Abstract: Transformers using self-attention mechanisms have recently advanced medical imaging by modeling long-range semantic dependencies, though they lack the ability of convolutional neural ...