WO2021091092A1 - 소장 정결도를 진단하는 시스템 및 방법 - Google Patents
소장 정결도를 진단하는 시스템 및 방법 Download PDFInfo
- Publication number
- WO2021091092A1 WO2021091092A1 PCT/KR2020/013636 KR2020013636W WO2021091092A1 WO 2021091092 A1 WO2021091092 A1 WO 2021091092A1 KR 2020013636 W KR2020013636 W KR 2020013636W WO 2021091092 A1 WO2021091092 A1 WO 2021091092A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- small intestine
- cleanliness
- images
- image
- collection
- Prior art date
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
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
-
- 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
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/1032—Determining colour for diagnostic purposes
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/42—Detecting, measuring or recording for evaluating the gastrointestinal, the endocrine or the exocrine systems
- A61B5/4222—Evaluating particular parts, e.g. particular organs
- A61B5/4255—Intestines, colon or appendix
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/74—Details of notification to user or communication with user or patient ; user input means
- A61B5/742—Details of notification to user or communication with user or patient ; user input means using visual displays
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/90—Determination of colour characteristics
-
- 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
-
- 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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- 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
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- 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
- A61B1/00002—Operational features of endoscopes
- A61B1/00004—Operational features of endoscopes characterised by electronic signal processing
- A61B1/00009—Operational features of endoscopes characterised by electronic signal processing of image signals during a use of endoscope
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2576/00—Medical imaging apparatus involving image processing or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06T2207/30028—Colon; Small intestine
Definitions
- the present invention relates to a system and method for diagnosing small intestine cleanliness, and more specifically, to a system and method for displaying in a UI by automatically diagnosing the cleanliness of small intestine by learning in advance of the cleanliness of the small intestine. .
- the small intestine is the digestive tract of 6-7m in length between the stomach and the large intestine. It is an important part that digests and absorbs nutrients while exercising. From the top, the small intestine is divided into three parts: duodenum, jejunum, and ileum.
- Such cleanliness of the small intestine should be described as an essential item when preparing a diagnostic report after diagnosing the condition of the small intestine.
- specialists are evaluating the degree of cleanliness according to the judgment of specialists based on the learned videos.
- the present invention is to solve the problems of the prior art as described above, and an object thereof is to objectively and consistently diagnose small intestine cleanliness.
- a system for diagnosing small bowel cleanliness for achieving the above object, comprising: a similarity analysis unit for analyzing to select representative images of similar small bowel images among a plurality of small bowel images; In the state that a plurality of small intestine images are learned, when a series of small intestine images to be diagnosed are received, the representative image is applied to the learning result to predict the small intestine cleanliness, thereby classifying the small intestine cleanliness by score. part; And a cleanliness diagnosis unit that calculates a final small intestine cleanliness for the series of small intestine images based on a score related to the small intestine cleanliness of the representative image and the number of small intestine images similar to the representative image. .
- the similarity analysis unit selects the representative image by comparing the color histograms of the two collection images based on the Bhattacharyya Distance algorithm, and measures the similarity, and the multiple selected by similarity according to a predetermined similarity criterion. It is characterized in that the representative image is selected from the collection of images.
- the similarity criterion is characterized in that the smaller the similarity of the two small intestine images converges to 0, and the higher the similarity, the converges to 1.
- the image classification unit learns the color and texture characteristics of the plurality of collection images based on a CNN (Convolution Neural Network) model, and predicts the collection cleanliness by applying it to the learning result based on landmark information. It is characterized.
- CNN Convolution Neural Network
- the image classification unit is characterized in that it learns after unifying color standards of the plurality of collection images output from a plurality of image sensors.
- the color standard is characterized in that at least one of RGB, Hue Saturation Value (HSV), and Lab (color coordinates) is used.
- the plurality of small intestine images are characterized in that scoring for small intestine cleanliness from 1 to 10 is completed.
- the final collectible cleanliness is characterized by dividing a value obtained by multiplying the number of similar images included in the representative image from 1 to n by the cleanliness score by the total number of the plurality of collectible images.
