CN110867242A - Capsule endoscope image intelligent screening system - Google Patents
Capsule endoscope image intelligent screening system Download PDFInfo
- Publication number
- CN110867242A CN110867242A CN201910980905.2A CN201910980905A CN110867242A CN 110867242 A CN110867242 A CN 110867242A CN 201910980905 A CN201910980905 A CN 201910980905A CN 110867242 A CN110867242 A CN 110867242A
- Authority
- CN
- China
- Prior art keywords
- image
- screening
- module
- report
- capsule endoscope
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
Landscapes
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Endoscopes (AREA)
Abstract
The invention relates to the field of medical imaging, and discloses a capsule endoscope image intelligent screening system, which comprises an image control module: the system is used for loading capsule endoscope image data, and carrying out format standardization processing and initialization management; the image screening module: the system is used for screening the capsule endoscope image data through a deep learning algorithm model; the screening result processing module: the system comprises a module for executing inquiry and processing operation on the screening result of the image screening module, wherein the processing operation comprises confirmation and modification of positive images and confirmation of negative images; a report editing module: the report editing device is used for executing report editing operation, wherein the report editing operation comprises the importing of report images, the importing of case information and the editing of report characters; a report derivation module: for generating electronic reports or printing into paper reports. The invention can improve the diagnosis efficiency of medical endoscope images and can be widely applied to medical clinical diagnosis.
Description
Technical Field
The invention relates to the field of medical imaging, in particular to an intelligent capsule endoscope image screening system.
Background
In recent 20 years, the total population of China is continuously increased, so that the demand base quantity is increased; in the population structure, the population of the elderly with high incidence of diseases is continuously increased, so that the medical needs of everyone are increased; the improvement of the GDP for everyone brings the upgrading of consumption, residents need better and more medical services to improve the quality of life, and the demand of medical images is increased. Thus, the ever-rising medical needs drive the expanding medical imaging market.
Capsule endoscope case image has characteristics such as large in quantity, normal image quantity accounts for than big, leads to the doctor to read the piece time long, goes out report cycle length scheduling problem, along with artificial intelligence's more and more extensive application, consequently, needs to provide a capsule endoscope image intelligence screening system to improve the screening efficiency of capsule endoscope image.
Disclosure of Invention
Aiming at the defects that the capsule image screening efficiency is low and manual screening is needed in the prior art, the invention overcomes the defects of the prior art and solves the technical problems that: the utility model provides a capsule endoscope image intelligence screening system to improve the screening efficiency of capsule endoscope image.
In order to solve the technical problems, the invention adopts the technical scheme that: a capsule endoscope image intelligent screening system comprises:
the image control module: the capsule endoscope image data loading system is used for loading capsule endoscope image data, and performing format standardization processing and initialization management on the image data;
the image screening module: the system is used for screening capsule endoscope image data through a deep learning algorithm model, judging whether the image is a focus image or a suspected focus image according to morphological characteristics and edge smoothness characteristics on the capsule endoscope image, and classifying the focus type;
the screening result processing module: the system comprises a module for executing inquiry and processing operation on the screening result of the image screening module, wherein the processing operation comprises confirmation and modification of positive images and confirmation of negative images;
a report editing module: the report editing device is used for executing report editing operation, wherein the report editing operation comprises the importing of report images, the importing of case information and the editing of report characters;
a report derivation module: for generating electronic reports or printing into paper reports.
The intelligent capsule endoscope image screening system further comprises a human-computer interaction interface, wherein the human-computer interaction interface is used for displaying the screening result of the image screening module, inputting a screening command, and inputting a processing command and a report editing operation command of the screening result.
The deep learning algorithm model in the image screening module comprises a capsule enteroscopy algorithm model and a capsule gastroscopy algorithm model which are respectively used for screening the capsule enteroscopy image and the capsule gastroscopy image to obtain a focus image and a suspected focus image.
The network model used by the deep learning algorithm model in the image screening module is CaffeNet, GoogleNet, VGG Net or ResNet network, and supervised learning is carried out through an SVM algorithm so as to screen the capsule endoscope image data.
The specific method for training the deep learning algorithm model comprises the following steps: the method comprises the steps of pre-training a model on a general image classification big data set IMAGENET, and then performing fine tuning training on the trained model by using capsule endoscopy training set data.
