“…To facilitate information interaction in the entire image space, combining a deep neural network-based framework with channel attention holds promise. At present, the dominant framework still relies on convolutional operations, such as CNN [36], [37], FCN [38]- [40], and advanced complexvalued [41] and 3-D networks [42], which have demonstrated commendable performance. However, although the design of multi-level attempts to incrementally enlarge the receptive field to improve global modeling capacity, resource overconsumption also appears.…”