- the image classification unit is characterized in that it learns a collection section from the plurality of collection images based on color and texture, and classifies the plurality of collection images according to the collection section based on landmark information set in the collection section. do.
- the method comprising: receiving a plurality of small bowel images for which scoring has been completed; Learning a plurality of collection images for which the scoring has been completed; Analyzing to select representative images of similar small intestine images from among the plurality of small intestine images when a series of small intestine images for which small intestine cleanliness is to be diagnosed are received; Classifying a small intestine cleanliness by score by predicting a small intestine cleanliness by applying the representative image to a learning result; And calculating a final small intestine cleanliness for the plurality of small intestine images based on the representative image.
- the system and method for diagnosing small intestine cleanliness according to the present invention has the following effects.
- the system and method for diagnosing small intestine cleanliness according to the present invention is not consistent according to the subjective judgment of a specialist as in the prior art because the system automatically diagnoses small intestine cleanliness by learning pictures or images of a large amount of small intestine in advance. It is more consistent and can provide the degree of cleanliness of the small intestine, which was diagnosed as not being possible.
- the system and method for diagnosing small intestine cleanliness according to the present invention immediately diagnose small intestine cleanliness according to the learned algorithm when a user inputs a picture of the small intestine to be diagnosed based on a previously learned picture or image of the small intestine. Therefore, it is possible to diagnose small intestine cleanliness in a short time, and accordingly, time for diagnosing small intestine cleanliness can be saved.
- the small intestine cleanliness is immediately diagnosed according to the learned algorithm, so that the user can easily diagnose the small intestine cleanliness.
- FIG. 1 is a block diagram of a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 2 is a flow chart of a method for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 3 is a diagram exemplarily showing that scoring of small intestine cleanliness for a small intestine image is completed in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 4 is a diagram schematically illustrating unification of a plurality of color criteria before learning a small intestine image based on a CNN model in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 5 is a diagram schematically illustrating learning of a small intestine image based on a CNN model in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 6 is a diagram schematically illustrating comparing color histograms of two small intestine images using a batacharya distance algorithm in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 7 is a diagram schematically illustrating selection of representative images based on similarity in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 8 is a diagram schematically showing the calculation of the final small intestine cleanliness in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- FIG. 1 is a block diagram of a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- the system for diagnosing small intestine cleanliness includes a data receiving unit 100, an image classifying unit 110, a similarity analysis unit 120, and a cleanness diagnosis unit 130. And a database 140.
- the data receiving unit 100 is configured to receive a plurality of collection images from a specialist.
- the data receiving unit 100 may receive information by being connected to a terminal of a specialist through a wired or wireless network.
- the specialist can score small intestine cleanliness from 1 to 10 depending on the ease of review and diagnosis for multiple small intestine images.
- the plurality of collection images may include the number of tens of thousands of collection images.
- the image classification unit 110 learns the cleanliness of the small intestine based on the plurality of small intestine images received from the data receiving unit 100, and then receives a series of small intestine images in real time. It is a composition that is classified by scoring.
- the image classification unit 110 may learn a plurality of small intestine images through an artificial neural network. In this case, learning a plurality of small intestine images can be learned based on a Convolution Neural Network (CNN) model.
- Collectible images are scored from 1 to 10 regarding the cleanliness of the collectible, and the image classification unit 110 is the color and texture of the collectible images for each score in relation to the collectible images scored from 1 to 10.
- the similarity analysis unit 120 is configured to analyze to select representative images of similar small intestine images among a series of small intestine images received from the data receiving unit 100 in real time.
- the similarity analysis unit 120 analyzes whether two small intestine images are similar small intestine images among a plurality of small intestine images and selects a representative image, so that the image classification unit 110 classifies the small intestine cleanliness for the plurality of small intestine images. Accordingly, the cleanliness diagnosis unit 130 allows the final small intestine cleanliness to be quickly diagnosed.
