June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Internal and external validation of a deep learning model for estimating non-perfusion areas from ultra-widefield fundus images
Hidenori Takahashi; Satoru Inoda; Yoichiro Hisadome; Yusuke Kondo; Hironobu Tampo; Takuya Takayama; Kosuke Nagaoka; Shinichi Sakamoto; Yusaku Katada; Toshihide Kurihara; Hidetoshi Kawashima; Yasuo Yanagi
Author Affiliations & Notes
  • Hidenori Takahashi
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
    DeepEyeVision Inc., Japan
  • Satoru Inoda
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Yoichiro Hisadome
    DeepEyeVision Inc., Japan
  • Yusuke Kondo
    DeepEyeVision Inc., Japan
  • Hironobu Tampo
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Takuya Takayama
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Kosuke Nagaoka
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Shinichi Sakamoto
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Yusaku Katada
    Laboratory of Photobiology, Keio Gijuku Daigaku Igakubu Daigakuin Igaku Kenkyuka, Shinjuku-ku, Tokyo, Japan
    Department of Ophthalmology, Keio Gijuku Daigaku Igakubu Daigakuin Igaku Kenkyuka, Shinjuku-ku, Tokyo, Japan
  • Toshihide Kurihara
    Laboratory of Photobiology, Keio Gijuku Daigaku Igakubu Daigakuin Igaku Kenkyuka, Shinjuku-ku, Tokyo, Japan
    Department of Ophthalmology, Keio Gijuku Daigaku Igakubu Daigakuin Igaku Kenkyuka, Shinjuku-ku, Tokyo, Japan
  • Hidetoshi Kawashima
    Jichi Ika Daigaku, Shimotsuke, Tochigi, Japan
  • Yasuo Yanagi
    DeepEyeVision Inc., Japan
    Yokohama Shiritsu Daigaku Igakubu Daigakuin Igaku Kenkyuka, Yokohama, Kanagawa, Japan
  • Footnotes
    Commercial Relationships   Hidenori Takahashi Boehringer, Code C (Consultant/Contractor), Kyowa Kirin, Senju, Bayer, Novartis, Code F (Financial Support), DeepEyeVision, Code I (Personal Financial Interest), DeepEyeVision, Code O (Owner), DeepEyeVision, Code P (Patent), Santen, Alcon, Kowa, Viatoris, Senju, Novartis, Bayer, Boehringer, MSD, Chugai, Code R (Recipient); Satoru Inoda Novartis, Code F (Financial Support), Kowa, Chugai, Code R (Recipient); Yoichiro Hisadome DeepEyeVision, Code E (Employment), DeepEyeVision, Code P (Patent); Yusuke Kondo DeepEyeVision, Code P (Patent), DeepEyeVision, Code S (non-remunerative); Hironobu Tampo None; Takuya Takayama None; Kosuke Nagaoka Novartis, Code R (Recipient); Shinichi Sakamoto None; Yusaku Katada Restore Vision, Code I (Personal Financial Interest), Nikon, Code P (Patent), Restore Vision, Code R (Recipient); Toshihide Kurihara Rohto, SEED, Kowa, Tsubota Labo, Code F (Financial Support), Tsubota Labo, Restore Vision, Code I (Personal Financial Interest), Nikon, Code P (Patent); Hidetoshi Kawashima HOYA, Heiwa Iyo, Code C (Consultant/Contractor), Novartis, Linical, DeepEyeVision, HOYA, Santen, Senju, Bayer, Code F (Financial Support), Senju, Mitsubishi Tanabe, Kowa, Alcon, Santen, Novartis, Carl Zeiss, Chugai, Code R (Recipient); Yasuo Yanagi Roche, Boehringer, Code C (Consultant/Contractor), Alcon, Sanbio, Code F (Financial Support), Chugai, Novartis, Bayer, Santen, Boehringer, Senju, Roche, Code R (Recipient), DeepEyeVision, Code S (non-remunerative)
  • Footnotes
    Support  DeepEyeVision Inc.
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 229. doi:
  • Views
  • Share
  • Tools
    • Alerts
      User Alerts
      You are adding an alert for:
      Internal and external validation of a deep learning model for estimating non-perfusion areas from ultra-widefield fundus images
      You will receive an email whenever this article is corrected, updated, or cited in the literature. You can manage this and all other alerts in My Account
      The alert will be sent to:
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation
      Citation

      Hidenori Takahashi, Satoru Inoda, Yoichiro Hisadome, Yusuke Kondo, Hironobu Tampo, Takuya Takayama, Kosuke Nagaoka, Shinichi Sakamoto, Yusaku Katada, Toshihide Kurihara, Hidetoshi Kawashima, Yasuo Yanagi; Internal and external validation of a deep learning model for estimating non-perfusion areas from ultra-widefield fundus images. Invest. Ophthalmol. Vis. Sci. 2023;64(8):229.

      Download citation file:

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

      ×
    • Get Permissions
  • Supplements
Abstract

Purpose : To verify whether an artificial intelligence (AI) that estimates non-perfusion areas (NPA) from ultra-widefield color fundus images (UWF) without angiography remains accurate in cross-validation using training data from our institution and a different institution. The AI was constructed using UWFs (17,600 for ischemic retinal disease and 16,000 for non-ischemic retinal disease) and NPA obtained by annotating UWF fluorescent angiography (UWF-FA) images acquired using an Optos California system (Nikon, Tokyo) at Jichi Medical University Hospital.

Methods : For internal validation, all 650 UWF images with NPA on UWF-FA were used in a 5-fold cross-validation. For external validation, 89 UWFs from 44 eyes (23 patients) who underwent UWF-FA using an Optos California system at another institution were used. For validation, a Bland-Altman plot was used to evaluate the fixed and proportional errors and 95% limits of agreement (LOA) between the ground truth (GT) and estimated NPA sizes.

Results : In the internal and external Bland-Altman plot, the fixed error, assessed as the mean difference in area between the estimated NPA and GT (42, 136 mm2, P = 0.0004, < 0.0001), was smaller than the LOA (350, 654 mm2), and outlier UWF images were found in 6.5, 6.5% of the cases, respectively.

Conclusions : The estimated NPA was interpreted as being consistent with the GT in the internal cross-validation and external validation.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

A Bland-Altman plot for all internal data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.
View OriginalDownload Slide

A Bland-Altman plot for all internal data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.

A Bland-Altman plot for all internal data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.

A Bland-Altman plot for all internal data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.
View OriginalDownload Slide
 
A Bland-Altman plot for external data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.
View OriginalDownload Slide

A Bland-Altman plot for external data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.

A Bland-Altman plot for external data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.

A Bland-Altman plot for external data. Non-perfusion areas (NPA) were confirmed by fluorescein angiography in all eyes with ischemic disease, and there was smaller difference between the estimated NPA and ground truth compared to limitation of agreement.
View OriginalDownload Slide

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Attribution-NonCommercial-NoDerivatives
Copyright © 2024 Association for Research in Vision and Ophthalmology.
×
×

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.

×
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy.Accept