AI-based lesion diagnosis method to improve lesion diagnosis speed and precision

An AI-based lesion diagnosis method to improve lesion diagnosis speed and precision, according to an embodiment, may comprise the steps of: obtaining patch images which are made by dividing an original image by a predetermined size; determining an optimum location for detecting a first lesion object...

Full description

Saved in:
Bibliographic Details
Main Authors JANG JAE MYUNG, CHUNG HEE WON, SEO JANG WON
Format Patent
LanguageEnglish
Korean
Published 08.03.2023
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:An AI-based lesion diagnosis method to improve lesion diagnosis speed and precision, according to an embodiment, may comprise the steps of: obtaining patch images which are made by dividing an original image by a predetermined size; determining an optimum location for detecting a first lesion object from the patch images; outputting diagnosis information related to the first lesion object from a first patch image corresponding to the optimum location through a first lesion diagnosis model; generating a second patch image larger than the first patch image by referring to lesion distribution information corresponding to the first lesion object; and outputting diagnosis information related to a second lesion object from the second patch image through a second lesion diagnosis model. 실시예에 따른 병변 진단 속도와 정밀도를 향상시키기 위한 인공 지능 기반의 병변 진단 방법은, 원본이미지가 소정의 크기로 분할된 패치이미지들을 획득하는 단계; 상기 패치이미지들로부터 제1 병변오브젝트를 검출하기 위한 최적의 위치를 판별하는 단계; 제1 병변진단모델을 통해 상기 최적의 위치에 대응하는 제1 패치이미지로부터 상기 제1 병변오브젝트와 관련된 진단 정보를 출력하는 단계; 상기 제1 병변오브젝트에 대응하는 병변분포도 정보를 참조하여 상기 제1 패치이미지보다 큰 제2 패치이미지를 생성하는 단계; 및 제2 병변진단모델을 통해 상기 제2 패치이미지로부터 제2 병변오브젝트와 관련된 진단 정보를 출력하는 단계;를 포함할 수 있다.
Bibliography:Application Number: KR20210114910