WO2021189849A1 - 基于神经网络的杯盘比确定方法、装置、设备及存储介质 - Google Patents

基于神经网络的杯盘比确定方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021189849A1
WO2021189849A1 PCT/CN2020/125009 CN2020125009W WO2021189849A1 WO 2021189849 A1 WO2021189849 A1 WO 2021189849A1 CN 2020125009 W CN2020125009 W CN 2020125009W WO 2021189849 A1 WO2021189849 A1 WO 2021189849A1
Authority
WO
WIPO (PCT)
Prior art keywords
cup
ratio
optic
disk
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/125009
Other languages
English (en)
French (fr)
Inventor
李葛
成冠举
曾婵
高鹏
谢国彤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
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 Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Priority to US17/612,566 priority Critical patent/US12086982B2/en
Priority to JP2021568938A priority patent/JP7260674B2/ja
Publication of WO2021189849A1 publication Critical patent/WO2021189849A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
    • A61B3/0016—Operational features thereof
    • A61B3/0025—Operational features thereof characterised by electronic signal processing, e.g. eye models
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
    • A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
    • A61B3/12—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
    • A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
    • A61B3/1005—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for measuring distances inside the eye, e.g. thickness of the cornea
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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 OR CALCULATING; 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
    • G06T2207/30041—Eye; Retina; Ophthalmic

