Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
BREAST TUMOR IMAGE CLASSIFICATION AND PREDICTION METHOD AND APPARATUS FOR MULTI-SOURCE DATA
Document Type and Number:
WIPO Patent Application WO/2022/147940
Kind Code:
A1
Abstract:
Provided is a breast tumor image classification and prediction method for multi-source data. The method comprises: acquiring multi-source data of a breast tumor image to be subjected to detection, wherein the multi-source data comprises labeled image data and unlabeled image data; performing, by using the multi-source data, semi-supervised training on a teacher-student network segmentation model with a preset encoder/decoder structure, so as to obtain an ROI segmentation result; and performing feature extraction on the ROI segmentation result, so as to obtain an image geometric texture feature vector, and inputting the ROI segmentation result into an encoder part in the teacher-student network segmentation model, so as to obtain an image feature vector of deep learning, and in combination with a non-image data vector obtained by pre-processing non-image data, performing classification and prediction by using a pre-trained cascaded random forest classification model, so as to obtain a corresponding classification result which indicates either malignancy or benignancy. By implementing the present invention, the heterogeneity between different sources of data sets can be reduced, and the breast tumor image classification and prediction for multi-source data can be realized.

Inventors:
PAN ZHIFANG (CN)
RU JINTAO (CN)
CHEN GAOXIANG (CN)
LIN YEZHI (CN)
Application Number:
PCT/CN2021/094088
Publication Date:
July 14, 2022
Filing Date:
May 17, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
UNIV WENZHOU MEDICAL (CN)
International Classes:
G06T7/00
Foreign References:
CN112734723A2021-04-30
CN111695644A2020-09-22
CN108304889A2018-07-20
US20180214105A12018-08-02
Attorney, Agent or Firm:
WENZHOU MINGCHUANG INTELLECTUAL PROPERTY AGENCY CO., LTD (CN)
Download PDF: