN112200773A (en) * | 2020-09-17 | 2021-01-08 | 苏州慧维智能医疗科技有限公司 | Large intestine polyp detection method based on encoder and decoder of cavity convolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110335230A (en) | A kind of endoscopic image lesion real-time detection method and device | |
| Pogorelov et al. | Deep learning and hand-crafted feature based approaches for polyp detection in medical videos | |
| CN110490856B (en) | Method, system, machine device, and medium for processing medical endoscope image | |
| CN106530279B (en) | A kind of image processing method and system | |
| CN108765392B (en) | Digestive tract endoscope lesion detection and identification method based on sliding window | |
| CN108615236A (en) | A kind of image processing method and electronic equipment | |
| CN109431523B (en) | Autism primary screening device based on non-social voice stimulation behavior paradigm | |
| CN110600122A (en) | Digestive tract image processing method and device and medical system | |
| CN108596237B (en) | A kind of endoscopic polyp of colon sorter of LCI laser based on color and blood vessel | |
| CN112102332A (en) | Cancer WSI segmentation method based on local classification neural network | |
| CN112750132A (en) | White blood cell image segmentation method based on dual-path network and channel attention | |
| Jin et al. | Countering spoof: towards detecting deepfake with multidimensional biological signals | |
| CN116030306A (en) | Pulmonary tissue pathology image type auxiliary classification method based on multilayer perceptron | |
| Su et al. | FeDNet: feature decoupled network for polyp segmentation from endoscopy images | |
| CN112885464B (en) | Internal nasal disease real-time auxiliary diagnosis and treatment system based on Att-Res2-CE-Net | |
| CN110473176A (en) | Image processing method and device, method for processing fundus images, electronic equipment | |
| WO2022005336A1 (en) | Noise-resilient vasculature localization method with regularized segmentation | |
| CN112308827A (en) | Hair follicle detection method based on deep convolutional neural network | |
| Garcia-Peraza-Herrera et al. | Interpretable fully convolutional classification of intrapapillary capillary loops for real-time detection of early squamous neoplasia | |
| CN111881803A (en) | Livestock face recognition method based on improved YOLOv3 | |
| CN104850861B (en) | Based on the fungal keratitis image-recognizing method of RX abnormality detection and texture analysis | |
| WO2020246676A1 (en) | System for automatic diagnosis of uterine cervical cancer | |
| Venkatalakshmi et al. | Graphical user interface for enhanced retinal image analysis for diagnosing diabetic retinopathy | |
| CN106778751A (en) | A kind of non-face ROI recognition methods and device | |
| Seetah et al. | A convolutional neural network approach to diabetic retinopathy detection and its automated classification |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191015 |