Early recurrence prediction and risk stratification of hepatocellular carcinoma after transarterial chemoembolization achieving radiological complete response based on contrast-enhanced CT machine learning

Early recurrence prediction and risk stratification of hepatocellular carcinoma after transarterial chemoembolization achieving radiological complete response based on contrast-enhanced CT machine learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Early recurrence prediction and risk stratification of hepatocellular carcinoma after transarterial chemoembolization achieving radiological complete response based on contrast-enhanced CT machine learning luhao liu, Yiyang Liu, Dongxi Lin, Ke Meng, Xiaoman Yang, Jiliang Zhou, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7611742/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Mar, 2026 Read the published version in Abdominal Radiology → Version 1 posted 10 You are reading this latest preprint version Abstract Objectives To develop and validate machine learning (ML) models using clinical and contrast-enhanced CT (CECT) parameters to assess recurrence risk in hepatocellular carcinoma (HCC) after transarterial chemoembolization (TACE) achieving radiological complete response (CR). Methods 122 HCC patients who underwent TACE and achieved radiological CR from two centers were divided into the development (n = 100) and external validation dataset (n = 22). Recurrence free survival (RFS) was tracked, and patients were categorized into early recurrence (ER) and non-ER groups based on a 1-year cutoff. Forty clinical and CECT parameters were collected and screened. Six ML models were constructed and compared using the area under the curve (AUC) and decision curve analysis (DCA). Key parameters were used to construct a Cox regression nomogram and stratify recurrence risk using log-rank test. Results The extreme gradient boosting (XGBoost) model demonstrated the best predictive performance based on 13 parameters, with AUCs of 0.913 and 0.812 for the internal and external validation datasets. SHapley Additive exPlanations (SHAP) analysis identified the top 10 parameters. The Cox regression nomogram was constructed with ECV, complete capsule, FIB-4 index, tumor size, platelet-to-neutrophil ratio, and delayed phase tumor CT value. Log-rank test demonstrated significant risk stratification between the two datasets (both p < 0.01). Conclusion The XGBoost-based ER prediction model identifies 1-year recurrence following TACE with radiological CR. The Cox regression nomogram enables risk stratification, dividing patients into three subgroups. Hepatocellular carcinoma Transarterial chemoembolization X-Ray Computed Machine learning Recurrence free survival Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Liver cancer is the sixth most common cancer and the third leading cause of cancer-related mortality globally[ 1 , 2 ]. Hepatocellular carcinoma (HCC) is the dominant type of primary liver cancer. HCC progresses insidiously, with 70% of patients diagnosed at intermediate or advanced stages, resulting in a 5-year survival rate of only 20.8%[ 3 ]. Although 30% of patients may benefit from curative treatments like surgical resection, transplantation, or ablation, transarterial chemoembolization (TACE) remains a widely used option for those ineligible for these approaches due to various contraindications[ 4 ]. Recent developments in TACE have emphasized precision TACE, which involves superselective catheterization of tumor-feeding arteries to minimize damage to healthy liver tissue. For non-massive or non-diffuse HCC, precision TACE aims for complete (CR) or substantial partial response (PR) based on the modified Response Evaluation Criteria in Solid Tumors (mRECIST)[ 5 ]. Despite its effectiveness, recurrence remains a concern, which is often overlooked. A study of 134 patients achieving CR after TACE found a 1-year local recurrence rate of 44%, with 3- and 5-year survival rates of 40% and 27%, respectively, for those who recurred within 1 year[ 6 ]. These results underscore the need for accurate identification of high-risk patients, even those with radiological CR, to guide treatment and maximize survival. Most studies focus on recurrence and risk factors post-HCC resection or ablation, with limited evidence on recurrence following TACE[ 7 – 9 ]. Some factors, such as tumor number, size, location, and serum alpha-fetoprotein (AFP), have been linked to recurrence after TACE[ 10 – 12 ]. However, these studies mainly rely on clinical data and do not fully explore the predictive value of imaging features. Recent research has shown that several MRI features can predict ER after TACE[ 13 ]. Yet, conventional statistical methods in such studies may not meet the growing demand for more advanced predictive models. The rapid development of artificial intelligence in medical imaging has highlighted the power of machine learning (ML) in capturing complex, nonlinear relationships in high-dimensional datasets. These algorithms offer enhanced predictive accuracy and generalization, making them ideal for model development. Liu et al demonstrated the effectiveness of ML models in risk stratification following intra-arterial therapies for HCC[ 14 ]. Therefore, our study aimed to develop and validate predictive models using six ML algorithms based on clinical and contrast-enhanced CT (CECT) parameters to identify ER after TACE with radiological CR. Additionally, we constructed a Cox regression nomogram incorporating recurrence free survival (RFS) to enable robust risk stratification, offering a personalized, imaging-guided framework for patient management. Materials and methods Study population This retrospective, dual-center study was approved by the institutional review board (No.