WO2020114037A1 - 基于计算机视觉的肠镜退镜速度实时监测方法和系统 - Google Patents

基于计算机视觉的肠镜退镜速度实时监测方法和系统 Download PDF

Info

Publication number
WO2020114037A1
WO2020114037A1 PCT/CN2019/106102 CN2019106102W WO2020114037A1 WO 2020114037 A1 WO2020114037 A1 WO 2020114037A1 CN 2019106102 W CN2019106102 W CN 2019106102W WO 2020114037 A1 WO2020114037 A1 WO 2020114037A1
Authority
WO
WIPO (PCT)
Prior art keywords
picture
enteroscopy
time
video
pictures
Prior art date
Application number
PCT/CN2019/106102
Other languages
English (en)
French (fr)
Inventor
于红刚
刘斌
胡珊
吴练练
Original Assignee
武汉楚精灵医疗科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 武汉楚精灵医疗科技有限公司 filed Critical 武汉楚精灵医疗科技有限公司
Priority to EP19893880.5A priority Critical patent/EP3848891B1/en
Publication of WO2020114037A1 publication Critical patent/WO2020114037A1/zh
Priority to US16/986,232 priority patent/US11800969B2/en

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B1/00—Instruments for performing medical examinations of the interior of cavities or tubes of the body by visual or photographical inspection, e.g. endoscopes; Illuminating arrangements therefor
    • A61B1/00002—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
    • 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
    • A61B1/00043—Operational features of endoscopes provided with output arrangements
    • A61B1/00045—Display arrangement
    • 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
    • A61B1/00043—Operational features of endoscopes provided with output arrangements
    • A61B1/00055—Operational features of endoscopes provided with output arrangements for alerting the user
    • 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
    • A61B1/31—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 for the rectum, e.g. proctoscopes, sigmoidoscopes, colonoscopes
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0077—Devices for viewing the surface of the body, e.g. camera, magnifying lens
    • 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
    • A61B5/48—Other medical applications
    • A61B5/4836—Diagnosis combined with treatment in closed-loop systems or methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/486—Bio-feedback
    • 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
    • A61B5/7465—Arrangements for interactive communication between patient and care services, e.g. by using a telephone network
    • A61B5/747—Arrangements for interactive communication between patient and care services, e.g. by using a telephone network in case of emergency, i.e. alerting emergency services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/60—Editing figures and text; Combining figures or text
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformation in the plane of the image
    • G06T3/40—Scaling the whole image or part thereof
    • G06T3/4007—Interpolation-based scaling, e.g. bilinear interpolation
    • G06T5/92—
    • 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
    • G06T7/0014—Biomedical image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • 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
    • G06T2207/10068—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
    • G06T2207/20112—Image segmentation details
    • G06T2207/20132—Image cropping
    • 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 belongs to the field of information technology, and in particular relates to a method and system for monitoring the speed of colonoscope withdrawal under colonoscopy.
  • Colonoscopy is the most common method of finding lower gastrointestinal lesions such as colorectal polyps and tumors.
  • Colonoscopy withdrawal time refers to the actual time from the beginning of the colonoscope to the cecum to the withdrawal of the mirror to the anal canal during the colonoscopy, minus the time for additional operations such as staining or biopsy. Studies have shown that with the extension of the time of withdrawal, the detection rate of polyps, adenoma detection rate and the average number of polyps found in each subject of colonoscopy patients increased significantly. Therefore, in the operating guidelines for enteroscopy in various countries, the time for withdrawal is an important quality control index.
  • the technical problem to be solved by the present invention is to provide a computer vision-based real-time monitoring method and system for enteroscope endoscope removal speed, proactively prompting the doctor for the speed and stability of the exit mirror, and reminding the doctor when overspeed or unstable picture, supervising Physicians always control the speed of withdrawal of the mirror within a safe range to improve the quality of enteroscopy.
  • the present invention adopts the perceptual hash algorithm technology, and the specific technical solution is: a real-time monitoring method for the speed of bowel mirror withdrawal based on computer vision, including the following steps,
  • Step 1 Obtain the enteroscopy real-time video through the endoscopy equipment, decode the video into a picture, crop the picture first, and then further shrink the cropped picture, only retaining the structural information of the picture;
  • Step 2 Convert the picture to grayscale
  • Step 3 Obtain the hash fingerprint of the picture
  • Step 4. Calculate the Hamming distance between different pictures
  • Step 5 Compare the hash fingerprint of the current enteroscopy picture with the n-frame enteroscopy picture before the current picture to obtain the overlap rate between the current picture and any picture in the previous n-frame pictures, that is, the current enteroscopy picture is similar to any picture degree;
  • Step 6 Calculate the weighted similarity of the image at time t;
  • Step 7 convert the weighted overlap rate at time t into a stable coefficient
  • Step 8 Calculate the average stability coefficient of the enteroscopy image during the 0-t period
  • Step 9 Analyze the existing enteroscopy video to obtain the dividing line 1 between the standard enteroscopy video and the substandard enteroscopy video, and the dividing line 2 between the substandard enteroscopy video and the poor quality enteroscopy video;
  • Step 10 Follow steps S1-S8 to monitor the stability coefficient of the endoscopist’s withdrawal operation in real time and feed it back to the physician.
  • the mirror speed is not within the normal range; when the stability factor is between the dividing line 1 and the dividing line 2, a warning signal is issued; when the stability factor exceeds the dividing line 2, a danger alarm signal is issued.
  • step 1 a bicubic (cubic convolution) interpolation method is used to further reduce the cropped picture.
  • R, G, and B represent the information values of red light, green light, and blue light, respectively.
  • step 3 a hash algorithm of the difference value is used to obtain a hash fingerprint of the picture.
  • d (x, y) represents the Hamming distance between different pictures
  • x and y respectively represent the string corresponding to different pictures, that is, the hash fingerprint of the picture, Indicates XOR.
  • the formula for converting the weighted overlap rate at time t into a stability coefficient in step 7 is among them Represents the weighted similarity of images at time t.
  • the present invention also provides a computer vision-based real-time monitoring system for bowel mirror withdrawal speed, which is characterized by comprising the following modules:
  • Picture acquisition module used to obtain real-time enteroscopy video through endoscopy equipment, decode the video into pictures, crop the pictures first, and then further shrink the cropped pictures, only retain the structural information of the pictures;
  • Gray image conversion module used to convert the picture into gray image
  • Hash fingerprint acquisition module used to obtain the hash fingerprint of the picture
  • Hamming distance calculation module used to calculate the Hamming distance between different pictures
  • the similarity calculation module is used to compare the hash fingerprint of the current enteroscopy picture and the n-frame enteroscopy picture before the current picture to obtain the overlap rate between the current picture and any picture in the previous n-frame pictures, that is, the current enteroscopy picture and any The similarity of a picture;
  • Weighted similarity calculation module used to calculate the weighted similarity of images at time t;
  • Stability coefficient conversion module used to convert the weighted overlap rate at time t into a stability coefficient
  • the average stability coefficient conversion module is used to calculate the average stability coefficient of the colonoscopy image in the 0-t time period
  • Enteroscopy video analysis module used to analyze the existing enteroscopy video, to obtain the dividing line 1 between standard enteroscopy video and substandard enteroscopy video, and the dividing line 2 between substandard enteroscopy video and poor quality enteroscopy video;
  • the exit mirror speed feedback module is used to monitor the stability coefficient of the endoscope doctor's exit mirror operation in real time and feed it back to the physician.
  • the stability coefficient is less than or equal to the boundary line 1, it means that the exit mirror speed is within the normal range, and when it is greater than the boundary line 1, it means that The exit speed is not within the normal range; when the stability factor is between the boundary line 1 and the boundary line 2, a warning signal is issued; when the stability factor exceeds the boundary line 2, a danger alarm signal is issued.
  • the calculation formula for calculating the overlapping ratio between the current enteroscopy picture and any picture in the previous n frames of enteroscopy pictures is:
  • d (x, y) represents the Hamming distance between different pictures
  • x and y respectively represent the string corresponding to different pictures, that is, the hash fingerprint of the picture, Indicates XOR.
