Deep learning-based methods have achieved state-of-the-art performance in various medical image analysis tasks. Most of these models are trained under the assumption that the training data (source domain) and testing data (target domain) are independent and identically distributed (i.i.d.). In practice, variations in image acquisition, such as differences in medical centers, equipment manufacturers, scanning protocols, and imaging modalities, render this assumption unrealistic, often resulting in significant performance degradation. This challenge is commonly referred to as domain shift or domain gap.