RPS 501 - Interventional oncology and treatment assessment in liver cancer

Author Block: S-H. Zhang¹, Y. Song², Y-C. Wang¹; ¹Jiangsu/CN, ²Shanghai/CN
Purpose: To develop a predictive model that integrates radiomics features from contrast-enhanced MRI with conventional radiologic features to identify early recurrence locations in nonviable hepatocellular carcinoma (HCC) after transarterial chemoembolization (TACE).
Methods or Background: This multicenter retrospective study included HCC patients treated with TACE whose contrast-enhanced MRI at 1–2 months post-treatment showed LR-TR nonviable lesions ≥1.0 cm with at least one-year follow-up. A 1-cm peritumoral ring was divided into eight sectors for radiomics feature extraction to build a radiomics model. A fusion model combined these features with two radiologic characteristics (non-smooth margin and peritumoral hyperintensity on T2- or diffusion-weighted images). Model performance was evaluated using ROC curves, and differences were assessed by the DeLong test.
Results or Findings: A total of 616 sectors from 77 lesions were included in the study, including 96 (15.6%) sectors with recurrence within one year of follow-up. The radiomics model achieved AUC of 0.738 and 0.718 in the training and test cohorts. The radiologic model, constructed using non-smooth margin and peritumoral hyperintensity on T2-weighted or diffusion-weighted images, achieved AUC of 0.751 and 0.744 in the training and test cohorts, respectively. The fusion model combining three radiomics features and two radiologic features achieved AUC of 0.794 and 0.765 in the training and test cohorts, respectively. The DeLong test demonstrated a statistically significant difference between the fusion model and the radiologic model in the training cohort (P = 0.026).
Conclusion: The fusion model combining selected radiomics features and conventional radiologic features demonstrated excellent performance in predicting the site of recurrence in LR-TR nonviable HCC lesions following TACE.
Limitations: Its retrospective design may introduce selection bias, and the limited sample size raises concerns about potential overfitting and restricted generalizability, highlighting the need for prospective validation.
Funding for this study: This study has received funding by National Natural Science Foundation of China (NSFC, No. 82271978).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This multicenter, retrospective study was reviewed and approved by Institutional Ethics Committee for clinical research of the hospital, approval number [2024ZDSYLL149-P01], with a waiver for informed consent.

Author Block: B. Li, S-H. Zhang, Y-C. Wang; Nanjing/CN
Purpose: Current prognostic prediction models of hepatocellular carcinoma (HCC) who treated with transarterial chemoembolization (TACE) include only the preoperative imaging and clinical features, with their available repeated postoperative MRI features measurements underutilized. HCC prognostic prediction models were constructed in this study to clarify whether and to what extent the inclusion of postoperative longitudinal measurements of longest diameter of active lesions, number of tumors, longest diameter of target lesions and Apparent Diffusion Coefficient (ADC) value of target lesions can improve the model performance and perform a dynamic prediction.
Methods or Background: The training cohort study included 56 HCC patients who underwent TACE, with preoperative measurement and three or more measurements within 24 months after TACE, respectively. Prediction models to predict HCC untreatable tumor-aggressive progression (macrovascular invasion, extrahepatic spread, >50% liver involvement) were constructed with clinical variables, by incorporating preoperative quantitative MRI imaging features , as well as their postoperative longitudinal measurements.
Results or Findings: The prediction model with postoperative longitudinal measurements of longest diameter of active lesions and number of tumors within 24 months after TACE outperformed the model including preoperative imaging and clinical features only, with the better area under the receiver operating characteristic curves (AUCs: 0.859 vs 0.797), (AUCs: 0.923 vs 0.815) and (AUCs: 0.824 vs 0.786) at 12, 24, and 36 months after TACE. The proposed longitudinal prediction model can provide a personalized dynamic prediction for a new patient, with estimated survival probability updated when a new measurement is collected during 24 months after TACE.
Conclusion: Prediction models including longitudinal measurements of longest diameter of active lesions and number of tumors have improved accuracy in predicting the prognosis of HCC patients.
Limitations: The proposed prediction models were developed solely on a single-institution cohort without external validation.
Funding for this study: This study has received funding by National Natural Science Foundation of China (NSFC, No. 82271978, 92359304, 82330060)
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: P. A. Bonaffini, P. Majoro, A. Celestino, A. Barbaro, C. Gargiulo, P. Marra, S. Sironi; Monza/IT
Purpose: To assess the prognostic value of iodine concentration differences measured on iodine maps derived by dual energy CT (DECT) in hepatocellular carcinoma (HCC) in cirrhosis, comparing untreated nodules (disease progression/stability) and those with recurrence after loco-regional therapies.
