Related Experiment Video For arterial phase
Updated: Nov 30, 2025
Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018
Clinical and contrast-enhanced image features in the prediction model for the detection of small hepatocellular
Ming-Feng Chiang1, Tse-Kai Tseng2, Chia-Wen Shih3
1Division of Gastroenterology and Hepatology, Department of Internal Medicine, Lo-Hsu Medical Foundation, Lotung Poh-Ai Hospital, Yilan, Taiwan.
Abstract:
Purpose: To identify novel radiological features and clinical characteristics to improve diagnostic criteria for early detection of small hepatocellular carcinoma (HCC). Patients and Methods: We retrospectively recruited asymptomatic patients with no history of HCC but a high risk of HCC in whom a new, solitary, well-defined, solid nodule between 10 and 20 mm was detected through a screening ultrasound. We retrospectively collected all clinical data, and patients were examined using dynamic contrast-enhanced computed tomography or magnetic resonance imaging; subsequently, fine-needle biopsy was performed. A multivariate analysis of the predictors of small HCCs was performed by fitting a multiple logistic regression model with the stepwise variable selection method. Results: In total, 392 and 347 patients with a small liver nodule received a final pathologic confirmation of HCC and non-HCC, respectively. The estimated odds ratios and 95% confidence intervals of tumor size > 12.45 mm, age > 56.61 years, liver cirrhosis, hepatitis C virus (HCV) carrier status, ln alpha-fetoprotein (AFP) > 1.954, arterial phase enhancement, and portal or venous phase washout appearance without arterial phase enhancement were 2.0735 (1.4746-2.9155), 1.8878 (1.2949-2.7521), 1.6927 (1.1294-2.5369), 1.6186 (1.0347-2.5321), 2.0297 (1.3342-3.0876), 3.7451 (2.3845-5.8821), and 2.0327 (1.3500-3.0608), respectively. The area under the receiver operating characteristic curves for the diagnosis of small HCCs was 0.79 for arterial phase enhancement and 0.75 for portal or venous phase washout appearance without arterial phase enhancement. Conclusion: Clinical and contrast-enhanced image features are valuable in the prediction model for the detection and early diagnosis of small HCCs in patients with a high risk of HCC. In addition to negative portal or venous washout and negative arterial enhancement in images, age > 56.61 years, tumor size > 12.45 mm, HCV carrier status, and ln(AFP) > 1.954, are useful indicators for the early detection of small HCCs.
