US 12,446,765 B2
Artificial intelligence-based gastroscopic image analysis method
Kyung Nam Kim, Suwon-si (KR); and Jie-Hyun Kim, Seoul (KR)
Assigned to WAYCEN INC., Seoul (KR); and INDUSTRY-ACADEMIC COOPERATION FOUNDATION, YONSEI UNIVERSITY, Seoul (KR)
Appl. No. 18/020,232
Filed by WAYCEN INC., Seoul (KR); and INDUSTRY-ACADEMIC COOPERATION FOUNDATION, YONSEI UNIVERSITY, Seoul (KR)
PCT Filed Aug. 4, 2021, PCT No. PCT/KR2021/010186
§ 371(c)(1), (2) Date Feb. 7, 2023,
PCT Pub. No. WO2022/035119, PCT Pub. Date Feb. 17, 2022.
Claims priority of application No. 10-2020-0100030 (KR), filed on Aug. 10, 2020.
Prior Publication US 2023/0301503 A1, Sep. 28, 2023
Int. Cl. A61B 1/273 (2006.01); G06T 7/00 (2017.01); G16H 30/20 (2018.01); G16H 30/40 (2018.01)
CPC A61B 1/2736 (2013.01) [G06T 7/0012 (2013.01); G16H 30/20 (2018.01); G06T 2207/10068 (2013.01); G06T 2207/20084 (2013.01); G16H 30/40 (2018.01)] 12 Claims
OG exemplary drawing
 
1. An artificial intelligence (AI)-based gastroscopic image analysis method, wherein a captured gastroscopic image is input to a computer and the gastroscopic image is analyzed by an artificial intelligence training model installed in the computer, the artificial intelligence (AI)-based gastroscopic image analysis method comprising:
a) configuring a plurality of artificial intelligence-based image classification models for the captured gastroscopic image;
b) configuring training data for each of the plurality of image classification models and training the plurality of image classification models;
c) observing or photographing respective parts of a stomach from an oral cavity and a laryngopharynx to a second part of a duodenum or from the second part of the duodenum to the oral cavity and the laryngopharynx using a gastroscopic probe;
d) automatically classifying and recognizing an anatomical location of the stomach with respect to an image captured during photographing using the plurality of image classification models, automatically storing or reporting the location of a lesion, and verifying whether parts, images of which are recommended to be captured and stored, in gastroscopy have been photographed; and
e) segmenting an image of a specific part in the captured gastroscopic image for each region by a region segmentation model and outputting the same number of segmented maps as the number of target classes as a result of segmentation, wherein
when the plurality of image classification models is configured in step a), the image classification models are configured so as to be classified into parts adjacent to target parts, and the image classification models are configured so as to be classified into a first classification model to a tenth classification model,
the image of the specific part comprises images of a body of the stomach, an antrum of the stomach, and a fundus of the stomach in step e), and
each of the images of the body of the stomach, the antrum of the stomach, and the fundus of the stomach is segmented for each region by the region segmentation model and comprises an anterior wall, a posterior wall, a lesser curvature, and a greater curvature as a result of segmentation.