AUTHOR=Lin Ting-Yi , Chen Hung-Ruei , Huang Hsin-Yi , Hsiao Yu-Ier , Kao Zih-Kai , Chang Kao-Jung , Lin Tai-Chi , Yang Chang-Hao , Kao Chung-Lan , Chen Po-Yin , Huang Shih-En , Hsu Chih-Chien , Chou Yu-Bai , Jheng Ying-Chun , Chen Shih-Jen , Chiou Shih-Hwa , Hwang De-Kuang 

TITLE=Deep learning to infer visual acuity from optical coherence tomography in diabetic macular edema

JOURNAL=Frontiers in Medicine

VOLUME=Volume 9 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2022.1008950

DOI=10.3389/fmed.2022.1008950

ISSN=2296-858X

ABSTRACT=Purpose: Diabetic macular edema (DME) is one of the leading causes of visual impairment in diabetic retinopathy (DR). Best-corrected visual acuity (BCVA) from chart-based examinations may not wholly reflect DME status. Current clinical trials recommend treatment only for baseline visual acuity (VA) lower than 0.50, with those greater than 0.80 to undergo follow-up. Therefore, the ability to infer VA from objective optical coherence tomography (OCT) images will allow greater accommodation of patients to follow clinical trial established recommendations 
Methods: We enrolled a retrospective cohort of 251 DME patients from Big Data Center (BDC) of Taipei Veteran General Hospital (TVGH) from February 2011 and August 2019. A total of 3920 OCT images, labeled as “visualy impaired” or “adequate” according to baseline VA, were grouped into training (2826), validation (779), and testing cohort (315). We applied confusion matrix and receiver operating characteristic (ROC) curve to evaluate the performance.
Results: We developed an OCT-based convolutional neuronal network (CNN) model that could classify two VA classes by the threshold of 0.50 (decimal notation) with an accuracy of 75.9%, a sensitivity of 78.9%, and an area under the ROC curve of 80.1% on the testing cohort.
Conclusions: This study demonstrated the feasibility of inferring VA from routined retinal images and allow greater accommodation of real-world patient to enjoy the guidelines vigorously evaluated by clinical trials. 
Translational relevance: Serves as a pilot study to encourage further use of deep learning in deriving functional outcomes and secondary surrogate endpoints for retinal diseases.