LUNA16-DiagnosticPreservation-2D (LUNA16-DP2D): 2D CT Slices with Anatomical Masks for Diagnostic-Preservation Image Compression

LUNA16-DP2D is a 2D axial CT slice dataset derived from the complete LUNA16 challenge dataset (888 scans, subsets 0–9), designed for research on anatomical attention-based medical image compression. Contents:- 6,216 raw CT slices (512×512, int16 HU values)- 6,216 binary lung segmentation masks (*_lung_mask.npy, uint8 0/1)- 6,216 binary nodule masks (*_nodule_mask.npy, uint8 0/1)- metadata.csv (per-slice: subset, seriesUID, percentile, slice_index, HU stats) Slices extracted at percentiles [10, 25, 40, 50, 60, 75, 90]: idx = int(round(p / 100.0 * (depth - 1))) Mask sources:- Lung masks: binarized from seg-lungs-LUNA16- Nodule masks: ellipsoid projection from annotations.csv Full dataset available at:https://huggingface.co/datasets/nourbourai/luna16-dp2d Verified: binary ✓ | lung coverage p25–p75: 12–25% ✓ | nodule hit rate 3.5% ✓