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

Tongue image texture quantitative analysis method based on multi-scale convolutional neural network

A convolutional neural network and quantitative analysis technology, applied in the field of quantitative analysis of tongue image texture based on multi-scale convolutional neural network, can solve the problems of difficult judgment and low judgment accuracy for doctors, and achieve enhanced pixel recognition ability and exclusive Strong performance and improved detection ability

Pending Publication Date: 2019-09-06
合肥云诊信息科技有限公司
View PDF2 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Benefits of technology

This patented technology allows for faster and more accurate identification of objects by creating multiple layers on top of existing images that have been previously analyzed or processed beforehand. These layers are then combined together into a single layer called a deep learning model (DL). Each new layer helps distinguish differences from previous ones based solely upon its own characteristics. Overall, this innovation improves efficiency and reliance over current methods used for identifying these areas.

Problems solved by technology

Technics: In order to improve understanding how tongues look like during medical tests or physical examinations it's necessary to analyze their appearance by measuring them manually. While this method has limitations due to subjective factors that affect results, there may exist methods to automatedly measure these qualities accurately without requiring any specialized training.

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
  • Tongue image texture quantitative analysis method based on multi-scale convolutional neural network
  • Tongue image texture quantitative analysis method based on multi-scale convolutional neural network
  • Tongue image texture quantitative analysis method based on multi-scale convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0084] Example 1: Consistency verification of detection results for small-scale prick targets

[0085] Experimental results such as Figure 4Shown: the present invention extracts 5,000 prick-like tongue images from the tongue image library through random collection of tongue images and expert review, and establishes a test data set according to four categories: point, prick, petechiae and ecchymosis, and no prick. Three methods of image algorithm, MaskRCNN instance segmentation network and multi-scale convolutional neural network are used to detect tongue-like feature items. The detection rate of the multi-scale convolutional neural network algorithm is higher than 80%, and the contour of the detected point and the data marked point overlap by more than 80%, and the situation of missed detection and false detection is significantly lower than other algorithms; tongue surface The analysis of the experimental results of the small-scale feature detection results of pricks is sho...

Embodiment 2

[0090] Example 2: Consistency verification of detection results for medium-scale targets such as tooth marks

[0091] Experimental results such as Figure 5 As shown, the item of the present invention extracts 5000 cases of tooth marks and no tooth marks from the tongue image database through random collection of tongue images and expert review, according to mild tooth marks, moderate tooth marks, severe tooth marks, and no tooth marks Four classes build test datasets. Using three methods of image algorithm, MaskRCNN instance segmentation network and multi-scale convolutional neural network, the tongue and face tooth mark feature item detection experiment is carried out. The detection rate of the multi-scale convolutional neural network algorithm is higher than 85%, and the detected tooth marks and the contours of the tooth marks marked in the data overlap by more than 80%, and the cases of missed detection and false detection are significantly lower than other algorithms; T...

Embodiment 3

[0096] Example 3: Verification of the consistency of detection results for cracked multi-scale targets

[0097] Experimental results such as Figure 6 As shown, the present invention extracts 5,000 cases of cracked and non-cracked tongue images from the tongue image library through randomly collected tongue images and expert reviews, and establishes test data sets according to four categories: mild cracks, moderate cracks, severe cracks, and no cracks. . Three methods of image algorithm, MaskRCNN instance segmentation network and multi-scale convolutional neural network are used to detect the feature items of tongue surface cracks. The detection rate of the multi-scale convolutional neural network algorithm is higher than 85%, and the detected cracks coincide with the contours of the cracks marked in the data by more than 80%, and the cases of missed detection and false detection are significantly lower than other algorithms; The analysis of the experimental results of the m...

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 discloses a tongue image texture quantitative analysis method based on a multi-scale convolutional neural network, and relates to the technical field of image recognition processing. Themethod comprises the following steps: S1, constructing a multi-scale convolutional neural network; S2, collecting and preprocessing tongue picture big data; S3, training a neural network model; and S4, carrying out pricking, speckle and tooth mark detection on the tongue image. According to the invention, by constructing the multi-scale convolutional neural network, three types of targets with different scales, such as spurs, tooth marks and cracks, in the tongue surface image can be detected at one time; pixel-level contour areas are extracted for detection targets, point pricks, tooth marksand cracks are quantified and graded according to the depth degree and the number index of each detection target, the method has the advantages of being high in specificity, good in repeatability andhigh in accuracy, and the detection capacity can be remarkably improved.

Description

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

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
Owner 合肥云诊信息科技有限公司
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