××) with informed consent waived. The development cohort comprised 100 HCC patients who underwent initial TACE and achieved radiological CR at ××Hospital (Center 1) between March 2015 and August 2023. The external validation cohort included 22 patients from ××Hospital (Center 2) between August 2022 and August 2023, meeting identical eligibility criteria. Inclusion criteria were: (1) HCC diagnosis according to the American Association for the Study of Liver Diseases (AASLD) guidelines[ 15 ]; (2) liver CECT examination within 2 weeks prior to TACE; (3) Eastern Cooperative Oncology Group performance status (ECOG PS) 0–2; and (4) standardized conventional TACE with post-treatment radiological CR. The TACE protocol is detailed in Supplementary Text. Exclusion criteria and the patient selection flowchart are shown in Fig. 1 . CT protocol CT examinations was performed using Siemens dual-source scanners (Somatom Definition, Siemens Healthcare, Germany) or GE spectral CT scanners (Discovery CT 750 HD, GE Healthcare, USA). Imaging included non-contrast (NP), arterial (AP), portal venous (PVP), and delayed (DP) phases. After NP scanning, iodinated contrast (350 mgI/mL, GE Healthcare, China) was administered intravenously at 1.5 mL/kg and 3.5 mL/s, followed by a 40-mL saline flush. Detailed scanning parameters are provided in Table 1 . Table 1 Scanning parameters and acquisition protocols of two center Center 1 Center 2 CT scanner Somatom Definition CT Discovery CT 750 HD Manufacture Siemens Healthcare GE Healthcare Tube voltage 120 kVp 120 kVp Tube current Automated Tube Current Modulation Automated Tube Current Modulation Detector configuration 64 × 0.600 mm 64 × 0.625 mm Pitch 0.7 1.375 Rotation time 0.33 s/r 0.5 s/r Image reconstruction thickness 1 mm 1.375 mm AP initiation time 20–25 s 25–30 s PVP initiation time 40–60 s 60–70 s DP initiation time 90–120 s 100–120 s Definition of radiological CR Follow-up CECT or contrast-enhanced MRI was performed 4–6 weeks post-TACE. According to mRECIST criteria, CR was defined as complete disappearance of intratumoral arterial enhancement in all target lesions[ 16 ]. Additional embolization was performed when residual enhancement or incomplete lipiodol deposition was present, repeated until CR was achieved. CR assessments were conducted by a radiologist with over 20 years’ experience in liver tumor imaging. Date collection and image analysis Clinical and laboratory parameters included age, sex, hepatitis B (HBV) and C (HCV) infection, ECOG PS, Barcelona Clinic Liver Cancer (BCLC) and China Liver Cancer Staging (CNLC) grades, Child-Pugh grade, albumin-bilirubin (ALBI) score, serum AFP, fibrosis-4 (FIB-4) index, aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin (ALB), total bilirubin (TBIL), C-reactive protein (CRP), prothrombin time (PT), platelet count, neutrophil count, lymphocyte count, neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), platelet-to-neutrophil ratio (PNR), and hematocrit (HCT). The ALBI score was calculated as: (log10 TBIL [µmol/L] × 0.66) + (ALB [g/L] × -0.085) and classified into three grades: Grade 1 (≤-2.60), Grade 2 (-2.60 to -1.39), and Grade 3 (>-1.39). The FIB-4 index was computed as: (age × AST [U/L]) / (platelet count [×10⁹/L] × ALT [U/L] 1/2 ). CECT image analysis was performed independently by two radiologists with 3 and 5 years of liver tumor imaging experience. Discrepancies were resolved by a senior radiologist with 12 years of experience. In cases with multiple lesions, the largest tumor was evaluated. Qualitative features assessed included tumor number, location, cirrhosis, ascites, tumor margin, intratumoral necrosis, peritumoral enhancement, capsule appearance, and intratumoral arteries. For quantitative analysis, the radiologist with 3 years of experience selected the largest cross-section of the tumor, avoiding the area of vessels, necrosis, and heterogeneous enhancement. A circular region of interest (ROI) measuring 20–40 mm² was manually delineated to obtain CT values for each tumor phase. Additionally, aortic CT values were measured by placing an ROI in the abdominal aorta during NP and DP. A schematic of ROI delineation is shown in Fig. 2 . Tumor CT values in NP、AP、PVP、DP (T0, Ta, Tp, Td) and aortic CT values in NP and DP (A0, Ad) were recorded. Four CECT-derived quantitative parameters were calculated: Arterial enhancement ratio (AER) = (Ta - T0) / T0 Portal venous enhancement ratio (PER) = (Tp - T0) / T0 Arterial enhancement fraction (AEF) = [(Ta - T0) / (Tp - T0)] × 100% Extracellular volume fraction (ECV) = (1 - HCT) × [(Td - T0) / (Ad - A0)] × 100% Tumor size was defined as the longest axial measurement on the largest cross-section and recorded in millimeters. To assess reproducibility, a randomly selected subset of 30 cases was reanalyzed by the radiologist with 5 years of experience after a 2-week interval. Interobserver consistency was evaluated using the intraclass correlation coefficient (ICC), with ICC > 0.75 indicating excellent consistency. Follow-up Post-treatment surveillance with CECT or contrast-enhanced MRI was performed every 3–6 months. RFS was defined as the interval from the final TACE achieving radiological CR to imaging-confirmed local recurrence, intrahepatic recurrence, or distant metastasis. The final follow-up date was August 1, 2024. Recurrence date and pattern were recorded, with RFS calculated in days. Patients were categorized by 1-year recurrence status: early recurrence (ER) (RFS ≤ 365 days) and non-ER (late recurrence (RFS > 365 days) or recurrence-free at last follow-up). ML modeling workflow ML modeling was conducted on the ××Platform following these steps: Feature selection: Variables with p 0.75), the variable less associated with the outcome was excluded. Remaining features were refined using least absolute shrinkage and selection operator (LASSO) regression. Model construction: Using Center 1 data, six ML algorithms—logistic regression (LR), random forest (RF), support vector machine (SVM), decision tree (DT), extreme gradient boosting (XGBoost), and Gaussian naive Bayes (GNB)—were trained with 10-fold cross-validation to predict ER after TACE achieving CR. Center 2 served for external validation. Model evaluation: Performance was assessed using receiver operating characteristic (ROC) curves and metrics including area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score. Decision curve analysis (DCA) evaluated clinical utility. The best-performing model was selected based on overall performance and calibration. Interpretability analysis: Shapley Additive exPlanations (SHAP) analysis identified the most influential predictors in the optimal model. Cox regression nomogram: A multivariate Cox proportional hazards model incorporating key SHAP-identified parameters was used to develop a nomogram predicting 6, 12, and 18-month recurrence. Risk stratification: Patients were classified into low-, middle-, and high-risk groups by nomogram-derived scores, with stratification assessed by Kaplan-Meier analysis and log-rank test. Statistical analysis Analyses were performed using SPSS 26.0 (IBM, Armonk, NY, USA) and R 4.4.2 ( https://www.r-project.org ). Parameters with > 15% missing data were excluded; those with ≤ 15% missingness were imputed using the RF algorithm. In Center 1, group comparisons between ER and non-ER were as follows: normally distributed continuous variables were expressed as mean ± standard deviation and compared with independent t -tests; non-normally distributed data were expressed as median (interquartile range) and compared using Mann-Whitney U tests; categorical variables were reported as counts (%) and compared using chi-square or Fisher’s exact tests. A