跳到正文
arXiv cs.AI· Yosuke Yamagishi, Tomohiro Kikuchi, Shouhei Hanaoka, Takeharu Yoshikawa, Osamu Abe·· 10 小时前AI 评分31

通用 CT 基础模型 RADAR 与 Merlin 能否区分肝脏肿瘤?

How Well Do Generalist CT Foundation Models Differentiate Liver Tumors?

AI 导读

研究评估了 RADAR 和 Merlin 两个通用 CT 基础模型区分肝脏肿瘤的能力,在零样本设置下 RADAR 四分类准确率仅 54.4%(κ=0.37),并将 31%(129/421)的恶性肿瘤预测为血管瘤。仅用线性探针在四期特征上训练线性分类器后,RADAR 准确率提升至 79.9%(κ=0.72),误判为血管瘤的恶性肿瘤降至 1%(5/421),Merlin 准确率为 60.9%。

正文 · 原文

View PDF

Abstract:Generalist CT foundation models detect many findings with high performance, but their ability to differentiate detected lesions remains largely unevaluated. This retrospective study evaluated two generalist models, RADAR and Merlin, for differentiating liver tumors using a public single-institution multiphase CT dataset (517 cases: hepatocellular carcinoma, intrahepatic cholangiocarcinoma, breast and colorectal cancer liver metastases, and hemangioma). The models were evaluated in a zero-shot setting using their existing outputs and with linear probing, in which only a linear classifier was trained on frozen image features (stratified fivefold cross-validation repeated five times). In the zero-shot setting, RADAR achieved an accuracy of 54.4% (95% CI: 49.9, 58.8) and a $\kappa$ of 0.37 for four-class differentiation at the portal venous phase and predicted 31% (129 of 421) of malignant tumors as hemangioma. With linear probing of four-phase features, its accuracy improved to 79.9% (95% CI: 76.6, 83.4) with a $\kappa$ of 0.72, and 1% (five of 421) of malignant tumors were predicted as hemangioma; Merlin achieved an accuracy of 60.9% (95% CI: 56.9, 65.0). Generalist CT foundation models showed limited zero-shot ability to differentiate liver tumors, which improved substantially with linear probing alone.
Comments: 12 pages, 3 figures. Code: this https URL
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.24122 [cs.CV]
  (or arXiv:2611.24122v1 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2611.24122

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Yosuke Yamagishi [view email]
[v1] Wed, 18 Nov 2026 09:41:41 UTC (974 KB)

来源:arXiv cs.AI · arxiv.org