CN111710394A - Artificial intelligence assisted early gastric cancer screening system - Google Patents
Artificial intelligence assisted early gastric cancer screening system Download PDFInfo
- Publication number
- CN111710394A CN111710394A CN202010506897.0A CN202010506897A CN111710394A CN 111710394 A CN111710394 A CN 111710394A CN 202010506897 A CN202010506897 A CN 202010506897A CN 111710394 A CN111710394 A CN 111710394A
- Authority
- CN
- China
- Prior art keywords
- image
- benign
- labels
- artificial intelligence
- malignant
- 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
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
- 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
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Biomedical Technology (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Investigating Or Analysing Biological Materials (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The embodiment of the invention discloses an artificial intelligence auxiliary early gastric cancer screening system, which comprises: a slice image acquisition module for acquiring a plurality of gastroscope slice images; an image segment generation module for generating a plurality of image segments with benign tags and a plurality of image segments with malignant tags from the plurality of gastroscopic slice images; the model training module is used for obtaining a training set and a testing set according to the image fragments with benign labels and the image fragments with malignant labels, and generating a recognition model for recognizing whether the image fragments are malignant or benign according to the training set and the testing set; and the identification output module is used for identifying the target gastroscope section image according to the identification model and outputting an identification result. The invention can effectively assist pathologists in filtering negative specimens which do not need to be reviewed by professional pathologists, and is beneficial to quality control of routine pathological diagnosis.
Description
Technical Field
The embodiment of the invention relates to the field of diagnosis equipment, in particular to an artificial intelligence assisted early gastric cancer screening system.
Background
Gastric cancer is a common malignant tumor of the digestive tract. The incidence of gastric cancer in men is second to the malignancy in China, and second to lung cancer, and in women is third to the malignancy, after breast cancer and lung cancer. Early diagnosis is an important means to prevent tumor progression, guide therapy, and improve patient survival, especially gastric cancer.
Artificial intelligence is a rule for researching human intelligence activities, an artificial system with certain intelligence is constructed, and how to construct a set of equipment capable of screening early gastric cancer based on artificial intelligence is a problem to be solved urgently.
Disclosure of Invention
The embodiment of the invention aims to provide an artificial intelligence auxiliary early gastric cancer screening system, which is used for solving the problem that the workload for manually analyzing a gastroscope section image to determine gastric cancer is large.
In order to achieve the above object, the embodiments of the present invention mainly provide the following technical solutions:
the embodiment of the invention provides an artificial intelligence assisted early gastric cancer screening system, which comprises: a slice image acquisition module for acquiring a plurality of gastroscopic slice images, wherein the cancer regions in the plurality of gastroscopic slice images have marking information; an image segment generation module for generating a plurality of image segments with benign labels and a plurality of image segments with malignant labels according to the plurality of gastroscope slice images, wherein the image segment generation module respectively sets the malignant labels or the benign labels for each image segment according to the proportion of cancer parts in the image segment; the model training module is used for obtaining a training set and a testing set according to the image fragments with benign labels and the image fragments with malignant labels, and generating a recognition model for recognizing whether the image fragments are malignant or benign according to the training set and the testing set; and the identification output module is used for identifying the target gastroscope section image according to the identification model and outputting an identification result.
According to one embodiment of the invention, the image segment generation module scans and slices the plurality of gastroscopic slice images using a digital slice scanner at a preset magnification, converts each image segment into a preset format, and separately labels each image segment using a preset labeling tool.
According to one embodiment of the invention, the digital slice scanner is a KFBIO digital slice scanner.
According to an embodiment of the invention, the magnification is between 30 and 50 times.
According to an embodiment of the present invention, the preset labeling tool is ASAP optical software.
According to one embodiment of the present invention, the cancer region in the plurality of gastroscopic slice images is extracted by binarizing and filtering the plurality of gastroscopic slice images.
According to one embodiment of the invention, the image segment generation module is configured to add a malignant label to image segments having more than fifty percent of cancerous portions of the image segments and to add a benign label to image segments having no cancerous portions.
According to one embodiment of the invention, the model training module uses a neural network for training, testing and parameter adjustment to obtain the recognition model.
The technical scheme provided by the embodiment of the invention at least has the following advantages:
according to the artificial intelligence auxiliary early gastric cancer screening system provided by the embodiment of the invention, in a technical aspect, the subjective factors influencing the diagnosis of the gastric biopsy specimen are numerous, a pathologist can only judge the result through subjective experience, the artificial intelligence technical diagnosis standards are consistent, and the influence of the subjective factors can be avoided. In addition, the invention is not influenced by environmental conditions and fatigue degree, has good repeatability and high diagnosis efficiency, relieves doctors from a large amount of simple and repeated fussy work, not only lightens the workload of the doctors, but also relieves the pressure of patients for seeing a doctor, and solves the problems of uneven distribution of high-quality medical resources and the like.
From the aspect of screening, China is a high-incidence country of stomach cancer, the number of stomach biopsy specimens is large, the culture period of pathologists is long, and the number of people is seriously insufficient.
Drawings
Fig. 1 is a block diagram illustrating an artificial intelligence assisted early gastric cancer screening system according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided for illustrative purposes, and other advantages and effects of the present invention will become apparent to those skilled in the art from the present disclosure.
In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, interfaces, techniques, etc. in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
Fig. 1 is a block diagram illustrating an artificial intelligence assisted early gastric cancer screening system according to an embodiment of the present invention. As shown in fig. 1, the artificial intelligence assisted early gastric cancer screening system of the embodiment of the present invention comprises: a slice image acquisition module 100, an image segment generation module 200, a model training module 300, and a recognition output module 400.
The slice image acquisition module 100 is configured to acquire a plurality of gastroscopic slice images. The cancer regions in the plurality of gastroscopic slice images have labeling information.
In one example of the present invention, the plurality of gastroscopic slice images obtained include 665 cases of gastroscopic biopsy tissue slices, 288 cases of benign, 335 cases of malignant (mainly adenocarcinoma), including 135 cases of tubular adenocarcinoma, 40 cases of papillary adenocarcinoma, 60 cases of low-adhesive cancer (including 6 cases of in situ signet ring cell carcinoma), 20 cases of mucinous adenocarcinoma, and 50 cases of mixed adenocarcinoma (2 cases of tubular adenocarcinoma + hepatoid adenocarcinoma); 60 additional tumors of a specific type (such as undifferentiated carcinoma, lymphoma, etc.) were collected for further use. The pathological diagnosis of each section was evaluated by at least two experienced pathologists according to the revised wiener classification, determining the final consistent diagnosis for each case.
The invention establishes a definite classification standard of gastroscope biopsy specimens: benign specimens: including inflammation, polyps, low-grade adenomas/dysplasias; malignancy: high grade adenoma/dysplasia, intra-mucosal carcinoma, invasive carcinoma, indeterminate as tumor/dysplasia and specific types of tumor. Table 1 summarizes the correspondence of the revised vienna classification with the artificial intelligence output results.
TABLE 1 revised Vienna classifications and Artificial Intelligence ratings
| Revised Vienna classification | Description of diagnosis | Artificial intelligence grading |
| Classification 1 | Non-neoplastic/dysplasia (Negative for neoplasma/dysplasia) | 1 |
| Class 2 | Indeterminate as tumor/dysplasia (Indexinite for neoplasma ™) dysplasia) | 2 |
| Class 3 | Low grade adenoma/dysplasia (Low-grade adenoma/d)ysplasia) | 1 |
| Classification 4.1 | High grade adenoma/dysplasia (High-grade adenoma/dyssplasia) | 2 |
| Classification 4.2 | Non-invasive carcinoma (nonanvasive carcinosoma) | 2 |
| Classification 4.3 | Suspicious invasive carcinoma (Suspion for invasive carcinoma) | 2 |
| Classification 5.1 | Mucosal carcinoma (Intramucosal carcinosoma) | 2 |
| Classification 5.2 | Submucosal invasive carcinoma (Submucosal invasive carcinoma) | 2 |
The image segment generation module 200 is configured to generate a plurality of image segments with benign tags and a plurality of image segments with malignant tags from a plurality of gastroscopic slice images. The image segment generation module is used for setting a malignant label or a benign label for each image segment according to the proportion of cancer parts in the image segments.
