Multi-level perturbations in image and feature spaces for semi-supervised medical image segmentation

Open Webpage

Feiniu Yuan, Biao Xiang, Zhengxiao Zhang, Changhong Xie, Yuming Fang

Published: 01 Jan 2025, Last Modified: 16 Nov 2025Displays 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We propose a three-level perturbation strategy in image space for significantly deepening the level of perturbation hierarchy.•We design a random indicative function to present a one-level perturbation in feature space for effectively enhancing the model’s resistance to large perturbations across various levels.•We propose a novel Semi-supervised Learning framework combined with Multi-level Perturbations (SLMP) in both image and feature spaces.