npj Digital Medicine (Aug 2024)

A deep learning system for myopia onset prediction and intervention effectiveness evaluation in children

  • Ziyi Qi,
  • Tingyao Li,
  • Jun Chen,
  • Jason C. Yam,
  • Yang Wen,
  • Gengyou Huang,
  • Hua Zhong,
  • Mingguang He,
  • Dan Zhu,
  • Rongping Dai,
  • Bo Qian,
  • Jingjing Wang,
  • Chaoxu Qian,
  • Wei Wang,
  • Yanfei Zheng,
  • Jian Zhang,
  • Xianglong Yi,
  • Zheyuan Wang,
  • Bo Zhang,
  • Chunyu Liu,
  • Tianyu Cheng,
  • Xiaokang Yang,
  • Jun Li,
  • Yan-Ting Pan,
  • Xiaohu Ding,
  • Ruilin Xiong,
  • Yan Wang,
  • Yan Zhou,
  • Dagan Feng,
  • Sichen Liu,
  • Linlin Du,
  • Jinliuxing Yang,
  • Zhuoting Zhu,
  • Lei Bi,
  • Jinman Kim,
  • Fangyao Tang,
  • Yuzhou Zhang,
  • Xiujuan Zhang,
  • Haidong Zou,
  • Marcus Ang,
  • Clement C. Tham,
  • Carol Y. Cheung,
  • Chi Pui Pang,
  • Bin Sheng,
  • Xiangui He,
  • Xun Xu

DOI
https://doi.org/10.1038/s41746-024-01204-7
Journal volume & issue
Vol. 7, no. 1
pp. 1 – 10

Abstract

Read online

Abstract The increasing prevalence of myopia worldwide presents a significant public health challenge. A key strategy to combat myopia is with early detection and prediction in children as such examination allows for effective intervention using readily accessible imaging technique. To this end, we introduced DeepMyopia, an artificial intelligence (AI)-enabled decision support system to detect and predict myopia onset and facilitate targeted interventions for children at risk using routine retinal fundus images. Based on deep learning architecture, DeepMyopia had been trained and internally validated on a large cohort of retinal fundus images (n = 1,638,315) and then externally tested on datasets from seven sites in China (n = 22,060). Our results demonstrated robustness of DeepMyopia, with AUCs of 0.908, 0.813, and 0.810 for 1-, 2-, and 3-year myopia onset prediction with the internal test set, and AUCs of 0.796, 0.808, and 0.767 with the external test set. DeepMyopia also effectively stratified children into low- and high-risk groups (p < 0.001) in both test sets. In an emulated randomized controlled trial (eRCT) on the Shanghai outdoor cohort (n = 3303) where DeepMyopia showed effectiveness in myopia prevention compared to NonCyc-based model, with an adjusted relative reduction (ARR) of −17.8%, 95% CI: −29.4%, −6.4%. DeepMyopia-assisted interventions attained quality-adjusted life years (QALYs) of 0.75 (95% CI: 0.53, 1.04) per person and avoided blindness years of 13.54 (95% CI: 9.57, 18.83) per 1 million persons compared to natural lifestyle with no active intervention. Our findings demonstrated DeepMyopia as a reliable and efficient AI-based decision support system for intervention guidance for children.