Specifically, the image slice generation module 200 digitizes the entire collection of slices and scans the slices using a KFBIO digital slice scanner (model KF-Pro-005) at between 30-50 times magnification, preferably 40 times magnification (0.23 microns/pixel) to generate a plurality of image slices.
The image segment generating module 200 converts each image segment into a preset format, and labels each image segment with a preset labeling tool. Illustratively, the image segments are in the format of KFB, which is converted into standard image file grid TIF, which applies an annotation tool (ASAP) to label the image for classification, to calibrate tumor cells of various histological types, and to enter the database.
The model training module 300 is configured to obtain a training set and a testing set according to a plurality of image segments with benign labels and a plurality of image segments with malignant labels, and generate a recognition model for recognizing whether an image segment is malignant or benign according to the training set and the testing set, for example, by using a neural network for training and testing.
Specifically, the model training module 300 selects 274 stomach cancer pathological section images with annotations and 178 benign section images. From these, a portion of gastric cancer slices and benign slices were randomly selected as a test set to verify the classification ability of the model.
After the benign slice is subjected to image preprocessing, the labels of the generated image segments are benign. After image preprocessing of cancer slices, image segments generated in the labeled regions and having cancer portions accounting for 50% or more are labeled as malignant, and image segments generated in the unlabeled normal regions (not including cancer portions) are labeled as benign.
The number of benign image segments generated from the above data was 173329, and the number of cancer image segments was 915636. And randomly selecting 90% of the total number of the image fragments as a training set for training the model. 10% of the total number of image segments was used as the validation set for tuning parameters.
Randomly selected from the data set, a part of gastric cancer slices and benign slices were used as a test set to verify the classification ability of the model. And (3) taking the image segments generated after the test set slices are preprocessed as input, further carrying out model classification processing to obtain a classification result of each image segment, if the classification result meets the requirements, obtaining a recognition model, and if the classification result does not meet the requirements, continuing training the model until the requirements are met.
The recognition output module 400 is used for recognizing the target gastroscope section image according to the recognition model and outputting the recognition result.
In one example of the present invention, the specific process of the artificial intelligence assisted early gastric cancer screening system is as follows:
pretreatment: extracting the effective area of the case image through an image processing algorithm such as binarization, filtering and the like, segmenting the effective area according to coordinates to obtain image segments, and adopting the image segments of pixels 608 × 608 to be scaled into pixels 304 × 304 for training.
When the image segments are cut, attention should be paid to the use of an overlapping method for slicing, that is, the step size is selected to be half the length and width (152 pixels) of the image segment, so that the situation that the effective area in the image segment is too small can be avoided to a great extent and the data set can be enlarged, and then each image segment is subjected to color equalization to remove the influence of the dyeing degree on the image.
The prediction analysis process comprises the following steps: and cutting the predicted pictures into image segments, sending the image segments into a network, screening out pictures which are possibly cancer according to a certain threshold value, and recording the pictures into an xml file to finally obtain a predicted result picture.
Through practical tests, the false negative rate of the artificial intelligence assisted early gastric cancer screening system is 0.012, and the false positive rate of the artificial intelligence assisted early gastric cancer screening system is 0.001. Most of the stroma is misrecognized as cancer, probably because the stroma (such as vascular endothelial cells, histiocytes, lymphocytes and the like) is similar to the characteristics of low-differentiation cancer, and the artificial intelligence assisted early gastric cancer screening system is sometimes difficult to identify. Relevant cases were sought for the existing problems and these lesions were trained specifically to reduce the false negative and false positive rates.
According to the artificial intelligence auxiliary early gastric cancer screening system provided by the embodiment of the invention, in a technical aspect, the subjective factors influencing the diagnosis of the gastric biopsy specimen are numerous, a pathologist can only judge the result through subjective experience, the artificial intelligence technical diagnosis standards are consistent, and the influence of the subjective factors can be avoided. In addition, the invention is not influenced by environmental conditions and fatigue degree, has good repeatability and high diagnosis efficiency, relieves doctors from a large amount of simple and repeated fussy work, not only lightens the workload of the doctors, but also relieves the pressure of patients for seeing a doctor, and solves the problems of uneven distribution of high-quality medical resources and the like.
From the aspect of screening, China is a high-incidence country of stomach cancer, the number of stomach biopsy specimens is large, the culture period of pathologists is long, and the number of people is seriously insufficient.
