Investigating the structural cost of spectral equivariance in 3D dense medical segmentation. Implementing state-of-the-art Volumetric SegResNets natively on BraTS 2021 to mathematically evaluate BlurPool 3D anti-aliasing (~50% AVR reduction) against boundary topological precision.
computer-vision deep-learning signal-processing pytorch medical-imaging resnet-50 brain-tumor-segmentation monai brats2021 boundary-f1 blurpool spectral-aliasing nyquist-shannon
-
Updated
Apr 14, 2026 - Python