Methods or Background: DECT of suspected HCC (LI-RADS 3-5) were retrospectively reviewed (March 2022-September 2023; minimum 3 months follow-up) . Lesions were divided in untreated (group 1) and undergoing loco-regional treatments (group 2). Progression was defined as ≥20% size increase (mRECIST). Iodine maps were generated in the hepatic arterial and portal venous phases. In each phase iodine concentration density (ICD, mg/ml) was measured by ROI placed in homogeneously enhancing portion of the target nodule (aICDnodule, vICDnodule) and in the non-nodular parenchyma (aICDliver, vICDliver). Lesion-to-normal liver ICD ratio (aLNR, vLNR) to quantify nodule enhancement behavior and ratio of LNR during the arterial to the venous phase (ravLNR) were calculated and compared.
Results or Findings: Twenty-two patients (mean age 69.5 years) with 41 lesions (19 LI-RADS 3; 6 LI-RADS 4; 16 LI-RADS 5) were included in group 1. In lesions demonstrating progression (14/41) baseline ICD values were significantly lower compared to stable nodules: aICDnodule 26.0 (IQR 20.5-30.5) vs 33 (IQR 27-40), vICDnodule 22 (IQR 19-25) vs 25.5 (IQR 27-40), respectively. Twenty-two patients (mean age 70.3 years) with 31 lesions (8 LI-RADS 3; 2 LI-RADS 4; 21 LI-RADS 5) were included in group 2. The comparison of the median iodine concentrations (aICD, vICD, aLNR, vLNR, ravLNR) at baseline between lesions relapsing after treatment and those without recurrence showed no significant differences.
Conclusion: Quantification of iodine concentration on DECT may support the prognostic stratification of suspected HCC in cirrhotic patients, helping to predict the risk of progression in non-treated lesions.
Limitations: Retrospective
Small sample size
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: P. Buscemi¹, M. D'Amico¹, M. G. Vallone¹, F. Meloni²; ¹Palermo/IT, ²Milan/IT
Purpose: Hepatocellular carcinoma (HCC) is the most common primary liver tumor, with an increasing global incidence. Early recurrence detection post-ablation improves retreatment outcomes and survival. This study evaluates the role of CEUS in identifying early CC recurrence through the characterization of patent arteries observed 24 hours post-ablation via CECT and follow-up imaging.
Methods or Background: Ninety-eight patients (66 men, 32 women) with 121 HCCs treated with RFA (62) or MWA (59) between January 2016 and December 2019 were analyzed.
Post-treatment imaging included CT and CEUS at 24 hours and CT/MRI with CEUS every four months. Peri-ablative vessels (5-10 mm from the periphery) were identified in 51/121 lesions (42.1%), while intranecrotic vessels were found in 28/51 lesions (54.9%).
Results or Findings: Intralesional vessels increased in caliber and length in 75% of cases, leading to the development of enhancement with wash-in and wash-out consistent with intrazonal recurrence. CEUS detected intralesional vascular recurrence earlier than CT/MRI, with only 2/21 cases (9.5%) visible on CT/MRI. A significant correlation was found between local recurrence and perilesional vessels (p<0.01), as well as between periablation and intralesional vessels (p<0.01), highlighting a microenvironment favorable to neo-angiogenesis. Perilesional vascularization was significantly associated with RFA (p<0.01).
Conclusion: Persistent vascularization near the ablation site may promote intra- and perizonal recurrence by sustaining residual tumor cells. CEUS demonstrated superior sensitivity in detecting early intralesional recurrence compared to CT/MRI, supporting its role in post-ablation surveillance.
Limitations: Single-center retrospective design
Funding for this study: This study received no funding
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: R. Zhang; Harbin, Heilongjiang/CN
Purpose: This study aimed to develop predictive models for local recurrence after HCC ablation by integrating multidimensional clinical data and imaging features with multiple machine learning algorithms.
Methods or Background: A retrospective cohort study was conducted, including 460 patients with 1,130 HCC lesions who underwent ablation between June 2016 and June 2024. After applying strict inclusion and exclusion criteria, 459 lesions were enrolled and divided into a training cohort (n = 321) and a validation cohort (n = 138). Multiple machine learning algorithms, including logistic regression, random forest, support vector machine, multilayer perceptron, LightGBM, and LASSO, were applied for feature selection and model construction. Predictive performance was compared across models using the area under the receiver operating characteristic curve (AUC), and the optimal model was identified. A visual nomogram was then developed. Calibration curves and decision curve analysis (DCA) were employed to evaluate model performance, while Kaplan–Meier (KM) survival analysis was used to assess recurrence-free survival (RFS) differences among different risk groups.
Results or Findings: Performance comparison demonstrated that the predictive model based on LASSO-selected features achieved the best performance, with AUC values of 0.894 in the training cohort and 0.775 in the validation cohort. The constructed nomogram further confirmed the model’s feasibility and practicality in clinical applications. KM survival analysis showed significant differences in recurrence time across risk groups (p < 0.001), validating the model’s effectiveness and clinical value.