- the similarity analysis unit 120 does not analyze all tens of thousands of collection images by applying a similarity algorithm, and allows the image classification unit 110 to classify the small collection purity by taking only representative images of similar collection images. In this way, the similarity analysis unit 120 can help optimize the small intestine cleanliness diagnosis system 1 by increasing the speed of classifying small intestine images to diagnose small intestine cleanliness.
- the cleanliness diagnosis unit 130 is a component that calculates the final small intestine cleanliness for small intestine images while a series of small intestine images are received in real time.
- the final collection cleanliness should be calculated by considering how many similar images the representative image represents. That is, in the small intestine image, the purity of the small intestine will be different according to the section of the small intestine. Therefore, the small intestine cleanliness is finally diagnosed by calculating the average of the small intestine cleanliness for a series of small intestine images received by collecting scores for each small intestine section.
- the database 140 is a component that stores and stores a plurality of collection images received from a specialist according to scores for small collection cleanliness.
- the database 140 may store and store details of the cleanliness diagnosis unit 130 diagnosed in the small intestine.
- FIG. 2 is a flow chart of a method for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- a plurality of small intestine images for which scoring has been completed are initially received.
- FIG. 3 is a diagram exemplarily showing that scoring of small intestine cleanliness for a small intestine image is completed in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- the small intestine cleanliness for the small intestine image may be scored from 1 to 10.
- the degree of cleanliness of the small intestine may be scored by a specialist viewing a collection of images based on various experiences and subjective judgment.
- the specialist determines the degree of cleanliness of the small intestine according to the degree of clarity of the small intestine on the image of the small intestine, the proportion of the area representing the small intestine among the total size of the image, and the degree of floating matter on the image.
- the specialist may determine that the small intestine cleanliness of the corresponding small intestine image is 1 point.
- the specialist can determine that the small intestine cleanliness of the corresponding small intestine image is 2 points.
- the cleanliness of the small intestine is 10 points
- the small intestine is very clear on the image of the small intestine, so that the shape of the inner wall of the small intestine and even the ridges can be observed. It can be determined that the purity of the small intestine is 10 points.
- the data receiving unit 100 may receive information on the collection images scored by a specialist for the plurality of collection images and store them in the database 140 for each score of the collection cleanliness.
- the image classifying unit 110 learns a plurality of scoring-completed collection of images. ⁇ S21>
- the image classification unit 110 may learn that a specialist scores a small intestine cleanliness for a small intestine image through an artificial neural network. This process will be described in detail with reference to FIGS. 4 and 5.
- FIG. 4 is a schematic diagram showing unification of a plurality of color criteria before learning a small intestine image based on a CNN model in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention
- FIG. 5 is an embodiment of the present invention. It is a diagram schematically showing the learning of a small intestine image based on a CNN model in a system for diagnosing small intestine cleanliness according to.
- the image classifier 110 may undergo a pre-processing process for each small intestine image before learning a plurality of small intestine images.
- the plurality of collection images are images output from various image sensors, each of the collection images has a different color standard. That is, even if it is a collection image stored with the same collection purity score, RGB (Red, Blue, Green), HSV (Hue Saturation Value), and Lab (color coordinates) can be set differently and output. In order to learn, the color standards must be the same so that the standards for learning can become clear. This becomes the learning input.
- each small intestine image can be output from various image sensors, so the color standards of images with a small intestine purity of 6 points may not be the same.
- the image classification unit 110 converts and inputs images having 6 points of small intestine purity into gray scale for RGB, selects and inputs the S component for HSV, and selects and inputs the b component for Lab. Collectible images can be preprocessed.
- the image classification unit 110 may unify and learn a color standard through a process of pre-processing a plurality of collection images stored for each score of the collection degree of cleanliness for each score of the collection degree of purity.
- the image classification unit 110 may perform a pre-processing process on a plurality of collection images for all the collection cleanliness scores from 1 point to 10 points based on the same color.
- the image classifier 110 may learn a plurality of small intestine images through an artificial neural network. Specifically, the image classification unit 110 learns by matching the state of the small intestine image and the corresponding small intestine cleanliness score with respect to a plurality of small intestine images stored according to the score of the small intestine cleanliness.
