Metraipolis › archive › dc_d7e90ea258d2

Big, Bright, or Invisible: A Frozen-Feature Benchmark of 3D CT Foundation Models — arXiv:2608.05960

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.05960
License (as archived)
CC0-1.0
Archived
2026-10-31T10:36:25.198Z
SHA-256 of the archived text
b21ce721fcc4ba429a3666788b16a1748799126e2da0bf3bf9031fa993a99342
Stable URL of this page
https://metraipolis.com/documento/dc_d7e90ea258d2

How to cite

Big, Bright, or Invisible: A Frozen-Feature Benchmark of 3D CT Foundation Models — arXiv:2608.05960 — arXiv (en), https://arxiv.org/abs/2608.05960 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d7e90ea258d2 (Metraipolis archive dc_d7e90ea258d2, sha256 b21ce721fcc4ba429a3666788b16a1748799126e2da0bf3bf9031fa993a99342)
BibTeX
@misc{dcd7e90ea258d2,
  title = {Big, Bright, or Invisible: A Frozen-Feature Benchmark of 3D CT Foundation Models — arXiv:2608.05960},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.05960},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d7e90ea258d2; sha256 b21ce721fcc4ba429a3666788b16a1748799126e2da0bf3bf9031fa993a99342},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.05960. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Routine CT interpretation is inherently comprehensive, capturing incidental findings across the entire scan volume. 3D CT foundation models could assist this process by providing generalizable representations of anatomy and pathology. To evaluate their diagnostic breadth, we benchmark ten frozen CT encoders across three cohorts of thoracic CT scans, including an unseen internal clinical dataset, using $k$-nearest neighbors, zero-shot prompting, and linear probing. We find no universal state-of-the-art, with rankings fluctuating significantly depending on the evaluation context. While models combining fine-grained image tokenization with vision-language alignment generally perform best, a lightweight supervised encoder remains highly competitive, demonstrating that explicit labels can effectively substitute for scale. Crucially, rather than model architecture, we observe that the primary determinant of performance is a physical bottleneck: a finding's detectability scales with its contrast against surrounding tissue and its spatial extent. Through controlled within-organ comparisons, we empirically demonstrate that widespread or high-contrast abnormalities, such as devices and effusions, are reliably recovered. Conversely, small, low-contrast focal lesions remain a persistent challenge across all evaluated encoders. We attribute this to the inherent limitations of globally pooled embeddings, suggesting that accurately representing small, low-contrast structures will require region- or lesion-level pretraining.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-07 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.05960 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-31T10:23:31.038Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-08-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).