Deep learning models for screening of high myopia using optical coherence tomography

Abstract

This study aimed to validate and evaluate deep learning (DL) models for screening of high myopia using spectral-domain optical coherence tomography (OCT). This retrospective cross-sectional study included 690 eyes in 492 patients with OCT images and axial length measurement. Eyes were divided into three groups based on axial length: a “normal group,” a “high myopia group,” and an “other retinal disease” group. The researchers trained and validated three DL models to classify the three groups based on horizontal and vertical OCT images of the 600 eyes. For evaluation, OCT images of 90 eyes were used. Diagnostic agreements of human doctors and DL models were analyzed. The area under the receiver operating characteristic curve of the three DL models was evaluated. Absolute agreement of retina specialists was 99.11% (range: 97.78–100%). Absolute agreement of the DL models with multiple-column model was 100.0% (ResNet 50), 90.0% (Inception V3), and 72.22% (VGG 16). Areas under the receiver operating characteristic curves of the DL models with multiple-column model were 0.99 (ResNet 50), 0.97 (Inception V3), and 0.86 (VGG 16). The DL model based on ResNet 50 showed comparable diagnostic performance with retinal specialists. The DL model using OCT images demonstrated reliable diagnostic performance to identify high myopia.

Predicting Axial Length From Choroidal Thickness on Optical Coherence Tomography Images With Machine Learning Based Algorithms

“…Machine learning and deep learning have been successfully applied in OCT images for biomarker identification in AMD (24). Since myopia is a rising problem in ophthalmology, OCT images have been used for AI prediction in myopic eyes recently (25)(26)(27).…”

Accuracy and feasibility with AI-assisted OCT in retinal disorder community screening

“…Previous studies also involved auto-detection of myopic macular diseases using AI algorithm due to the rising prevalence of myopia and multiple vision-threatening retinal damages ( Sogawa et al, 2020 ; Choi et al, 2021 ; Ye et al, 2021 ; Li et al, 2022a ). Ye et al (2021 ) engineered the deep-learning (DL) model to identify myopic maculopathy, including macular choroidal thinning, macular Bruch membrane (BM) defects, sub-retinal hyper-reflective material (SHRM), myopic traction maculopathy (MTM), and dome-shaped macula (DSM), and the result showed that the AUC was 0.927–0.974 for five myopic maculopathies.…”

“… Ye et al (2021 ) engineered the deep-learning (DL) model to identify myopic maculopathy, including macular choroidal thinning, macular Bruch membrane (BM) defects, sub-retinal hyper-reflective material (SHRM), myopic traction maculopathy (MTM), and dome-shaped macula (DSM), and the result showed that the AUC was 0.927–0.974 for five myopic maculopathies. Choi et al (2021 ) trained and validated three DL models to identify myopia and generated a result of the absolute agreement with retina specialists which was 99.11%. However, the specific lesion associated with myopia could not be detected, and validation with an external dataset was needed.…”

Automatic Prediction of Conductive Hearing Loss Using Video Pneumatic Otoscopy and Deep Learning Algorithm

“…It was natural to come up with a deep learning model architecture that takes both POS and NEG images within a cycle simultaneously as inputs because clinicians can obtain information from dynamic movements of TM between those images (Cho et al 2009; Lee et al 2011). As a method of feeding multiple images in parallel into a deep learning model, a multi-column CNN has been suggested and showed its effectiveness (Ciresan et al 2012; Zhang et al 2016; Jin et al 2019; Choi et al 2021). Therefore, we devised multi-column CNN models to train both POS and NEG images simultaneously, where the features from dynamic input images of each column are abstracted independently and combined by various arithmetic operations (concatenate, summation, or subtraction) at the near-endpoint.…”