  • the formula for converting the weighted overlap rate at time t into the stability coefficient in the stability coefficient conversion module is among them Represents the weighted similarity of images at time t.
  • the present invention reflects the speed of colonoscopy withdrawal in real time by analyzing the stability of the enteroscopy image in an instant or a period of time to remind the doctor to always control the speed of the withdrawal of the mirror during the colonoscopy. Improve the comprehensiveness and effectiveness of testing.
  • FIG. 1 is a flowchart of an embodiment of the present invention.
  • Figure 2 is a schematic diagram of scaling an image using bicubic interpolation
  • Fig. 3 is a schematic diagram of the closest mapping point of a pixel (x, y) in the target interpolation image in the original image.
  • FIG. 5a is a schematic diagram of picture processing in the 10th to 11th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5a is a schematic diagram of picture processing in the 10th to 11th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5b is a schematic diagram of picture processing in the 12th to 13th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5b is a schematic diagram of picture processing in the 12th to 13th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5c is a schematic diagram of picture processing in the 14th to 15th frames of steps 1 to 3 in Embodiment 2.
  • FIG. 5d is a schematic diagram of picture processing in the 16th to 17th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5d is a schematic diagram of picture processing in the 16th to 17th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5e is a schematic diagram of picture processing of the 18th to 19th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • FIG. 5e is a schematic diagram of picture processing of the 18th to 19th frames of the pictures in steps 1 to 3 in Embodiment 2.
  • the present invention discloses a computer vision-based real-time monitoring method of enteroscope retraction speed, which includes the following steps:
  • a 360*360 picture with more than 100,000 pixels, contains a huge amount of information, and a lot of details need to be processed. Therefore, we need to zoom the picture to a very small size.
  • the function is to remove the details of the picture, only retain the basic information such as structure, light and dark, and discard the differences in pictures caused by different sizes and proportions.
  • the zoomed picture adopts bicubic (cubic convolution) interpolation method. Although the amount of calculation is large, the image quality after zooming is high and it is not easy to be distorted.
  • the pixel value corresponding to the (i', j') coordinate point in the reduced image after interpolation is the 16 pixels adjacent to (i, j) in the original image.
  • the sum of the weighted convolution, P00 in Fig. 3 represents the closest mapping point of a certain pixel (x, y) in the target interpolation image in the original image. If the expression of the pixel value of each (i,j) coordinate point in the original image is f(i,j), then the pixel value of the corresponding coordinate after interpolation is F(i',j'), which can be obtained by the following formula :
  • v represents the deviation of the number of rows
  • u represents the deviation of the number of columns
  • row represents a certain row
  • col represents a certain column
  • S(x) represents an interpolation expression, which can be selected according to different needs.
  • B-spline curve expression Commonly based on triangular values and Bell distribution Expression, B-spline curve expression. In the embodiment of the present invention, the Bell distribution expression is selected.
  • the embodiment of the present invention reduces the picture to 9*8, a total of 72 pixels.
  • S2 convert the picture into a grayscale image; usually the similarity and color relationship of the contrast image is not very large, so it is processed as a grayscale image to reduce the complexity of later calculations.
  • the weighted average method is used: for each pixel of the picture, due to the different sensitivity of the human eye to red light, green light, and blue light, different weights are given to obtain the gray value of the point.
  • the formula is as follows:
  • the hash fingerprint of the picture that is, get the hash string corresponding to the picture
  • commonly used perceptual hash algorithms include aHash, pHash, and dHash, where aHash (average hash) is faster, but often less accurate ; PHash (perceived hash) accuracy is relatively high, but the speed is worse; dHash (difference value hash) accuracy is high, and the speed is also very fast. Therefore, in the embodiment of the present invention, the difference value hash algorithm is selected to obtain the hash fingerprint of the picture.
  • the Hamming distance represents the number of different characters of two equal-length character strings at corresponding positions, and we denote the character strings x and y by d(x,y) The distance between Hamming.
  • the Hamming distance measures the minimum number of replacements required to change the character string x into y by replacing characters.
  • the Hamming distance indicates how many steps are required to change A into B. For example, the character strings "abc" and “ab3”, the Hamming distance is 1, because you only need to modify "c" to "3".
  • the Hamming distance in dHash is the modified number of digits by calculating the difference value.
  • Our difference value is represented by 0 and 1, which can be regarded as binary.
  • the Hamming distance between binary 0110 and 1111 is 2. Convert the dHash values of the two pictures to a binary difference, and take the exclusive OR. Calculate the number of "1" digits of the XOR result, that is, the number of different digits, which is the Hamming distance.
  • Sim i refers to the similarity between the current image at time t and the image of the previous i-th frame (value range is 1-9);
  • the average stability coefficient is the average value of the stability coefficient at each time point
  • Embodiment 2 of the present invention provides specific implementation steps for the real-time monitoring method of colonoscopy withdrawal speed based on computer vision described in Embodiment 1, as shown in FIGS. 5a to 5e, including:
  • S1 Obtain an example of a 6-minute bowel mirror resection video through an endoscopy device, decode the video as a picture at a rate of 2 frames per second, and crop it to a size of 360 ⁇ 360 pixels, thereby obtaining 720 360 ⁇ 360 pixels 360 pixel picture.
  • bicubic (cubic convolution) interpolation method the picture is reduced to 9*8, and 720 72-pixel pictures are obtained.
  • Hash value (n 11): 0011001100010001001011000001010000000010000000000001010000011001
  • Hash value (n 12): 0001001100011001100010011000001110000111100001111000111110011111
  • Hash value (n 13): 0001011101001011001000010010000100001101000010110001001100101010
  • Hash value (n 14): 1000110111000101110010100100000001000010011001101110110111111011
  • Hash value (n 15): 1001110111011110110001101101001011000100110001001101100101011001
  • Hash value (n 16): 1101100110000100100000101100000011000000110001011100101111010110
  • Hash value (n 17): 1100101110100011100100011000000110000110100010101001111010110100
  • Hash value (n 18): 1100011010110010000010001100010111000001110000111101001111100111
  • Hash value (n 19): 1100010010000010100000011100000011001010110010101101010001100101
  • the real-time exit mirror stability coefficient of the enteroscope is 39.1, which falls within the exit mirror warning speed range (>30 and ⁇ 45).
  • the endoscopist issued a safety warning that the endoscopist was in the warning speed range, and should slow down the speed of the mirror.
  • the present invention also provides a computer vision-based real-time monitoring system for bowel mirror withdrawal speed, which is characterized by comprising the following modules:
  • Picture acquisition module used to obtain real-time video of enteroscopy through endoscopy equipment, decode the video into pictures, crop the pictures first, and then further reduce the cropped pictures, only retain the structural information of the pictures;
  • Gray image conversion module used to convert the picture into gray image
  • Hash fingerprint acquisition module used to obtain the hash fingerprint of the picture
  • Hamming distance calculation module used to calculate the Hamming distance between different pictures
  • the similarity calculation module is used to compare the hash fingerprint of the current enteroscopy picture and the n-frame enteroscopy picture before the current picture to obtain the overlap rate between the current picture and any picture in the previous n-frame pictures, that is, the current enteroscopy picture and any The similarity of a picture;
  • Weighted similarity calculation module used to calculate the weighted similarity of images at time t;
  • Stability coefficient conversion module used to convert the weighted overlap rate at time t into a stability coefficient
  • the average stability coefficient conversion module is used to calculate the average stability coefficient of the colonoscopy image in the 0-t time period
  • Enteroscopy video analysis module used to analyze the existing enteroscopy video, to obtain the dividing line 1 between standard enteroscopy video and substandard enteroscopy video, and the dividing line 2 between substandard enteroscopy video and poor quality enteroscopy video;
  • the exit mirror speed feedback module is used to monitor the stability coefficient of the endoscope doctor's exit mirror operation in real time and feed it back to the physician.
  • the stability coefficient is less than or equal to the boundary line 1, it means that the exit mirror speed is within the normal range, and when it is greater than the boundary line 1, it means that The exit speed is not within the normal range; when the stability factor is between the boundary line 1 and the boundary line 2, a warning signal is issued; when the stability factor exceeds the boundary line 2, a danger alarm signal is issued.