The capsule endoscopy data set training set data is divided into four types of data sets, namely normal, mild, moderate and severe.
Compared with the prior art, the invention has the following beneficial effects: the invention provides an intelligent capsule endoscope image screening system, which can automatically screen capsule endoscope images through a deep learning model to obtain suspected focuses and focus influences, can also perform query and processing operations on screening results of an image screening module, edit reports, generate electronic reports and print paper reports, and improves the efficiency of medical image diagnosis;
in addition, in the process of establishing the deep learning model, irrelevant features can be learned through overfitting, so that the model lacks generalization capability and is always a great problem in machine learning. Large data sets can mitigate overfitting, which requires as much labeled data as possible, while pictures from various angles can help, except most normal. But in the medical field, it is very difficult to acquire large labeled data sets. In the invention, the problem can be solved to a certain extent by adopting transfer learning, and the problem that a large amount of medical training data with labels cannot be obtained is solved by pre-training on a general image classification big data set IMAGENET and fine-tuning a model obtained by pre-training by using a capsule endoscope picture.
Drawings
Fig. 1 is a schematic structural diagram of a capsule endoscope image intelligent screening system according to the present invention;
FIG. 2 is a flowchart illustrating the operation of an image screening module according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a human-machine interface according to an embodiment of the invention;
FIG. 4 is a diagram of another human-machine interface according to an embodiment of the invention.
Detailed Description
In order to make the technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to specific embodiments and accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention; all other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, an embodiment of the present invention provides a capsule endoscope image intelligent screening system, which includes an image control module, an image screening module, a screening result processing module, a report editing module, and a report exporting module.
The image control module is used for loading capsule endoscope image data, and performing format standardization processing and initialization management on the image data; the image screening module is used for screening the capsule endoscope image data through the deep learning algorithm model, judging whether the image is a focus image or a suspected focus image according to morphological characteristics and edge smoothness characteristics on the capsule endoscope image, and classifying the focus types; the screening result processing module is used for inquiring and processing the screening result of the image screening module, and the processing operation comprises confirmation and modification of positive images and confirmation of negative images; the report editing module is used for executing report editing operation, wherein the report editing operation comprises the importing of report images, the importing of case information and the editing of report characters; the report export module is used for generating an electronic report or printing a paper report.
Specifically, in the embodiment of the present invention, a workflow diagram of the image screening module is shown in fig. 2, the image screening module firstly performs two-class screening on the image, determines whether the image is negative or positive, and then performs multi-class fine screening and accurate location of a lesion on the positive image, where the multi-class fine screening is to determine what type of disease, such as polyp and ulcer, is based on the characteristics of the disease type, the number of screened lesions, and the location of the positive image; accurate location of the focus means: and after the suspected focus image is obtained, recording the site of the suspected focus, and judging the disease category of the case according to the focus type and the site after the whole image is screened. In addition, the screening result is displayed through a human-computer interaction interface, and a doctor can perform review or modification on the screening result through the screening result processing module.
Specifically, the intelligent capsule endoscope image screening system provided by the embodiment further includes a human-computer interaction interface, where the human-computer interaction interface is used to display the screening result of the image screening module, and is also used to input a screening command, and to input a processing command and a report editing operation command of the screening result. As shown in fig. 3 to 4, which are schematic diagrams of a human-computer interaction interface in this embodiment, a screening result may be displayed through the interface of fig. 3, and an image may also be marked through the interface of fig. 4.
Specifically, since the capsule endoscopy image includes a enteroscope and a gastroscope, in this embodiment, the deep learning algorithm model in the image screening module includes a capsule enteroscope algorithm model and a capsule gastroscope algorithm model, which are respectively used for screening the capsule enteroscope image and the capsule gastroscope image to obtain an image with a focus and a suspected focus.
Specifically, in this embodiment, the network model used by the deep learning algorithm model in the image screening module may be a cafnenet, GoogleNet, VGG Net or ResNet network, which all have superior performance in computer vision classification, recognition and other problems; the weights learned from the large-scale image classification data set ImageNet are used as initial values of network weights through a transfer learning method, and then the pre-trained networks are finely adjusted on the images of the capsule endoscopy, so that the capsule endoscopy influence classification and identification tasks can be performed. That is to say, the specific method for training the deep learning algorithm model is as follows: the method comprises the steps of pre-training a model on a general image classification big data set IMAGENET, and then performing fine tuning training on the trained model by using capsule endoscopy training set data.