Definitions

  • This application relates to the field of image processing, and in particular to a method, device, device, and storage medium for determining the cup-to-disk ratio based on a neural network.
  • Glaucoma is one of the three major blind ophthalmological diseases in the world. Its irreversibility leads to its early diagnosis and treatment which play a vital role in improving the quality of life of patients.
  • the cup-to-disk ratio is usually used as an evaluation index, and the segmentation method is used to segment the optic cup and the optic disk in the fundus image, and then the cup-to-disk ratio is calculated.
  • the inventor realizes that existing methods for segmentation of optic discs are easily affected by the image quality of fundus images, such as illumination, occlusion, etc., resulting in reduced segmentation accuracy, resulting in lower accuracy of the calculated cup-to-disk ratio, and prone to multiple screening or The case of a leaking sieve.
  • a method for determining a cup-to-disk ratio based on a neural network comprising:
  • the segmented image of the video disc determines the arithmetic cup-to-disk ratio; and the actual cup-to-disk ratio is determined according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • a device for determining a cup-to-tray ratio based on a neural network comprising:
  • the image detection module is used to obtain the fundus image and perform optic disc area detection on the fundus image to obtain the optic disc area;
  • the network prediction module is used to input the optic disc area into the pre-trained neural network to obtain the predicted cup-to-disk ratio and The segmented image of the cup disc;
  • the arithmetic determination module is used to determine the arithmetic cup-to-disk ratio based on the segmented image of the cup-to-disk Cup to plate ratio.
  • a computer device includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the following steps when the computer program is executed:
  • the segmented image of the video disc determines the arithmetic cup-to-disk ratio; and the actual cup-to-disk ratio is determined according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the following steps:
  • the segmented image of the video disc determines the arithmetic cup-to-disk ratio; and the actual cup-to-disk ratio is determined according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • FIG. 1 is a schematic flowchart of a method for determining a cup-to-disk ratio based on a neural network according to an embodiment of the present application
  • FIG. 2 is a schematic flowchart of obtaining the optic disc area by detecting the optic disc area on the fundus image according to an embodiment of the present application
  • FIG. 3 is a schematic diagram of the optic disc area in the fundus image provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of a video disc area provided by an embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of a pre-trained neural network provided by an embodiment of the present application.
  • FIG. 6 is a schematic flowchart of determining arithmetic cup-to-disk ratio provided by an embodiment of the present application
  • FIG. 7 is a schematic block diagram of a device for determining a cup-to-plate ratio based on a neural network according to an embodiment of the application
  • FIG. 8 is a schematic block diagram of the structure of a computer device according to an embodiment of the application.
  • the embodiments of the present application provide a method, a device, a computer device, and a storage medium for determining a cup-to-tray ratio based on a neural network.
  • the cup-to-disk ratio determination method based on neural network and artificial intelligence can be used to screen glaucoma diseases, and reduce the multiple screening and missed screening of glaucoma diseases.
  • FIG. 1 is a schematic flowchart of a method for determining a cup-to-plate ratio based on a neural network according to an embodiment of the present application.
  • This neural network-based cup-to-disk ratio determination method outputs the predicted cup-to-disk ratio and the segmented images of the cup-to-disk ratio through a pre-trained neural network to calculate the arithmetic cup-to-disk ratio based on the segmented image of the cup-to-disk, and finally calculates the cup-to-disk ratio based on the arithmetic cup-to-disk ratio
  • the ratio and predicted cup-to-disk ratio further determine the actual cup-to-disk ratio, and improve the accuracy of the determined cup-to-disk ratio.
  • the method for determining the cup-to-plate ratio based on the neural network specifically includes: step S101 to step S104.
  • the fundus image is acquired by the device that collects the fundus image, and the optic disc area is detected on the acquired fundus image, so as to obtain the optic disc area.
  • detecting the optic disc area on the fundus image to obtain the optic disc area includes step S1011 and step S1012.
  • the MaskRCNN model can be used to detect the fundus image, so as to obtain the boundary coordinates of the optic disc area.
  • the boundary coordinates can refer to the two diagonal coordinates of the optic disc area.
  • Figure 3 it is a schematic diagram of the optic disc area in the fundus image.
  • the rectangular box is the optic disc area obtained by detecting the fundus image.
  • the boundary coordinates of the optic disc area It can be the coordinates of the upper left corner and the lower right corner of the rectangular box.
  • the fundus image is cropped based on the boundary coordinates to obtain an image of the optic disc area.
  • the fundus image can be cropped according to the coordinates of the upper left corner and the lower right corner of the rectangular frame, so as to obtain the optic disc area as shown in FIG. 4.
  • the obtained optic disc area is input into a pre-trained neural network, and the neural network makes predictions based on the input optic disc area, and outputs the predicted cup-to-disk ratio and the segmented image of the optic disc.
  • the method before inputting the optic disc area into a pre-trained neural network, the method includes: preprocessing the optic disc area, the preprocessing including scaling processing.
  • the scaling process is to process the image of the optic disc area into a specific size, for example, 256*256, so that the pre-trained neural network can perform feature extraction on the image of the optic disc area.
  • the pre-trained neural network includes a feature extraction layer, a convolutional layer, a connection layer, an image segmentation layer, and a regression prediction layer.
  • the pre-trained neural network provided in this embodiment of the application Schematic diagram of the network structure of the network.
  • the inputting the optic disc region into a pre-trained neural network to obtain the predicted cup-to-disk ratio and the segmented image of the optic disc includes: performing feature extraction on the optic disc region through the feature extraction layer to obtain the optic disc region Corresponding first feature image; performing convolution operation on the first feature image through the convolution layer to obtain a second feature image representing the predicted cup-to-disk ratio and a third feature image representing the segmented image of the cup-to-disk Connect the second feature image and the third feature image through the connection layer to obtain a connected fourth feature image; input the fourth feature image into the regression prediction layer to obtain a prediction cup Disc ratio; input the fourth characteristic image into the image segmentation layer to obtain a segmented image of the cup disc.
  • the feature extraction layer in the neural network is used to extract the features of the optic disc area.
  • the feature extraction layer can use the mobilenetv2 network to perform feature extraction on the optic disc area to obtain the first corresponding to the optic disc area. Feature image.
  • the convolutional layer After obtaining the first feature image, the convolutional layer performs convolution and dilation convolution operations on the first feature image. For example, 1*1 convolution and dilation parameter 6 dilation convolution can be used to perform convolution operations to obtain a representative The second feature image of the predicted cup-to-disk ratio and the third feature image of the segmented image representing the cup-to-disk ratio.
  • connection layer the second feature image representing the predicted cup-to-disk ratio and the third feature image representing the segmented image of the cup disc are connected to obtain the fourth feature image after the connection is completed.
  • the fourth feature image is input into the regression prediction layer for regression prediction to obtain the predicted cup-to-disk ratio; and the fourth feature image is input into the image segmentation layer for image segmentation to obtain the segmented image of the cup and disc.
  • the loss function of the regression prediction layer is the minimum mean square error loss function
  • the loss function of the image segmentation layer is the cross entropy loss function.
  • S103 Determine an arithmetic cup-to-disk ratio based on the segmented image of the cup-to-disk.
  • the cup-to-disk ratio is calculated based on the segmented image of the cup and optic disc.
  • determining the arithmetic cup-to-plate ratio includes step S1031 and step S1032.
  • S1031 Determine the diameter of the optic cup and the diameter of the optic disc based on the segmented image of the optic disc;
  • S1032 Determine the arithmetic cup-to-disk ratio according to the diameter of the optic cup and the optic disc.
  • the cup diameter and the optic disc diameter are respectively determined.
  • the cup disc ratio can be calculated according to the calculation formula of the cup disc ratio.
  • the determining the diameter of the optic cup and the diameter of the optic disc based on the segmented image of the optic disc includes: determining the outer contour of the optic cup and the outer contour of the optic disc based on the segmented image of the optic disc; The outer contour of the cup and the outer contour of the optic disc respectively determine the smallest circumscribed rectangle of the optic cup and the smallest circumscribed rectangle of the optic disc; the diameter of the optic cup and the optic disc are respectively determined based on the smallest circumscribed rectangle of the optic cup and the smallest circumscribed rectangle of the optic disc.
  • the method of finding the maximum outer contour can be used to extract the outer contour of the optic cup and the outer contour of the optic disc from the segmented image of the optic disc, and then use the rotating caliper algorithm to obtain the smallest outer rectangle of the outer contour of the optic cup and the outer contour of the optic disc.
  • the length of the side in the vertical direction of the smallest circumscribed rectangle of the optic cup is the diameter of the optic cup (VCD).
  • the length of the side in the vertical direction of the smallest circumscribed rectangle of the optic disc is the diameter of the optic disc (VDD).
  • the calculated cup-to-disk ratio and the predicted cup-to-disk ratio are compared and judged, and then the actual cup-to-disk ratio is determined.
  • step S104 specifically includes: calculating the distance difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio; if the distance difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio When it is less than or equal to the first threshold, the arithmetic cup-to-disk ratio is taken as the actual cup-to-disk ratio; if the distance difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio is greater than the second threshold, the prediction The cup-to-disk ratio is used as the actual cup-to-disk ratio; if the distance difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio is greater than the first threshold and less than or equal to the second threshold, calculate the arithmetic cup-to-disk ratio and the total The average value of the predicted cup-to-disk ratio is described, and the average value is used as the actual cup
