Energy-Based Slice Selection
Energy-based slice selection is used instead of central-slice heuristics, all-slice voting, sequence modeling, segmentation, or manual annotation. The top ten slices by energy score are selected for each patient. The slice energy score is…
1 sources - 5 claims
Energy-based slice selection is used instead of central-slice heuristics, all-slice voting, sequence modeling, segmentation, or manual annotation. The top ten slices by energy score are selected for each patient. The slice energy score is computed as mean pixel intensity multiplied by pixel-intensity standard deviation. Each normalized T1Gd slice is resized to 64 by 64 before computing its energy score. Patient labels are repeated ten times to align with the selected slice-level labels.