Based on long-term practical experience, the method uses an SVM (support Vector machine) algorithm for supervised learning so as to realize classification, and cases of patients are classified into A, B, C grades according to a Child-Pugh improved classification standard so as to classify a data set into four categories of normal, mild, moderate and severe. In the actual classification process, cases are compared and sorted according to the matching degree and the similarity degree of different indexes by the same method, and the most suspected focus site is selected for marking.
The test is carried out by selecting 30 normal cases and about 160 abnormal cases, and selecting 10628 marked images in total, wherein the training data: 6540; and (3) verifying data: 2126; test data: 1062, in a ratio of 7: 2: 1. the algorithm results are shown in table 1, and in the test stage, the accuracy rates of GoogleNet, VGG Net and ResNet are 99.8%, 99.15% and 99.07%, that is, the accuracy rates are stably improved on the existing data as the network model is continuously updated and advanced.
TABLE 1 test data for each network model at the test stage
| Network model name | Accuracy of measurement | Speed of rotation | Size of model |
| GoogleNet | 99.8％ | 9-10fps | 51M |
| VGG Net | 99.15％ | 13-14fps | 527M |
| ResNet | 99.07％ | 9-10fps | 97M |
In addition, through statistics, the operation efficiency of the classification algorithm is about 90 pieces/second, and the operation time is about 11 minutes for small intestine capsule cases with the quantity of 6 ten thousand; for capsule gastroscope cases with the number of about 3 ten thousand, the operation time is about 5 minutes.
Algorithm sensitivity:
clinical data tests of the digestive endoscopy center of the Changhai hospital affiliated to the second military medical university show that the intelligent diagnosis performance of the Shinenet (Version1.1.0) in the capsule endoscopy is shown in Table 2.
TABLE 2 test sensitivity for various intestinal disorders
| Numbering | Diagnosis of | Testing the number of pictures | Sensitivity (true positive rate) |
| 1 | Ulcer of small intestine | 351 | 96.43％ |
| 2 | Duodenal bulbar inflammation small intestinal lymphangioma | 303 | 80％ |
| 3 | Malformation of small intestine enterocoel bleeding blood vessel | 286 | 75％ |
| 4 | Hookworm of small intestine | 296 | 71.43％ |
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
Claims (6)
1. The utility model provides a capsule endoscope image intelligence screening system which characterized in that includes:
the image control module: the capsule endoscope image data loading system is used for loading capsule endoscope image data, and performing format standardization processing and initialization management on the image data;
the image screening module: the system is used for screening capsule endoscope image data through a deep learning algorithm model, judging whether the image is a focus image or a suspected focus image according to morphological characteristics and edge smoothness characteristics on the capsule endoscope image, and classifying the focus type;
the screening result processing module: the system comprises a module for executing inquiry and processing operation on the screening result of the image screening module, wherein the processing operation comprises confirmation and modification of positive images and confirmation of negative images;
a report editing module: the report editing device is used for executing report editing operation, wherein the report editing operation comprises the importing of report images, the importing of case information and the editing of report characters;
a report derivation module: for generating electronic reports or printing into paper reports.
2. The system according to claim 1, further comprising a human-computer interface for displaying the screening result of the image screening module, inputting a screening command, and inputting a processing command and a report editing operation command of the screening result.
3. The system according to claim 1, wherein the deep learning algorithm model in the image screening module comprises a capsule enteroscopy algorithm model and a capsule gastroscopy algorithm model, which are respectively used for screening the capsule enteroscopy image and the capsule gastroscopy image to obtain the image with the focus and the image with the suspected focus.
4. The system of claim 1, wherein the network model used by the deep learning algorithm model in the image screening module is a CaffeNet, GoogleNet, VGG Net or ResNet network, which performs supervised learning by SVM algorithm to screen the image data of the capsule endoscope.
5. The system according to claim 1, wherein the deep learning algorithm model is trained by the following specific method: the method comprises the steps of pre-training a model on a general image classification big data set IMAGENET, and then performing fine tuning training on the trained model by using capsule endoscopy training set data.
