Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Diabetic Macular Edema Treatment Responses Using OCT: Dataset and Methods of APTOS Competition

Weiyi Zhang, Peranut Chotcomwongse, Yinwen Li, Pusheng Xu, Ruijie Yao, Lianhao Zhou, Yuxuan Zhou, Hui Feng, Qiping Zhou, Xinyue Wang, Shoujin Huang, Zihao Jin, Florence H. T. Chung, Shujun Wang, Yalin Zheng, Mingguang He, Danli Shi, Paisan Ruamviboonsuk

TL;DR

This work introduces OCT4DME, a large public OCT dataset paired with the APTOS 2021 Big Data Competition to predict anti-VEGF treatment responses in DME. It details dataset collection, a two-round competition, and multi-task evaluation (IRF/SRF/PED/HRF presence, CST, VA, and CI) using OCT images. Top solutions leverage MIL, weak supervision, and ensemble strategies to achieve high AUC on PED/SRF/IRF detections and meaningful CST/VA predictions, demonstrating AI’s potential to support personalized DME management. The study also discusses data distribution shifts, preprocessing tricks (e.g., gamma transforms), and clinical considerations for deploying AI in ophthalmology, underscoring both the promise and the challenges of clinical translation.

Abstract

Diabetic macular edema (DME) significantly contributes to visual impairment in diabetic patients. Treatment responses to intravitreal therapies vary, highlighting the need for patient stratification to predict therapeutic benefits and enable personalized strategies. To our knowledge, this study is the first to explore pre-treatment stratification for predicting DME treatment responses. To advance this research, we organized the 2nd Asia-Pacific Tele-Ophthalmology Society (APTOS) Big Data Competition in 2021. The competition focused on improving predictive accuracy for anti-VEGF therapy responses using ophthalmic OCT images. We provided a dataset containing tens of thousands of OCT images from 2,000 patients with labels across four sub-tasks. This paper details the competition's structure, dataset, leading methods, and evaluation metrics. The competition attracted strong scientific community participation, with 170 teams initially registering and 41 reaching the final round. The top-performing team achieved an AUC of 80.06%, highlighting the potential of AI in personalized DME treatment and clinical decision-making.

Predicting Diabetic Macular Edema Treatment Responses Using OCT: Dataset and Methods of APTOS Competition

TL;DR

This work introduces OCT4DME, a large public OCT dataset paired with the APTOS 2021 Big Data Competition to predict anti-VEGF treatment responses in DME. It details dataset collection, a two-round competition, and multi-task evaluation (IRF/SRF/PED/HRF presence, CST, VA, and CI) using OCT images. Top solutions leverage MIL, weak supervision, and ensemble strategies to achieve high AUC on PED/SRF/IRF detections and meaningful CST/VA predictions, demonstrating AI’s potential to support personalized DME management. The study also discusses data distribution shifts, preprocessing tricks (e.g., gamma transforms), and clinical considerations for deploying AI in ophthalmology, underscoring both the promise and the challenges of clinical translation.

Abstract

Diabetic macular edema (DME) significantly contributes to visual impairment in diabetic patients. Treatment responses to intravitreal therapies vary, highlighting the need for patient stratification to predict therapeutic benefits and enable personalized strategies. To our knowledge, this study is the first to explore pre-treatment stratification for predicting DME treatment responses. To advance this research, we organized the 2nd Asia-Pacific Tele-Ophthalmology Society (APTOS) Big Data Competition in 2021. The competition focused on improving predictive accuracy for anti-VEGF therapy responses using ophthalmic OCT images. We provided a dataset containing tens of thousands of OCT images from 2,000 patients with labels across four sub-tasks. This paper details the competition's structure, dataset, leading methods, and evaluation metrics. The competition attracted strong scientific community participation, with 170 teams initially registering and 41 reaching the final round. The top-performing team achieved an AUC of 80.06%, highlighting the potential of AI in personalized DME treatment and clinical decision-making.
Paper Structure (42 sections, 3 equations, 12 figures, 5 tables)

This paper contains 42 sections, 3 equations, 12 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Review of Previous Challenges
  4. AI Methods for OCT Image Analysis
  5. Dataset: OCT4DME
  6. Data Overview
  7. Data Acquisition
  8. Data Statistics
  9. Data Annotation
  10. Competition Setup
  11. Challenge Organizers
  12. Competition Overview
  13. Competition Tasks
  14. The Preliminary Round:
  15. The Second Round:
  16. ...and 27 more sections

Figures (12)

  • Figure 1: An example from the dataset OCT4DME.
  • Figure 2: The overall organization workflow of the 2021 Asia Pacific Tele-Ophthalmology Society (APTOS) Big Data Competition.
  • Figure 3: Three sub-tasks of the competition, which are diagnosing the symptoms of the eye, predicting the Central Subfield Thickness, and judging whether to continue treatment.
  • Figure 4: Solution for Task 1 by the champion team BlueSky. This illustrates the architecture of the Multi-Instance Learning network proposed by the team. Given a bag containing n images, the network outputs an N x 4-dimensional vector, where 4 represents the categories: IRF, SRF, PED, and HRF.
  • Figure 5: Solution for Task 1 by the runner-up team LightRain. They used a weakly supervised learning framework for classification.
  • ...and 7 more figures