Abstract

一种基于计算机视觉的肠镜退镜速度实时监测方法和系统，主动提示医师退镜的速度及稳定性，并在超速或画面不稳时提醒医生，监督医师始终将退镜速度控制在安全范围内，提高肠镜检查质量。采用感知哈希算法技术，通过分析瞬时或一段时间内肠镜图像的稳定性，实时反映肠镜退镜速度，以提醒医生在肠镜检查时始终将退镜速度控制在安全范围内，提高检测的全面性和有效性。

Description

基于计算机视觉的肠镜退镜速度实时监测方法和系统 技术领域
本发明属于信息技术领域，具体涉及一种肠镜下监测肠镜退镜速度的方法和系统。
背景技术
肠镜检查是发现结直肠息肉、肿瘤等下消化道病变最常见的检查方法。肠镜退镜时间是指在结肠镜检查过程中，从进镜到达盲肠开始到退镜至肛管之间的实际时间，减去行染色检查或取活检等额外操作的时间。研究表明，随着退镜时间的延长，肠镜检查患者的息肉检出率、腺瘤检出率以及平均每个受检者的息肉发现数目明显增加。因此，各国的肠镜操作指南中，均将退镜时间做为重要的质量控制指标。美国指南建议结肠镜退镜时间为6-10min，2015年《中国早期结直肠癌筛查及内镜诊治指南》建议的退镜时间应不少于6min。然而，尽管指南对退镜时间进行了明确的规范，在实际的临床实践中，由于缺乏有效监管和实用的监督工具，加上我国病人数量庞大，内镜检查质量往往达不到指南规定的标准。一项研究表明，结肠镜检查退镜时间大部分在2-6分钟，和指南的规定有较大差距，严重影响我国患者的生命健康。因此，在如何在已有技术条件下保证退镜时间以提高结肠镜检查质量十分重要。
为提高肠镜检查质量，截止至2018年10月，国家卫生健康委员会已在全国范围内建立18个省级和21个地市级内镜质控中心，定期对各医院内镜检查进行质量监督。部分医院会定期记录医生的退镜时间，统计并将数据反馈给医生和管理层。然而，这种监测制度一方面只能监测特定时间段、部分样本的肠镜质量，无法覆盖到每例检查；另一方面，由于指南只要求总的退镜时间大于6min，部分医生在肠镜开始时退镜很快，即将结束时才放慢速度以达到标准，这导致医生对患者的远端结肠观察不到位，而在近端结肠浪费大量不必要的时间。因此，我们拟通过现有技术，发明一种可以实时监控肠镜退镜速度的方法，以提醒医生在日常工作中始终将退镜速度控制在安全范围内，提高肠镜检查质量。
发明内容
本发明要解决的技术问题是：提供一种基于计算机视觉的肠镜退镜速度实时监测方法和系统，主动提示医师退镜的速度及稳定性，并在超速或画面不稳时提醒医生，监督医师始终将退镜速度控制在安全范围内，提高肠镜检查质量。
为实现上述目的，本发明采用感知哈希算法技术，具体技术方案为：一种基于计算机视 觉的肠镜退镜速度实时监测方法，包括如下步骤，
步骤1，通过内镜检查设备获取肠镜实时视频，将视频解码为图片，先对图片进行裁剪，然后对裁剪后的图片进一步缩小，只保留图片的结构信息；
步骤2，将图片转化成灰度图；
步骤3，获得图片的哈希指纹；
步骤4，计算不同图片之间的汉明距离；
步骤5，比较当前肠镜图片与当前图片之前n帧肠镜图片的哈希指纹，分别得到当前图片与前n帧图片中任一图片的重叠率，即当前肠镜图片与任一图片的相似度；
步骤6，计算t时间点图像的加权相似度；
步骤7，将t时间点的加权重叠率转换成稳定系数；
步骤8，计算0-t时间段内肠镜图像的平均稳定系数；
步骤9，分析现有的肠镜视频，分别获得标准肠镜视频与次标准肠镜视频的分界线1，次标准的肠镜视频与差质量肠镜视频的分界线2；
步骤10，按步骤S1-S8实时监测内镜医师退镜操作的稳定系数并反馈给医师，当稳定系数小于等于分界线1时，说明操作速度在正常范围内，大于分界线1时，说明退镜速度不在正常范围内；当稳定系数在分界线1和分界线2之间时，发出警告信号；当稳定系数超过分界线2时，发出危险报警信号。
进一步的，步骤1中采用双立方(三次卷积)插值法对裁剪后的图片进一步缩小。
进一步的，步骤2中将图片转化成灰度图的计算公式如下，
Gray＝0.30*R+0.59*G+0.11*B
其中，R、G、B分别表示红色光，绿色光，蓝色光的信息值。
进一步的，步骤3中利用差异值哈希算法获取图片的哈希指纹。
进一步的，步骤5中计算当前肠镜图片与之前n帧肠镜图片中任一图片的重叠率的计算公式为，
Sim＝100*(64-d(x,y))/64
其中，d(x,y)表示不同图片之间的汉明距离，
Figure PCTCN2019106102-appb-000001
x和y分别表示不同图片对应的字符串，即图片的哈希指纹，
Figure PCTCN2019106102-appb-000002
表示异或。
进一步的，步骤7中将t时间点的加权重叠率转换成稳定系数的公式为
Figure PCTCN2019106102-appb-000003
其中
Figure PCTCN2019106102-appb-000004
表示t时间点图像的加权相似度。
本发明还提供一种基于计算机视觉的肠镜退镜速度实时监测系统，其特征在于，包括如下模块：
图片获取模块，用于通过内镜检查设备获取肠镜实时视频，将视频解码为图片，先对图片进行裁剪，然后对裁剪后的图片进一步缩小，只保留图片的结构信息；
灰度图转化模块，用于将图片转化成灰度图；
哈希指纹获取模块，用于获得图片的哈希指纹；
汉明距离计算模块，用于计算不同图片之间的汉明距离；
相似度计算模块，用于比较当前肠镜图片与当前图片之前n帧肠镜图片的哈希指纹，分别得到当前图片与前n帧图片中任一图片的重叠率，即当前肠镜图片与任一图片的相似度；
加权相似度计算模块，用于计算t时间点图像的加权相似度；
稳定系数转化模块，用于将t时间点的加权重叠率转换成稳定系数；
平均稳定系数转换模块，用于计算0-t时间段内肠镜图像的平均稳定系数；
肠镜视频分析模块，用于分析现有的肠镜视频，分别获得标准肠镜视频与次标准肠镜视频的分界线1，次标准的肠镜视频与差质量肠镜视频的分界线2；
退镜速度反馈模块，用于实时监测内镜医师退镜操作的稳定系数并反馈给医师，当稳定系数小于等于分界线1时，说明退镜速度在正常范围内，大于分界线1时，说明退镜速度不在正常范围内；当稳定系数在分界线1和分界线2之间时，发出警告信号；当稳定系数超过分界线2时，发出危险报警信号。
