This paper proposes a novel method for automatic diagnosis of Helicobacter pylori (H. pylori) infection based on self-supervised learning and self-knowledge distillation. Our method consists of two phases, the first is the self-supervised learning phase for learning discriminative representations from gastric endoscopic images, and the second is the self-knowledge distillation based fine-tuning phase for accurate automatic diagnosis of H. pylori infection. Our method achieves high diagnosis performance in a complex endoscopic image dataset.
(英)
This paper proposes a novel method for automatic diagnosis of Helicobacter pylori (H. pylori) infection based on self-supervised learning and self-knowledge distillation. Our method consists of two phases, the first is the self-supervised learning phase for learning discriminative representations from gastric endoscopic images, and the second is the self-knowledge distillation based fine-tuning phase for accurate automatic diagnosis of H. pylori infection. Our method achieves high diagnosis performance in a complex endoscopic image dataset.