대한안과학회 학술대회 발표 연제 초록

발표일자: 2019년 11월 1일(금)~3(일) 발표번호: P(e-poster)-006 발표장소: B3 Parking Area
하이브리드 딥러닝 모델을 이용한 무적색광 안저사진에서의 황반부 신경절세포-내망상층 두께 예측
1. 서울대학교 의과대학 안과학교실 2. 서울대학교병원 안과 3. 서울대학교 의과대학 의공학교실 학제간 융합연구 프로그램 4. 한림대학교춘천성심병원 안과
이진호(1,2), 김영국(1,2), 하아늘(1,2), 선석규(3), 김용우(1,2), 김진수(4), 정진욱(1,2), 박기호(1,2)
목적 : To propose a hybrid deep learning model (HDLM) algorithm that quantitatively predicts macular ganglion cell-inner plexiform layer (mGCIPL) thickness from red-free retinal nerve fiber layer photographs (RNFLPs). 방법 : A total of 789 pairs of RNFLPs (Vx-10a; Kowa Optimed) and macular spectral domain-optical coherence tomography (SD-OCT; Cirrus, Carl Zeiss Meditec) scans from 431 eyes of 259 participants were enrolled. An HDLM was built by combining a deep learning network and support vector machine. The correlation and agreement between the predicted and observed mGCIPL thicknesses were calculated. 결과 : The predicted mGCIPL thickness was strongly correlated to the actual thickness obtained from SD-OCT (r = 0.739; R2 = 0.545). Even when the peripapillary area was masked, the strong correlation was not changed significantly (r = 0.713; R2 = 0.508; P = 0.378). 결론 : The trained HDLM algorithm showed a great capability for mGCIPL thickness prediction from RNFLPs.

[돌아가기]