进一步的，相似度计算模块中计算当前肠镜图片与之前n帧肠镜图片中任一图片的重叠率的计算公式为，
Sim＝100*(64-d(x,y))/64
其中，d(x,y)表示不同图片之间的汉明距离，
Figure PCTCN2019106102-appb-000005
x和y分别表示不同图片对应的字符串，即图片的哈希指纹，
Figure PCTCN2019106102-appb-000006
表示异或。
进一步的，稳定系数转化模块中将t时间点的加权重叠率转换成稳定系数的公式为
Figure PCTCN2019106102-appb-000007
其中
Figure PCTCN2019106102-appb-000008
表示t时间点图像的加权相似度。
本发明的有益效果为：本发明通过分析瞬时或一段时间内肠镜图像的稳定性，实时反映肠镜退镜速度，以提醒医生在肠镜检查时始终将退镜速度控制在安全范围内，提高检测的全面性和有效性。
附图说明
图1为本发明实施例流程图。
图2为利用双立方插值缩放图像的原理图；
图3为目标插值图中的某像素点(x,y)在原图中最接近的映射点示意图。
图4为实施例1中步骤1的流程图。
图5a为实施例2中步骤1至步骤3图片第10至11帧图片处理示意图。
图5b为实施例2中步骤1至步骤3图片第12至13帧图片处理示意图。
图5c为实施例2中步骤1至步骤3图片第14至15帧图片处理示意图。
图5d为实施例2中步骤1至步骤3图片第16至17帧图片处理示意图。
图5e为实施例2中步骤1至步骤3图片第18至19帧图片处理示意图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本发明，并不用于限定本发明。
实施例1：
请参阅图1，本发明公开了一种基于计算机视觉的肠镜退镜速度实时监测方法，其包括以下步骤：
S1，通过内镜检查设备获取肠镜实时视频，将视频解码为图片(每秒钟2帧)，将图片裁剪为360×360像素大小，为了只指保留图片的结构信息，还需要将图片进一步缩小；
一张360*360的图片，有10万多个像素点，包含相当庞大的信息量，非常多的细节需要处理。因此，我们需要把图片缩放到非常小，作用是去除图片的细节，只保留结构、明暗等基本信息，摒弃不同尺寸、比例带来的图片差异。
缩放图片采用双立方(三次卷积)插值法，虽然计算量大点但缩放后图像质量高，不易失真。根据图2与双立方插值的数学表达式可以看出，插值后的缩小图中(i’,j’)坐标点对应的像素值是原图中(i,j)处邻近16个像素点的权重卷积之和，图3中的P00代表目标插值图中的某像素点(x,y)在原图中最接近的映射点。设原图中每一个(i,j)坐标点的像素值的表达式为f(i,j)，则插值后对应坐标的像素值为F(i’,j’)，可由以下公式得出：
Figure PCTCN2019106102-appb-000009
其中v代表行数偏差，u代表列数偏差；row代表某一行，col代表某一列；S(x)表示插值表达式，可以根据需要选择的表达式不同，常见有基于三角取值、Bell分布表达、B样条曲线表达式。本发明实施例中选取Bell分布表达式。
为了更好的计算图片的dHash值，本发明实施例将图片缩小至9*8，共72个像素。
S2，将图片转化为灰度图；通常对比图像相似度和颜色关系不是很大，所以处理为灰度图，减少后期计算的复杂度。采用了加权平均法：对于图片每一个像素点，由于人眼对红色光，绿色光，蓝色光的敏感程度不同而赋予不同的权重，从而得到该点的灰度值，公式如下:
Gray＝0.30*R+0.59*G+0.11*B
S3，获得图片的哈希指纹，即获得图片对应的哈希字符串；常用的感知哈希算法包括aHash、pHash、dHash，其中aHash(平均值哈希)的速度比较快，但是常常不太精确；pHash(感知哈希)精确度比较高，但是速度方面较差一些；dHash(差异值哈希)精确度较高，且速度也非常快。因此，本发明实施例中选择差异值哈希算法获取图片的哈希指纹。
S4，计算不同图片之间的汉明距离；在信息理论中，汉明距离表示两个等长字符串在对应位置上不同字符的数目，我们以d(x,y)表示字符串x和y之间的汉明距离。
Figure PCTCN2019106102-appb-000010
其中，
Figure PCTCN2019106102-appb-000011
表示异或，从另外一个方面看，汉明距离度量了通过替换字符的方式将字符串x变成y所需要的最小的替换次数。汉明距离表示将A修改成为B，需要多少个步骤。比如字符串“abc”与“ab3”，汉明距离为1，因为只需要修改“c”为“3”即可。
dHash中的汉明距离是通过计算差异值的修改位数。我们的差异值是用0、1表示的，可以看做二进制。二进制0110与1111的汉明距离为2。将两张图片的dHash值转换为二进制difference，并取异或。计算异或结果的“1”的位数，也就是不相同的位数，这就是汉明距离。
S5，比较当前肠镜图片与当前图片之前9帧肠镜图片的哈希指纹，分别得到当前图片与前9帧图片的重叠率，计算当前肠镜图片与任一图片的相似度：Sim＝100*(64-d(x,y))/64；
S6，计算t时间点图像的加权相似度：
Figure PCTCN2019106102-appb-000012
Sim i是指t时间点的当前图像与前面的第i帧(取值范围为1-9)图像之间的相似度；
S7，将t时间点的加权重叠率转换成稳定系数：
Figure PCTCN2019106102-appb-000013
S8，计算0-t时间段内肠镜图像的平均稳定系数，平均稳定系数为各时间点稳定系数的平均值；
S9，通过分析50段退镜时间>6min的标准肠镜视频、50段退镜时间为4-6min次标准的肠镜视频，50段退镜时间<4min的差质量肠镜视频，得出结果：
Figure PCTCN2019106102-appb-000014