- a convolution neural network CNN may be used as a model in which the artificial neural network learns a plurality of small intestine images.
- a CNN is an artificial neural network that uses convolution operations, consisting of input and output layers as well as multiple hidden layers.
- the hidden layers of the CNN are usually made up of a series of convolutional layers that are related to multiplication or other dot products.
- the active function is usually a ReLU (Rectified Linear Unit) layer, and since the inputs and outputs are masked by the active function and the final convolution, additional convolutions such as pooling layers, fully connected layers, and normalization layers, later referred to as hidden layers.
- This CNN has a structure suitable for learning 2D data, and can be trained through the (Backpropagation algorithm). It is one of the representative models widely used in various application fields such as object classification and object detection in images. Accordingly, in the present invention, it is also possible to classify a plurality of collection images by learning based on the CNN model in the present invention.
- CNN is a known technology, so a detailed description thereof will be omitted. It should be.
- the image classifying unit 110 performs convolution 1 using the CNN, and then undergoes Max Pooling or Avg Pooling through the ReLU layer. Convolutions 2, 3, 4 and 5 are performed for the collection 2, 3, 4, and 5, respectively, as in the collection 1 image to perform ReLU layer and Max Pooling or Avg Pooling. Will go through.
- RGB HSV
- Lab set as learning inputs
- Avg Pooling is used when HSV is S scale
- Lab is b scale
- Max Pooling is used when RGB is gray.
- the image classification learns by classifying the color and texture of the collection image at 1 point, the color and texture of the collection image at 2 points, and the color and texture of the collection at 10 points. You can learn.
- a landmark is set based on the features of the ridges of the small intestine or the inner wall, and is used to divide the collection section.
- the image classification unit 110 may classify the collection cleanliness score for each collection section. For such landmark information, a specific mark set by a system user may be used.
- the image classification unit 110 may learn an image having a small intestine cleanliness of 6 points when the color is Gray, S, and b.
- FIG. 6 is a diagram schematically illustrating comparing color histograms of two small intestine images using a batacharya distance algorithm in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- the similarity analysis unit 120 may select a representative image by analyzing small intestine images with similar small intestine cleanliness among a plurality of small intestine images. Selecting a representative image from among the plurality of small intestine images in this way increases the speed of the process of diagnosing small intestine cleanliness, and makes it possible to diagnose small intestine cleanliness in a short time.
- the similarity analysis unit 120 may analyze the similarity of the small intestine image using the Batacharya distance algorithm.
- the Batacharya distance algorithm is known as a numerical image processing technique that measures the statistical segregation of spectral grades given a probability value of an accurate classification.
- This Batacharya distance algorithm is the most robust among algorithms to find the distance between two distributions. That is, the similarity analysis unit 120 may measure the similarity by comparing the color histograms of the two collection images. Accordingly, the color histograms of the two images are compared, and the smaller the similarity is, the converges to "0", and the higher the similarity, the converges to "1".
- Fig. 6 shows such a color histogram.
- the color histogram may use a gray scale as a comparison type, or may use a Local Binary Pattern (LBP) or Multi-Block Local Binary Pattern (MB-LBP) image method.
- LBP is a visual description type method used in computer vision classification. Since LBP is a known technology, detailed description will be omitted. Accordingly, the similarity analysis unit 120 may transform a plurality of collection images into a comparison type in order to compare them with a color histogram.
- the color histogram will be described in detail.
- the two small intestine images shown above were converted into color histograms H1 and H2 in gray scale by the similarity analysis unit 120, respectively.
- the color histograms of these two collection images rise high on the left, and show a lower shape as it progresses to the right.
- the two small intestine images shown below in FIG. 6 were converted into color histograms H1 and H2 by the similarity analysis unit 120, respectively, in the MB-LBP method.
- the color histograms of these two collection images may be formed in the form of approximately' ⁇ '.
- the similarity analysis unit 120 may measure the similarity by comparing the color histograms based on the gray scale color histogram of the two holding images or the LBP image method color histogram.
