Fast visual field measurement using the variational Bayes linear regression

2019 Fiscal Year Final Research Report

• Project/Area Number: 17K11418
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Ophthalmology
• Research Institution: The University of Tokyo
• Principal Investigator: Asaoka Ryo 東京大学, 医学部附属病院, 助教 (00362202)
• Project Period (FY): 2017-04-01 – 2020-03-31
• Keywords: 視野

Outline of Final Research Achievements

eyes of 45 patients with open angle glaucoma. Visual field measurement was carried out using the conventional SITA standard algorithm and the variational Bayes linear regression model using . As a result, the measurement duration with the latter was 5.4±1.7 minutes which was significantly shorter than that with SITA standard.(p＜0.001)

Pre-training was conducted using 10-2 visual field of 772 eyes. Then, using optical coherence tomography measurements from 86 eyes of 43 normal subjects and 505 eyes of 304 patients with open angle glaucoma, 10-2 visual field was predicted using the convolutional neural network The prediction error with root mean squared error was 6.32 ± 3.76 dB.

Free Research Field

緑内障

Academic Significance and Societal Importance of the Research Achievements

緑内障患者において、variational Bayes linear regressionを用いて視野を推測しつつ測定することで、無駄な視標提示を省き、視野を高速に測定することが可能となった。また１０－２視野を、光干渉断層計による網膜層厚から畳み込みニューラルネットワークを用いて推測することが可能となった。