Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Algorithm for automatically identifying non-perfusion area of fundus fluorescence angiography image and recommending laser photocoagulation area

A technology of contrast-enhanced images and automatic recognition, which is applied in the field of intelligent medical care, can solve problems that have not been reported, and achieve the effect of improving disease analysis and diagnosis ability

Pending Publication Date: 2021-06-15
山西省眼科医院 +1
View PDF11 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, there is no report on the FFA imaging of retinal vein occlusion

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Algorithm for automatically identifying non-perfusion area of fundus fluorescence angiography image and recommending laser photocoagulation area
  • Algorithm for automatically identifying non-perfusion area of fundus fluorescence angiography image and recommending laser photocoagulation area
  • Algorithm for automatically identifying non-perfusion area of fundus fluorescence angiography image and recommending laser photocoagulation area

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0042] Construction of Neural Network Model

[0043] Three standardized Unet convolutional neural network models were constructed based on the Keras deep learning framework with TensorFlow as the backend, which were used to identify the optic disc area, macular area, and non-perfusion area. Build various layers of the network model through the layers module in the keras framework. The method of building the convolution layer is layers.Conv2D, the method of building the pooling layer is layers.MaxPooling2D, and the method of building the transposed convolution layer is layers.Upsampling2D. The splicing of high-level features and low-level features uses the layers.concatenate method.

[0044] Network model structure description: The network model has a total of 34 layers, including 4 parts: input layer, downsampling layer, upsampling layer, and output layer.

[0045] The input is all contrast images with a size of 768*768.

[0046] The downsampling layer consists of the first ...

Embodiment 2

[0050] The optimization of the network model includes the following steps:

[0051] Step 1: A professional physician collects a 55-degree fundus angiography image of the posterior pole of the fundus, and the collection device is Heidelberg SPECTRALIS HRA+OCT.

[0052] Step 2: Image desensitization preprocessing: Use the cv2.findContours method under the opencv-python module to extract all the contours of the fundus area, among which the retinal contour has the largest contour area, and remove the patient information on the image and set the background area to 0.

[0053] For desensitization pretreatment results, see figure 1 ,in figure 1 -a is the extraction result of the fundus area contour, and the dotted line range is the extraction result of the fundus area; figure 1 -b is the output result of desensitization, which removes patient information and other content; figure 1 -c is the final result after preprocessing.

[0054] Step 3: Use the Contrast Limited Adaptive Hist...

Embodiment 3

[0062] Recognition and Segmentation of Fundus Imaging Based on Multi-Stream Convolutional Neural Network Model

[0063]Load the collected patient's fundus contrast image into the trained network model to obtain the probability that each pixel of the contrast image belongs to each category. The fundus contrast images are identified and segmented as follows.

[0064] 1. Positioning of the center of the disc:

[0065] (1) Select the maximum value p_max and the minimum value p_min of all pixel prediction probabilities; (2) determine the dynamic threshold for separating the optic disc and the background, p_thresh=(1-k)*p_min+k*p_max, where p_tresh represents the separation of the optic disc and the background The threshold value of k is 0.8; (3) Select the pixel points greater than the threshold value as the candidate area of ​​the optic disc, such as Figure 4 -shown in -a; (4) open operation to the disc region; (5) traverse each connected region S in the candidate region, and j...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention belongs to the technical field of intelligent medical treatment, in particular to an algorithm for automatically identifying non-perfusion area of Fundus Fluorescence Angiography (FFA) image and recommending laser photocoagulation area. Aiming at the condition that non-perfusion area quantitative analysis is carried out on FFA images without Artificial Intelligence (AI) technology at home and abroad at present, the algorithm adopts a fundus fluorescence angiography machine with a 55-degree visual field to mark non-perfusion area of retina of an Retinal Vein Occlusion (RVO) patient, and tracks natural disease course, treatment and prognosis of the RVO patient by combining powerful and efficient image analysis and judgment capability as well as clinical big data processing capability of AI, thereby achieving accurate quantitative analysis of the non-perfusion area of the FFA image; and moreover, planning can also be provided for a laser operation with optic disc macular area intelligently avoided.

Description

technical field [0001] The invention belongs to the technical field of intelligent medical treatment, and in particular relates to an algorithm for automatic identification of non-perfusion areas in fundus contrast images and recommendation of laser photocoagulation areas. Background technique [0002] Thrombosis in the retinal venous system or obstruction of the corresponding veins due to inflammation is called retinal vein occlusion (RVO), which is the second most common retinal vascular disease after diabetic retinopathy. Important cause of vision loss. Clinically, RVO can be divided into central retinal vein occlusion (Central Retinal Vein Occlusion, CRVO) and branch retinal vein occlusion (Branch Retinal Vein Occlusion, BRVO). [0003] Fundus fluorescein angiography (FFA) can be used to evaluate the extent of RVO vascular obstruction, degree of ischemia, and type of macular edema, and is an important tool and gold standard for evaluating retinal circulation status. Is...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): A61B3/12A61B3/14
CPCA61B3/1241A61B3/14
Inventor 张喜梅侯军军谢娟孙斌张光华马非刘汉王龙
Owner 山西省眼科医院
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com