Conclusion: The machine learning prediction model constructed using LASSO feature selection exhibited high accuracy and stability in predicting local recurrence after HCC ablation. This provides a reliable tool for early identification of high-risk patients and formulation of individualized treatment strategies in clinical practice.
Limitations: This study is that only Chinese patients were included.
Funding for this study: Harbin Medical University Cancer HospitalClimbing program (PDYS2024-10).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: the ethics committee of HMUCH（KY2024‑33）

Author Block: S-H. Chiu¹, D. Delitto², L. S. Yoon², A. Kamaya², J. Tse²; ¹Taipei/TW, ²Palo alto, CA/US
Purpose: The LI-RADS CT/MRI Nonradiation Treatment Response Algorithm (v2024) extends its application to assessing the surgical margin after hepatocellular carcinoma (HCC) resection. This study aims to apply v2024 to surgically-resected HCCs, determine rates of local recurrence defined by v2024, and evaluate the timing of recurrence during postoperative surveillance.
Methods or Background: This retrospective study included patients with surgically-resected HCC and postoperative liver CT/MRI for ≥6 months. v2024 and LI-RADS CT/MRI Diagnosis Version 2018 (v2018) were applied to the surgical margin. Two abdominal radiologists assessed for local recurrence (LR-TR Viable meeting ≥LR-4 by v2018 at diagnosis or follow-up), intrahepatic recurrence away from the margin (new lesion meeting ≥LR-4), and extrahepatic disease. Discrepancies were resolved by a third radiologist.
Results or Findings: Two hundred patients (142 men and 58 women; median age 67 years) were followed for a median of 30.5 months [IQR 13.7-60.1]. LR-TR Viable at the surgical margin occurred in 57 patients (28.5%) after a median of 5.1 months. At diagnosis, 26 LR-TR Viable observations (45.6%) already met v2018 criteria for ≥LR-4, while 31 (54.4%) did not; of these, 30 progressed to ≥LR-4 after a median of 3.2 months, and 1 resolved with no residual HCC at transplantation. Intrahepatic recurrence away from the margin developed in 77 patients (38.5%) after a median of 10.9 months, and extrahepatic recurrence in 37 patients (18.5%) after 5.4 months. Inter-reader agreement for LR-TR category was excellent (κ=0.93), and its positive predictive value for local recurrence was 98.2%.
Conclusion: In about half of cases, v2024 detects local recurrence at the surgical margin roughly 3 months before HCC criteria are met, enabling earlier diagnosis of local recurrence.
Limitations: Single-center, retrospective study, and relied on v2018 rather than histologic confirmation as the reference standard for recurrence.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Stanford University IRB eProtocol 77441

Author Block: J. N. Nonninger¹, J. Slavicek¹, N. Fezoulidis¹, K. Hergan², S. Zandieh¹; ¹Vienna/AT, ²Salzburg/AT
Purpose: Hepatocellular carcinoma (HCC), the most common primary liver cancer, is often treated with transarterial chemoembolization (TACE). Standard criteria (RECIST, mRECIST) may miss early changes, underestimating therapeutic effects. Quantitative imaging may improve early assessment and guide therapy. This study analyzed CT perfusion and radiomic features of HCC before and after TACE and their association with tumor burden.
Methods or Background: In this retrospective, single-center study, 32 patients with histologically confirmed HCC underwent CT perfusion and radiomic analysis prior to and following TACE. Multiple quantitative perfusion parameters (arterial flow, perfusion flow, perfusion index) and radiomic features were extracted. Statistical comparisons were performed using the Wilcoxon signed-rank test and Spearman’s correlation.
Results or Findings: Arterial flow significantly decreased from a median of 56.5 to 47.7 mL/100 mL/min after TACE (p = 0.009), while nonsignificant increases in perfusion flow (from 101.3 to 107.8 mL/100 mL/min, p = 0.44) and decreases in the perfusion index (from 38.6% to 35.7%, p = 0.25) were observed. Perfusion flow was strongly and positively correlated with tumor size (ρ = 0.94, p < 0.001). Five radiomic texture features especially ShortRunHighGrayLevelEmphasis (Δ = +2.11, p = 0.0001) and LargeAreaHighGrayLevelEmphasis (Δ = +75,706, p = 0.0006)—changed significantly after treatment. These changes were more pronounced in tumors ≥50 mm.
Conclusion: CT perfusion and radiomics offer complementary insights into the treatment response of patients with HCC. While perfusion parameters reflect macroscopic vascular changes and are correlated with tumor burden, radiomic features can indicate microstructural changes after TACE. This combined imaging approach may improve early therapeutic assessment and support precision oncology strategies.
Limitations: Despite its strengths, the study has limitations, including its retrospective design and potential observer bias from manual segmentation, which was mitigated by consensus reading.
Funding for this study: No funds were used in this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics Committee of the City of Vienna