two-sided p < 0.05 was considered statistically significant. Results Patient recurrence and baseline characteristics At final follow-up, among 100 patients from Center 1, 54 (54.0%) had ER, 40 (40.0%) late recurrence, and 6 (6.0%) remained recurrence-free. Median RFS was 304 days (95% CI: 248.14-359.86). In the ER group, 34 (62.96%) developed local recurrence and 20 (37.04%) intrahepatic or distant metastases. In late recurrence, 10 (25.0%) were local and 30 (75.0%) intrahepatic or distant metastases. In Center 2, 13 (59.09%) had ER, 2 (9.09%) late recurrence, and 7 (31.82%) no recurrence; median RFS was 326 days (95% CI: 275.44–376.56). ER in this cohort comprised 7 (53.85%) local and 6 (46.15%) intrahepatic or distant metastases; all late recurrences were intrahepatic or distant metastases. Baseline characteristics in Center 1 included 76 (76.0%) males and 24 (24.0%) females, aged 30–78 (57.96 ± 10.41) years. Center 2 had 19 (86.36%) males and 3 (13.64%) females, aged 43–74 years (62.05 ± 9.35). No significant intercenter differences were observed for age, sex, HBV or HCV infection, BCLC grade, and CNLC grade (all p > 0.05). Interobserver consistency ICCs for CECT quantitative parameters (T0, Ta, Tp, Td, A0, Ad, AER, PER, AEF, ECV) ranged from 0.945 to 0.998, all > 0.75, indicating excellent agreement and measurement reliability. Univariate analysis in the development dataset As shown in Table 2 , ER was associated with advanced CNLC grade ( p = 0.032), higher FIB-4 index ( p = 0.025), elevated CRP ( p < 0.001), higher neutrophil count ( p = 0.049), higher NLR ( p = 0.003), and lower PNR ( p = 0.004). Age also differed significantly ( p 0.05). As shown in Table 3 , significant CECT differences included lower Td ( p = 0.021), reduced ECV ( p < 0.001), larger tumor size ( p < 0.001), more frequent intratumoral necrosis ( p < 0.001) and intratumoral arteries ( p = 0.012), and fewer complete capsules ( p = 0.007) in ER. No differences were noted for T0, Ta, Tp, AER, PER, AEF, tumor number, location, cirrhosis, ascites, tumor margin, or arterial peritumoral enhancement (all p > 0.05). Table 2 Univariate analysis of clinical and laboratory parameters between ER and non-ER groups in the development dataset Parameters ER (n = 54) Non-ER (n = 46) t/Z/χ 2 p Age (years) 60.28 ± 10.40 55.24 ± 9.85 2.475 a 0.015 Sex 0.000 c 0.985 Male 41 (75.93%) 35 (76.09%) Female 13 (24.07%) 11 (23.91%) HBV infection 47 (87.04%) 38 (82.61%) 0.382 c 0.537 HCV infection 5 (9.26%) 5 (10.87%) - > 0.999 BCLC grade 3.382 c 0.066 A 42 (77.78%) 42 (91.30%) B 12 (22.22%) 4 (8.70%) CNLC grade 8.420 d 0.032 Ia 28 (51.85%) 35 (76.09%) Ib 17 (31.48%) 8 (17.39%) IIa 4 (7.41%) 3 (6.52%) IIb 5 (9.26%) 0 (0.00%) Child-Pugh grade 0.756 c 0.385 A 27 (50.00%) 27 (58.70%) B 27 (50.00%) 19 (41.30%) ALBI score 0.090 c 0.956 1 11 (20.37%) 9 (19.57%) 2 37 (68.52%) 31 (67.39%) 3 6 (11.11%) 6 (13.04%) Serum AFP (ng/mL) 1.418 c 0.234 ≤400 42 (77.78%) 40 (86.96%) >400 12 (22.22%) 6 (13.04%) FIB-4 index 6.29 (3.15, 9.56) 3.97 (2.55, 6.25) -2.234 b 0.025 AST (U/L) 44.35 (28.10, 71.06) 42.10 (30.78, 53.17) -1.477 b 0.140 ALT (U/L) 39.20 (24.48, 62.95) 37.11 (24.10, 53.38) -0.916 b 0.359 ALB (g/L) 34.95 (31.50, 38.15) 36.66 (33.50, 38.95) -0.896 b 0.370 TBIL (µmol/L) 21.40 (15.18, 35.93) 25.35 (15.30, 33.10) -0.048 b 0.961 CRP (ng/ml) 8.40 (2.23, 24.61) 2.85 (1.06, 6.69) -3.320 b < 0.001 PT (s) 12.60 (12.00, 13.90) 12.78 (11.85, 13.70) -0.246 b 0.806 Platelet count (×10 9 /L) 75.00 (56.00, 114.25) 87.50 (61.00, 119.75) -0.955 b 0.340 Neutrophil count (×10 9 /L) 2.12 (2.01, 3.79) 2.12 (1.59, 2.15) -1.967 b 0.049 Lymphocyte count (×10 9 /L) 0.85 (0.62, 1.10) 0.85 (0.81, 1.10) -1.589 b 0.112 NLR 2.87 (2.29, 4.46) 2.38 (1.56, 2.70) -2.988 b 0.003 PLR 95.63 (71.03, 137.37) 93.28 (62.42, 136.27) -0.660 b 0.509 PNR 31.57 (21.72, 44.91) 43.11 (27.99, 61.72) -2.877 b 0.004 HBV hepatitis B virus; HCV hepatitis C virus; BCLC Barcelona Clinic Liver Cancer; CNLC China Liver Cancer Staging; ALBI albumin-bilirubin; AFP alpha-fetoprotein; AST aspartate aminotransferase; ALT alanine aminotransferase; ALB albumin; TBIL total bilirubin; CRP C-reactive protein; PT prothrombin time; NLR neutrophil-to-lymphocyte ratio; PLR platelet-to-lymphocyte ratio; PNR platelet-to-neutrophil ratio a t value, b Z value, c χ 2 value by chi-square test, d χ2 value by Fisher's exact test, - indicates no statistical value Table 3 Univariate analysis of CECT parameters between ER and non-ER groups in the development dataset Parameters ER (n = 54) Non-ER (n = 46) t/Z/χ 2 p T0 (Hu) 44.19 ± 5.61 44.45 ± 6.65 -0.215 a 0.830 Ta (Hu) 86.92 ± 19.51 87.78 ± 22.25 -0.207 a 0.836 Tp (Hu) 89.98 ± 18.94 89.78 ± 20.66 0.049 a 0.961 Td (Hu) 75.57 ± 10.86 81.13 ± 12.90 -2.340 a 0.021 AER 0.98 ± 0.43 1.00 ± 0.49 -0.219 a 0.827 PER 1.05 ± 0.40 1.02 ± 0.39 0.264 a 0.792 AEF (%) 99.06 ± 38.00 98.90 ± 39.00 0.021 a 0.983 ECV (%) 21.45 ± 5.01 29.33 ± 6.51 -6.836 a < 0.001 Tumor size (mm) 40.00 (28.50, 55.00) 22.00 (18.25, 33.00) -4.432 b 2 6 (11.11%) 2 (4.35%) Tumor location 4.833 d 0.145 Left lobe 12 (22.22%) 8 (17.39%) Right lobe 38 (70.37%) 37 (80.43%) Caudate lobe 0 (0.00%) 1 (2.17%) Junction area 4 (7.41%) 0 (0.00%) Cirrhosis 49 (90.74%) 42 (91.30%) - > 0.999 Ascites 19 (35.19%) 11 (23.91%) 1.503 c 0.220 Tumor margin 1.307 c 0.253 Smooth 22 (40.74%) 24 (52.17%) Non-smooth 32 (59.26%) 22 (47.83%) Intratumoral necrosis 26.235 c < 0.001 Absence 17 (31.48%) 38 (82.61%) Presence 37 (68.52%) 8 (17.39%) Arterial peritumoral enhancement - 0.723 Absence 49 (90.74%) 43 (93.48%) Presence 5 (9.26%) 3 (6.52%) Capsule appearance 7.369 c 0.007 Complete 10 (18.52%) 20 (43.48%) Incomplete 44 (81.48%) 26 (56.52%) Intratumoral arteries 6.355 c 0.012 Absence 17 (31.48%) 26 (56.52%) Presence 37 (68.52%) 20 (43.48%) T0, Ta, TP, Td Tumor CT values of non-contrast phase, arterial phase, portal venous phase and delayed phase; AER arterial enhancement ratio; PER portal venous enhancement ratio; AEF arterial enhancement fraction; ECV extracellular volume fraction The note is the same as in Table 2 . ML feature selection A total of 40 clinical, laboratory, and CECT parameters were analyzed using a stepwise selection strategy. Parameters with p < 0.1 in univariate analysis were subjected to Spearman correlation analysis, followed by the LASSO regression (Supplementary Fig. 1). Thirteen parameters were retained for model development: age, BCLC grade, CNLC grade, FIB-4 index, PNR, NLR, CRP, Td, ECV, tumor size, capsule appearance, intratumoral necrosis, and intratumoral arteries. A correlation heatmap was generated to illustrate associations between these variables and ER (Supplementary Fig. 2). Model construction and evaluation Six ML algorithms—LR, RF, SVM, DT, XGBoost, and GNB—were trained with 10-fold cross-validation on the Center 1 dataset, with external validation in Center 2. Internal validation AUCs were: LR 0.934, RF 0.815, SVM 0.912, DT 0.815, XGBoost 0.913, GNB 0.865; external validation AUCs were 0.769, 0.791, 0.744, 0.650, 0.812, and 0.769, respectively (Fig. 3 A-B; Table 4 ). DCA confirmed clinical utility across threshold probabilities (Fig. 3 C-D). Although LR yielded the highest internal AUC, its external performance was inferior. XGBoost, with the highest external AUC (0.812) and strong internal performance, demonstrated greater robustness and was selected as the optimal model. Calibration analysis showed excellent agreement between predicted and observed outcomes (internal: χ² = 22.867, p = 0.196; external: χ² = 22.060, p = 0.229) (Fig. 3 E-F). Table 4 Performance of six ML algorithms in the internal validation dataset and external validation dataset Dataset AUC (95% CI) Accuracy Sensitivity Specificity PPV NPV F1 score LR Internal 0.934 (0.877–0.991) 0.900 0.907 0.891 0.907 0.891 0.907 External 0.769 (0.517-1.000) 0.773 0.846 0.667 0.786 0.750 0.815 RF Internal 0.815 (0.726–0.904) 0.810 0.889 0.717 0.787 0.846 0.835 External 0.791 (0.587–0.994) 0.818 1.000 0.556 0.765 1.000 0.867 SVM Internal 0.912 (0.857–0.968) 0.810 0.852 0.761 0.807 0.814 0.829 External 0.744 (0.498–0.989) 0.818 0.923 0.667 0.800 0.857 0.857 DT Internal 0.815 (0.726–0.904) 0.790 0.778 0.804 0.824 0.755 0.800 External 0.650 (0.402–0.897) 0.545 0.462 0.667 0.667 0.462 0.546 XGBoost Internal 0.913 (0.861–0.966) 0.830 0.870 0.783 0.825 0.837 0.847 External 0.812 (0.624-1.000) 0.818 0.923 0.667 0.800 0.857 0.857 GNB Internal 0.865 (0.789–0.941) 0.550 0.222 0.935 0.800 0.506 0.348 External 0.769 (0.517-1.000) 0.364 0.000 0.889 0.000 0.381 - - indicates no statistical value Model interpretability SHAP analysis of the XGBoost model (Fig. 4 ) identified the top 10 contributors to ER prediction: ECV, complete capsule, CRP, intratumoral necrosis, FIB-4 index, tumor size, age, PNR, Td, and NLR. Multivariate Cox regression analysis Ten SHAP-derived parameters were entered into multivariate Cox analysis, yielding six independent RFS predictors: ECV ( HR = 0.845, 95% CI 0.803–0.890, p < 0.001), complete capsule ( HR = 0.273, 95% CI 0.151–0.492, p < 0.001), FIB-4 index ( HR = 1.048, 95% CI 1.004–1.094, p = 0.032), tumor size ( HR = 1.020, 95% CI 1.003–1.038, p = 0.020), PNR ( HR = 0.984, 95% CI 0.972–0.996, p = 0.011), and Td ( HR = 1.033, 95% CI 1.008–1.057, p = 0.008) (Table 5 ). Table 5 Multivariate Cox regression analysis of RFS β value HR (95% CI) p ECV -0.146 0.864 (0.823–0.907) <0.001 Complete capsule -1.148 0.317 (0.177–0.568) <0.001 CRP 0.007 1.007 (0.997–1.016) 0.187 Intratumoral necrosis 0.583 1.792 (0.883–3.633) 0.106 FIB-4 index 0.052 1.054 (1.008–1.102) 0.021 Tumor size 0.018 1.018 (1.002–1.036) 0.032 Age 0.012 1.012 (0.990–1.034) 0.296 PNR -0.013 0.987 (0.975–0.998) 0.025 Td 0.031 1.032 (1.008–1.057) 0.010 NLR -0.035 0.965 (0.917–1.017) 0.183 Cox regression nomogram and risk stratification A nomogram incorporating these six predictors was developed to estimate 6, 12, and 18-month recurrence risk after TACE with radiological CR (Fig. 5 A). Risk scores stratified patients into low- (0-33rd percentile), middle- (34th-67th percentile), and high-risk (68th-100th percentile) groups: 34, 33, and 33 patients in Center 1; 8, 7, and 7 patients in Center 2, respectively. Kaplan–Meier curves with log-rank test demonstrated significant RFS differences among subgroups (Center 1: p < 0.0001; Center 2: p = 0.0026) (Fig. 5 B-C). Discussion TACE is a widely used minimally invasive treatment for unresectable HCC, offering tumor control and survival benefits. While ML is often applied to predict TACE treatment response, post-TACE recurrence risk assessment remains underexplored[ 17 ]. To address this, we decided to develop the recurrence assessment model to guide clinical decision-making. Specifically, we created and validated an ER prediction model using the XGBoost algorithm with 13 parameters, including CECT-derived ECV, complete capsule, CRP, et al. The model showed strong performance, with AUC values of 0.913 and 0.812 in internal and external validation datasets. Additionally, we constructed a Cox regression nomogram to predict post-TACE recurrence risk, successfully stratifying patients into three subgroups, demonstrating its potential for clinical implementation. Previous studies reported 1-year recurrence or progression rates post-TACE ranging from 23% to 76%, while our study observed rates of 54%-59%, with a median recurrence time of 304–326 days[ 10 , 11 , 18 – 20 ]. Furthermore, we found that over 50% of ER cases involved local recurrence, whereas more than 75% of late recurrence cases developed intrahepatic or distant metastases. ER likely arises from undetectable residual or satellite tumors, while late recurrence may result from de novo tumors in cirrhotic livers or underlying liver disease. This supports Ilagan's view that radiologic CR does not equate to pathologic CR, with local progression often preceding distant progression[ 19 ]. We identified 13 recurrence-associated predictors in the XGBoost model and used SHAP analysis to interpret their relationship with ER. Among these, CECT-derived ECV had the highest contribution, underscoring its critical role in post-TACE recurrence assessment. Our findings showed that low ECV values were associated with higher ER risk, and ECV was an independent predictor of RFS. Previous studies support ECV as a predictive imaging biomarker for HCC. Fu et al reported longer PFS and OS in patients with higher tumor ECV receiving immune checkpoint inhibitors[ 21 ]. Li et al integrated ECV into a LightGBM model for non-invasive pathological grading[ 22 ]. ECV has also been used to stage liver fibrosis and predict liver-related events such as post-hepatectomy liver failure[ 23 , 24 ]. By quantifying extracellular matrix, ECV reflects key tumor microenvironment features—including angiogenesis, fibroblast activation, immune infiltration, and tumor-stromal interactions—that influence treatment response and prognosis[ 25 , 26 ], though the underlying mechanisms require further clarification. Complete capsule and intratumoral necrosis were also pivotal qualitative CT features. The HCC capsule, composed of an outer collagen fiber layer and inner neovascularization, may limit tumor invasion and enhance embolization efficacy, explaining its association with lower ER risk, consistent with Lan et al[ 27 ]. In contrast, tumors with necrotic components exhibited higher ER risk, likely reflecting aggressive biology, ischemia and hypoxia-induced necrosis, and impaired anti-tumor immunity[ 28 , 29 ]. Tumor size was another significant