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

Sample evaluation method and model training method of coronary artery segmentation model

A technology of segmentation model and evaluation method, applied in image analysis, character and pattern recognition, image data processing, etc., can solve the problem of unrobust post-processing, and achieve the effect of robust results

Active Publication Date: 2019-01-25
数坤(上海)医疗科技有限公司
View PDF10 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Since the loss function used in the training of the 3D coronary artery model is DiceLoss, this function cannot allow the segmentation to learn connected information, and the coronary artery is a continuum. If the prediction result of the continuum breaks, the prediction result will be affected by the post-processing link is very unrobust

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
  • Sample evaluation method and model training method of coronary artery segmentation model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] The invention discloses a sample evaluation method of a coronary artery segmentation model, comprising:

[0027] S1. Input samples to train the 3D coronary artery segmentation model to obtain volume prediction data for coronary artery segmentation;

[0028] S2. Slice the sample and its corresponding volume prediction data along the Z direction, and perform score calculation on each slice:

[0029] S21. Marking the sample slices;

[0030] S22. Comparing the sample slice marking result with the corresponding volume prediction data slice result to obtain the score of the sample slice: if the volume prediction data slice is segmented at the mark position corresponding to the sample slice, the true positive number is increased by 1; If the data slice does not appear to be segmented at the marked position corresponding to the sample slice, then the false negative number is increased by 1; if the volume prediction data slice is segmented at the marked position not correspondi...

Embodiment 2

[0033] The present invention also provides a training method for a coronary artery segmentation model, comprising:

[0034] S1, utilize the sample evaluation method of coronary artery segmentation model as claimed in claim 1 to evaluate all samples of current training cycle, obtain the scoring of each sample;

[0035] S2. Classify samples with different scores, and adopt different training strategies for samples of different classes.

[0036] Sample classification can be based on the following ideas:

[0037] a. Calculate the mean value of each sample score, evaluate the samples whose scores are lower than the mean value as difficult samples, and evaluate the samples whose scores are greater than or equal to the mean value as easy samples;

[0038] b. Arrange the samples according to the level of their scores, and select the sample with the lowest score according to the set ratio (set to 5% in this embodiment), and evaluate it as a difficult sample, and evaluate the remaining...

Embodiment 3

[0049] The present invention also provides another training method for a coronary artery segmentation model, including:

[0050] S1. Perform multiple trainings under different conditions for each sample;

[0051] S2, utilize the sample evaluation method of coronary artery segmentation model as claimed in claim 1 to carry out scoring to each training to certain cycle or the sample that training is finished, make each sample obtain multinomial scoring;

[0052] S3. Perform mean value calculation or voting on each sample to obtain a comprehensive score for each sample;

[0053] S4. Classify the comprehensive score of each sample, and adopt different training strategies for different types of samples.

[0054] The different training strategies are carried out with reference to Example 2.

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 sample evaluation method of a coronary artery segmentation model, comprising the following steps: S1, inputting a sample to train a 3D coronary artery segmentation model to obtain volume prediction data; S2, slicing the sample and the corresponding volume prediction data along the Z direction, and calculating the score of each slice; S21, marking the sample slice; S22, comparing the sample slice mark result with the corresponding volume prediction data slice result to obtain the score of the sample slice; S23, taking the mean value of the scores of each sample slice to obtain the scores of the samples. At the same time, the invention also discloses two training methods of the coronary artery segmentation model, namely, utilizing the sample score to evaluate the samples of the coronary artery segmentation model for single training or multiple training respectively, and endowing different training strategies according to the difficulty degree of the samples, soas to obtain a better training model.

Description

technical field [0001] The invention relates to the field of coronary artery image segmentation, in particular to a sample evaluation method and a model training method of a coronary artery segmentation model. Background technique [0002] Automated coronary reconstruction has important clinical value and practical significance for doctors. Because the loss function used in the training of the 3D coronary artery model is DiceLoss, this function cannot allow the segmentation to learn connected information, and the coronary artery is a continuum. If the prediction result of the continuum breaks, the prediction result will be affected by the post-processing Links are very unrobust. [0003] Therefore, it is necessary to evaluate the training results of the coronary artery segmentation model, and propose corresponding model training methods to improve the robustness of the prediction results. Contents of the invention [0004] The object of the present invention is to provid...

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): G06T7/11G06K9/34G06K9/62
CPCG06T7/11G06V10/267G06F18/2193
Inventor 肖月庭阳光郑超
Owner 数坤(上海)医疗科技有限公司
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