Unlock instant, AI-driven research and patent intelligence for your innovation.

Method and system for automatic marking of small lesions in high-definition medical images

A medical image, small technology, applied in the field of image processing, can solve the problems of image information loss, excessive resolution difference, high cost, and achieve the effect of improving accuracy, ensuring accuracy, and enhancing sensitivity

Active Publication Date: 2021-06-22
HIGHWISE CO LTD
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The direct compression of images generally adopts the pixel sampling method. If the resolution difference between the original image and the compressed image is too large, it will often cause a large loss of image information. In the field of high-definition medical imaging, there are even negative effects such as loss of details in small lesion areas.
On the other hand, machine learning algorithms based on convolutional neural networks often require a large amount of manually labeled data as training samples. Due to the particularity of the medical field, the task of labeling data requires high-cost medical staff with relevant medical backgrounds. , obtaining a large amount of medical image annotation data is another challenge that is different from ordinary image tasks

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
  • Method and system for automatic marking of small lesions in high-definition medical images
  • Method and system for automatic marking of small lesions in high-definition medical images
  • Method and system for automatic marking of small lesions in high-definition medical images

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0173] Embodiment 1: The implementation process of an automatic marking method for small lesions in a high-definition chest X-ray film image based on semi-supervised deep learning can be as follows figure 1 As shown, it includes the following steps:

[0174] Step 1: Construct a large-scale image input network for docking with a general convolutional neural network when high-definition medical images are used as data input. The implementation methods may include:

[0175] Select a large-scale convolution kernel C with a size of 13×13×64 0 , will be dimensioned as W I ×H I The high-definition medical image of ×3 is used as the input I for convolution operation, the convolution step is set to 6, and the smaller-scale convolution feature matrix T is obtained through pooling Pooling(·) and activation function A(·) 0 :

[0176]

[0177] where T 0 is of size Represents a convolution operation;

[0178] Select a convolution kernel C with a size of 1×1×128 1 As the convol...

Embodiment 2

[0216] Example 2: The implementation process of an automatic marking method for small lesions in diabetic retinal images based on semi-supervised deep learning can also refer to figure 1 , which includes the following steps:

[0217] Step 1: Construct a large-scale image input network for docking with a general convolutional neural network when using high-definition diabetic retinal images as data input. The implementation methods may include:

[0218] Select a large-scale convolution kernel C with a size of 13×13×64 0 , will be dimensioned as W I ×H I The ×3 high-definition diabetic retinal image is used as the input I for convolution operation, the convolution step is set to 6, and the smaller-scale convolution feature matrix T is obtained through pooling Pooling(·) and activation function A(·) 0 :

[0219]

[0220] where T 0 is of size Represents a convolution operation;

[0221] Select a convolution kernel C with a size of 1×1×128 1 As the convolution of the ...

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

This application discloses a method for automatic marking of small lesions in high-definition medical images, including: constructing a large-scale image input network, a deep cascaded regional feature extraction network model, and setting a central target focus detection method, and then combining the three It is jointly incorporated into the convolutional neural network target detection model based on the feature pyramid and regional hypothesis methods, so as to construct a high-definition image convolutional neural network model for small object detection that can be trained and implemented; set the semi-supervised learning based The automatic screening and iterative training method of hidden positive samples is used to automatically obtain the unmarked positive sample data in the training data set to the maximum extent, add them to the hidden positive sample data set that can be used for training, and then use the hidden positive samples to improve the model training effect. Using the method of the present application, small lesion areas can be automatically, quickly and accurately marked based on high-definition medical 2D scan images, which is beneficial to assist doctors in subsequent operations to improve diagnostic accuracy.

Description

technical field [0001] The present application relates to an image processing method, in particular to an automatic marking method for small lesions in high-definition medical images, which belongs to the field of computer technology. Background technique [0002] With the development of medical imaging equipment, the quality and functions of medical imaging are constantly updated and improved, and more and more medical imaging equipment has been popularized in hospitals at all levels. Medical imaging examination has become an important diagnostic basis and reference in the process of diagnosing various diseases because it can intuitively provide important information such as the location, structure and function of lesions, as well as provide intuitive information and reference for the diagnosis and treatment of diseases. means of inspection. 2D high-definition biomedical scanning image inspection is an important part of medical imaging. Because of its clear representation ...

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 Patents(China)
IPC IPC(8): G06K9/62G06K9/66G06N3/04G06N3/08
CPCG06N3/08G06V30/194G06N3/045G06F18/214
Inventor 曹鱼陈齐磊刘本渊
Owner HIGHWISE CO LTD