DeepAMD: Detect Early Age-Related Macular Degeneration by Applying Deep Learning in a Multiple Instance Learning Framework

Huiying Liu, Damon W.K. Wong, Huazhu Fu, Yanwu Xu, Jiang Liu

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

6 Citations (Scopus)

Abstract

Automatic screening of Age-related Macular Degeneration (AMD) is important for both patients and ophthalmologists. In this paper, we focus on the task of AMD detection at the very early stage from fundus images. The difficulty of this task is that at the very early stage, the signs, e.g., drusen, are too tiny and subtle to be detected by most of the current methods. To address this issue, we apply deep learning in a multiple instance learning framework to catch these subtle features to detect AMD at the very early stage. The deep networks is able to learn a discriminative representation of the subtle signs of AMD. The multiple instance learning framework helps in two ways. First, It is able to choose the location where AMD happens because it works on image patches instead of the whole image. Second, It works on the image of high resolution instead of down sampling the image which may lead to invisibility of the tiny drusen. The experiments are carried out on a dataset consists of 3596 AMD and 1129 normal fundus images. The final average AUC is 0.79, compared with 0.74 of the same neural network but without multiple instance learning.

Original languageEnglish
Title of host publicationComputer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers
EditorsKonrad Schindler, C.V. Jawahar, Hongdong Li, Greg Mori
PublisherSpringer Verlag
Pages625-640
Number of pages16
ISBN (Print)9783030208721
DOIs
Publication statusPublished - 2019
Externally publishedYes
Event14th Asian Conference on Computer Vision, ACCV 2018 - Perth, Australia
Duration: 2 Dec 20186 Dec 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11365 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th Asian Conference on Computer Vision, ACCV 2018
Country/TerritoryAustralia
CityPerth
Period2/12/186/12/18

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science (all)

Fingerprint

Dive into the research topics of 'DeepAMD: Detect Early Age-Related Macular Degeneration by Applying Deep Learning in a Multiple Instance Learning Framework'. Together they form a unique fingerprint.

Cite this