  • the distance difference is also the numerical difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio, based on the calculated arithmetic cup-to-disk ratio and the prediction
  • the difference in the distance between the cup-to-plate ratio adopts different methods to determine the actual cup-to-plate ratio.
  • the distance difference between the calculated cup-to-disk ratio and the predicted cup-to-disk ratio is less than or equal to the first threshold value, it can be considered that the arithmetic cup-to-disk ratio calculated by the segmented image of the cup optic disc is more accurate, and the arithmetic cup The disc ratio is used as the actual cup-to-plate ratio, where the first threshold may be 0.05, for example.
  • the distance difference between the calculated cup-to-disk ratio and the predicted cup-to-disk ratio is calculated to be greater than the second threshold, it can be considered that the predicted cup-to-disk ratio obtained by the pre-trained neural network is more accurate, and the predicted cup-to-disk ratio is taken as the actual The cup-to-plate ratio, where the second threshold may be 0.1, for example.
  • the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio need to be fused, for example, Calculate the average of the cup-to-disk ratio and the predicted cup-to-disk ratio, and use the calculated average value as the actual cup-to-disk ratio.
  • the formula for determining the actual cup-to-plate ratio may be as follows:
  • C represents the actual cup-to-disk ratio
  • C 1 represents the arithmetic cup-to-disk ratio
  • C 2 represents the predicted cup-to-disk ratio
  • the method for determining the cup-to-disk ratio based on the neural network obtains the optic disc area by acquiring the fundus image and detecting the optic disc area on the fundus image, and then inputs the optic disc area into the pre-trained neural network to obtain the predicted cup-to-disk ratio and the optic disc ratio.
  • the segmented image of the cup disc is then determined based on the segmented image of the cup disc to determine the arithmetic cup-to-disk ratio, and finally the value of the actual cup-to-disk ratio is determined based on the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • the same neural network is used to separately output the segmented images of the predicted cup-to-disk ratio and the cup-to-disk ratio, and finally the actual cup-to-disk ratio is further determined according to the calculated cup-to-disk ratio and the predicted cup-to-disk ratio, which reduces the number of cup-to-disk ratios when determining the cup-to-disk ratio.
  • the accuracy of the determined cup-to-plate ratio is further improved.
  • FIG. 7 is a schematic block diagram of an apparatus for determining a cup-to-disk ratio based on a neural network according to an embodiment of the present application. How to determine the cup-to-plate ratio.
  • the device for determining the cup-to-tray ratio based on the neural network can be configured in a server or a terminal.
  • the server can be an independent server or a server cluster.
  • the terminal can be an electronic device such as a mobile phone, a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, and a wearable device.
  • the device 200 for determining the cup-to-plate ratio based on the neural network includes: an image detection module 201, a network prediction module 202, an arithmetic determination module 203 and an actual determination module 204.
  • the image detection module 201 is configured to obtain a fundus image and perform optic disc area detection on the fundus image to obtain the optic disc area.
  • the image detection module 201 includes a boundary determination sub-module 2011 and an image cropping sub-module 2012.
  • the boundary determination sub-module 2011 is configured to detect the fundus image to obtain the boundary coordinates of the optic disc area;
  • the image cropping sub-module 2012 is configured to crop the fundus image based on the boundary coordinates to obtain the optic disc area .
  • the network prediction module 202 is configured to input the region of the optic disc into a pre-trained neural network to obtain a predicted cup-to-disk ratio and a segmented image of the optic disc.
  • the arithmetic determination module 203 is configured to determine the arithmetic cup-to-disk ratio based on the segmented image of the cup optic disc.
  • the arithmetic determination module 203 includes a diameter determination sub-module 2031 and an arithmetic calculation sub-module 2032.
  • the diameter determination sub-module 2031 is used to determine the diameter of the optic cup and the optic disc based on the segmented image of the optic disc; the arithmetic calculation sub-module 2032 is used to determine the arithmetic cup-to-disk ratio according to the diameter of the optic cup and the optic disc .
  • the actual determination module 204 is configured to determine the actual cup-to-disk ratio according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • the above-mentioned apparatus for determining the cup-to-tray ratio based on the neural network may be implemented in the form of a computer program, and the computer program may be run on a computer device as shown in FIG. 8.
  • FIG. 8 is a schematic block diagram of a structure of a computer device provided by an embodiment of the present application.
  • the computer equipment can be a server or a terminal.
  • the computer device includes a processor, a memory, and a network interface connected through a system bus, where the memory may include a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium can store an operating system and a computer program.
  • the computer program includes program instructions, and when the program instructions are executed, the processor can execute any method for determining the cup-to-disk ratio based on a neural network.
  • the processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.
  • the internal memory provides an environment for the operation of the computer program in the non-volatile storage medium.
  • the processor can execute any method for determining the cup-to-disk ratio based on a neural network.
  • the network interface is used for network communication, such as sending assigned tasks.
  • FIG. 8 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
  • the processor may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), and application specific integrated circuits (Application Specific Integrated Circuits). Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor.
  • the processor is configured to run a computer program stored in a memory to implement the following steps: obtain a fundus image, and perform optic disc area detection on the fundus image to obtain the optic disc area; Input the pre-trained neural network into the optic disc area to obtain the predicted cup-to-disk ratio and the segmented images of the cup-to-disk; determine the arithmetic cup-to-disk ratio based on the segmented images of the cup-to-disk; according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio The ratio determines the actual cup-to-plate ratio.
  • the processor when the processor realizes the detection of the optic disc area on the fundus image to obtain the optic disc area, the processor is configured to: detect the fundus image to obtain the boundary coordinates of the optic disc area; The boundary coordinates crop the fundus image to obtain the optic disc area.
  • the processor before implementing the input of the optical disc area into the pre-trained neural network, is configured to implement: preprocessing the optical disc area, where the preprocessing includes scaling processing.
  • the pre-trained neural network includes a feature extraction layer, a convolutional layer, a connection layer, an image segmentation layer, and a regression prediction layer;
  • the neural network obtains the predicted cup-to-disk ratio and the segmented image of the cup-to-disc, it is used to implement: feature extraction of the optic disc region through the feature extraction layer to obtain the first feature image corresponding to the optic disc region;
  • the convolution layer performs a convolution operation on the first feature image to obtain a second feature image representing the predicted cup-to-disk ratio and a third feature image representing the segmented image of the cup-to-disk;
  • the second feature image and the third feature image are connected to obtain a connected fourth feature image;
  • the fourth feature image is input to the regression prediction layer to obtain a predicted cup-to-disc ratio;
  • the characteristic image is input to the image segmentation layer to obtain a segmented image of the cup disc.
  • the processor when the processor implements the determination of the arithmetic cup-to-disk ratio based on the segmented image of the cup optic disc, it is configured to: determine the cup diameter and the optic disc diameter based on the segmented image of the cup optic disc ; Determine the arithmetic cup-disc ratio according to the diameter of the optic cup and the diameter of the optic disc.
  • the processor when the processor implements the determination of the diameter of the optic cup and the diameter of the optic disc based on the segmented image of the optic disc, the processor is configured to: determine the outer contour of the optic cup based on the segmented image of the optic disc And the outer contour of the optic disc; respectively determine the smallest circumscribed rectangle of the optic cup and the smallest circumscribed rectangle of the optic disc according to the outer contour of the optic cup and the outer contour of the optic disc; Diameter and optic disc diameter.
  • the processor when the processor realizes the determination of the actual cup-to-disk ratio based on the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio, the processor is configured to: calculate the arithmetic cup-to-disk ratio and the prediction The distance difference between the cup-to-plate ratio; if the distance difference between the arithmetic cup-to-plate ratio and the predicted cup-to-plate ratio is less than or equal to the first threshold, the arithmetic cup-to-plate ratio is used as the actual cup-to-plate ratio; if When the distance difference between the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio is greater than a second threshold, the predicted cup-to-disk ratio is used as the actual cup-to-disk ratio; When the distance difference between the ratios is greater than the first threshold and less than or equal to the second threshold, the average value of the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio
  • the embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium may be volatile or non-volatile, and the computer-readable storage medium stores a computer program,
  • the computer program includes program instructions, and the processor executes the program instructions to implement the following steps:
  • the segmented image of the video disc determines the arithmetic cup-to-disk ratio; and the actual cup-to-disk ratio is determined according to the arithmetic cup-to-disk ratio and the predicted cup-to-disk ratio.
  • the computer-readable storage medium may be the internal storage unit of the computer device described in the foregoing embodiment, for example, the hard disk or memory of the computer device.
  • the computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), and a Secure Digital (SD) ) Card, Flash Card, etc.
  • a plug-in hard disk equipped on the computer device such as a smart memory card (Smart Media Card, SMC), and a Secure Digital (SD) ) Card, Flash Card, etc.
  • SD Secure Digital
  • all the above-mentioned data can also be stored in a node of a blockchain .
  • fundus images and segmented images, etc. these data can be stored in the blockchain node.
  • the blockchain referred to in this application is a new application mode of computer technology such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Quality & Reliability (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Ophthalmology & Optometry (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Surgery (AREA)
  • Geometry (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Image Analysis (AREA)
  • Eye Examination Apparatus (AREA)

