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

Method for establishing and detecting detection model of pathological image with missing annotation as training data

A technology of pathological images and detection models, applied in the field of medical image analysis, can solve problems such as wrong assignment of suggestion frame labels, inaccurate detection results, etc., and achieve the effect of improving precision and detection accuracy

Active Publication Date: 2020-12-18
NORTHWEST UNIV
View PDF6 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to solve the deficiencies in the prior art, the present invention provides a detection model establishment and detection method in which the training data is a pathological image with missing annotations, so as to solve the problem of wrong distribution of the suggestion frame label in the existing detection method, which makes the detection result inaccurate. exact question

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 for establishing and detecting detection model of pathological image with missing annotation as training data
  • Method for establishing and detecting detection model of pathological image with missing annotation as training data
  • Method for establishing and detecting detection model of pathological image with missing annotation as training data

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0056] This embodiment discloses a method for establishing a detection model in which the training data is pathological images with missing annotations. The process is as follows: figure 1 shown. Specifically include the following steps:

[0057] Step 1, image preprocessing

[0058] Step 1.1, expand the pathological image data to obtain the expanded image;

[0059] Pathological images will be affected by various factors during the production process, such as the concentration of stains used in making glass slides, and the brightness difference of digital scanning, which will cause different slices in the pathological image or even areas of the same semantic category in the same slice Pixels vary greatly. In order to make the model adapt to the problem of inconsistent pixel values ​​in the data during the training process, the present invention expands the data through random color changes, so that the color information of the image in the training data is more abundant.

...

Embodiment 2

[0081] This embodiment discloses a system for establishing a detection model whose training data is pathological images with missing annotations. The system includes an image preprocessing module, a detection network construction module and a detection model training module. Its structure is as figure 2 shown.

[0082] (1) image preprocessing module, including image expansion module and re-encoding module, wherein,

[0083](1.1) Image expansion module, used to expand the image data to obtain the expanded image;

[0084] (1.2) The re-encoding module is used to fill the instance-level annotation box in the expanded image to generate a mask image, and re-encode the mask image to obtain the re-encoded mask image.

[0085] The re-encoding and image extension methods in this embodiment are the same as those in Embodiment 1.

[0086] (2) The detection network building block is used to construct the detection network. Specifically, the detection network in this embodiment include...

Embodiment 3

[0112] This embodiment discloses a detection method of a pathological image, the detection method comprising the following steps:

[0113] Step 1, the pathological image to be processed, such as Figure 5 As shown in (a) figure, carry out the pretreatment of the step 1 of embodiment 1, obtain the image after the expansion;

[0114] Step 2. Input the image obtained in Step 1 into the detection model obtained in Example 1 to obtain the category of the suggestion frame. Such as Figure 5 As shown in (b) figure.

[0115] According to the category of the suggestion frame, it can be judged whether the target to be detected is contained in the suggestion frame. If the suggestion box is marked with a positive label, it means that the features of the area surrounded by the suggestion box are closer to the features of the target to be detected. Through this result, a certain auxiliary effect can be given when manually judging pathological images.

[0116] In addition, the present i...

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 method for establishing and detecting a detection model of a pathological image with missing annotation as training data, which comprises the following steps: firstly, preprocessing a case image, then constructing a detection network, and finally training the detection network to obtain the detection model; when a to-be-processed case image is detected, inputting the to-be-processed pathological image into the detection model to obtain a suggestion box category, thereby judging whether the suggestion box contains the to-be-detected object or not. According to the method, the segmentation model and the suggestion box label updating module are introduced into the existing region-based target detection network, the label of the suggestion box is corrected through theoutput of the segmentation model, and the potential positive suggestion box is mined. The problem that the labels of the suggestion box are wrongly distributed due to missing of the labels is solved.By adopting the collaborative supervision training method, the precision and the detection accuracy of the model are improved.

Description

technical field [0001] The invention belongs to the technical field of medical image analysis, and relates to a detection model establishment and detection method in which the training data is a missing and marked pathological image. Background technique [0002] With the development of convolutional neural networks, the accuracy and efficiency of object detection networks have been continuously improved in recent years, among which region-based fully supervised object detection methods have always shown the best performance, and at the same time, these methods are also widely used. It is used in pathological image analysis, such as abnormal cell detection or red blood cell detection. On the other hand, the rapid development of region-based fully supervised detection methods benefits from the availability of large-scale data with accurate instance-level annotation boxes, however, collecting such data takes a lot of time and labor, especially for pathological data, which requ...

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
IPC IPC(8): G06T7/00G06T7/11G06T7/136G06K9/62
CPCG06T7/0012G06T7/11G06T7/136G06T2207/20081G06T2207/20084G06T2207/30096G06F18/2415G06F18/214
Inventor 冯筠韩鑫李涵生
Owner NORTHWEST UNIV