June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Automatic Classification of Slit Lamp Photographs by Illumination Technique
Author Affiliations & Notes
  • Ming-Chen Lu
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Callie Deng
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Miles Greenwald
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Sina Farsiu
    Department of Ophthalmology, Duke University Medical Center, Durham, North Carolina, United States
  • Venkatesh Prajna
    Aravind Eye Care System, Madurai, Tamil Nadu, India
  • Nambi Nallasamy
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
    Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, Michigan, United States
  • Mercy Pawar
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Jenna N. Hart
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Sumithra S.R.
    Aravind Eye Care System, Madurai, Tamil Nadu, India
  • Prabhleen Kochar
    Aravind Eye Care System, Madurai, Tamil Nadu, India
  • Suvitha Selvaraj
    Aravind Eye Care System, Madurai, Tamil Nadu, India
  • Harry Levine
    Department of Ophthalmology, University of Miami Health System Bascom Palmer Eye Institute, Miami, Florida, United States
  • Guillermo Amescua
    Department of Ophthalmology, University of Miami Health System Bascom Palmer Eye Institute, Miami, Florida, United States
  • Paula A. Sepulveda-Beltran
    Department of Ophthalmology, University of Miami Health System Bascom Palmer Eye Institute, Miami, Florida, United States
  • Leslie M. Niziol
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Maria A. Woodward
    Department of Ophthalmology and Visual Sciences, University of Michigan Michigan Medicine, Ann Arbor, Michigan, United States
  • Footnotes
    Commercial Relationships   Ming-Chen Lu None; Callie Deng None; Miles Greenwald None; Sina Farsiu None; Venkatesh Prajna None; Nambi Nallasamy None; Mercy Pawar None; Jenna Hart None; Sumithra S.R. None; Prabhleen Kochar None; Suvitha Selvaraj None; Harry Levine None; Guillermo Amescua None; Paula A. Sepulveda-Beltran None; Leslie Niziol None; Maria Woodward None
  • Footnotes
    Support  NEI R01EY031033 (M.A.W.); Research to Prevent Blindness Career Advancement Award (M.A.W.)
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 4019. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ming-Chen Lu, Callie Deng, Miles Greenwald, Sina Farsiu, Venkatesh Prajna, Nambi Nallasamy, Mercy Pawar, Jenna N. Hart, Sumithra S.R., Prabhleen Kochar, Suvitha Selvaraj, Harry Levine, Guillermo Amescua, Paula A. Sepulveda-Beltran, Leslie M. Niziol, Maria A. Woodward; Automatic Classification of Slit Lamp Photographs by Illumination Technique. Invest. Ophthalmol. Vis. Sci. 2023;64(8):4019.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Deep learning approaches for diagnosing microbial keratitis (MK) type require organization of large collections of medical images, a time-consuming and often manual process. To address this bottleneck, we developed an automated algorithm that classifies corneal slit lamp photographs (SLPs) based on illumination technique.

Methods : Within the Automated Quantitative Ulcer Analysis (AQUA) study, SLPs were collected from patients with MK using Haag-Streit biomicroscopes with Canon 7D cameras at the Kellogg Eye Center, Bascom Palmer Eye Institute, and Aravind Eye Care Systems. Illumination techniques included slit beam, diffuse white light, diffuse blue light with fluorescein, and sclerotic scatter. Images were manually labeled for illumination and randomly split into training, validation, and testing datasets (70%:15%:15%). Classification algorithms including k-nearest neighborhood, multilayer perceptron, MobilNetV2, ResNet50, LeNet, and AlexNet were trained to distinguish the four illumination techniques. Algorithm performances on the test dataset were evaluated with 95% confidence intervals (CI) for accuracy, F1 score, precision, recall, and area under the curve (AUC), overall and by illumination technique (one-versus-rest).

Results : A total of 12,132 SLPs from 409 patients were analyzed, including 41.8% (n=5069) slit beam photographs, 21.2% (2571) diffuse white light, 19.5% (2364) diffuse blue light, and 17.5% (2128) sclerotic scatter. Figure 1 displays representative SLPs for each illumination technique. MobilNetV2 achieved the highest overall F1 score of 97.95% (CI, 97.94-97.97%), AUC of 99.83% (99.72-99.9%), and accuracy of 98.98% (98.97-98.98%). The F1 score for slit beam photographs, diffuse white light, diffuse blue light, and sclerotic scatter were 97.82% (97.80-97.84%), 96.62% (96.58-96.66%), 99.88% (99.87-99.89%), 97.59% (97.55-97.62%), respectively (Figure 2). Slit beam and sclerotic scatter were the two most frequently misclassified illumination.

Conclusions : MobilNetV2 accurately classified SLPs by illumination technique using a large dataset of MK images. Effective, automatic classification of SLPs is key to integrating deep learning systems for clinical decision support into practice workflows.

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

 

Slit lamp photograph illumination type, including: A. Diffuse blue light. B. Slit beam. C. Sclerotic scatter. D. Diffuse white light.

Slit lamp photograph illumination type, including: A. Diffuse blue light. B. Slit beam. C. Sclerotic scatter. D. Diffuse white light.

 

Class-wise performance metrics on the test dataset for MobilNetV2.

Class-wise performance metrics on the test dataset for MobilNetV2.

×
×

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.

×