Abstract

一种基于神经网络的杯盘比确定方法、装置、设备及存储介质，所述方法包括：获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域(S101)；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像(S102)；基于所述视杯视盘的分割图像确定算数杯盘比(S103)；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比(S104)。该方法提高了确定出的杯盘比的准确度，减少疾病筛查过程中的多筛、漏筛情况，适用于智慧医疗领域。

Description

基于神经网络的杯盘比确定方法、装置、设备及存储介质
本申请要求于2020年9月21日提交中国专利局、申请号为CN202010992346X，发明名称为“基于神经网络的杯盘比确定方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图像处理领域，尤其涉及一种基于神经网络的杯盘比确定方法、装置、设备及存储介质。
背景技术
青光眼是一种全球三大致盲的眼科疾病之一，其不可逆性导致它的早期诊断和治疗对于提高患者的生活质量有至关重要的作用。在对青光眼进行自动筛查时，通常使用杯盘比作为评估指标，采用分割方法对眼底图像中的视杯和视盘进行分割，然后计算杯盘比。发明人意识到现有的视盘分割方法容易受到眼底图像的图像质量的影响，例如光照、遮挡等，导致分割精度降低，进而导致计算出的杯盘比的准确度较低，容易产生多筛或漏筛的情况。
因此，如何提高确定出的杯盘比的准确度，减少疾病筛查过程中的多筛、漏筛情况成为亟待解决的问题。
发明内容
一种基于神经网络的杯盘比确定方法，所述方法包括：
获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；基于所述视杯视盘的分割图像确定算数杯盘比；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
一种基于神经网络的杯盘比确定装置，所述装置包括：
图像检测模块，用于获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；网络预测模块，用于将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；算数确定模块，用于基于所述视杯视盘的分割图像确定算数杯盘比；实际确定模块，用于根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
一种计算机设备，所述计算机设备包括存储器和处理器；所述存储器用于存储计算机程序；所述处理器，用于执行所述计算机程序并在执行所述计算机程序时实现如下步骤：
获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；基于所述视杯视盘的分割图像确定算数杯盘比；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时使所述处理器实现如下步骤：
获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；基于所述视杯视盘的分割图像确定算数杯盘比；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例提供的一种基于神经网络的杯盘比确定方法的示意流程图；
图2是本申请实施例提供的对眼底图像进行视盘区域检测得到视盘区域的示意流程图；
图3是本申请实施例提供的眼底图像中视盘区域的示意图；
图4是本申请实施例提供的视盘区域的示意图；
图5是本申请实施例提供的预先训练的神经网络的结构示意图；
图6是本申请实施例提供的确定算数杯盘比的示意流程图；
图7为本申请实施例提供的一种基于神经网络的杯盘比确定装置的示意性框图；
图8为本申请一实施例提供的一种计算机设备的结构示意性框图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
附图中所示的流程图仅是示例说明，不是必须包括所有的内容和操作/步骤，也不是必须按所描述的顺序执行。例如，有的操作/步骤还可以分解、组合或部分合并，因此实际执行的顺序有可能根据实际情况改变。
应当理解，在此本申请说明书中所使用的术语仅仅是出于描述特定实施例的目的而并不意在限制本申请。如在本申请说明书和所附权利要求书中所使用的那样，除非上下文清楚地指明其它情况，否则单数形式的“一”、“一个”及“该”意在包括复数形式。
还应当理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。
本申请的实施例提供了一种基于神经网络的杯盘比确定方法、装置、计算机设备及存储介质。基于神经网络的杯盘比确定方法利用人工智能可用于筛查青光眼疾病，降低青光眼疾病的多筛、漏筛情况。
下面结合附图，对本申请的一些实施方式作详细说明。在不冲突的情况下，下述的实施例及实施例中的特征可以相互组合。
请参阅图1，图1是本申请实施例提供的一种基于神经网络的杯盘比确定方法的示意流程图。该基于神经网络的杯盘比确定方法通过预先训练的神经网络分别输出预测杯盘比和视杯视盘的分割图像，以根据视杯视盘的分割图像计算算数杯盘比，并最终根据算数杯盘比和预测杯盘比进一步确定实际杯盘比，提高确定出的杯盘比的准确度。
如图1所示，该基于神经网络的杯盘比确定方法，具体包括：步骤S101至步骤S104。
S101、获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域。
通过采集眼底图像的器械获取眼底图像，并对获取到的眼底图像进行视盘区域的检测，从而得到视盘区域。
在一些实施例中，如图2所示，对眼底图像进行视盘区域检测得到视盘区域，包括步骤S1011和步骤S1012。
S1011、对所述眼底图像进行检测，得到视盘区域的边界坐标。
在对眼底图像进行检测时，可以采用多种目标检测技术，例如，可以是采用MaskRCNN模型对眼底图像进行检测，从而得到视盘区域的边界坐标。其中，边界坐标可以是指视盘区域的两个对角坐标，如图3所示，为眼底图像中视盘区域的示意图，矩形框内为对眼底图像进行检测得到的视盘区域，视盘区域的边界坐标可以是矩形框的左上角与右下角坐标。
S1012、基于所述边界坐标对所述眼底图像进行裁剪，得到视盘区域。
基于边界坐标对眼底图像进行裁剪，得到视盘区域的图像。在进行裁剪时，可根据矩形框的左上角与右下角坐标对眼底图像进行裁剪，从而得到如图4所示的视盘区域。
S102、将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像。
将得到的视盘区域输入预先训练的神经网络中，神经网络基于输入的视盘区域进行预测，输出预测杯盘比和视杯视盘的分割图像。
在一些实施例中，在所述将所述视盘区域输入预先训练的神经网络之前，所述方法包括：对所述视盘区域进行预处理，所述预处理包括伸缩处理。
伸缩处理是为了将视盘区域的图像处理为特定的尺寸，例如可以是256*256，便于所述预先训练的神经网络对视盘区域的图像进行特征提取。
在一些实施例中，所述预先训练的神经网络包括特征提取层、卷积层、连接层、图像分割层和回归预测层，如图5所示，为本申请实施例提供的预先训练的神经网络的网络结构示意图。
所述将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像，包括：通过所述特征提取层对所述视盘区域进行特征提取，以得到所述视盘区域对应的第一特征图像；通过所述卷积层对所述第一特征图像进行卷积操作，以得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像；通过所述连接层将所述第二特征图像和所述第三特征图像进行连接，得到连接后的第四特征图像；将所述第四特征图像输入所述回归预测层，以得到预测杯盘比；将所述第四特征图像输入所述图像分割层，以得到视杯视盘的分割图像。
将视盘区域输入预先训练的神经网络后，通过神经网络中的特征提取层对视盘区域进行特征提取，例如，特征提取层可以使用mobilenetv2网络对视盘区域进行特征提取，进而得到视盘区域对应的第一特征图像。
得到第一特征图像后，卷积层对第一特征图像进行卷积和膨胀卷积操作，例如，可以使用1*1卷积和dilation参数为6的膨胀卷积进行卷积操作，进而得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像。
在连接层将代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像进行连接，得到连接完成后的第四特征图像。
最终将第四特征图像输入回归预测层进行回归预测，得到预测杯盘比；以及将第四特征图像输入图像分割层进行图像分割，得到视杯视盘的分割图像。其中，回归预测层的损失函数为最小均方误差损失函数，图像分割层的损失函数为交叉熵损失函数。
S103、基于所述视杯视盘的分割图像确定算数杯盘比。
对于预先训练的神经网络输出的视杯视盘的分割图像，基于该视杯视盘的分割图像计算算数杯盘比。
在一些实施例中，请参阅图6，确定算数杯盘比包括步骤S1031和步骤S1032。
S1031、基于所述视杯视盘的分割图像确定视杯直径和视盘直径；S1032、根据所述视杯直径和视盘直径确定算数杯盘比。
基于视杯视盘的分割图像分别确定视杯直径和视盘直径，在得到视杯直径和视盘直径后，即可根据杯盘比的计算公式计算算数杯盘比。其中，杯盘比的计算公式为：CDR＝VCD/VDD，其中，CDR为计算出的算数杯盘比，VCD为视杯直径，VDD为视盘直径。
在一些实施例中，所述基于所述视杯视盘的分割图像确定视杯直径和视盘直径，包括：基于所述视杯视盘的分割图像确定视杯外轮廓和视盘外轮廓；根据所述视杯外轮廓和视盘外轮廓分别确定视杯的最小外接矩形和视盘的最小外接矩形；基于所述视杯的最小外接矩形和视盘的最小外接矩形分别确定视杯直径和视盘直径。
可以采用寻找最大外轮廓的方法，从视杯视盘的分割图像中分别提取出视杯外轮廓和视盘外轮廓，然后再利用旋转卡尺算法分别对视杯外轮廓和视盘外轮廓求得最小外接矩形，得到视杯的最小外接矩形和视盘的最小外接矩形。视杯的最小外接矩形的垂直方向上的边长即是视杯直径(VCD)，同样的，视盘的最小外接矩形的垂直方向上的边长即是视盘直 径(VDD)。
