Accuracy of a deep convolutional neural network in detection of retinitis pigmentosa on ultrawide-field images.

Hideharu Ohsugi,Hiroki Masumoto,Eiko Ohsugi,Naofumi Ishitobi,Hitoshi Tabuchi,Hiroki Enno,Shunsuke Nakakura,Yoshinori Mitamura

doi:10.7717/peerj.6900

Hideharu Ohsugi, Hiroki Masumoto + Show 6 more

Open Access

https://doi.org/10.7717/peerj.6900

Copy DOI

Abstract

Evaluating the discrimination ability of a deep convolution neural network for ultrawide-field pseudocolor imaging and ultrawide-field autofluorescence of retinitis pigmentosa. In total, the 373 ultrawide-field pseudocolor and ultrawide-field autofluorescence images (150, retinitis pigmentosa; 223, normal) obtained from the patients who visited the Department of Ophthalmology, Tsukazaki Hospital were used. Training with a convolutional neural network on these learning data objects was conducted. We examined the K-fold cross validation (K = 5). The mean area under the curve of the ultrawide-field pseudocolor group was 0.998 (95% confidence interval (CI) [0.9953–1.0]) and that of the ultrawide-field autofluorescence group was 1.0 (95% CI [0.9994–1.0]). The sensitivity and specificity of the ultrawide-field pseudocolor group were 99.3% (95% CI [96.3%–100.0%]) and 99.1% (95% CI [96.1%–99.7%]), and those of the ultrawide-field autofluorescence group were 100% (95% CI [97.6%–100%]) and 99.5% (95% CI [96.8%–99.9%]), respectively. Heatmaps were in accordance with the clinician’s observations. Using the proposed deep neural network model, retinitis pigmentosa can be distinguished from healthy eyes with high sensitivity and specificity on ultrawide-field pseudocolor and ultrawide-field autofluorescence images.

Highlights

Retinitis pigmentosa (RP) is a group of hereditary diseases causing progressive visual impairments because of retinal photoreceptor cell degradation (Fahim, Daiger & Weleber, 2017)
In 2017, voretigene neparvovec-rzyl demonstrated the efficacy in patients with RPE65-mediated inherited retinal dystrophy (Russell et al, 2017) and it was approved by Food and Drug Administration but it was not used worldwide
The results of the present study indicated that the proposed deep neural network (DNN) model could sufficiently distinguish RP from a normal fundus with high sensitivity and specificity (UWPC: sensitivity 1⁄4 99.3%, specificity 1⁄4 99.1%; ultrawide-field autofluorescence (UWAF): sensitivity 1⁄4 100%, specificity 1⁄4 99.5%) with both Ultrawide-field pseudocolor (UWPC) and UWAF images

Summary

Introduction

Retinitis pigmentosa (RP) is a group of hereditary diseases causing progressive visual impairments because of retinal photoreceptor cell degradation (rods and cones) (Fahim, Daiger & Weleber, 2017). Night blindness is a typical clinical feature of the early stage of RP, which is exacerbated by peripheral visual field narrowing and eventually results in loss of central vision (Fahim, Daiger & Weleber, 2017). RP occurs at a frequency of one case per 3,000–7,000 persons with no ethnic preference and is the third leading cause of vision loss in Japan, which experienced a notable increase in the number of. Accuracy of a deep convolutional neural network in detection of retinitis pigmentosa on ultrawide-field images. Induced pluripotent stem cells have been identified as therapeutic tools for treatment of RP (Yoshida et al, 2014); study results have remained inconclusive

Objectives

Methods

Conclusion