Primus V2 Transformer architecture sets new state-of-the-art in 3D medical image segmentation

Researchers have developed Primus and PrimusV2, novel Transformer-centric architectures for 3D medical image segmentation that outperform hybrid models. These new architectures address shortcomings in current Transformer-based methods by optimizing the use of Transformer blocks with high-resolution tokens and advanced positional embeddings. PrimusV2, in particular, achieves state-of-the-art performance, rivaling leading CNNs on multiple public datasets and establishing Transformers as a competitive approach in this domain. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Establishes Transformer-centric models as competitive for 3D medical image segmentation, potentially shifting research focus from hybrid approaches.

RANK_REASON This is a research paper introducing new model architectures for a specific AI task.

Read on arXiv cs.CV →

• Primus
• PrimusV2
• Transformer
• CNN
• nnU-Net
• ResEnc-L
• MedNeXt
• arXiv
• Saikat Roy

• paper
• model release