Sun Yat-sen/Canon framework segments body tissues on CPU in 44 seconds

Automated 3D segmentation of muscle and adipose tissue from CT scans is critical for body composition analysis, but multi-source data variability and high memory demands limit clinical adoption. Researchers from Sun Yat-sen University and Canon Medical Systems China address this with a coarse-to-fine hierarchical framework that segments ten tissue structures. Efficiency gains come from Dynamic Spacing and Anisotropic Patching, a Group Inference mechanism for low-memory sliding-window processing, and Topology-Aware Asymmetric Resampling for fast post-processing. The framework was trained on 1,558 CT volumes from seven public and two private datasets.

On a 12-core CPU workstation, the GPU-free pipeline averaged 44.5 seconds per volume with 4.73 GB peak memory. Evaluation on an independent cohort of 105 scans yielded per-structure Dice coefficients from 0.924 to 0.982, with eight major structures meeting the ±10% relative error clinical acceptance limit. This balance of accuracy and efficiency allows robust, large-scale body composition analysis on standard CPU workstations, removing the need for expensive GPU hardware and enabling wider clinical deployment.