OrganLens conditions a shared CT encoder on an organ identity and uses mask-supervised pooling to produce 11 organ-specific representations from the same volume, with organ-matched representations improving downstream detection and prognosis.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
OrganLens: Organ-Specific Representation Learning for CT Foundation Models
OrganLens conditions a shared CT encoder on an organ identity and uses mask-supervised pooling to produce 11 organ-specific representations from the same volume, with organ-matched representations improving downstream detection and prognosis.