July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Automatic Deep Learning OCT Analysis Algorithm Reliably Reproduces Expert-Evaluated Outcome of a Randomized Clinical Trial for Macular Telangiectasia Type 2 Treatment
Author Affiliations & Notes
  • Jessica Loo
    Biomedical Engineering, Duke University, Durham, North Carolina, United States
  • Traci E Clemons
    Ophthalmology, Emmes, Rockville, Maryland, United States
  • Emily Y Chew
    Epidemiology and Clinical Applications, National Eye Institute, Bethesda, Maryland, United States
  • Martin Friedlander
    Molecular Medicine, The Scripps Research Institute, La Jolla, California, United States
    The Lowy Medical Research Institute, La Jolla, California, United States
  • Glenn J Jaffe
    Ophthalmology, Duke University Eye Center, Durham, North Carolina, United States
  • Sina Farsiu
    Biomedical Engineering, Duke University, Durham, North Carolina, United States
    Ophthalmology, Duke University Eye Center, Durham, North Carolina, United States
  • Footnotes
    Commercial Relationships   Jessica Loo, None; Traci Clemons, None; Emily Chew, None; Martin Friedlander, None; Glenn Jaffe, Heidelberg Engineering (C); Sina Farsiu, Google (R)
  • Footnotes
    Support  The Lowy Medical Research Institute, National Institutes of Health (R01 EY022691 and P30 EY005722), Google Faculty Research Award, and 2018 Unrestricted Grant from Research to Prevent Blindness
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 1449. doi:https://doi.org/
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Jessica Loo, Traci E Clemons, Emily Y Chew, Martin Friedlander, Glenn J Jaffe, Sina Farsiu; Automatic Deep Learning OCT Analysis Algorithm Reliably Reproduces Expert-Evaluated Outcome of a Randomized Clinical Trial for Macular Telangiectasia Type 2 Treatment. Invest. Ophthalmol. Vis. Sci. 2019;60(9):1449. doi: https://doi.org/.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To evaluate the efficacy of an automatic OCT image analysis algorithm for measuring the outcome of a clinical trial for MacTel2 treatment.

Methods : The dataset consisted of 99 eyes from 67 participants enrolled in the international, multicenter, randomized, phase 2 clinical trial of ciliary neurotrophic factor (CNTF) for the treatment of MacTel2 (NCT01949324) [1]. Each eye received either a CNTF implant or sham treatment.

The ellipsoid zone (EZ) defects area were measured using both manual (by expert Readers) and automatic (by our recently-published algorithm [2]) segmentation methods, on OCT images captured at Baseline and Month 24. The Readers and the algorithm development team were masked to treatment groups. The primary outcome measure was the change in EZ defects area from Baseline to Month 24 for each eye. A mixed effects model was used to compute the mean difference in change between the two treatment groups, and the corresponding one-sided p-values. This analysis excluded seven eyes from five participants that were later found to meet the clinical trial exclusion criteria [1].

[1] E. Y. Chew, et al. Ophthalmology, (IN PRESS), 2018. https://doi.org/10.1016/j.ophtha.2018.09.041
[2] J. Loo, et al. Biomed. Opt. Express 9(6), 2681-2698, 2018. https://doi.org/10.1364/BOE.9.002681

Results : There was excellent correlation (Pearson’s r = 0.90) between the change in EZ defects area measured by both methods across all eyes. The mean difference between the CNTF and sham treatment groups measured by manual and automatic segmentation methods were 0.065 ± 0.033 mm2 (p = 0.025) and 0.072 ± 0.035 mm2 (p = 0.021), respectively.

Conclusions : The mean difference between CNTF and sham groups measured by both methods were similar and statistically significant at the 95% confidence level, indicating that the CNTF implant slowed down the progression of EZ defects. We have demonstrated for the first time that a fully-automatic algorithm can reliably reproduce the expert-evaluated outcome in a clinical trial for an ophthalmic treatment. This automatic method has the potential to replace manual EZ area measurement, which is expensive, laborious, subjective, and prone to human error.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

×
×

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.

×