June 2021
Volume 62, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2021
Innovative artificial intelligence-based cataract diagnostic method uses a slit-lamp video recording device and multiple machine-learning
Author Affiliations & Notes
  • Eisuke Shimizu
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • HIROYUKI YAZU
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Naohiko Aketa
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Makoto Tanji
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Akito Sakasegawa
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Shintaro Nakayama
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Toshiki Ishikawa
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
    OUI Inc., Tokyo, Japan
  • Ryota Yokoiwa
    OUI Inc., Tokyo, Japan
  • Shinri Sato
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Taiichiro Katayama
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Akiko Hanyuda
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Yasunori Sato
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Kazumi Fukagawa
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Hiroshi Fujishima
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Yoko Ogawa
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Kazuo Tsubota
    Keio Gijuku Daigaku, Minato-ku, Tokyo, Japan
  • Footnotes
    Commercial Relationships   Eisuke Shimizu, OUI Inc. (P); HIROYUKI YAZU, OUI Inc. (P); Naohiko Aketa, OUI Inc. (P); Makoto Tanji, None; Akito Sakasegawa, None; Shintaro Nakayama, None; Toshiki Ishikawa, None; Ryota Yokoiwa, None; Shinri Sato, None; Taiichiro Katayama, None; Akiko Hanyuda, None; Yasunori Sato, None; Kazumi Fukagawa, None; Hiroshi Fujishima, None; Yoko Ogawa, None; Kazuo Tsubota, None
  • Footnotes
    Support  This work was supported by the Japan Agency for Medical Research and Development (20he1022003h0001 and 20hk0302008h0001), Uehara Memorial Foundation, Hitachi Global Foundation, Kondo Memorial Foundation, Eustylelab, and Kowa Life Science Foundation. There are no other funding statements to declare including the company and patent. Role of the Funders/Sponsors: The study sponsors had no role in the study design, collection, analysis, and interpretation of data; writing of the report, and decision to submit the paper for publication.
Investigative Ophthalmology & Visual Science June 2021, Vol.62, 1031. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Eisuke Shimizu, HIROYUKI YAZU, Naohiko Aketa, Makoto Tanji, Akito Sakasegawa, Shintaro Nakayama, Toshiki Ishikawa, Ryota Yokoiwa, Shinri Sato, Taiichiro Katayama, Akiko Hanyuda, Yasunori Sato, Kazumi Fukagawa, Hiroshi Fujishima, Yoko Ogawa, Kazuo Tsubota; Innovative artificial intelligence-based cataract diagnostic method uses a slit-lamp video recording device and multiple machine-learning. Invest. Ophthalmol. Vis. Sci. 2021;62(8):1031.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : The use of artificial intelligence (AI) remains limited in cataract imaging diagnosis because there is a lack of numerous standardized images and analysis models. An AI model for cataract diagnosis could potentially be created by slit-light images and machine-learning. We aimed to determine whether our machine-learning model based on images recorded with a slit-lamp device would have the comparable diagnostic performance to that of ophthalmologists.

Methods : A dataset comprising 18,596 frames collected retrospectively were used for training and cross-validation of a machine-learning algorithm. Cataract diagnosis, cataract severity grading, and surgical indication prediction between our model and evaluations performed by ophthalmologists were assessed the use of a slit-lamp device to record cataract video and machine-learning system. A sensitivity, specificity, positive predictive value, and negative predictive value for cataract diagnosis and surgical indication. The area under the receiver operating characteristic curve for each cataract grading. Weighted kappa statistics for cataract video analysis and cross-validation.

Results : Our model could diagnose cataract with sensitivity, 99.60% (95% confidence interval [CI], 99.40–99.70), and specificity, 96.00% (95% CI, 83.40–99.30), compared to the diagnostic performance of ophthalmologists. The results of cataract severity grading were nuclear cataract (NUC) 0: Area under curve (AUC), 0.987 (95% CI, 0.947–1.000); NUC1: AUC, 0.916 (95% CI, 0.888–0.945); NUC2: AUC, 0.862 (95% CI, 0.844–0.879); and NUC3: AUC, 0.943 (95% CI, 0.931–0.955). For overall cataract grading, the accuracy was 87.80% (kappa, 0.811 [95% CI, 0.791–0.831]). The surgical indication prediction with a sensitivity of 91.80% (95% CI, 82.00–95.10) and specificity of 92.30% (95% CI, 86.10–94.40). The cross-validation accuracy was 86.0% (kappa, 0.800 [95% CI, 0.780–0.820]).

Conclusions : We successfully created a high-performance cataract diagnostic AI using machine-learning from images recorded with a portable slit-lamp device, which can simplify the cataract diagnostic process and be of particular use in settings where access to ophthalmologic services is not available.

This is a 2021 ARVO Annual Meeting abstract.

 

×
×

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.

×