SimCroP learns chest-CT representations by aligning each report sentence to its most similar visual patches and fusing whole-scan and word-patch features, reporting higher classification and segmentation scores than six prior methods.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SimCroP: Radiograph Representation Learning with Similarity-driven Cross-granularity Pre-training
SimCroP learns chest-CT representations by aligning each report sentence to its most similar visual patches and fusing whole-scan and word-patch features, reporting higher classification and segmentation scores than six prior methods.