S104、根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
将算数杯盘比和预测杯盘比进行比较判断，进而确定出实际杯盘比。
在一些实施例中，步骤S104具体包括：计算所述算数杯盘比和所述预测杯盘比之间的距离差；若所述算数杯盘比和所述预测杯盘比之间的距离差小于或等于第一阈值时，将所述算数杯盘比作为实际杯盘比；若所述算数杯盘比和所述预测杯盘比之间的距离差大于第二阈值时，将所述预测杯盘比作为实际杯盘比；若所述算数杯盘比和所述预测杯盘比之间的距离差大于第一阈值且小于或等于第二阈值时，计算所述算数杯盘比和所述预测杯盘比的平均值，并将所述平均值作为实际杯盘比。
计算算数杯盘比和预测杯盘比之间的距离差，其中，所述距离差也即算数杯盘比和预测杯盘比之间的数值差值，根据计算出的算数杯盘比和预测杯盘比之间的距离差采用采用不同的方式确定实际杯盘比。
当计算出算数杯盘比和预测杯盘比之间的距离差小于或等于第一阈值时，则可以认为利用视杯视盘的分割图像计算出的算数杯盘比更为准确，则将算数杯盘比作为实际杯盘比，其中，第一阈值例如可以是0.05。
当计算出算数杯盘比和预测杯盘比之间的距离差大于第二阈值时，则可以认为利用预先训练的神经网络得到的预测杯盘比更为准确，则将预测杯盘比作为实际杯盘比，其中，第二阈值例如可以是0.1。
当计算出算数杯盘比和预测杯盘比之间的距离差大于第一阈值且小于或等于第二阈值时，则认为需要对算数杯盘比和预测杯盘比进行融合，例如可以是取算数杯盘比和预测杯盘比的平均值，并将计算出的平均值作为实际杯盘比。
也即，确定实际杯盘比的公式例如可以如下所示：
Figure PCTCN2020125009-appb-000001
其中，C表示实际杯盘比，C 1表示算数杯盘比，C 2表示预测杯盘比。
上述实施例提供的基于神经网络的杯盘比确定方法，通过获取眼底图像，并对眼底图像进行视盘区域检测得到视盘区域，然后将视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像，再基于视杯视盘的分割图像确定算数杯盘比，最终根据算数杯盘比和预测杯盘比确定实际杯盘比的值。采用同一神经网络分别输出预测杯盘比和视杯视盘的分割图像，并最终根据算数杯盘比和预测杯盘比进一步确定实际杯盘比，降低了在确定杯盘比时，对视杯视盘的分割图像的依赖，进而提高确定出的杯盘比的准确度。
请参阅图7，图7是本申请的实施例还提供一种基于神经网络的杯盘比确定装置的示意性框图，该基于神经网络的杯盘比确定装置用于执行前述的基于神经网络的杯盘比确定方法。其中，该基于神经网络的杯盘比确定装置可以配置于服务器或终端中。
其中，服务器可以为独立的服务器，也可以为服务器集群。该终端可以是手机、平板电脑、笔记本电脑、台式电脑、个人数字助理和穿戴式设备等电子设备。
如图7所示，基于神经网络的杯盘比确定装置200包括：图像检测模块201、网络预测模块202、算数确定模块203和实际确定模块204。
图像检测模块201，用于获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域。
其中，图像检测模块201包括边界确定子模块2011和图像裁剪子模块2012。
具体地，边界确定子模块2011，用于对所述眼底图像进行检测，得到视盘区域的边界坐标；图像裁剪子模块2012，用于基于所述边界坐标对所述眼底图像进行裁剪，得到视盘 区域。
网络预测模块202，用于将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像。
算数确定模块203，用于基于所述视杯视盘的分割图像确定算数杯盘比。
其中，算数确定模块203包括直径确定子模块2031和算数计算子模块2032。
具体地，直径确定子模块2031，用于基于所述视杯视盘的分割图像确定视杯直径和视盘直径；算数计算子模块2032，用于根据所述视杯直径和视盘直径确定算数杯盘比。
实际确定模块204，用于根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
需要说明的是，所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的基于神经网络的杯盘比确定装置和各模块的具体工作过程，可以参考前述基于神经网络的杯盘比确定方法实施例中的对应过程，在此不再赘述。
上述的基于神经网络的杯盘比确定装置可以实现为一种计算机程序的形式，该计算机程序可以在如图8所示的计算机设备上运行。
请参阅图8，图8是本申请实施例提供的一种计算机设备的结构示意性框图。该计算机设备可以是服务器或终端。
参阅图8，该计算机设备包括通过系统总线连接的处理器、存储器和网络接口，其中，存储器可以包括非易失性存储介质和内存储器。
非易失性存储介质可存储操作系统和计算机程序。该计算机程序包括程序指令，该程序指令被执行时，可使得处理器执行任意一种基于神经网络的杯盘比确定方法。
处理器用于提供计算和控制能力，支撑整个计算机设备的运行。
内存储器为非易失性存储介质中的计算机程序的运行提供环境，该计算机程序被处理器执行时，可使得处理器执行任意一种基于神经网络的杯盘比确定方法。
该网络接口用于进行网络通信，如发送分配的任务等。本领域技术人员可以理解，图8中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
应当理解的是，处理器可以是中央处理单元(Central Processing Unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。其中，通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
其中，在一个实施例中，所述处理器用于运行存储在存储器中的计算机程序，以实现如下步骤：获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；基于所述视杯视盘的分割图像确定算数杯盘比；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
在一个实施例中，所述处理器在实现所述对所述眼底图像进行视盘区域检测，以得到视盘区域时，用于实现：对所述眼底图像进行检测，得到视盘区域的边界坐标；基于所述边界坐标对所述眼底图像进行裁剪，得到视盘区域。
在一个实施例中，所述处理器在实现所述将所述视盘区域输入预先训练的神经网络之前，用于实现：对所述视盘区域进行预处理，所述预处理包括伸缩处理。
在一个实施例中，所述预先训练的神经网络包括特征提取层、卷积层、连接层、图像分割层和回归预测层；所述处理器在实现所述将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像时，用于实现：通过所述特征提取层对所述视盘区 域进行特征提取，以得到所述视盘区域对应的第一特征图像；通过所述卷积层对所述第一特征图像进行卷积操作，以得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像；通过所述连接层将所述第二特征图像和所述第三特征图像进行连接，得到连接后的第四特征图像；将所述第四特征图像输入所述回归预测层，以得到预测杯盘比；将所述第四特征图像输入所述图像分割层，以得到视杯视盘的分割图像。
在一个实施例中，所述处理器在实现所述基于所述视杯视盘的分割图像确定算数杯盘比时，用于实现：基于所述视杯视盘的分割图像确定视杯直径和视盘直径；根据所述视杯直径和视盘直径确定算数杯盘比。
在一个实施例中，所述处理器在实现所述基于所述视杯视盘的分割图像确定视杯直径和视盘直径时，用于实现：基于所述视杯视盘的分割图像确定视杯外轮廓和视盘外轮廓；根据所述视杯外轮廓和视盘外轮廓分别确定视杯的最小外接矩形和视盘的最小外接矩形；基于所述视杯的最小外接矩形和视盘的最小外接矩形分别确定视杯直径和视盘直径。
在一个实施例中，所述处理器在实现所述根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比时，用于实现：计算所述算数杯盘比和所述预测杯盘比之间的距离差；若所述算数杯盘比和所述预测杯盘比之间的距离差小于或等于第一阈值时，将所述算数杯盘比作为实际杯盘比；若所述算数杯盘比和所述预测杯盘比之间的距离差大于第二阈值时，将所述预测杯盘比作为实际杯盘比；若所述算数杯盘比和所述预测杯盘比之间的距离差大于第一阈值且小于或等于第二阈值时，计算所述算数杯盘比和所述预测杯盘比的平均值，并将所述平均值作为实际杯盘比。
本申请的实施例中还提供一种计算机可读存储介质，所述计算机可读存储介质可以是易失性的，也可以是非易失性的，所述计算机可读存储介质存储有计算机程序，所述计算机程序中包括程序指令，所述处理器执行所述程序指令，实现如下步骤：
获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；基于所述视杯视盘的分割图像确定算数杯盘比；根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
其中，所述计算机可读存储介质可以是前述实施例所述的计算机设备的内部存储单元，例如所述计算机设备的硬盘或内存。所述计算机可读存储介质也可以是所述计算机设备的外部存储设备，例如所述计算机设备上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。所述程序指令被处理器执行时所实现的方法的实施例可参照本申请基于神经网络的杯盘比确定方法的各个实施例，此处不再赘述。
在另一实施例中，本申请所提供的基于神经网络的杯盘比确定方法，为进一步保证上述所有出现的数据的私密和安全性，上述所有数据还可以存储于一区块链的节点中。例如眼底图像及分割图像等，这些数据均可存储在区块链节点中。
需要说明的是，本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到各种等效的修改或替换，这些修改或替换都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (20)