predictor, as larger tumors correlate with higher tumor burden, microvascular invasion, and advanced stage[ 30 ]. Notably, tumor number was not considered a predictive factor in this study, whereas Cerban et al identified multiple tumors as a risk factor for recurrence after TACE treatment[ 11 ]. The reason may be that our study mainly included early-stage single-nodule HCC, which did not reflect the impact of multinodular lesions on recurrence. Clinical and laboratory markers are key in predictive model development. CRP, an acute-phase protein produced by hepatocytes, reflects tumor-induced inflammation and correlates with tumor progression[ 31 ]. We found higher CRP levels linked to increased ER risk. Ota et al. also identified CRP as an independent risk factor for overall and recurrence-free survival post-hepatectomy, showing that CRP-based ML models aid in effective risk stratification[ 32 ]. Low PNR and high NLR, indicative of chronic inflammation, were associated with higher ER risk, as elevated neutrophil levels promote a pro-tumor microenvironment[ 33 ]. The FIB-4 index, which gauges liver fibrosis, predicts ER and RFS post-TACE. Recent studies have shown FIB-4 can predict recurrence in HBV-related HCC after stereotactic body radiation therapy[ 34 ]. Although serum AFP is a key HCC marker, we found no significant difference in AFP levels between ER and non-ER groups, likely due to the inclusion of early-stage patients with smaller tumors. Previous studies have suggested serum AFP > 20 ng/mL predicts recurrence after TACE, warranting further research[ 10 ]. Although ML models achieve strong predictive performance, their reliance on specialized computational tools. In contrast, nomograms—visual tools derived from multivariate statistical analysis—are more accessible and practical. In this study, we performed multivariate Cox regression on key variables identified by SHAP analysis of the XGBoost model to determine independent predictors of RFS. A Cox regression-based nomogram was then developed to estimate recurrence risk at 6, 12, and 18 months after TACE achieving radiological CR, offering a more comprehensive survival prediction than the ER model. The nomogram stratified patients into low-, middle-, and high-risk groups, enabling tailored clinical management. High-risk patients may benefit from early, intensive interventions such as adjuvant resection, combination ablation, or more frequent surveillance with shorter follow-up intervals. In contrast, follow-up for low- and middle-risk patients may be extended, with greater emphasis on quality of life. However, this does not mean the lesions will remain in CR status, as late recurrence is more likely to involve new lesions in the liver or distant organs like the lungs or bones. Clinicians should therefore closely evaluate liver regions beyond 1 cm from the primary lesion and screen organs susceptible to distant spread[ 35 ]. This dual-center study, despite differences in equipment and scanning parameters, demonstrated good performance and effective risk stratification in the external validation set, confirming the model's generalizability. However, there are several limitations. First, the small sample size due to the limited number of high-quality cases achieving radiological CR after TACE warrants further validation. Second, CR determination based on CECT may be suboptimal due to dense lipiodol interference. Third, the study focused only on TACE monotherapy, excluding patients who received combination therapies like ablation, radiotherapy, targeted therapy, or immunotherapy, which may help reduce recurrence risk. Fourth, manual ROI measurement for CT quantitative parameters may affect robustness. Although senior radiologists validated measurements with ICC values above 0.9, future studies should adopt automated segmentation or advanced techniques like dual-energy CT iodine density or MRI T1 mapping for more stable tumor parameter values[ 36 ]. In conclusion, we developed and externally validated an ER prediction model based on the XGBoost algorithm and a Cox regression nomogram to predict recurrence after TACE achieving radiological CR in unresectable HCC patients. These easily obtainable clinical and imaging parameters, along with the visual model, offer a reliable, practical solution for identifying high-risk recurrence patients and guiding personalized management. Abbreviations AFP Alpha-fetoprotein ALBI Albumin-bilirubin AUC Area under the curve BCLC Barcelona Clinic Liver Cancer CECT Contrast-enhanced CT CNLC China Liver Cancer Staging CR Complete response ECV Extracellular volume fraction ER Early recurrence GNB Gaussian naive Bayes HCC Hepatocellular carcinoma ICC Intraclass correlation coefficient LASSO Least absolute shrinkage and selection operator LR Logistic regression RF Random forest RFS Recurrence free survival ROI Region of interest SHAP Shapley Additive Explanations SVM Support vector machine TACE Transarterial chemoembolization XGBoost Extreme gradient boosting Declarations Author Contribution LHL, YYL, DXL, KM, and XMY were involved in the design of the article study. LHL, DXL, JLZ, and XRN contributed to the collection, processing, analysis, or interpretation of the data. LHL, CLY, and ZZ contributed to the drafting and revision of the manuscript. All authors read and approved the final version of the manuscript. Acknowledgement The authors acknowledge the expert team of Deepwise Intelligent Technology Co. for their technical support. Data Availability The datasets uesd or analysed during this study are available from the corresonding author on reasonable request. References Bray F, Laversanne M, Sung H et al (2024) Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 74:229–263 Hwang SY, Danpanichkul P, Agopian V et al (2025) Hepatocellular carcinoma: updates on epidemiology, surveillance, diagnosis and treatment. Clin Mol Hepatol 31:S228–S254 Torimura T, Iwamoto H (2022) Treatment and the prognosis of hepatocellular carcinoma in Asia. Liver Int 42:2042–2054 Llovet JM, Pinyol R, Yarchoan M et al (2024) Adjuvant and neoadjuvant immunotherapies in hepatocellular carcinoma. Nat Rev Clin Oncol 21:294–311 Zhong BY, Jin ZC, Chen JJ, Zhu HD, Zhu XL (2023) Role of transarterial chemoembolization in the treatment of hepatocellular carcinoma. J Clin Transl Hepatol 11:480–489 Rou WS, Lee BS, Moon HS, Lee ES, Kim SH, Lee HY (2014) Risk factors and therapeutic results of early local recurrence after transcatheter arterial chemoembolization. World J Gastroenterol 20:6995–7004 Jiang H, Yang C, Chen Y et al (2023) Development of a model including MRI features for predicting advanced-stage recurrence of hepatocellular carcinoma after liver resection. Radiology 309:e230527 Kraglund F, Skou