June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
A device-agnostic deep learning model for detecting keratoconus based on anterior elevation corneal maps
Author Affiliations & Notes
  • Ali Al-Timemy
    Biomedical Engineering, University of Baghdad Al-Jaderyia Campus Al-Khwarizmi College of Engineering, Baghdad, Baghdad , Iraq
    University of Plymouth, Plymouth, Devon, United Kingdom
  • Laith Al-Zubaidi
    School of Mechanical, Medical and Process Engineering, Queensland University of Technology, Brisbane, Queensland, Australia
  • Nebras Ghaeb
    Biomedical Engineering, University of Baghdad Al-Jaderyia Campus Al-Khwarizmi College of Engineering, Baghdad, Baghdad , Iraq
  • Hidenori Takahashi
    Department of Ophthalmology, Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Alexandru Lavric
    Computers, Electronics and Automation, Universitatea Stefan cel Mare din Suceava, Suceava, Romania
  • Zahraa Mosa
    College of Pharmacy, Uruk University, Baghdad, Iraq
  • Rossen M Hazarbassanov
    Department of Ophthalmology and Visual Sciences, Universidade Federal de Sao Paulo Departamento de Diagnostico por Imagem, Sao Paulo, SP, Brazil
  • Zaid Abdi Alkareem Alyasseri
    ECE Department-Faculty of Engineering,, University of Kufa, Kufa, Najaf, Iraq
    Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia, Bangi, Selangor, Malaysia
  • Siamak Yousefi
    Department of Ophthalmology, The University of Tennessee Health Science Center, Memphis, Tennessee, United States
    Department of Genetics, Genomics, and Informatics, The University of Tennessee Health Science Center, Memphis, Tennessee, United States
  • Footnotes
    Commercial Relationships   Ali Al-Timemy None; Laith Al-Zubaidi None; Nebras Ghaeb None; Hidenori Takahashi None; Alexandru Lavric None; Zahraa Mosa None; Rossen Hazarbassanov None; Zaid Abdi Alkareem Alyasseri None; Siamak Yousefi None
  • Footnotes
    Support  ARVO Foundation Collaborative Research Fellowship
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 2101 – F0090. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ali Al-Timemy, Laith Al-Zubaidi, Nebras Ghaeb, Hidenori Takahashi, Alexandru Lavric, Zahraa Mosa, Rossen M Hazarbassanov, Zaid Abdi Alkareem Alyasseri, Siamak Yousefi; A device-agnostic deep learning model for detecting keratoconus based on anterior elevation corneal maps. Invest. Ophthalmol. Vis. Sci. 2022;63(7):2101 – F0090.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : To develop and validate a deep learning model for detecting keratoconus based on corneal maps from CASIA and Pentacam instruments.

Methods : We collected 3428 corneal maps from CASIA optical coherence tomography (OCT)-based imaging instrument (SS-1000, Tomey, Japan) at Jichi Medical University, Japan, and collected 217 corneal maps from Pentacam Scheimpflug-based imaging instrument (Oculus Optikgerate GmbH) at Al-Amal eye clinic in Baghdad, Iraq. We developed a deep learning framework based on AlexNet architecture to detect keratoconus based on anterior elevation maps in CASIA dataset only. We then evaluated the developed model based on corneal maps in the Pentacam dataset using accuracy, specificity, sensitivity, and area under the receiver operating characteristic curve (AUC) (Fig.1)

Results : The CASIA dataset included corneal maps from 1845 normal eyes and 1583 eyes with keratoconus (annotated based on the CASIA built-in Ectasia Screening Index;ESI). The Pentacam dataset included 114 normal eyes and 103 eyes with keratoconus (annotated based on clinical evaluations through two specialists). The accuracy and AUC of the proposed deep learning model in detecting keratoconus from anterior elevation maps tested on Pentacam dataset was 86.7% and 0.93, respectively (Fig.2). The sensitivity, specificity, and F-score of the model were 0.74, 0.98, and 0.88, respectively.

Conclusions : We developed a device-agnostic deep learning model to detect keratoconus and evaluated it based on corneal maps collected from two different cohorts (Japan and Iraq) based on two different instruments (CASIA and Pentacam). The accuracy of the proposed model in detecting keratoconus from two different instruments was reasonable. This novel approach demonstrates the generalization of the developed deep learning model. This model may augment clinical evaluations of keratoconus in day-to-day ophthalmological care.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

 

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×