- the similarity analysis unit 120 may calculate the batatcharya distance d(H1, H2) for the color histograms of the two collection images.
- H1 and H2 are color histograms for the two small intestine images as described above, , Is the average of the color histograms for the two small images, I is the pixel value, and N is the total number of color histogram bins.
- the similarity analysis unit 120 may calculate the batacharya distance d(H1, H2) based on such factors as follows.
- the distance between the two distributions is calculated, and accordingly, the similarity can be measured by comparing the color histograms of the two collection images.
- This similarity is calculated as a value between 0 and 1, and when the color histograms of two images are compared, the smaller the similarity is, the convergent to "0", and the larger the similarity is, the converged to "1".
- FIG. 7 is a diagram schematically illustrating selection of representative images based on similarity in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- comparing the similarity of the two small intestine images with a reference value of the similarity may be repeatedly performed on consecutive frames of the small intestine to select a representative image. For example, assume that the standard of similarity is 0.850.
- the similarity analysis unit 120 selects one of the first compared small intestine images as a representative image, and compares and analyzes the small intestine images and later small intestine images as shown in FIG. 7 to select small intestine images similar to the representative image. I can.
- the similarity analysis unit 120 may compare and analyze the small intestine images later by selecting A as the representative image of the first small intestine image.
- the similarity analysis unit 120 compares and analyzes the histograms of the two small intestine images by applying the Batacharya distance algorithm to the first and third small intestine images, thereby calculating a similarity of 0.921, Since this is higher than the reference similarity of 0.850, the first small intestine image and the third small intestine image can be analyzed as being similar to each other. In addition, when comparing the 4th small intestine image and the first small intestine image, since the 4th small intestine image and the first small intestine image have a similarity of 0.918, which is higher than the standard similarity of 0.850, the first small intestine image and the fourth small intestine image are analyzed as similar. can do.
- the similarity analysis unit 120 calculates the similarity of the n-th small intestine image by comparing and analyzing the representative image and other images, respectively, and similar small intestine images may be included as a small intestine image group starting with the representative image. At this time, if a small intestine image with a similarity lower than the reference similarity of 0.850 appears, a new representative image is selected.
- representative images A and B are lower than the reference similarity of 0.850 because the similarity of the two small intestine images is 0.754, and accordingly, the two small intestine images are analyzed to be not similar. Accordingly, the similarity analysis unit 120 may select similar images by using B as a new representative image. When comparing the small intestine images after the representative image B in the same manner as the images similar to the representative image A, the similarities are 0.874 and 0.856, respectively, so the two small intestine images are analyzed as similar to the representative image B. Accordingly, a group of small intestine images starting with the representative image B may be selected.
- the image classification unit 110 is based on the fact that the similarity analysis unit 120 first selects representative images for the plurality of collection images when a plurality of collection images are received in real time based on the learned plurality of collection images. Classify according to the degree of cleanliness. That is, the collection purity is scored from 1 point to 10 points, and the image classification unit 110 classifies representative images from 1 point to 10 points by applying the learning result, and selects a collection image that is similar to the representative image. They can also be classified by score along with the representative video.
- the cleanliness diagnosis unit 130 calculates the final small intestine cleanliness.
- FIG. 8 is a diagram schematically showing the calculation of the final small intestine cleanliness in a system for diagnosing small intestine cleanliness according to an embodiment of the present invention.
- the final small intestine cleanliness can be diagnosed by considering how many small intestine images are representative of similar small intestine images.
- the final collection cleanliness is a value obtained by multiplying the number of similar collection images represented by the representative image by the small collection cleanliness of the representative image and dividing the sum by the total number of collection images. That is, the final small intestine cleanliness is to diagnose the final small intestine cleanliness by calculating an average value of the small intestine cleanliness of the entire small intestine image. This can be expressed as an equation as follows.
- the image classification unit 110 classifies the scores of the small intestine cleanliness as shown in FIG. 8 on a plurality of collection images received by the data receiving unit 100 in real time.
- the representative image A has 9 points for the cleanliness of the small intestine, and the small intestine images similar to the representative image A have a total of 7 including the representative image A.