N, Villadsen GE, Jepsen P (2024) Landmark analysis of the risk of recurrence after resection or ablation for HCC: A nationwide study. Hepatol Commun 8:e0472 Qin S, Chen M, Cheng AL et al (2023) Atezolizumab plus bevacizumab versus active surveillance in patients with resected or ablated high-risk hepatocellular carcinoma (IMbrave050): a randomised, open-label, multicentre, phase 3 trial. Lancet 402:1835–1847 Jeong SO, Kim EB, Jeong SW et al (2017) Predictive Factors for Complete Response and Recurrence after Transarterial Chemoembolization in Hepatocellular Carcinoma. Gut Liver 11:409–416 Cerban R, Ester C, Iacob S et al (2018) Predictive Factors of Tumor Recurrence and Survival in Patients with Hepatocellular Carcinoma treated with Transarterial Chemoembolization. J Gastrointestin Liver Dis 27:409–417 Asano K, Kageyama K, Yamamoto A et al (2023) Transcatheter arterial chemoembolization for treatment-naive hepatocellular carcinoma has different treatment effects depending on central or peripheral tumor location. Liver Cancer 12:576–589 Sheng Y, Wang Q, Liu HF, Chen WH, He ZM, Wang Q (2023) Preoperative nomogram incorporating clinical factors, serological markers and LI-RADS MRI features to predict early recurrence of hepatocellular carcinoma treated with transarterial chemoembolization. Acad Radiol 30:1288–1297 Liu W, Wei R, Chen J et al (2024) Prognosis prediction and risk stratification of transarterial chemoembolization or intraarterial chemotherapy for unresectable hepatocellular carcinoma based on machine learning. Eur Radiol 34:5094–5107 Singal AG, Llovet JM, Yarchoan M et al (2023) AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology 78:1922–1965 Llovet JM, Lencioni R (2020) mRECIST for HCC: Performance and novel refinements. J Hepatol 72:288–306 Soni L, Soopramanien J, Acharya A, Ashrafian H, Giannarou S, Fotiadis N, Darzi A (2025) The use of machine learning in transarterial chemoembolisation/transarterial embolisation for patients with intermediate-stage hepatocellular carcinoma: a systematic review. Radiol Med 130:1124–1138 Young S, Sanghvi T, Sharma S et al (2022) Local recurrence following complete radiologic response in patients treated with transarterial chemoembolization for hepatocellular carcinoma. Diagn Interv Imaging 103:143–149 Ilagan CH, Goldman DA, Gönen M et al (2022) Recurrence of hepatocellular carcinoma after complete radiologic response to trans-arterial embolization: A retrospective study on patterns, treatments, and prognoses. Ann Surg Oncol 29:6815–6826 Kinugasa H, Nouso K, Takeuchi Y et al (2012) Risk factors for recurrence after transarterial chemoembolization for early-stage hepatocellular carcinoma. J Gastroenterol 47:421–426 Fu X, Guo Y, Zhang K et al (2025) Prognostic impact of extracellular volume fraction derived from equilibrium contrast-enhanced CT in HCC patients receiving immune checkpoint inhibitors. Sci Rep 15:13643 Li J, Zou L, Ma H et al (2024) Interpretable machine learning based on CT-derived extracellular volume fraction to predict pathological grading of hepatocellular carcinoma. Abdom Radiol (NY) 49:3383–3396 Peng Y, Shen H, Tang H et al (2022) Nomogram based on CT-derived extracellular volume for the prediction of post-hepatectomy liver failure in patients with resectable hepatocellular carcinoma. Eur Radiol 32:8529–8539 Peng Y, Tang H, Huang Y et al (2023) CT-derived extracellular volume and liver volumetry can predict posthepatectomy liver failure in hepatocellular carcinoma. Insights Imaging 14:145 Maehara J, Masugi Y, Abe T et al (2020) Quantification of intratumoral collagen and elastin fibers within hepatocellular carcinoma tissues finds correlations with clinico-patho-radiological features. Hepatol Res 50:607–619 Hu Q, Zhu Y, Mei J, Liu Y, Zhou G (2025) Extracellular matrix dynamics in tumor immunoregulation: from tumor microenvironment to immunotherapy. J Hematol Oncol 18:65 Lan YF,Chen LX,Tang LL,Pan X,Wu SJ,Fang ZT (2025) MRI imaging features for evaluating the efficacy and prognosis of TACE combined with targeted immunotherapy in unresectable hepatocellular carcinoma. Eur J Radiol 191:112296 Yen YH, Kuo FY, Eng HL et al (2023) Tumor necrosis as a predictor of early tumor recurrence after resection in patients with hepatoma. PLoS One 18:e0292144 Hou J, Zhao R, Xia W et al (2020) PD-L1-mediated gasdermin C expression switches apoptosis to pyroptosis in cancer cells and facilitates tumour necrosis. Nat Cell Biol 22:1264–1275 Shinkawa H, Tanaka S, Kabata D et al (2021) The prognostic impact of tumor differentiation on recurrence and survival after resection of hepatocellular carcinoma is dependent on tumor size. Liver Cancer 10:461–472 Okumura T, Kimura T, Iwadare T et al (2023) Prognostic significance of C-reactive protein in Lenvatinib-treated unresectable hepatocellular carcinoma: A multi-institutional study. Cancers (Basel) 15:5343 Ota M, Komeda K, Iida H et al (2023) The prognostic value of preoperative serum markers and risk classification in patients with hepatocellular carcinoma. Ann Surg Oncol 30:2807–2815 Geh D, Leslie J, Rumney R, Reeves HL, Bird TG, Mann DA (2022) Neutrophils as potential therapeutic targets in hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol 19:257–273 Zeng Z, Han Y, Li W, Chen H, Lin N, Yu Y, Xu X (2025) Baseline FIB-4 may be a risk factor of recurrence after SBRT in patients with HBV-related small HCC. Cancer Med 14:e70535 Hsiao CY, Ho CM, Ho MC, Cheng HY, Wu YM, Lee PH, Hu RH (2024) Risk factors, patterns, and outcome predictors of late recurrence in patients with hepatocellular carcinoma after curative resection: A large cohort study with long-term follow-up results. Surgery 176:2–10 Li S, Zhou D, Sirajuddin A et al (2022) T1 mapping and extracellular volume fraction in dilated cardiomyopathy: A prognosis study. JACC Cardiovasc Imaging 15:578–590 Additional Declarations No competing interests reported. Supplementary Files Supplementaryappendix.docx Cite Share Download PDF Status: Published Journal Publication published 03 Mar, 2026 Read the published version in Abdominal Radiology → Version 1 posted Editorial decision: Revision requested 15 Jan, 2026 Reviews received at journal 08 Jan, 2026 Reviewers agreed at journal 07 Jan, 2026 Reviewers agreed at journal 04 Jan, 2026 Reviews received at journal 03 Oct, 2025 Reviewers agreed at journal 23 Sep, 2025 Reviewers invited by journal 21 Sep, 2025 Editor assigned by journal 19 Sep, 2025 Submission checks completed at journal 19 Sep, 2025 First submitted to journal 14 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7611742","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":523597944,"identity":"9a51c692-4cba-4943-ad60-9eb24e98de87","order_by":0,"name":"luhao