RF-Deep trains random forests on tumor-anchored deep features from segmentation backbones to detect near-OOD and far-OOD CT scans with AUROC above 93 and 99 respectively, outperforming prior methods by 4-7 points on near-OOD cases.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Tumor-anchored deep feature random forests for out-of-distribution detection in lung cancer segmentation
RF-Deep trains random forests on tumor-anchored deep features from segmentation backbones to detect near-OOD and far-OOD CT scans with AUROC above 93 and 99 respectively, outperforming prior methods by 4-7 points on near-OOD cases.