An attention-based context-informed deep framework for infant brain subcortical segmentation

The first infant-dedicated coarse-to-fine deep framework for subcortical segmentation from original multi-modal images, including T1w, T2w, and the ratio of T1w and T2w images.A spatial context-guided and spatial and channel attention-based framework.Delineation of inner and outer subcortical boundaries as extra labels to help precisely estimate the ambiguous boundaries.State-of-the-art performance and wide applicability of our framework on both neonates and infants.