  1. 一种基于神经网络的杯盘比确定方法，其中，包括：
    获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；
    将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；
    基于所述视杯视盘的分割图像确定算数杯盘比；
    根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
  2. 根据权利要求1所述的基于神经网络的杯盘比确定方法，其中，所述对所述眼底图像进行视盘区域检测，以得到视盘区域，包括：
    对所述眼底图像进行检测，得到视盘区域的边界坐标；
    基于所述边界坐标对所述眼底图像进行裁剪，得到视盘区域。
  3. 根据权利要求1所述的基于神经网络的杯盘比确定方法，其中，在所述将所述视盘区域输入预先训练的神经网络之前，所述方法包括：
    对所述视盘区域进行预处理，所述预处理包括伸缩处理。
  4. 根据权利要求1所述的基于神经网络的杯盘比确定方法，其中，所述预先训练的神经网络包括特征提取层、卷积层、连接层、图像分割层和回归预测层；所述将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像，包括：
    通过所述特征提取层对所述视盘区域进行特征提取，以得到所述视盘区域对应的第一特征图像；
    通过所述卷积层对所述第一特征图像进行卷积操作，以得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像；
    通过所述连接层将所述第二特征图像和所述第三特征图像进行连接，得到连接后的第四特征图像；
    将所述第四特征图像输入所述回归预测层，以得到预测杯盘比；
    将所述第四特征图像输入所述图像分割层，以得到视杯视盘的分割图像。
  5. 根据权利要求1所述的基于神经网络的杯盘比确定方法，其中，所述基于所述视杯视盘的分割图像确定算数杯盘比，包括：
    基于所述视杯视盘的分割图像确定视杯直径和视盘直径；
    根据所述视杯直径和视盘直径确定算数杯盘比。
  6. 根据权利要求5所述的基于神经网络的杯盘比确定方法，其中，所述基于所述视杯视盘的分割图像确定视杯直径和视盘直径，包括：
    基于所述视杯视盘的分割图像确定视杯外轮廓和视盘外轮廓；
    根据所述视杯外轮廓和视盘外轮廓分别确定视杯的最小外接矩形和视盘的最小外接矩形；
    基于所述视杯的最小外接矩形和视盘的最小外接矩形分别确定视杯直径和视盘直径。
  7. 根据权利要求1所述的基于神经网络的杯盘比确定方法，其中，所述根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比，包括：
    计算所述算数杯盘比和所述预测杯盘比之间的距离差；
    若所述算数杯盘比和所述预测杯盘比之间的距离差小于或等于第一阈值时，将所述算数杯盘比作为实际杯盘比；
    若所述算数杯盘比和所述预测杯盘比之间的距离差大于第二阈值时，将所述预测杯盘比作为实际杯盘比；
    若所述算数杯盘比和所述预测杯盘比之间的距离差大于第一阈值且小于或等于第二阈值时，计算所述算数杯盘比和所述预测杯盘比的平均值，并将所述平均值作为实际杯盘比。
  8. 一种基于神经网络的杯盘比确定装置，其中，包括：
    图像检测模块，用于获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；
    网络预测模块，用于将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；
    算数确定模块，用于基于所述视杯视盘的分割图像确定算数杯盘比；
    实际确定模块，用于根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
  9. 一种计算机设备，其中，所述计算机设备包括存储器和处理器；
    所述存储器用于存储计算机程序；
    所述处理器，用于执行所述计算机程序并在执行所述计算机程序时实现如下步骤：
    获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；
    将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；
    基于所述视杯视盘的分割图像确定算数杯盘比；
    根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
  10. 根据权利要求9所述的计算机设备，其中，所述对所述眼底图像进行视盘区域检测，以得到视盘区域，包括：
    对所述眼底图像进行检测，得到视盘区域的边界坐标；
    基于所述边界坐标对所述眼底图像进行裁剪，得到视盘区域。
  11. 根据权利要求9所述的计算机设备，其中，在所述将所述视盘区域输入预先训练的神经网络之前，所述处理器执行所述计算机程序还实现如下步骤：
    对所述视盘区域进行预处理，所述预处理包括伸缩处理。
  12. 根据权利要求9所述的计算机设备，其中，所述预先训练的神经网络包括特征提取层、卷积层、连接层、图像分割层和回归预测层；所述将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像，包括：
    通过所述特征提取层对所述视盘区域进行特征提取，以得到所述视盘区域对应的第一特征图像；
    通过所述卷积层对所述第一特征图像进行卷积操作，以得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像；
    通过所述连接层将所述第二特征图像和所述第三特征图像进行连接，得到连接后的第四特征图像；
    将所述第四特征图像输入所述回归预测层，以得到预测杯盘比；
    将所述第四特征图像输入所述图像分割层，以得到视杯视盘的分割图像。
  13. 根据权利要求9所述的计算机设备，其中，所述基于所述视杯视盘的分割图像确定算数杯盘比，包括：
    基于所述视杯视盘的分割图像确定视杯直径和视盘直径；
    根据所述视杯直径和视盘直径确定算数杯盘比。
  14. 根据权利要求13所述的计算机设备，其中，所述基于所述视杯视盘的分割图像确定视杯直径和视盘直径，包括：
    基于所述视杯视盘的分割图像确定视杯外轮廓和视盘外轮廓；
    根据所述视杯外轮廓和视盘外轮廓分别确定视杯的最小外接矩形和视盘的最小外接矩形；
    基于所述视杯的最小外接矩形和视盘的最小外接矩形分别确定视杯直径和视盘直径。
  15. 根据权利要求9所述的计算机设备，其中，所述根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比，包括：
    计算所述算数杯盘比和所述预测杯盘比之间的距离差；
    若所述算数杯盘比和所述预测杯盘比之间的距离差小于或等于第一阈值时，将所述算数杯盘比作为实际杯盘比；
    若所述算数杯盘比和所述预测杯盘比之间的距离差大于第二阈值时，将所述预测杯盘比作为实际杯盘比；
    若所述算数杯盘比和所述预测杯盘比之间的距离差大于第一阈值且小于或等于第二阈值时，计算所述算数杯盘比和所述预测杯盘比的平均值，并将所述平均值作为实际杯盘比。
  16. 一种计算机可读存储介质，其中，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时使所述处理器实现如下步骤：
    获取眼底图像，并对所述眼底图像进行视盘区域检测，以得到视盘区域；
    将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像；
    基于所述视杯视盘的分割图像确定算数杯盘比；
    根据所述算数杯盘比和所述预测杯盘比确定实际杯盘比。
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述对所述眼底图像进行视盘区域检测，以得到视盘区域，包括：
    对所述眼底图像进行检测，得到视盘区域的边界坐标；
    基于所述边界坐标对所述眼底图像进行裁剪，得到视盘区域。
  18. 根据权利要求16所述的计算机可读存储介质，其中，在所述将所述视盘区域输入预先训练的神经网络之前，所述计算机程序被处理器执行时使所述处理器还实现如下步骤：
    对所述视盘区域进行预处理，所述预处理包括伸缩处理。
  19. 根据权利要求16所述的计算机可读存储介质，其中，所述预先训练的神经网络包括特征提取层、卷积层、连接层、图像分割层和回归预测层；所述将所述视盘区域输入预先训练的神经网络，得到预测杯盘比和视杯视盘的分割图像，包括：
    通过所述特征提取层对所述视盘区域进行特征提取，以得到所述视盘区域对应的第一特征图像；
    通过所述卷积层对所述第一特征图像进行卷积操作，以得到代表预测杯盘比的第二特征图像和代表视杯视盘的分割图像的第三特征图像；
    通过所述连接层将所述第二特征图像和所述第三特征图像进行连接，得到连接后的第四特征图像；
    将所述第四特征图像输入所述回归预测层，以得到预测杯盘比；
    将所述第四特征图像输入所述图像分割层，以得到视杯视盘的分割图像。
  20. 根据权利要求16所述的计算机可读存储介质，其中，所述基于所述视杯视盘的分割图像确定算数杯盘比，包括：
    基于所述视杯视盘的分割图像确定视杯直径和视盘直径；
    根据所述视杯直径和视盘直径确定算数杯盘比。
PCT/CN2020/125009 2020-09-21 2020-10-30 基于神经网络的杯盘比确定方法、装置、设备及存储介质 Ceased WO2021189849A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US17/612,566 US12086982B2 (en) 2020-09-21 2020-10-30 Method for confirming cup-disc ratio based on neural network, apparatus, device, and computer readable storage medium
JP2021568938A JP7260674B2 (ja) 2020-09-21 2020-10-30 ニューラルネットワークに基づくｃ／ｄ比決定方法、コンピュータ機器及び記憶媒体