- the representative image B has 8 points for the small intestine cleanliness, and the small intestine images similar to the representative image B include a total of 3 images including the representative image B.
- the final collection cleanliness is a value obtained by multiplying the number of similar collection images included in the representative image A by 9 points, which is the collection cleanliness score of the representative image A, and the number of similar collection images included in the representative image B.
- the cleanliness diagnosis unit 130 diagnoses the final small intestine cleanliness by adding a value multiplied by 8 points, which is the small intestine cleanliness score of the representative image B, and dividing the total small intestine images by 10, which is the total number of small intestine images. This can be expressed as an equation as follows.
- the cleanliness diagnosis unit 130 when the cleanliness diagnosis unit 130 finally diagnoses the score for the small intestine cleanliness, the small intestine cleanliness for a series of small intestine images received in real time may be diagnosed as 8.7 points.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Surgery (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Heart & Thoracic Surgery (AREA)
- Evolutionary Computation (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Physiology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Psychiatry (AREA)
- Fuzzy Systems (AREA)
- Optics & Photonics (AREA)
- Gastroenterology & Hepatology (AREA)
Abstract
Description
Claims (10)
- 소장(small bowel) 정결도를 진단하는 시스템에 있어서,다수의 소장 영상 중에서 유사한 소장 영상들의 대표 영상을 선별하도록 분석하는 유사도 분석부;다수의 소장 영상을 학습한 상태에서 정결도를 진단하고자 하는 일련의 다수의 소장 영상을 수신하였을 시 상기 대표 영상을 학습 결과물에 적용하여 소장 정결도를 예측함으로써 점수별로 소장 정결도를 분류하는 영상 분류부; 및상기 대표 영상의 소장 정결도에 관한 점수 및 상기 대표 영상과 유사한 소장 영상의 개수를 기반으로 상기 일련의 다수의 소장 영상에 관한 최종 소장 정결도를 산출하는 정결도 진단부;를 포함하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 유사도 분석부가 대표 영상을 선별하는 것은 바타챠랴 거리(Bhattacharyya Distance) 알고리즘을 기반으로 2개의 소장 영상의 컬러 히스토그램을 비교하는 것으로 유사도를 측정하고, 정해진 유사도 기준에 따라 유사도별로 선별된 상기 다수의 소장 영상 중에서 대표 영상을 선별하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제2항에 있어서,상기 유사도 기준은 상기 2개의 소장 영상이 유사도가 작을수록 0에 수렴하고, 유사도가 클수록 1에 수렴하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 영상 분류부는 CNN(Convolution Neural Network) 모델을 기반으로 상기 다수의 소장 영상의 컬러와 텍스처에 대한 특징을 학습하고, 랜드마크 정보를 기반으로 상기 학습 결과물에 적용하여 소장 정결도 예측하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 영상 분류부는 다수의 이미지 센서에서 출력된 상기 다수의 소장 영상의 컬러 기준을 통일한 후 학습하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제5항에 있어서,상기 컬러 기준은 RGB(Red, Green, Blue), HSV(Hue Saturation Value), Lab(색 좌표)를 적어도 하나 이상 사용하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 다수의 소장 영상은 1부터 10까지 소장 정결도에 대한 점수화가 완료되어 있는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 최종 소장 정결도는 1부터 n까지 각각 상기 대표 영상이 포함하는 유사 영상 수에 정결도 점수를 곱하여 더한 값을 상기 다수의 소장 영상의 총 개수로 나눈 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 제1항에 있어서,상기 영상 분류부는 상기 다수의 소장 영상에서 소장 구간을 컬러와 텍스처 기반으로 학습하고, 상기 소장 구간에 설정된 랜드마크 정보를 기반으로 상기 소장 구간에 따라 상기 다수의 소장 영상을 분류하는 것을 특징으로 하는 소장 정결도를 진단하는 시스템.
