Physically Aware Radiomics Without Interpolation: Disentangling Voxel Geometry and Signal Modification in CT and MRI
The episode discusses a paper introducing a voxel-spacing–aware radiomic framework for CT and MRI that extracts features without interpolating the original image signal. The hosts discuss findings showing near-native agreement with non-resampled extraction, suggesting this method provides more reliable quantitative biomarker analysis. They also explore future improvements involving physically aware convolutional kernels and geometry-invariant feature extraction modules.