S10，按步骤S1-S8实时监测内镜医师退镜操作的稳定系数并反馈给医师，当速度超出0-30的安全范围时，发出警告信号；当速度超出30-45的警告范围时，发出危险报警信号。
实施例2：
本发明实施例2针对实施例1中所述一种基于计算机视觉的肠镜退镜速度实时监测方法，提供了具体的实现步骤，如图5a至图5e所示，包括：
S1，通过内镜检查设备获取长为6min的肠镜退镜视频1例，将视频以每秒钟2帧的频率解码为图片，并裁剪为360×360像素大小，由此得到720张360×360像素的图片。采用双立方(三次卷积)插值法，将图片缩小至9*8，得到720张72个像素的图片。
S2，将图片转化为灰度图：对于图片每一个像素点，针对人眼对红色光，绿色光，蓝色光的敏感程度不同而分别赋予0.30、0.59和0.11的权重，从而得到该点的灰度值，公式如下:
Gray＝0.30*R+0.59*G+0.11*B
S3，选择差异值哈希算法获取图片的哈希指纹，计算每行相邻像素的差值，前一个像素大于后一个像素则为1，否则为0。计算得到第10-19帧图片的哈希指纹是：
哈希值(n＝10):0011111100111111100010111010010100010001000010010001011110100111
哈希值(n＝11):0011001100010001001011000001010000000010000000000001010000011001
哈希值(n＝12):0001001100011001100010011000001110000111100001111000111110011111
哈希值(n＝13):0001011101001011001000010010000100001101000010110001001100101010
哈希值(n＝14):1000110111000101110010100100000001000010011001101110110111111011
哈希值(n＝15):1001110111011110110001101101001011000100110001001101100101011001
哈希值(n＝16):1101100110000100100000101100000011000000110001011100101111010110
哈希值(n＝17):1100101110100011100100011000000110000110100010101001111010110100
哈希值(n＝18):1100011010110010000010001100010111000001110000111101001111100111
哈希值(n＝19):1100010010000010100000011100000011001010110010101101010001100101
S4，将两张图片的哈希指纹转换为二进制difference，并取异或。计算异或结果的“1”的位数，获得汉明距离[d(x,y)]。计算得到：
汉明距离[10,19]＝36
汉明距离[11,19]＝34
汉明距离[12,19]＝34
汉明距离[13,19]＝33
汉明距离[14,19]＝27
汉明距离[15,19]＝27
汉明距离[16,19]＝24
汉明距离[17,19]＝20
汉明距离[18,19]＝18
S5，按照公式Sim＝100*(64-d(x,y))/64计算第19帧肠镜图片与前9张图片的相似度：计算得到：
Sim[10,19]＝43.8
Sim[11,19]＝46.9
Sim[12,19]＝46.9
Sim[13,19]＝48.4
Sim[14,19]＝57.8
Sim[15,19]＝57.8
Sim[16,19]＝62.5
Sim[17,19]＝68.8
Sim[18,19]＝71.9
S6，按照公式
Figure PCTCN2019106102-appb-000015
计算第10-19帧图像的加权相似度为：Sim[10-19]＝60.9；
S7，按照公式
Figure PCTCN2019106102-appb-000016
将第10-19帧图像的加权相似度转换成稳定系数，计算得到：ESim[10-19]＝39.1；
S8，第19帧时(即退镜视频播放至第9.5s时)，肠镜的实时退镜稳定系数为39.1，落在退镜警告速度区间(＞30且＜45)内，系统此刻对内镜医师进行安全警告，提示内镜医师正处于警告速度区间，应放慢退镜速度。
本发明还提供一种基于计算机视觉的肠镜退镜速度实时监测系统，其特征在于，包括如下模块：
图片获取模块，用于通过内镜检查设备获取肠镜实时视频，将视频解码为图片，先对图片 进行裁剪，然后对裁剪后的图片进一步缩小，只保留图片的结构信息；
灰度图转化模块，用于将图片转化成灰度图；
哈希指纹获取模块，用于获得图片的哈希指纹；
汉明距离计算模块，用于计算不同图片之间的汉明距离；
相似度计算模块，用于比较当前肠镜图片与当前图片之前n帧肠镜图片的哈希指纹，分别得到当前图片与前n帧图片中任一图片的重叠率，即当前肠镜图片与任一图片的相似度；
加权相似度计算模块，用于计算t时间点图像的加权相似度；
稳定系数转化模块，用于将t时间点的加权重叠率转换成稳定系数；
平均稳定系数转换模块，用于计算0-t时间段内肠镜图像的平均稳定系数；
肠镜视频分析模块，用于分析现有的肠镜视频，分别获得标准肠镜视频与次标准肠镜视频的分界线1，次标准的肠镜视频与差质量肠镜视频的分界线2；
退镜速度反馈模块，用于实时监测内镜医师退镜操作的稳定系数并反馈给医师，当稳定系数小于等于分界线1时，说明退镜速度在正常范围内，大于分界线1时，说明退镜速度不在正常范围内；当稳定系数在分界线1和分界线2之间时，发出警告信号；当稳定系数超过分界线2时，发出危险报警信号。
各模块的具体实现方式和各步骤相对应，本发明不予撰述。
本文中所描述的具体实施例仅仅是对本发明精神作举例说明。本发明所属技术领域的技术人员可以对所描述的具体实施例做各种各样的修改或补充或采用类似的方式替代，但并不会偏离本发明的精神或者超越所附权利要求书所定义的范围。

Claims (9)

  1. 一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于，包括如下步骤：
    步骤1，通过内镜检查设备获取肠镜实时视频，将视频解码为图片，先对图片进行裁剪，然后对裁剪后的图片进一步缩小，只保留图片的结构信息；
    步骤2，将图片转化成灰度图；
    步骤3，获得图片的哈希指纹；
    步骤4，计算不同图片之间的汉明距离；
    步骤5，比较当前肠镜图片与当前图片之前n帧肠镜图片的哈希指纹，分别得到当前图片与前n帧图片中任一图片的重叠率，即当前肠镜图片与任一图片的相似度；
    ,步骤6，计算t时间点图像的加权相似度；
    步骤7，将t时间点的加权重叠率转换成稳定系数；
    步骤8，计算0-t时间段内肠镜图像的平均稳定系数；
    步骤9，分析现有的肠镜视频，分别获得标准肠镜视频与次标准肠镜视频的分界线1，次标准的肠镜视频与差质量肠镜视频的分界线2；
    步骤10，按步骤S1-S8实时监测内镜医师退镜操作的稳定系数并反馈给医师，当稳定系数小于等于分界线1时，说明操作速度在正常范围内，大于分界线1时，说明退镜速度不在正常范围内；当稳定系数在分界线1和分界线2之间时，发出警告信号；当稳定系数超过分界线2时，发出危险报警信号。
  2. 如权利要求1所述的一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于：步骤1中采用双立方(三次卷积)插值法对裁剪后的图片进一步缩小。
  3. 如权利要求1所述的一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于：步骤2中将图片转化成灰度图的计算公式如下，
    Gray＝0.30*R+0.59*G+0.11*B
    其中，R、G、B分别表示红色光，绿色光，蓝色光的信息值。
  4. 如权利要求1所述的一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于：步骤3中利用差异值哈希算法获取图片的哈希指纹。
  5. 如权利要求1所述的一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于：步骤5中计算当前肠镜图片与之前n帧肠镜图片中任一图片的重叠率的计算公式为，
    Sim＝100*(64-d(x,y))/64
    其中，d(x,y)表示不同图片之间的汉明距离，
    Figure PCTCN2019106102-appb-100001
    x和y分别表示不同图片对应的字符串，即图片的哈希指纹，
    Figure PCTCN2019106102-appb-100002
    表示异或。
  6. 如权利要求1所述的一种基于计算机视觉的肠镜退镜速度实时监测方法，其特征在于：步骤7中将t时间点的加权重叠率转换成稳定系数的公式为
    Figure PCTCN2019106102-appb-100003
    其中
    Figure PCTCN2019106102-appb-100004
    表示t时间点图像的加权相似度。
  7. 一种基于计算机视觉的肠镜退镜速度实时监测系统，其特征在于，包括如下模块：
    图片获取模块，用于通过内镜检查设备获取肠镜实时视频，将视频解码为图片，先对图片进行裁剪，然后对裁剪后的图片进一步缩小，只保留图片的结构信息；
    灰度图转化模块，用于将图片转化成灰度图；
    哈希指纹获取模块，用于获得图片的哈希指纹；
    汉明距离计算模块，用于计算不同图片之间的汉明距离；
    相似度计算模块，用于比较当前肠镜图片与当前图片之前n帧肠镜图片的哈希指纹，分别得到当前图片与前n帧图片中任一图片的重叠率，即当前肠镜图片与任一图片的相似度；
    加权相似度计算模块，用于计算t时间点图像的加权相似度；
    稳定系数转化模块，用于将t时间点的加权重叠率转换成稳定系数；
    平均稳定系数转换模块，用于计算0-t时间段内肠镜图像的平均稳定系数；
    肠镜视频分析模块，用于分析现有的肠镜视频，分别获得标准肠镜视频与次标准肠镜视频的分界线1，次标准的肠镜视频与差质量肠镜视频的分界线2；
    退镜速度反馈模块，用于实时监测内镜医师退镜操作的稳定系数并反馈给医师，当稳定系数小于等于分界线1时，说明退镜速度在正常范围内，大于分界线1时，说明退镜速度不在正常范围内；当稳定系数在分界线1和分界线2之间时，发出警告信号；当稳定系数超过分界线2时，发出危险报警信号。
  8. 如权利要求7所述的一种基于计算机视觉的肠镜退镜速度实时监测系统，其特征在于：相似度计算模块中计算当前肠镜图片与之前n帧肠镜图片中任一图片的重叠率的计算公式为，Sim＝100*(64-d(x,y))/64
    其中，d(x,y)表示不同图片之间的汉明距离，
    Figure PCTCN2019106102-appb-100005
    x和y分别表示不同图片对应的字符串，即图片的哈希指纹，
    Figure PCTCN2019106102-appb-100006
    表示异或。
  9. 如权利要求7所述的一种基于计算机视觉的肠镜退镜速度实时监测系统，其特征在于：稳定系数转化模块中将t时间点的加权重叠率转换成稳定系数的公式为
    Figure PCTCN2019106102-appb-100007
    其中
    Figure PCTCN2019106102-appb-100008
    表示t时间点图像的加权相似度。
PCT/CN2019/106102 2018-12-05 2019-09-17 基于计算机视觉的肠镜退镜速度实时监测方法和系统 WO2020114037A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP19893880.5A EP3848891B1 (en) 2018-12-05 2019-09-17 Computer vision-based method and system for monitoring speed of intestinal lens removal in real time
US16/986,232 US11800969B2 (en) 2018-12-05 2020-08-05 Method and device for monitoring colonoscope withdrawal speed

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811481234.7 2018-12-05
CN201811481234.7A CN109598716B (zh) 2018-12-05 2018-12-05 基于计算机视觉的肠镜退镜速度实时监测方法和系统

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/986,232 Continuation-In-Part US11800969B2 (en) 2018-12-05 2020-08-05 Method and device for monitoring colonoscope withdrawal speed

Publications (1)

Publication Number Publication Date
WO2020114037A1 true WO2020114037A1 (zh) 2020-06-11

Family

ID=65961227

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/106102 WO2020114037A1 (zh) 2018-12-05 2019-09-17 基于计算机视觉的肠镜退镜速度实时监测方法和系统

Country Status (4)

Country Link
US (1) US11800969B2 (zh)
EP (1) EP3848891B1 (zh)
CN (1) CN109598716B (zh)
WO (1) WO2020114037A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3933672A1 (en) * 2020-07-01 2022-01-05 Wuhan Endoangel Medical Technology Co., Ltd. Method of automatic image freezing of digestive endoscopy
CN115035152A (zh) * 2022-08-12 2022-09-09 武汉楚精灵医疗科技有限公司 医学图像处理方法、装置以及相关设备

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109598716B (zh) * 2018-12-05 2020-08-07 武汉楚精灵医疗科技有限公司 基于计算机视觉的肠镜退镜速度实时监测方法和系统
CN110097105A (zh) * 2019-04-22 2019-08-06 上海珍灵医疗科技有限公司 一种基于人工智能的消化内镜检查质量自动评估方法和系统
CN111863209B (zh) * 2019-04-25 2023-08-18 天津御锦人工智能医疗科技有限公司 基于图像识别的结肠镜检查质量评估工作站
CN111839444A (zh) * 2019-04-25 2020-10-30 天津御锦人工智能医疗科技有限公司 一种基于图像识别匹配的肠镜镜头静止检测方法
CN111861049A (zh) * 2019-04-25 2020-10-30 天津御锦人工智能医疗科技有限公司 基于图像识别的肠镜检查质量标准及打分系统
CN112529831A (zh) * 2019-08-28 2021-03-19 深圳市熠摄科技有限公司 利用图像处理技术的地貌潜变观测设备
CN112461206A (zh) * 2019-09-09 2021-03-09 深圳市熠摄科技有限公司 地貌潜变观测设备
CN111000633B (zh) * 2019-12-20 2020-11-03 山东大学齐鲁医院 一种内镜诊疗操作过程的监控方法及系统
CN111767958A (zh) * 2020-07-01 2020-10-13 武汉楚精灵医疗科技有限公司 基于随机森林算法的肠镜退镜时间的实时监测方法
CN111754503B (zh) * 2020-07-01 2023-12-08 武汉楚精灵医疗科技有限公司 基于两通道卷积神经网络的肠镜退镜超速占比监测方法
CN112785549B (zh) * 2020-12-29 2024-03-01 成都微识医疗设备有限公司 基于图像识别的肠镜检查质量评估方法、装置及存储介质
CN112597981B (zh) * 2021-03-04 2021-06-01 四川大学 基于深度神经网络的肠镜退镜质量智能监控系统及方法
CN113487553A (zh) * 2021-06-30 2021-10-08 武汉楚精灵医疗科技有限公司 一种肠道退镜速度平滑方法
CN113793334B (zh) * 2021-11-16 2022-02-08 武汉大学 设备监测方法和设备监测装置
CN113823400A (zh) * 2021-11-22 2021-12-21 武汉楚精灵医疗科技有限公司 肠道退镜速度监测方法、装置及计算机可读存储介质
CN113962998A (zh) * 2021-12-23 2022-01-21 天津御锦人工智能医疗科技有限公司 肠镜检查的有效退镜时间评估方法、装置及存储介质
CN114419521B (zh) * 2022-03-28 2022-07-01 武汉楚精灵医疗科技有限公司 肠道内镜检查的监控方法及装置
CN116977411B (zh) * 2022-12-01 2024-03-19 开立生物医疗科技(武汉)有限公司 内镜移动速度估计方法及装置、电子设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060293558A1 (en) * 2005-06-17 2006-12-28 De Groen Piet C Colonoscopy video processing for quality metrics determination
CN101674769A (zh) * 2007-09-06 2010-03-17 i3系统公司 能够控制图像的帧率的胶囊型内窥镜
CN104114077A (zh) * 2012-10-18 2014-10-22 奥林巴斯医疗株式会社 图像处理装置和图像处理方法
CN106056166A (zh) * 2016-06-29 2016-10-26 中科院合肥技术创新工程院 一种胶囊内窥镜相似图像的自适应筛除方法
CN109598716A (zh) * 2018-12-05 2019-04-09 上海珍灵医疗科技有限公司 基于计算机视觉的肠镜退镜速度实时监测方法和系统