In addition, other configurations and functions of the artificial intelligence assisted early gastric cancer screening system according to the embodiment of the present invention are known to those skilled in the art, and are not described in detail in order to reduce redundancy.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (8)
1. An artificial intelligence assisted early gastric cancer screening system, comprising:
a slice image acquisition module for acquiring a plurality of gastroscopic slice images, wherein the cancer regions in the plurality of gastroscopic slice images have marking information;
an image segment generation module for generating a plurality of image segments with benign labels and a plurality of image segments with malignant labels according to the plurality of gastroscope slice images, wherein the image segment generation module respectively sets the malignant labels or the benign labels for each image segment according to the proportion of cancer parts in the image segment;
the model training module is used for obtaining a training set and a testing set according to the image fragments with benign labels and the image fragments with malignant labels, and generating a recognition model for recognizing whether the image fragments are malignant or benign according to the training set and the testing set;
and the identification output module is used for identifying the target gastroscope section image according to the identification model and outputting an identification result.
2. The system of claim 1, wherein the image segment generation module scans and slices the plurality of gastroscopic slice images using a digital slice scanner at a preset magnification and converts each image segment into a preset format and individually labels each image segment using a preset labeling tool.
3. The artificial intelligence assisted early gastric cancer screening system of claim 2, wherein the digital slice scanner is a KFBIO digital slice scanner.
4. The artificial intelligence assisted early gastric cancer screening system of claim 2, wherein the magnification is between 30-fold and 50-fold.
5. The system of claim 2, wherein the predetermined annotation tool is ASAP optical software.
6. The artificial intelligence assisted early gastric cancer screening system of claim 1, wherein the cancer regions in the plurality of gastroscopic slice images are extracted by binarization and filtering processing of the plurality of gastroscopic slice images.
7. The system of claim 6, wherein the image segment generation module is configured to add a malignant label to image segments with more than fifty percent of cancer portions of the image segments and a benign label to image segments without cancer portions.
8. The artificial intelligence assisted early gastric cancer screening system of claim 1, wherein the model training module uses a neural network for training, testing and tuning to derive the recognition model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010506897.0A CN111710394A (en) | 2020-06-05 | 2020-06-05 | Artificial intelligence assisted early gastric cancer screening system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010506897.0A CN111710394A (en) | 2020-06-05 | 2020-06-05 | Artificial intelligence assisted early gastric cancer screening system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111710394A true CN111710394A (en) | 2020-09-25 |
Family
ID=72539331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010506897.0A Pending CN111710394A (en) | 2020-06-05 | 2020-06-05 | Artificial intelligence assisted early gastric cancer screening system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111710394A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309389A (en) * | 2023-02-28 | 2023-06-23 | 北京透彻未来科技有限公司 | Stomach tumor pathological typing recognition system based on deep learning |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975793A (en) * | 2016-05-23 | 2016-09-28 | 麦克奥迪（厦门）医疗诊断***有限公司 | Auxiliary cancer diagnosis method based on digital pathological images |
| CN107368670A (en) * | 2017-06-07 | 2017-11-21 | 万香波 | Stomach cancer pathology diagnostic support system and method based on big data deep learning |
| CN109300530A (en) * | 2018-08-08 | 2019-02-01 | 北京肿瘤医院 | The recognition methods of pathological picture and device |
| CN109493346A (en) * | 2018-10-31 | 2019-03-19 | 浙江大学 | It is a kind of based on the gastric cancer pathology sectioning image dividing method more lost and device |
| CN109512464A (en) * | 2018-11-23 | 2019-03-26 | 清华大学深圳研究生院 | A kind of disorder in screening and diagnostic system |
