Title:
IMAGE PROCESSING METHOD, IMAGE PROCESSING SYSTEM AND IMAGE PROCESSING PROGRAM
Document Type and Number:
Japanese Patent JP2023178943
Kind Code:
A
Abstract:
To achieve a balance between accuracy and speed during deep learning model inference to obtain more desirable inference results.SOLUTION: An image processing method according to an embodiment includes an identification step, an inference step, and an integration step. The identification step identifies a first portion of an image that includes a region corresponding to the anatomical site of a target and a second portion that includes a region different from the anatomical site. The inference step performs segmentation processing to identify a region corresponding to the anatomical site for the first portion and segmentation processing to identify a region different from the anatomical site for the second portion by using a deep learning model, or classification processing that classifies and detects an image that includes a region corresponding to the anatomical site for the first portion, and classification processing that classifies and detects an image that includes a region different from the anatomical site for the second portion. The integration step integrates the results of each process and outputs them.SELECTED DRAWING: Figure 3
Inventors:
WANG SHA
HAN BING
MENG FANJIE
XU CHI CHI
YUE YE
HAN BING
MENG FANJIE
XU CHI CHI
YUE YE
Application Number:
JP2023064768A
Publication Date:
December 18, 2023
Filing Date:
April 12, 2023
Export Citation:
Assignee:
CANON MEDICAL SYSTEMS CORP
International Classes:
G06T7/00; G06T7/11; G06V10/82
Attorney, Agent or Firm:
Toranomon Intellectual Property Office