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8341112B2 (en) * 2006-05-19 2012-12-25 Microsoft Corporation Annotation by search
DE602007007340D1 (de) * 2006-08-21 2010-08-05 Sti Medical Systems Llc Computergestützte analyse mit hilfe von videodaten aus endoskopen
KR20140055697A (ko) * 2012-11-01 2014-05-09 노틸러스효성 주식회사 원격 화상 금융거래처리 기기 및 이를 이용한 금융서비스 제공방법
WO2017201494A1 (en) * 2016-05-19 2017-11-23 Avantis Medical Systems, Inc. Methods for polyp detection
CN106056583A (zh) * 2016-05-24 2016-10-26 中国科学院苏州生物医学工程技术研究所 面向虚拟结肠镜的结肠息肉图像数据处理方法
CN108682013A (zh) * 2018-05-30 2018-10-19 广州众健医疗科技有限公司 一种胃镜图像智能处理装置
CN108897775A (zh) * 2018-06-01 2018-11-27 昆明理工大学 一种基于感知哈希的快速图像识别系统及方法
CN109166615B (zh) * 2018-07-11 2021-09-10 重庆邮电大学 一种随机森林哈希的医学ct图像存储与检索方法
CN114529742A (zh) * 2020-11-23 2022-05-24 中国移动通信集团重庆有限公司 图像相似度确定方法、装置、设备及计算机可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060293558A1 (en) * 2005-06-17 2006-12-28 De Groen Piet C Colonoscopy video processing for quality metrics determination
CN101674769A (zh) * 2007-09-06 2010-03-17 i3系统公司 能够控制图像的帧率的胶囊型内窥镜
CN104114077A (zh) * 2012-10-18 2014-10-22 奥林巴斯医疗株式会社 图像处理装置和图像处理方法
CN106056166A (zh) * 2016-06-29 2016-10-26 中科院合肥技术创新工程院 一种胶囊内窥镜相似图像的自适应筛除方法
CN109598716A (zh) * 2018-12-05 2019-04-09 上海珍灵医疗科技有限公司 基于计算机视觉的肠镜退镜速度实时监测方法和系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3933672A1 (en) * 2020-07-01 2022-01-05 Wuhan Endoangel Medical Technology Co., Ltd. Method of automatic image freezing of digestive endoscopy
CN115035152A (zh) * 2022-08-12 2022-09-09 武汉楚精灵医疗科技有限公司 医学图像处理方法、装置以及相关设备

Also Published As

Publication number Publication date
EP3848891A4 (en) 2021-12-29
US20200364880A1 (en) 2020-11-19
CN109598716A (zh) 2019-04-09
EP3848891A1 (en) 2021-07-14
US11800969B2 (en) 2023-10-31
EP3848891B1 (en) 2023-02-22
CN109598716B (zh) 2020-08-07

Similar Documents

Publication Publication Date Title
WO2020114037A1 (zh) 基于计算机视觉的肠镜退镜速度实时监测方法和系统
CN111128396B (zh) 一种基于深度学习的消化道疾病辅助诊断系统
EP3933672A1 (en) Method of automatic image freezing of digestive endoscopy
US20220172828A1 (en) Endoscopic image display method, apparatus, computer device, and storage medium
US20100183204A1 (en) Image processing device, image processing method, and computer readable storage medium storing image processing program
WO2021103938A1 (zh) 医学图像处理方法、装置、设备、介质以及内窥镜
US20120133753A1 (en) System, device, method, and computer program product for facial defect analysis using angular facial image
JP6716853B2 (ja) 情報処理装置、制御方法、及びプログラム
WO2020215810A1 (zh) 一种基于图像识别的结肠镜手术中窄带成像检测方法
WO2020211530A1 (zh) 用于检测眼底图片的模型训练方法和装置、眼底图片的检测方法和装置、计算机设备和介质
CN113379693B (zh) 基于视频摘要技术的胶囊内镜关键病灶图像检测方法
JP2013030105A (ja) 画像処理装置、画像処理方法、及び画像処理プログラム
CN113823400A (zh) 肠道退镜速度监测方法、装置及计算机可读存储介质
US20240005494A1 (en) Methods and systems for image quality assessment
CN113487553A (zh) 一种肠道退镜速度平滑方法
CN111784686A (zh) 一种内窥镜出血区域的动态智能检测方法、系统及可读存储介质
Xu et al. Artificial intelligence assisted identification of therapy history from periapical films for dental root canal
Wang et al. Detection and inpainting of specular reflection in colposcopic images with exemplar-based method
CN111754503B (zh) 基于两通道卷积神经网络的肠镜退镜超速占比监测方法
Li et al. Pixel-level diabetic retinopathy lesion detection using multi-scale convolutional neural network
CN117012344A (zh) 一种4cmos相机采集的图像分析方法
Arnold et al. Indistinct frame detection in colonoscopy videos
CN113870151B (zh) 内镜图像的矫正方法、装置、设备及存储介质
JP2023548226A (ja) 画像処理方法、電子機器および可読記憶媒体
Li et al. Detection of endoscope withdrawal time in colonoscopy videos

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: 19893880

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2019893880

Country of ref document: EP

Effective date: 20200928

NENP Non-entry into the national phase

Ref country code: DE