A deep learning model achieving ~95% accuracy in multi-organ segmentation for radiotherapy planning. Built on Swin-UNet, N2 integrates temporal context from prior scans through cross-attention fusion—blending present imagery with historical knowledge for precise anatomical delineation.