6. The system of claim 5, wherein the training set data of the endoscopic capsule data set is divided into four types of data sets, normal, mild, moderate and severe.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910980905.2A CN110867242A (en) | 2019-10-16 | 2019-10-16 | Capsule endoscope image intelligent screening system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910980905.2A CN110867242A (en) | 2019-10-16 | 2019-10-16 | Capsule endoscope image intelligent screening system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110867242A true CN110867242A (en) | 2020-03-06 |
Family
ID=69652787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910980905.2A Withdrawn CN110867242A (en) | 2019-10-16 | 2019-10-16 | Capsule endoscope image intelligent screening system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110867242A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364926A (en) * | 2020-11-17 | 2021-02-12 | 苏州大学 | Gastroscope picture classification method and device based on ResNet-50 time compression and storage medium |
| CN113035324A (en) * | 2021-03-19 | 2021-06-25 | 重庆金山医疗器械有限公司 | Capsule endoscope online film reading system and method |
| CN113808137A (en) * | 2021-11-19 | 2021-12-17 | 武汉楚精灵医疗科技有限公司 | Method, device, equipment and storage medium for screening image map of upper gastrointestinal endoscope |
-
2019
- 2019-10-16 CN CN201910980905.2A patent/CN110867242A/en not_active Withdrawn
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364926A (en) * | 2020-11-17 | 2021-02-12 | 苏州大学 | Gastroscope picture classification method and device based on ResNet-50 time compression and storage medium |
| CN113035324A (en) * | 2021-03-19 | 2021-06-25 | 重庆金山医疗器械有限公司 | Capsule endoscope online film reading system and method |
| CN113035324B (en) * | 2021-03-19 | 2024-05-24 | 重庆金山医疗技术研究院有限公司 | Online film reading system and film reading method for capsule endoscope |
| CN113808137A (en) * | 2021-11-19 | 2021-12-17 | 武汉楚精灵医疗科技有限公司 | Method, device, equipment and storage medium for screening image map of upper gastrointestinal endoscope |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886273B (en) | CMR image segmentation and classification system | |
| CN107591200B (en) | Bone age mark identification and evaluation method and system based on deep learning and image omics | |
| Chen et al. | SVIA dataset: A new dataset of microscopic videos and images for computer-aided sperm analysis | |
| Şevik et al. | Identification of suitable fundus images using automated quality assessment methods | |
| CN110245657B (en) | Pathological image similarity detection method and detection device | |
| CN110867242A (en) | Capsule endoscope image intelligent screening system | |
| Bhardwaj et al. | Diabetic retinopathy severity grading employing quadrant‐based Inception‐V3 convolution neural network architecture | |
| CN108922599A (en) | A kind of accurate mask method of medical image lesion point based on MIL | |
| Bibi et al. | Automated detection of diabetic retinopathy in fundus images using fused features | |
| Yue et al. | Automated endoscopic image classification via deep neural network with class imbalance loss | |
| CN111524124A (en) | Digestive endoscopy image artificial intelligence auxiliary system for inflammatory bowel disease | |
| Chen et al. | AI-PLAX: AI-based placental assessment and examination using photos | |
| CN110867233A (en) | System and method for generating electronic laryngoscope medical test reports | |
| CN109460717A (en) | Alimentary canal Laser scanning confocal microscope lesion image-recognizing method and device | |
| CN113222957A (en) | Multi-class focus high-speed detection method and system based on capsule lens image | |
| CN110009038A (en) | Training method, device and the storage medium of screening model | |
| CN114399465A (en) | Benign and malignant ulcer identification method and system | |
| Zhang et al. | A comprehensive review of methods based on deep learning for diabetes-related foot ulcers | |
| de Chauveron et al. | Artificial intelligence for oral squamous cell carcinoma detection based on oral photographs: A comprehensive literature review | |
| Vajravelu et al. | Machine learning techniques to detect bleeding frame and area in wireless capsule endoscopy video | |
| CN117350979A (en) | Arbitrary focus segmentation and tracking system based on medical ultrasonic image | |
| Arnold et al. | Indistinct frame detection in colonoscopy videos | |
| CN115457310A (en) | Traditional Chinese medicine tongue color gamut self-adaptive classification method based on meta-learning | |
| CN115760707A (en) | Skin damage image intelligent classification device based on self-supervision learning | |
| Giordano et al. | An AI-based framework for supporting large scale automated analysis of video capsule endoscopy |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200306 |