Back to all papers

Can machine learning predict vessels that encapsulates tumor cluster ptterns and histological differentiation in solitary small (< 5 cm) hepatocellular carcinoma?

May 26, 2026pubmed logopapers

Authors

Toshida K,Sugitani Y,Itoh S,Toshima T,Motomura T,Yoshiya S,Yugawa K,Iseda N,Iwasaki T,Aishima S,Ishigami K,Oda Y,Morooka K,Yoshizumi T

Affiliations (7)

  • Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, 812-8582, Japan.
  • Graduate School of Natural Science and Technology, Okayama University, Okayama, Japan.
  • Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, 812-8582, Japan. [email protected].
  • Department of Anatomic Pathology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan.
  • Department of Scientific Pathology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan.
  • Department of Clinical Radiology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan.
  • Faculty of Advanced Science and Technology, Kumamoto University, Kumamoto, Japan.

Abstract

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality. Machine learning (ML) may enable the noninvasive prediction of histopathological features using clinical and imaging data. This study evaluated the utility of ML in predicting vessels that encapsulate tumor clusters (VETC) and tumor differentiation in HCC. We retrospectively analyzed 232 patients who underwent hepatic resection for solitary HCC (≤ 5 cm). Histological assessment was used to determine the VETC status and tumor differentiation. Contrast-enhanced CT images were processed using BiomedCLIP to extract 512-dimensional image feature vectors, which were combined with clinical data and analyzed using a support vector machine classifier. The model performance was evaluated using five-fold cross-validation with precision, recall, and F1-score. VETC-positive tumors were significantly associated with a poor survival. Models using clinical data alone demonstrated limited predictive performance (F1 = 0.469 for VETC; F1 = 0.473 for differentiation). The incorporation of image features modestly improved VETC prediction (F1 = 0.599) but did not enhance the prediction of tumor differentiation. Image-based models did not outperform the clinical models. VETC is a histopathological marker associated with a poor prognosis in HCC. While ML models showed limited predictive accuracy, further optimization of imaging analysis and data integration may improve noninvasive histological prediction.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.