CT-VDETR: Semi-supervised 3D Trauma Detection in Computed Tomography (CT) scans using Dense Vertex Relative Position Encoding
Researchers have developed CT-VDETR, a novel framework for detecting traumatic injuries in CT scans, addressing the challenge of limited voxel-level annotations. The system combines self-supervised pretraining using Masked Image Modeling with a semi-supervised transformer-based detector. This approach achieved a 1.53x improvement over supervised-only training on the RSNA abdominal trauma detection task, utilizing only 78 labeled training volumes. AI
IMPACT Enhances label-efficient AI for medical imaging, potentially reducing annotation costs and improving diagnostic speed.