June 2017
Volume 58, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2017
Using Kalman Filtering to Personalize the Monitoring of Persons with Normal Tension Glaucoma
Author Affiliations & Notes
  • Mariel S Lavieri
    University of Michigan, Ann Arbor, Michigan, United States
  • Xiang Liu
    University of Michigan, Ann Arbor, Michigan, United States
  • Zhining Zhou
    University of Michigan, Ann Arbor, Michigan, United States
  • Jingyuan Wang
    University of Michigan, Ann Arbor, Michigan, United States
  • Kazuhisa Sugiyama
    Department of Ophthalmology, Kanazawa University Graduate School of Medical Science, Kanazawa, Japan
  • Koji Nitta
    Department of Ophthalmology , Fukui-ken Saiseikai Hospital, Fukui, Japan
  • Chris Andrews
    University of Michigan, Ann Arbor, Michigan, United States
  • Mark Van Oyen
    University of Michigan, Ann Arbor, Michigan, United States
  • Joshua D Stein
    University of Michigan, Ann Arbor, Michigan, United States
  • Footnotes
    Commercial Relationships   Mariel Lavieri, 13/668,280 (P); Xiang Liu, None; Zhining Zhou, None; Jingyuan Wang, None; Kazuhisa Sugiyama, None; Koji Nitta, None; Chris Andrews, None; Mark Van Oyen, 13/668,280 (P); Joshua Stein, 13/668,280 (P)
  • Footnotes
    Support  R01EY026641
Investigative Ophthalmology & Visual Science June 2017, Vol.58, 2870. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Mariel S Lavieri, Xiang Liu, Zhining Zhou, Jingyuan Wang, Kazuhisa Sugiyama, Koji Nitta, Chris Andrews, Mark Van Oyen, Joshua D Stein; Using Kalman Filtering to Personalize the Monitoring of Persons with Normal Tension Glaucoma. Invest. Ophthalmol. Vis. Sci. 2017;58(8):2870.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To evaluate whether the dynamic glaucoma forecasting and monitoring algorithms we previously calibrated and validated on patients with moderate to severe primary open angle glaucoma (POAG) can generate personalized forecasts of disease progression in a cohort of patients with normal tension glaucoma (NTG).

Methods : A Kalman filter was used to personalize the monitoring of patients with NTG to assess for disease progression. As additional VF and IOP tests are performed on each patient, the Kalman filter updates our knowledge about each patient’s unique disease trajectory to determine personalized schedules of the timing of the next set of VF and IOP tests for each patient to optimize progression detection. Parameterization of the models was performed using longitudinal data from 571 patients with POAG who were enrolled in the CIGTS and AGIS trials. The models were then applied (using bootstrapping with replacement with 1000 replications) to a cohort of Japanese patients with NTG who had ≥ 5 years of longitudinal follow-up. Our algorithm was compared against 1-, 1.5-, and 2-year fixed interval schedules for obtaining VFs and IOP measurements to determine which testing schedule was more efficient (i.e. performing IOP and VF tests at times when progression was observed) and had reduced diagnostic delay (i.e. length of time between when progression was first observed in the historical data and when progression was first detected) for identifying progression.

Results : Among the 168 eyes with NTG who were followed over an average of 6.0 ± 0.5 years, 63 (37.5%) experienced progression over a mean of 2.2 ± 1.4 years. Compared to 1 year fixed intervals of performing VFs and IOPs to check for progression of NTG, our algorithm using Kalman filtering achieved a 41.2% [41.1-41.3%] increased efficiency in detecting NTG progression and detected NTG progression 8% [6-9%] sooner (i.e., reduced diagnostic delay by 8%) using the same number of tests per patient.

Conclusions : Dynamic and personalized schedules for obtaining IOP and VF tests using a Kalman filter approach can improve the efficiency of detecting NTG progression and identify NTG progression a little sooner compared with a 1 year fixed interval schedule. Predictions are very likely to be enhanced by parameterizing the models with data on patients with NTG rather than those with POAG as NTG may behave differently than POAG.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.

×
×

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.

×