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010992346.X 2020-09-21
CN202010992346.XA CN111862187B (zh) 2020-09-21 2020-09-21 基于神经网络的杯盘比确定方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021189849A1 true WO2021189849A1 (zh) 2021-09-30

Family

ID=72967804

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/125009 Ceased WO2021189849A1 (zh) 2020-09-21 2020-10-30 基于神经网络的杯盘比确定方法、装置、设备及存储介质

Country Status (4)

Country Link
US (1) US12086982B2 (zh)
JP (1) JP7260674B2 (zh)
CN (1) CN111862187B (zh)
WO (1) WO2021189849A1 (zh)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111862187B (zh) * 2020-09-21 2021-01-01 平安科技（深圳）有限公司 基于神经网络的杯盘比确定方法、装置、设备及存储介质
CN112651938B (zh) * 2020-12-24 2023-12-19 平安科技（深圳）有限公司 视盘图像分类模型的训练方法、装置、设备及存储介质
CN113450341B (zh) * 2021-07-16 2024-12-06 依未科技(北京)有限公司 图像处理方法及装置、计算机可读存储介质及电子设备
CN113658165B (zh) * 2021-08-25 2023-06-20 平安科技（深圳）有限公司 杯盘比确定方法、装置、设备及存储介质
CN115778314B (zh) * 2021-09-09 2026-01-06 宏碁股份有限公司 眼睛状态评估方法及电子装置
CN115331050A (zh) * 2021-10-11 2022-11-11 深圳硅基智能科技有限公司 基于紧框标的眼底图像的测量方法、装置及网络训练
CN114627071B (zh) * 2022-03-10 2025-08-19 平安科技（深圳）有限公司 视杯视盘分割方法、装置、计算机及存储介质
CN115272333B (zh) * 2022-09-29 2023-08-04 西藏腾狮软件有限公司 一种杯盘比数据的存储系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102112044A (zh) * 2008-05-14 2011-06-29 科学、技术与研究机构 自动杯盘比测量系统
US20150363930A1 (en) * 2013-01-22 2015-12-17 Agency For Science, Technology And Research Cost-sensitive linear reconstruction based optic cup localization
CN107704886A (zh) * 2017-10-20 2018-02-16 北京工业大学 一种基于深度卷积神经网络的医疗图像分级系统和方法
CN109598733A (zh) * 2017-12-31 2019-04-09 南京航空航天大学 基于深度全卷积神经网络的视网膜眼底图像分割方法
CN109829877A (zh) * 2018-09-20 2019-05-31 中南大学 一种视网膜眼底图像杯盘比自动评估方法

