Real-world efficacy of radiomics versus clinical predictors for microvascular invasion in patients with hepatocellular carcinoma: Large cohort study

Shotaro Kinoshita ¹, Takeshi Nakaura ², Tomoharu Yoshizumi ³

¹Department of Gastroenterological Surgery, Graduate School of Life Sciences, Kumamoto University, Kumamoto, Japan.
²Department of Diagnostic Radiology, Graduate School of Life Sciences, Kumamoto University, Kumamoto, Japan.
³Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan.
⁴Department of Surgery, Saga University Faculty of Medicine, Saga, Japan.
⁵Department of Surgery, NHO Nagasaki Medical Center, Nagasaki, Japan.
⁶Department of Hepato-Biliary-Pancreatic Surgery, Clinical Research Institute, NHO Kyushu Medical Center, Fukuoka, Japan.
⁷Department of Gastroenterological, Breast and Endocrine Surgery, Yamaguchi University Graduate School of Medicine, Ube, Japan.
⁸Division of Hepato-Biliary-Pancreas Surgery, Department of Surgery, University of Miyazaki Faculty of Medicine, Miyazaki, Japan.
⁹Department of Gastroenterological and Pediatric Surgery, Oita University Faculty of Medicine, Oita, Japan.
¹⁰Department of Gastroenterological Surgery, Oita Prefectural Hospital, Oita, Japan.
¹¹Department of Gastroenterological Surgery, Faculty of Medicine, Fukuoka University, Fukuoka, Japan.
¹²Department of Surgery, Saga-Ken Medical Centre Koseikan, Saga, Japan.
¹³Department of Surgery, Kagoshima Kouseiren Hospital, Kagoshima, Japan.
¹⁴Department of Surgery, Gastroenterology and Hepatology Center, Kitakyushu City Yahata Hospital, Kitakyushu, Japan.
¹⁵Department of Surgery, Yamaga City Medical Center, Yamaga, Japan.
¹⁶Department of Digestive and General Surgery, Graduate School of Medicine, University of the Ryukyus, Nishihara, Japan.
¹⁷Department of Surgery, Fukuoka Tokushukai Hospital, Fukuoka, Japan.
¹⁸Department of Surgery, Nagasaki University Graduate School of Biomedical Sciences, Nagasaki, Japan.

⁰Department of Gastroenterological Surgery, Graduate School of Life Sciences, Kumamoto University, Kumamoto, Japan.

Hepatology Research : the Official Journal of the Japan Society of Hepatology +

|

May 3, 2025

View abstract on PubMed

Summary

A clinical model effectively predicts microvascular invasion (MVI) in hepatocellular carcinoma (HCC), performing similarly to CT radiomics. Combining clinical and radiomics data did not enhance prediction accuracy for MVI.

Area Of Science

Hepatocellular Carcinoma Research
Medical Imaging Analysis
Radiomics and Machine Learning

Background

Microvascular invasion (MVI) is a critical prognostic factor in hepatocellular carcinoma (HCC), yet its preoperative diagnosis remains challenging.
Computed tomography (CT) radiomics shows potential for MVI detection, but its performance is sensitive to imaging protocols.
This study addresses the need for reliable MVI prediction in HCC using real-world, nonstandardized CT data.

Purpose Of The Study

To compare the predictive performance of radiomics, clinical, and combined models for MVI in HCC.
To evaluate the efficacy of these models under nonstandardized CT scanning conditions.
To determine if integrating radiomics with clinical data improves MVI prediction accuracy.

Main Methods

A multicenter study involving 533 HCC patients undergoing hepatic resection.
Manual extraction of 3D CT features across hepatic arterial, portal venous, and venous phases.
Development of radiomics, clinical (logistic regression), and fused models, with performance assessed by AUC in test groups.

Main Results

The clinical model included HBV surface antigen, tumor diameter, and specific tumor markers.
Radiomics and clinical models demonstrated comparable predictive performance (p=0.76).
The fused model integrating radiomics did not significantly improve prediction accuracy over the clinical model alone (p=0.51).

Conclusions

A clinical model is as effective as a CT radiomics model for predicting MVI in HCC using real-world scanning data.
Integrating clinical information and radiomics does not enhance predictive performance beyond the clinical model alone.
The findings suggest that clinical factors are robust predictors of MVI in HCC, even with variable imaging protocols.

Keywords:

computed tomography hepatocellular carcinoma microvasculature nomogram radiomics