| 녹 F-011 |
| 딥러닝으로 구현한 슈퍼해상도 시신경유두 사진을 이용한 시신경출혈의 발견 |
| 1. 서울대학교 의과대학 안과학교실
2. 서울대학교병원 안과
3. 서울대학교 의과대학 의공학교실
4. 서울대학교병원 의공학과 |
| 하아늘(1,2), 선석규(3,4), 김영국(1,2), 이진호(1,2), 정진욱(1,2), 김희찬(3,4), 박기호(1,2) |
|
목적 : Proposal of deep-learning approach for improved resolution and legibility of optic-disc photography (ODP).
방법 : Each high-resolution original ODP was transformed into two counterparts: (1) down-scaled ‘low-resolution ODPs’, and (2) ‘compensated high-resolution ODPs’ produced via enhancement of the visibility of the optic disc margin and surrounding vessels using a customized image post-processing algorithm. Then, the differences between those two ODPs were directly learned through a super-resolution generative adversarial network (SR-GAN). Finally, by inputting the high-resolution ODPs into SR-GAN, 4-times-up-scaled and overall-color-and-brightness-transformed ‘accelerated ODPs’ could be obtained.
결과 : Datasets consisting of 50 original and 50 paired accelerated ODPs were evaluated, and 23 cases of disc hemorrhage (DH) were confirmed. Twelve general ophthalmologists were instructed (1) to assess each ODP’s image quality, and (2) to note any abnormal findings, at 1-month intervals. The quality grade was numbered between 1 and 5 (higher score indicating better quality). The score for accelerated ODPs was higher than that for the original ODP (4.36±0.4 vs. 3.15±0.9, P<0.001), and the overall DH-detection accuracy was significantly higher with the accelerated ODPs (90.7 vs. 76.3%, P<0.001).
결론 : We propose a novel deep-learning based accelerated-ODP that significantly enhances general ophthalmologists’ DH-detection accuracy relative to original ODPs.
|
| |