Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD, APPARATUS, AND DEVICE FOR DETECTING AND LOCATING LESION IN MEDICAL IMAGE, AND STORAGE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2020/215672
Kind Code:
A1
Abstract:
A method, apparatus, and device for detecting and locating a lesion in a medical image, and a storage medium. The invention applies to the technical field of deep learning, and solves the problem of the large amount of annotation work required in existing detection methods. The method comprises: acquiring a target medical image having a lesion to be detected (101); inputting the target medical image into a deep learning model to obtain a target sequence (102); then, extracting a first feature map outputted by the last convolution layer in the deep learning model (103); extracting from a fully connected layer a weight value for each network element corresponding to each predetermined lesion type (104); for each predetermined lesion type, on the basis of the first feature map and the corresponding weight values, calculating to obtain a fused feature map (105); re-sampling each fused feature map to the size of the target medical image, and generating class activation maps (106); determining the largest connected area in each class activation map, and establishing a marking frame surrounding the largest connected area (107); and adding to the target medical image a marking frame corresponding to each predetermined lesion type (108).

Inventors:
WANG YUE (CN)
LV BIN (CN)
LV CHUANFENG (CN)
Application Number:
PCT/CN2019/117098
Publication Date:
October 29, 2020
Filing Date:
November 11, 2019
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
PING AN TECH SHENZHEN CO LTD (CN)
International Classes:
G06T7/00; G06V10/764; G06V10/774; G06V10/776
Foreign References:
CN106780460A2017-05-31
CN109102491A2018-12-28
CN109949318A2019-06-28
US20160162757A12016-06-09
Attorney, Agent or Firm:
SHENZHEN ZHONGDING INTELLECTUAL PROPERTY AGENCY (CN)
Download PDF: