June 2020
Volume 61, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2020
Machine-learning-assisted Quantitative Analysis in Optical Coherence Tomography Angiography
Author Affiliations & Notes
  • Rongrong Liu
    Biomedical Engineering, Northwestern University, Chicago, Illinois, United States
    Topcon Medical Systems, Inc, Oakland, New Jersey, United States
  • Song Mei
    Topcon Medical Systems, Inc, Oakland, New Jersey, United States
  • Zaixing Mao
    Topcon Medical Systems, Inc, Oakland, New Jersey, United States
  • Zhenguo Wang
    Topcon Medical Systems, Inc, Oakland, New Jersey, United States
  • Kinpui Chan
    Topcon Medical Systems, Inc, Oakland, New Jersey, United States
  • Footnotes
    Commercial Relationships   Rongrong Liu, Topcon Medical Systems (E); Song Mei, Topcon Medical Systems (E); Zaixing Mao, Topcon Medical Systems (E); Zhenguo Wang, Topcon Medical Systems (E); Kinpui Chan, Topcon Medical Systems (E)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2020, Vol.61, 5346. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Rongrong Liu, Song Mei, Zaixing Mao, Zhenguo Wang, Kinpui Chan; Machine-learning-assisted Quantitative Analysis in Optical Coherence Tomography Angiography. Invest. Ophthalmol. Vis. Sci. 2020;61(7):5346.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : Changes in retinal vessel size and density are known to be related to the development of retinal diseases. This work presents the metrics developed for the quantitative analysis of optical coherence tomography angiography (OCTA), including vessel segmentation, size measurement, vessel area density (VAD), and vessel skeleton density (VSD). Repeatability of the present method is tested on multiple healthy eyes.

Methods : Major vessels and smaller vessels in OCTA enface images (superficial and deep plexuses) are segmented with a K-means clustering algorithm and a self-adjusting-threshold-based algorithm, respectively. Vessel size map is reconstructed by skeletonizing the vessel segmentation and then by applying a weighted nearest neighbor smoothing. VAD and VSD enface maps are reconstructed through nearest neighbor smoothing of the vessel segmentation and the vessel skeleton, respectively. The present metrics are illustrated in Fig. 1, where the 3×3 mm2 (320×320 pixels) OCTA images were acquired with a commercial swept-source OCT system (DRI-OCT Triton; Topcon, Tokyo).

Results : The repeatability of vessel segmentation is tested, where variances in VAD and VSD are calculated from regions within the equal-area concentric rings centered at the fovea for the co-registered repeated scans of the same eye, as well as for those across different eyes. Fig. 2 illustrates (a) concentric regions around the fovea, (b) mean VAD and VSD from different regions (error bar shows standard deviation), and (c) mean standard deviation (s) of VAD and VSD from 56 eyes.

Conclusions : Quantitative metrics including vessel segmentation, size measurement, and vessel density in OCTA enface images are provided with high image resolution through machine learning techniques. The robustness of current OCTA quantitative analysis is shown with repeatability study on multiple OCTA datasets.

This is a 2020 ARVO Annual Meeting abstract.

 

Fig. 1: (a) Vessel segmentation overlaid on enface image, (b) vessel size map, (c) VAD, and (d) VSD of the same scan. Scale bar: 500 m.

Fig. 1: (a) Vessel segmentation overlaid on enface image, (b) vessel size map, (c) VAD, and (d) VSD of the same scan. Scale bar: 500 m.

 

Fig. 2: (a) Concentric regions around the fovea, (b) mean VAD and VSD from different regions (error bar shows standard deviation), and (c) mean standard deviation ( ) of VAD and VSD from 56 eyes. Scale bar: 500 m.

Fig. 2: (a) Concentric regions around the fovea, (b) mean VAD and VSD from different regions (error bar shows standard deviation), and (c) mean standard deviation ( ) of VAD and VSD from 56 eyes. Scale bar: 500 m.

×
×

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.

×