June 2021
Volume 62, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2021
Automated analysis of large vessel blood flow using AOSLO
Author Affiliations & Notes
  • Ines Zamoun
    Department of Ophtalmology, University College London, London, London, United Kingdom
  • Sevim Ongun
    Department of Ophtalmology, University College London, London, London, United Kingdom
  • Sarah Houston
    Department of Ophtalmology, University College London, London, London, United Kingdom
  • John Greenwood
    Department of Ophtalmology, University College London, London, London, United Kingdom
  • Adam Dubis
    Department of Ophtalmology, University College London, London, London, United Kingdom
    Moorfields Eye Hospital NHS Foundation Trust, London, London, United Kingdom
  • Footnotes
    Commercial Relationships   Ines Zamoun, None; Sevim Ongun, None; Sarah Houston, None; John Greenwood, None; Adam Dubis, Boston Micromachines Corp (C)
  • Footnotes
    Support  Moorfields Eye Charity; Fight for Sight - 24NE172; Multiple Sclerosis Society
Investigative Ophthalmology & Visual Science June 2021, Vol.62, 1816. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ines Zamoun, Sevim Ongun, Sarah Houston, John Greenwood, Adam Dubis; Automated analysis of large vessel blood flow using AOSLO. Invest. Ophthalmol. Vis. Sci. 2021;62(8):1816.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Several studies have shown ocular blood flow anomalies in vascular disorders (diabetes, multiple sclerosis, and dementia) disorders using adaptive optics scanning ophthalmoscopy (AOSLO). One AOSLO method for studying vascular change been XT, where spatial data is acquired for part of the scan, then freezing the fast scanner to measure cells coming past the scanner (Figure 1). The challenge is that while data is becoming easier to acquire, analysis routines that stabilize eye motion, extract vessel features, and flow data are still manual and time-consuming.

Methods : Blood flow video sequences were acquired using a bespoke multimodal AOSLO (Boston Micromachines Corp, Boston, MA). The AOSLO simultaneously acquires up to four video sequences at 26Hz. Image motion stabilization and removal of bad frames is the first step, accomplished through a novel template matching algorithm. Next, line and edge detectors were employed to detect the vessel to scan direction angle. Cell speed was extracted by line detection techniques that detected the slope of the cell tracks in the Time portion (Figure 1). Lastly, to assess the validity of our extraction automated measures were compared to frames quantified manually by two graders.

Results : The comparison of manual and automated methods was made over 100 images. For this image set, 98% were manually measurable compared to 80% using automated methods. For this shared dataset, the most experienced grader had a slight bias (-2.61; limits: –12.5 to 7.3 mm/s, with greater variability at higher velocities) between trials, with poorer repeatability to the less experienced grader (bias: - 13.07; limits –33.8 to 7.7 mm/s). In comparison, automated methods showed results with 70% accuracy (bias: 2.42; limits -15 to 13 mm/s). The greatest benefit was in time-saving where manual processing took 43 ± 5 min to complete, automated methods took 2.83 sec to extract up to 96 cell profiles per frame.

Conclusions : The method presented here simplifies the processing pipeline from three steps to a single workflow. In addition, there is a drastic reduction in time consumption for data extraction, while providing comparable results. These processing improvements will allow this technology to proliferate and potentially. We aim at developing tools to quantitatively model the changes in blood velocity profiles in the human eye, non-invasively.

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.

×