July 2019
Volume 60, Issue 9
Open Access
ARVO Annual Meeting Abstract  |   July 2019
Characteristics of Best Corrected Visual Acuity with Machine Learning: In a Large Population
Author Affiliations & Notes
  • weiting hao
    Clinical College of Ophthalmology, Tianjin Medical University, Tianjin, China
    Tianjin Eye Institute, Tianjin Eye Hospital, Tianjin, China
  • shufan ji
    School of Computer Science and Engineering, Beihang University, Beijing, China
  • chenghao zheng
    School of Computer Science and Engineering, Beihang University, Beijing, China
  • tong cui
    Clinical College of Ophthalmology, Tianjin Medical University, Tianjin, China
    Tianjin Eye Institute, Tianjin Eye Hospital, Tianjin, China
  • Yan Wang
    Clinical College of Ophthalmology, Tianjin Medical University, Tianjin, China
    Tianjin Eye Institute, Tianjin Eye Hospital, Tianjin, China
  • Footnotes
    Commercial Relationships   weiting hao, None; shufan ji, None; chenghao zheng, None; tong cui, None; Yan Wang, None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science July 2019, Vol.60, 5823. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      weiting hao, shufan ji, chenghao zheng, tong cui, Yan Wang; Characteristics of Best Corrected Visual Acuity with Machine Learning: In a Large Population. Invest. Ophthalmol. Vis. Sci. 2019;60(9):5823.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To analyze characteristics of best corrected visual acuity(BCVA) in a large population aged 15-40 with machine learning technique.

Methods : This is a retrospective study of ocular aberrations with 1457 eyes. All participants are examined with BCVA, manifest refraction, and ocular aberrations. Participants are divided into two groups according to BCVA (BCVA=1.0 and BCVA=0.8). Ocular aberrations are measured by Hartmann-Shack aberrometer, with each feature divided into low, medium and high value ranges([0,0.3)μm, [0.3,0.5)μm, and [0.5,+∞)μm). Machine learning technique is employed to identify interesting association rules measured by Confidence, which is a specific value in the machine learning domain.

Results : As for lower-order aberrations, BCVA is identified to get worse with the increase of spherical diopter( Confidence of high myopia increase from 23.29% to 38.24%) and cylindrical diopter(Confidence of high astigmatism increase from 13.37% to 35.29%). Besides, shift of astigmatism axial position is found with the decrease of BCVA( Confidence of astigmatism against the rule increase from 8.23% to 11.76%). As for higher-order aberrations, associations are investigated between the root mean square(RMS) of 3rd/ 4th order aberrations and BCVA. That is, participants with 0.8 BCVA demonstrate higher 3rd order aberrations(Confidence of medium/high value range is 61.76%) than participants with 1.0 BCVA(Confidence of medium/high value range is 50.81%). However, participants with 0.8 BCVA demonstrate lower 4th order aberrations(Confidence of medium/high value range is 14.71%) than participants with 1.0 BCVA(Confidence of medium/high value range is 19.92%).

Conclusions : We have got some interesting findings regarding associations between BCVA and ocular aberrations. Both Lower-order aberrations and higher-order aberrations are confirmed to have impacts on BCVA. Participants with higher myopia/astigmatism and astigmatism against the rule tend to demonstrate worse BCVA. However, 3rd order aberrations and 4th order aberrations may have opposite effects on BCVA. Machine learning technique may be useful in medical big data analysis.

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.

×