MAGEFormer improves CT scan segmentation by using real-world space info
MAGEFormer: Learning Metric-Consistent Representations for Anisotropic CT Segmentation
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
CT scans often have different spacing in different directions, which can confuse computer programs that try to identify organs. The authors introduced MAGEFormer, a method that understands the true physical distances between points in the scan. This helps the program avoid mistakes caused by assuming all spaces are equal. Tests showed MAGEFormer was better at accurately outlining organs in abdominal CT scans than earlier methods.
What this means in practice
- •For medical imaging software developers: Build CT segmentation tools that respect physical spacing for more precise organ boundaries in anisotropic scans.
- •For radiology workflow engineers: Integrate MAGEFormer’s geometry-aware methods to improve accuracy of organ delineation in clinical CT image analysis pipelines.
Authors
Jiaying Li, Paolo Remagnino
Abstract
Vision Transformers (ViTs) have shown strong performance in volumetric segmentation, but their effectiveness on clinical CT is limited by an isotropic Euclidean lattice assumption. This conflicts with anisotropic CT acquisition, leading to two key issues: (1) a metric mismatch between voxel indices and physical anatomy, and (2) accuracy degradation from isotropic resampling. To address this, we propose MAGEFormer, a geometry-calibrated framework that embeds physical metric constraints directly into representation learning. Our method introduces Metric-Adaptive Spatial Embedding (MASE) to calibrate positional frequencies using voxel spacing, Geometry-Constrained Attention (GCA) to suppress physically implausible feature correlations, and Geometric View Voting (GVV) to reduce discretization bias during inference. We evaluate MAGEFormer on two multi-organ abdominal CT benchmarks, BTCV and FLARE 22, under a unified protocol against strong CNN and Transformer-based baselines. MAGEFormer achieves the strongest boundary accuracy among the compared methods, with 10.58 mm HD95 on BTCV and 3.40 mm HD95 on FLARE 22, and shows consistent gains in Dice under the same protocol. These results show that geometry-aware internal calibration is more effective than relying on conventional isotropic preprocessing alone for anisotropic CT segmentation.