Family Cites Families (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007095250A2 (en) * 2006-02-13 2007-08-23 The Cleveland Clinic Foundation Compositions and methods for inhibiting optic nerve damage
SG10201407330UA (en) * 2009-11-16 2015-01-29 Agency Science Tech & Res Obtaining data for automatic glaucoma screening, and screening and diagnostic techniques and systems using the data
US9545196B2 (en) * 2012-05-04 2017-01-17 University Of Iowa Research Foundation Automated assessment of glaucoma loss from optical coherence tomography
US9684959B2 (en) * 2012-08-24 2017-06-20 Agency For Science, Technology And Research Methods and systems for automatic location of optic structures in an image of an eye, and for automatic retina cup-to-disc ratio computation
US9675244B1 (en) * 2015-12-02 2017-06-13 Novartis Ag Location indicator for optical coherence tomography in ophthalmic visualization
EP3373798B1 (en) * 2016-11-22 2020-05-27 Delphinium Clinic Ltd. Method and system for classifying optic nerve head
WO2019083129A1 (ko) * 2017-10-27 2019-05-02 주식회사 뷰노 피검체에 대한 안저 영상의 판독을 지원하는 방법 및 이를 이용한 장치
WO2020149518A1 (ko) * 2019-01-14 2020-07-23 주식회사 에이아이인사이트 안저영상 자동판독을 위한 딥러닝 아키텍처 시스템 및 딥러닝 아키텍쳐 시스템을 이용한 안저영상 자동판독 방법
CN113646805A (zh) * 2019-03-29 2021-11-12 人工智能技术公司 眼科和全身性疾病的基于图像的检测
KR20210000021A (ko) * 2019-06-24 2021-01-04 삼성에스디에스 주식회사 녹내장 진단 장치
CN110992382B (zh) * 2019-12-30 2022-07-15 四川大学 用于辅助青光眼筛查的眼底图像视杯视盘分割方法及系统
US12400753B2 (en) * 2020-03-30 2025-08-26 University Of Florida Research Foundation, Inc. Collaborative feature ensembling adaptation for domain adaptation in unsupervised optic disc and cup segmentation
EP4199801A1 (en) * 2020-08-19 2023-06-28 Phelcom Technologies S/A System, device and method for portable, connected and intelligent eye imaging
CN111862187B (zh) * 2020-09-21 2021-01-01 平安科技（深圳）有限公司 基于神经网络的杯盘比确定方法、装置、设备及存储介质
CN112001920B (zh) * 2020-10-28 2021-02-05 北京至真互联网技术有限公司 眼底图像识别方法及装置和设备
TWI746287B (zh) * 2020-11-26 2021-11-11 宏碁股份有限公司 影像處理系統及影像處理方法
US20220198831A1 (en) * 2020-12-17 2022-06-23 Delphinium Clinic Ltd. System for determining one or more characteristics of a user based on an image of their eye using an ar/vr headset

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102112044A (zh) * 2008-05-14 2011-06-29 科学、技术与研究机构 自动杯盘比测量系统
US20150363930A1 (en) * 2013-01-22 2015-12-17 Agency For Science, Technology And Research Cost-sensitive linear reconstruction based optic cup localization
CN107704886A (zh) * 2017-10-20 2018-02-16 北京工业大学 一种基于深度卷积神经网络的医疗图像分级系统和方法
CN109598733A (zh) * 2017-12-31 2019-04-09 南京航空航天大学 基于深度全卷积神经网络的视网膜眼底图像分割方法
CN109829877A (zh) * 2018-09-20 2019-05-31 中南大学 一种视网膜眼底图像杯盘比自动评估方法

Also Published As

Publication number Publication date
US12086982B2 (en) 2024-09-10
JP2022533685A (ja) 2022-07-25
CN111862187B (zh) 2021-01-01
CN111862187A (zh) 2020-10-30
US20220309654A1 (en) 2022-09-29
JP7260674B2 (ja) 2023-04-18

Similar Documents

Publication Publication Date Title
WO2021189849A1 (zh) 基于神经网络的杯盘比确定方法、装置、设备及存储介质
WO2021189848A1 (zh) 模型训练方法、杯盘比确定方法、装置、设备及存储介质
WO2021057848A1 (zh) 网络的训练方法、图像处理方法、网络、终端设备及介质
CN110175995B (zh) 一种基于病理图像的图像状态确定方法、装置以及系统
JP7760700B2 (ja) 画像表示方法、装置、端末及び記憶媒体
CN110677585A (zh) 目标检测框的输出方法、装置、终端及存储介质
WO2021027325A1 (zh) 视频相似度获取方法、装置、计算机设备及存储介质
CN106231200B (zh) 一种拍照方法及装置
CN110335216A (zh) 图像处理方法、图像处理装置、终端设备及可读存储介质
WO2022100005A1 (zh) 牙齿图像的处理方法、装置、电子设备、存储介质及程序
WO2021147364A1 (zh) 图像曝光调整方法、装置、设备和介质
WO2021097595A1 (zh) 图像的病变区域分割方法、装置及服务器
CN115984228B (zh) 胃镜图像处理方法、装置、电子设备及存储介质
CN111814749A (zh) 人体特征点的筛选方法、装置、电子设备以及存储介质
CN111784607B (zh) 图像色调映射方法、装置、终端设备及存储介质
WO2020224118A1 (zh) 基于图片转换的病灶判断方法、装置、计算机设备
CN113077396B (zh) 直线段检测方法及装置、计算机可读介质和电子设备
CN111951214B (zh) 图像中可读区域的分割方法、装置、电子设备及存储介质
WO2026020742A1 (zh) 环境中的区域匹配方法、装置、电子设备、计算机可读存储介质及计算机程序产品
JP2017091530A (ja) 交通事故の検出方法、交通事故の検出装置及び電子機器
CN117522844A (zh) 一种基于边缘梯度方向的甲状腺结节轮廓特征分析方法、系统、设备及介质
CN115984551A (zh) 基于多尺度反向注意力的息肉分割方法、设备及存储介质
WO2022205606A1 (zh) 模具处理方法、装置、电子设备、系统和存储介质
CN116342457A (zh) 基于深度学习的睑板腺图像识别方法及相关设备
TWI922234B (zh) 圖像處理方法、電子設備及儲存介質

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: 20927896

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2021568938

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20927896

Country of ref document: EP

Kind code of ref document: A1