Primus: Enforcing Attention Usage for 3D Medical Image Segmentation

ArXi:2503.01835v2 Announce Type: replace Transformers have achieved remarkable success across multiple fields, yet their impact on 3D medical image segmentation remains limited with convolutional networks still dominating major benchmarks. In this work, (A) we analyze current Transformer-based segmentation models and identify critical shortcomings, particularly their over-reliance on convolutional blocks. Further, we nstrate that in some architectures, performance is unaffected by the absence of the Transformer, thereby nstrating their limited effectiveness.