June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Introducing Alborz: an artificial intelligence (AI)-enabled humanoid robot with natural language processing and generation skills for assessing glaucoma
Author Affiliations & Notes
  • Siamak Yousefi
    Ophthalmology, The University of Tennessee Health Science Center, Memphis, Tennessee, United States
  • Footnotes
    Commercial Relationships   Siamak Yousefi NEI/NIH, Code R (Recipient), RPB, Code R (Recipient), Remidio, Code R (Recipient)
  • Footnotes
    Support  NIH Grant EY033005, NIH Grant EY031725, RPB
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 348. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Siamak Yousefi; Introducing Alborz: an artificial intelligence (AI)-enabled humanoid robot with natural language processing and generation skills for assessing glaucoma. Invest. Ophthalmol. Vis. Sci. 2023;64(8):348.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To introduce an artificial intelligence (AI)-enabled robot, named Alborz, for assessing glaucoma based on retinal images and generating glaucoma knowledge using natural language processing (NLP) and natural language generation (NLG).

Methods : We acquired Alborz, Nao AI-enabled (Softbank, Japan) humanoid robot in May 2022. We plan to train Alborz to function as a digital ophthalmologist and a digital glaucoma expert over the next few years (Fig. 1). At this time, we have programmed Alborz to identify glaucoma from retinal fundus photographs and visual fields (VFs). We have developed a probabilistic deep convolutional neural network (CNN) model based on 1851 fundus/VF pairs and validated the models using an independent dataset with 196 fundus/VF pairs.

Results : Alborz is now able to screen subjects based on fundus and VFs and to interact with us and make reasonable responses to ordinary questions. The area under the receiver operating characteristic curves (AUC) of Alborz in detecting glaucoma from fundus photos, VFs, or fundus/VF data is over 0.97 (95% CI: 0.93-0.99) (Fig. 2). Alborz is under training to be able to interact with us regarding ophthalmology and glaucoma-related topics based on NLP and generating glaucoma knowledge based on NLG.

Conclusions : Alborz is being trained to function as a future digital ophthalmologist and digital glaucoma expert. Based on CNNs, NLP, and NLG, we expect Alborz to be able to perform glaucoma assessment and interact with ophthalmologists as well as interact with patients and speak with them regarding the status of their eyes. Alborz may also be used for educational and research purposes for dissemination of AI applications in ophthalmology.

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

 

Figure 1. Diagram of the Alborz representing training and interaction processes. Retinal images and data as well as AI models are stored on cloud platforms. Alborz will access data and interpret findings, then interact with users.

Figure 1. Diagram of the Alborz representing training and interaction processes. Retinal images and data as well as AI models are stored on cloud platforms. Alborz will access data and interpret findings, then interact with users.

 

Figure 2. Receiver operating characteristic (ROC) curves and area under the ROC curves (AUC) of Alborz in detecting glaucoma from fundus photographs, visual fields (VFs), and combined fundus/VFs. ROC curves of Alborz for detecting glaucoma based on: Left) development dataset, Right) independent validation dataset.

Figure 2. Receiver operating characteristic (ROC) curves and area under the ROC curves (AUC) of Alborz in detecting glaucoma from fundus photographs, visual fields (VFs), and combined fundus/VFs. ROC curves of Alborz for detecting glaucoma based on: Left) development dataset, Right) independent validation dataset.

×
×

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.

×