.png){kind=icon}
.png){kind=icon}
Return
DTG: Dual Transformers-based Generative Adversarial Networks for retinal 2D/3D OCT image classification
.png){kind=icon}
PRE
AI
DOI：10.1016/j.media.2025.103915.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We proposed the Dual Transformers-based Generative Adversarial Networks (DTG), the first model tailored to classify retinal pathologies using both 2D OCT images and 3D OCT images. • We devised a GAN-based strategy to infer high-quality semantic representations of 2D and 3D retinal images. • We designed an effective mechanism named patient instance-based data augmentation technique to yield higher classification accuracy. • We developed a weighted classifier that mimics the clinicians’ strategy to predict the diagnostic outcome. • The experimental results prove that our approach outperforms powerful Convolutional Neural Networks and Transformers widely used for 2D and 3D image classification. Additionally, it can improve the performance of various existing works for retinal data classification.