| CN109584218A (en) * | 2018-11-15 | 2019-04-05 | 首都医科大学附属北京友谊医院 | A kind of construction method of gastric cancer image recognition model and its application |
| CN109670510A (en) * | 2018-12-21 | 2019-04-23 | 万达信息股份有限公司 | A kind of gastroscopic biopsy pathological data screening system and method based on deep learning |
| CN110517256A (en) * | 2019-08-30 | 2019-11-29 | 重庆大学附属肿瘤医院 | A kind of early carcinoma assistant diagnosis system based on artificial intelligence |
-
2020
- 2020-06-05 CN CN202010506897.0A patent/CN111710394A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975793A (en) * | 2016-05-23 | 2016-09-28 | 麦克奥迪（厦门）医疗诊断***有限公司 | Auxiliary cancer diagnosis method based on digital pathological images |
| CN107368670A (en) * | 2017-06-07 | 2017-11-21 | 万香波 | Stomach cancer pathology diagnostic support system and method based on big data deep learning |
| CN109300530A (en) * | 2018-08-08 | 2019-02-01 | 北京肿瘤医院 | The recognition methods of pathological picture and device |
| CN109493346A (en) * | 2018-10-31 | 2019-03-19 | 浙江大学 | It is a kind of based on the gastric cancer pathology sectioning image dividing method more lost and device |
| CN109584218A (en) * | 2018-11-15 | 2019-04-05 | 首都医科大学附属北京友谊医院 | A kind of construction method of gastric cancer image recognition model and its application |
| CN109512464A (en) * | 2018-11-23 | 2019-03-26 | 清华大学深圳研究生院 | A kind of disorder in screening and diagnostic system |
| CN109670510A (en) * | 2018-12-21 | 2019-04-23 | 万达信息股份有限公司 | A kind of gastroscopic biopsy pathological data screening system and method based on deep learning |
| CN110517256A (en) * | 2019-08-30 | 2019-11-29 | 重庆大学附属肿瘤医院 | A kind of early carcinoma assistant diagnosis system based on artificial intelligence |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116309389A (en) * | 2023-02-28 | 2023-06-23 | 北京透彻未来科技有限公司 | Stomach tumor pathological typing recognition system based on deep learning |
| CN116309389B (en) * | 2023-02-28 | 2023-11-10 | 北京透彻未来科技有限公司 | Stomach tumor pathological typing recognition system based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11682118B2 (en) | Systems and methods for analysis of tissue images | |
| Nadkarni et al. | Detection of lung cancer in CT images using image processing | |
| US10192099B2 (en) | Systems and methods for automated screening and prognosis of cancer from whole-slide biopsy images | |
| US8712142B2 (en) | Method and apparatus for analysis of histopathology images and its application to cancer diagnosis and grading | |
| US9122897B2 (en) | Method for the detection and classification of microcalcification clusters in digital mammograms | |
| CN108416360B (en) | Cancer diagnosis system and method based on breast molybdenum target calcification features | |
| US20170053398A1 (en) | Methods and Systems for Human Tissue Analysis using Shearlet Transforms | |
| CN115087867A (en) | Pathological diagnosis assistance method and assistance device using AI | |
| Win et al. | Automated segmentation of cell nuclei in cytology pleural fluid images using OTSU thresholding | |
| CN111986150A (en) | Interactive marking refinement method for digital pathological image | |
| JPH09508815A (en) | Improved Computerized Automated Method and System for Mammography Finding and Classification | |
| Kharel et al. | Early diagnosis of breast cancer using contrast limited adaptive histogram equalization (CLAHE) and Morphology methods | |
| WO2018157381A1 (en) | Method and apparatus for intelligently classifying pathological slice image | |
| CN112990214A (en) | Medical image feature recognition prediction model | |
| Almuntashri et al. | Gleason grade-based automatic classification of prostate cancer pathological images | |
| Tarique et al. | Fourier transform based early detection of breast cancer by mammogram image processing | |
| CN111710394A (en) | Artificial intelligence assisted early gastric cancer screening system | |
| CN113256617A (en) | Pathological section virtual immunohistochemical staining method and system | |
| Sulaiman et al. | Overlapping cells separation method for cervical cell images | |
| Nanditha et al. | Texture analysis of color oral images for lesion detection | |
| Gautam et al. | Graphical User Interface based platform for the Lung Cancer Classification | |
| CN114638782B (en) | Method for detecting cervical exfoliated cell specimen | |
| MIC et al. | CANCER DETECTION USING MATLAB BASED IMAGE PROCESSING | |
| Antony et al. | Domain-specific approach for segmentation of nucleus-cytoplasm in bone cancer histopathology for malignancy analysis | |
| Toki et al. | Study of feature extraction for diagnosing prostate cancer |
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 |