- 소장(small bowel) 정결도를 진단하는 방법에 있어서,점수화가 완료된 다수의 소장 영상을 수신하는 단계;상기 점수화 완료된 다수의 소장 영상을 학습하는 단계;소장 정결도를 진단하고자 하는 일련의 다수의 소장 영상을 수신하였을 시 다수의 소장 영상 중에서 유사한 소장 영상들의 대표 영상을 선별하도록 분석하는 단계;상기 대표 영상을 학습 결과물에 적용하여 소장 정결도를 예측함으로써 점수별로 소장 정결도를 분류하는 단계; 및상기 대표 영상을 기반으로 상기 다수의 소장 영상에 관한 최종 소장 정결도를 산출하는 단계;를 포함하는 것을 특징으로 하는 소장 정결도를 진단하는 방법.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
GB2117924.7A GB2604706B (en) | 2019-11-08 | 2020-10-07 | System and method for diagnosing small bowel cleanliness |
CN202080048089.2A CN114080644A (zh) | 2019-11-08 | 2020-10-07 | 用于诊断小肠清洁度的系统和方法 |
US17/612,486 US20220304617A1 (en) | 2019-11-08 | 2020-10-07 | System and method for diagnosing small bowel cleanliness |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020190142203A KR102430946B1 (ko) | 2019-11-08 | 2019-11-08 | 소장 정결도를 진단하는 시스템 및 방법 |
KR10-2019-0142203 | 2019-11-08 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021091092A1 true WO2021091092A1 (ko) | 2021-05-14 |
Family
ID=75848002
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/KR2020/013636 WO2021091092A1 (ko) | 2019-11-08 | 2020-10-07 | 소장 정결도를 진단하는 시스템 및 방법 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20220304617A1 (ko) |
KR (1) | KR102430946B1 (ko) |
CN (1) | CN114080644A (ko) |
GB (1) | GB2604706B (ko) |
WO (1) | WO2021091092A1 (ko) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114332019B (zh) * | 2021-12-29 | 2023-07-04 | 小荷医疗器械(海南)有限公司 | 内窥镜图像检测辅助系统、方法、介质和电子设备 |
CN115082448B (zh) * | 2022-07-26 | 2023-05-23 | 青岛美迪康数字工程有限公司 | 肠道清洁度评分方法、装置和计算机设备 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8185185B2 (en) * | 2006-03-31 | 2012-05-22 | Given Imaging Ltd. | System and method for assessing a patient condition using tertiles derived from capsule endoscope images of the small bowel |
KR20160063128A (ko) * | 2014-11-26 | 2016-06-03 | 삼성전자주식회사 | 컴퓨터 보조 진단 장치 및 방법 |
KR101889722B1 (ko) * | 2017-02-10 | 2018-08-20 | 주식회사 루닛 | 악성 종양 진단 방법 및 장치 |
KR101928302B1 (ko) * | 2018-04-10 | 2018-12-13 | 주식회사 뷰노 | 콘텐츠 기반 의료 영상 검색 방법 및 시스템 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4991319B2 (ja) | 2007-01-11 | 2012-08-01 | 株式会社東芝 | 干渉検知装置及びそれを利用した医用画像診断装置 |
-
2019
- 2019-11-08 KR KR1020190142203A patent/KR102430946B1/ko active IP Right Grant
-
2020
- 2020-10-07 US US17/612,486 patent/US20220304617A1/en active Pending
- 2020-10-07 GB GB2117924.7A patent/GB2604706B/en active Active
- 2020-10-07 WO PCT/KR2020/013636 patent/WO2021091092A1/ko active Application Filing
- 2020-10-07 CN CN202080048089.2A patent/CN114080644A/zh active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8185185B2 (en) * | 2006-03-31 | 2012-05-22 | Given Imaging Ltd. | System and method for assessing a patient condition using tertiles derived from capsule endoscope images of the small bowel |
KR20160063128A (ko) * | 2014-11-26 | 2016-06-03 | 삼성전자주식회사 | 컴퓨터 보조 진단 장치 및 방법 |
KR101889722B1 (ko) * | 2017-02-10 | 2018-08-20 | 주식회사 루닛 | 악성 종양 진단 방법 및 장치 |
KR101928302B1 (ko) * | 2018-04-10 | 2018-12-13 | 주식회사 뷰노 | 콘텐츠 기반 의료 영상 검색 방법 및 시스템 |
Non-Patent Citations (1)
Title |
---|
POGORELOV KONSTANTIN, RANDEL KRISTIN RANHEIM, GRIWODZ CARSTEN, ESKELAND SIGRUN LOSADA, DE LANGE THOMAS, JOHANSEN DAG, SPAMPINATO C: "KVASIR : A Multi-Class Image Dataset for Computer Aided Gastrointestinal Disease Detection", PROCEEDINGS OF THE 8TH ACM ON MULTIMEDIA SYSTEMS CONFERENCE, ACM, NEW YORK, NEW YORK, USA, 20 June 2017 (2017-06-20), New York, New York, USA, pages 164 - 169, XP055810563, ISBN: 978-1-4503-5002-0, DOI: 10.1145/3083187.3083212 * |
Also Published As
Publication number | Publication date |
---|---|
CN114080644A (zh) | 2022-02-22 |
GB202117924D0 (en) | 2022-01-26 |
US20220304617A1 (en) | 2022-09-29 |
KR20210055881A (ko) | 2021-05-18 |
GB2604706A (en) | 2022-09-14 |
GB2604706B (en) | 2024-02-07 |
KR102430946B1 (ko) | 2022-08-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2019054638A1 (ko) | 영상 분석 방법, 장치 및 컴퓨터 프로그램 | |
WO2021091092A1 (ko) | 소장 정결도를 진단하는 시스템 및 방법 | |
US4965725A (en) | Neural network based automated cytological specimen classification system and method | |
CN110188615B (zh) | 一种人脸表情识别方法、装置、介质及系统 | |
CN107408211A (zh) | 用于物体的再识别的方法 | |
CN114283469B (zh) | 一种基于改进YOLOv4-tiny的轻量型目标检测方法及系统 | |
CN109685765B (zh) | 一种基于卷积神经网络的x光片肺炎结果预测装置 | |
CN108009560B (zh) | 商品图像相似类别判定方法及装置 | |
CN111815564B (zh) | 一种检测丝锭的方法、装置及丝锭分拣系统 | |
EP2557539A2 (en) | Image processing apparatus, image processing method, and image processing program | |
CN110287862B (zh) | 基于深度学习的防偷拍检测方法 | |
CN110705639A (zh) | 一种基于深度学习的医学精子图像识别系统 | |
CN114529839A (zh) | 一种面向无人机巡检的输电线金具异常检测方法及系统 | |
CN114627553A (zh) | 一种基于卷积神经网络的教室场景学生行为检测的方法 | |
CN109344758B (zh) | 基于改进局部二值模式的人脸识别方法 | |
CN106960188B (zh) | 天气图像分类方法及装置 | |
JP3020973B2 (ja) | 画像処理方法 | |
WO2023075303A1 (ko) | 인공지능 기반의 내시경 진단 보조 시스템 및 이의 제어방법 | |
JP2840347B2 (ja) | 基板実装検査装置 | |
Khavalko et al. | Classification and Recognition of Medical Images Based on the SGTM Neuroparadigm. | |
Zhang et al. | Blood cell image classification based on image segmentation preprocessing and CapsNet network model | |
WO2021141253A1 (ko) | 캡슐 내시경의 위치정보를 기반으로 캡슐 내시경의 위치를 파악하는 시스템 및 방법 | |
WO2020045903A1 (ko) | Cnn을 이용하여 크기 독립적으로 물체를 검출하는 방법 및 장치 | |
Iqbal et al. | Towards Efficient Segmentation and Classification of White Blood Cell Cancer Using Deep Learning | |
CN114022698A (zh) | 一种基于二叉树结构的多标签行为识别方法及装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20884433 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 202117924 Country of ref document: GB Kind code of ref document: A Free format text: PCT FILING DATE = 20201007 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20884433 Country of ref document: EP Kind code of ref document: A1 |