liu","email":"","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"luhao","middleName":"","lastName":"liu","suffix":""},{"id":523597945,"identity":"aee2d33f-a10f-4270-b139-b7aa07a5637f","order_by":1,"name":"Yiyang Liu","email":"","orcid":"","institution":"First Affiliated Hospital of Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Yiyang","middleName":"","lastName":"Liu","suffix":""},{"id":523597946,"identity":"71e89b15-8be8-486b-938a-fb1db4cf6aaa","order_by":2,"name":"Dongxi Lin","email":"","orcid":"","institution":"Hunan Provincial People's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Dongxi","middleName":"","lastName":"Lin","suffix":""},{"id":523597949,"identity":"a00bfe22-a317-43aa-bd10-f0279429b645","order_by":3,"name":"Ke Meng","email":"","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ke","middleName":"","lastName":"Meng","suffix":""},{"id":523597950,"identity":"b8dae433-e000-4773-be12-9e01d99c304b","order_by":4,"name":"Xiaoman Yang","email":"","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xiaoman","middleName":"","lastName":"Yang","suffix":""},{"id":523597952,"identity":"01d11395-e1b9-4daf-a734-c715e98ea380","order_by":5,"name":"Jiliang Zhou","email":"","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jiliang","middleName":"","lastName":"Zhou","suffix":""},{"id":523597955,"identity":"347cfdf6-36e3-4a0c-ad3f-f813d03234e1","order_by":6,"name":"Xinrui Ni","email":"","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xinrui","middleName":"","lastName":"Ni","suffix":""},{"id":523597956,"identity":"1c262b18-ff6c-4963-9cc1-0f478895f236","order_by":7,"name":"Chunlai Yu","email":"","orcid":"","institution":"Huanghe Science and Technology College","correspondingAuthor":false,"prefix":"","firstName":"Chunlai","middleName":"","lastName":"Yu","suffix":""},{"id":523597958,"identity":"f27fb616-68de-420c-bacf-3ec70cd2c281","order_by":8,"name":"Zhou Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4ElEQVRIie3PMQrCMBTG8VcKdSm6poh4hYggCkGvkhJ4UwePUCnURXCt6CG6qVs0WJceQMHZ2VHBwda5RN0c8odAhvcbPgCT6R8jxbNCgNF8erzfKBt+TyjJuZeMUfxAgNOme9uXf33tZZR5jw3z1yB5l1FpQ00dUh2xVhkSL0d/G02kCOilDi7iSUdsEvRIJ1Z+KCVXAb3aQNyeljgl8d+Ed6I+VVb4ibgl2RUkLabY8A0hBMVgEmN3keRozSgK59OWdiJ250fMWvPGNIP7kw0bNZVpScW6385NJpPJVNULuqhM3tY1LHMAAAAASUVORK5CYII=","orcid":"","institution":"First Affiliated Hospital of Henan University of Traditional Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Zhou","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2025-09-14 09:38:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7611742/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7611742/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00261-026-05436-x","type":"published","date":"2026-03-03T15:57:45+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":92802164,"identity":"a1a6935d-9494-4a4f-804e-2a92d0b93c59","added_by":"auto","created_at":"2025-10-05 11:40:41","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1029241,"visible":true,"origin":"","legend":"","description":"","filename":"Anonymisedmanuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/10a84e531d71a5d2024f64d7.docx"},{"id":92802161,"identity":"4e0f4892-6af0-47a4-8038-dd014c047d9d","added_by":"auto","created_at":"2025-10-05 11:40:40","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9854,"visible":true,"origin":"","legend":"","description":"","filename":"1545396f08774493a45883f3d09634dc.json","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/65af59751ee3334bae1364c7.json"},{"id":92800520,"identity":"7c645cf3-f4ad-403d-94bc-fbd9d09fb00f","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":550203,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryappendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/788cdb64054cb77e740d4493.docx"},{"id":92800522,"identity":"a5096576-2325-4a9b-9b66-6bc4fa3589f6","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136653,"visible":true,"origin":"","legend":"","description":"","filename":"1545396f08774493a45883f3d09634dc1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/0755be2297e64f4d2fdd173c.xml"},{"id":92800531,"identity":"4be6d8f4-36f6-4126-95df-e8328055dc93","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":135333,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/ce6069d88445894d52f5381c.png"},{"id":92800524,"identity":"770af075-65cf-4fdb-9913-23325565899b","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":169105,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/05ed1329b56da082e27cb2e8.png"},{"id":92802179,"identity":"2827b77e-ae68-49bd-8569-291ca54dfb6d","added_by":"auto","created_at":"2025-10-05 11:40:41","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":87201,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/a81215be1cb19038aafd5462.png"},{"id":92800534,"identity":"10183b3b-562a-423a-b7c7-6950adbc93a7","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30572,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/26f75aec4f53fd5ffe577fa2.png"},{"id":92802177,"identity":"199ba287-1728-4a62-a0cc-a4952866ac85","added_by":"auto","created_at":"2025-10-05 11:40:41","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":131973,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/2fcf7e5c612991f6467215b2.png"},{"id":92800525,"identity":"a626a757-f32b-47ac-95ea-8b3057f8cf42","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134733,"visible":true,"origin":"","legend":"","description":"","filename":"1545396f08774493a45883f3d09634dc1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/c19705ba536d51d5c9c12142.xml"},{"id":92800527,"identity":"9ab24188-073c-40d3-b95b-edca35d56239","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":141245,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/140f661d0a03ff8afd451f8d.html"},{"id":92800517,"identity":"6a236d61-4352-46e4-bf4d-943980f0107e","added_by":"auto","created_at":"2025-10-05 11:32:40","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":303239,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of patients selection and study design\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/d25babc55fb718d2ddf3d015.jpeg"},{"id":92802166,"identity":"adad1f87-e5b4-414e-b816-49547e887680","added_by":"auto","created_at":"2025-10-05 11:40:41","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":575086,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic diagram of ROI delineation for quantitative tumor and aortic CT value measurement\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/9b21e0485b70752b40390dda.jpeg"},{"id":92800530,"identity":"4a86619d-b457-46c3-af7e-c2028d884d69","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":463221,"visible":true,"origin":"","legend":"\u003cp\u003eEvaluation of ER models based on six ML algorithms. ROC curves for the internal (A) and external validation dataset (B). DCA for the internal (C) and external validation dataset (D). Calibration curves of XGBoost model in the internal (E) and external validation dataset (F)\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/c33a0930ea581dbae9e7bf7e.jpeg"},{"id":92800526,"identity":"b37740f8-ef96-41c0-8df5-3ef2a9f31ce9","added_by":"auto","created_at":"2025-10-05 11:32:41","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":143755,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP summary plot of the XGBoost model\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/3b3e1d4c5c71cd13b920b80c.jpeg"},{"id":92802176,"identity":"053b00a3-0945-4a6e-9e75-d01e3f04ca19","added_by":"auto","created_at":"2025-10-05 11:40:41","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":729703,"visible":true,"origin":"","legend":"\u003cp\u003eThe Cox regression nomogram for estimating the 6, 12, and 18-month recurrence risk in HCC patients after TACE achieving radiological CR (A). Kaplan–Meier survival curves of RFS stratified by low-, middle-, and high-risk groups in the internal (B) and external validation dataset (C)\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/dee39d1dac3a4b788b19721e.jpeg"},{"id":104252282,"identity":"fe6cd860-f093-4476-91ca-d20fa0a33e67","added_by":"auto","created_at":"2026-03-09 16:17:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3409737,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/6ed74eb0-544b-4944-93c3-ed8f6fdc6087.pdf"},{"id":92802946,"identity":"9f82477b-a3f9-41cf-af1d-42625279505d","added_by":"auto","created_at":"2025-10-05 11:48:40","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":550203,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryappendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7611742/v1/ca40efe1fe374c022992bd83.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Early recurrence prediction and risk stratification of hepatocellular carcinoma after transarterial chemoembolization achieving radiological complete response based on contrast-enhanced CT machine learning","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLiver cancer is the sixth most common cancer and the third leading cause of cancer-related mortality globally[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Hepatocellular carcinoma (HCC) is the dominant type of primary liver cancer. HCC progresses insidiously, with 70% of patients diagnosed at intermediate or advanced stages, resulting in a 5-year survival rate of only 20.8%[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Although 30% of patients may benefit from curative treatments like surgical resection, transplantation, or ablation, transarterial chemoembolization (TACE) remains a widely used option for those ineligible for these approaches due to various contraindications[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eRecent developments in TACE have emphasized precision TACE, which involves superselective catheterization of tumor-feeding arteries to minimize damage to healthy liver tissue. For non-massive or non-diffuse HCC, precision TACE aims for complete (CR) or substantial partial response (PR) based on the modified Response Evaluation Criteria in Solid Tumors (mRECIST)[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Despite its effectiveness, recurrence remains a concern, which is often overlooked. A study of 134 patients achieving CR after TACE found a 1-year local recurrence rate of 44%, with 3- and 5-year survival rates of 40% and 27%, respectively, for those who recurred within 1 year[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. These results underscore the need for accurate identification of high-risk patients, even those with radiological CR, to guide treatment and maximize survival.\u003c/p\u003e\u003cp\u003eMost studies focus on recurrence and risk factors post-HCC resection or ablation, with limited evidence on recurrence following TACE[\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Some factors, such as tumor number, size, location, and serum alpha-fetoprotein (AFP), have been linked to recurrence after TACE[\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. However, these studies mainly rely on clinical data and do not fully explore the predictive value of imaging features. Recent research has shown that several MRI features can predict ER after TACE[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Yet, conventional statistical methods in such studies may not meet the growing demand for more advanced predictive models.\u003c/p\u003e\u003cp\u003eThe rapid development of artificial intelligence in medical imaging has highlighted the power of machine learning (ML) in capturing complex, nonlinear relationships in high-dimensional datasets. These algorithms offer enhanced predictive accuracy and generalization, making them ideal for model development. Liu et al demonstrated the effectiveness of ML models in risk stratification following intra-arterial therapies for HCC[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTherefore, our study aimed to develop and validate predictive models using six ML algorithms based on clinical and contrast-enhanced CT (CECT) parameters to identify ER after TACE with radiological CR. Additionally, we constructed a Cox regression nomogram incorporating recurrence free survival (RFS) to enable robust risk stratification, offering a personalized, imaging-guided framework for patient management.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy population\u003c/h2\u003e\u003cp\u003eThis retrospective, dual-center study was approved by the institutional review board (No.\u0026times;\u0026times;) with informed consent waived. The development cohort comprised 100 HCC patients who underwent initial TACE and achieved radiological CR at \u0026times;\u0026times;Hospital (Center 1) between March 2015 and August 2023. The external validation cohort included 22 patients from \u0026times;\u0026times;Hospital (Center 2) between August 2022 and August 2023, meeting identical eligibility criteria.\u003c/p\u003e\u003cp\u003eInclusion criteria were: (1) HCC diagnosis according to the American Association for the Study of Liver Diseases (AASLD) guidelines[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]; (2) liver CECT examination within 2 weeks prior to TACE; (3) Eastern Cooperative Oncology Group performance status (ECOG PS) 0\u0026ndash;2; and (4) standardized conventional TACE with post-treatment radiological CR. The TACE protocol is detailed in Supplementary Text. Exclusion criteria and the patient selection flowchart are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eCT protocol\u003c/h3\u003e\n\u003cp\u003eCT examinations was performed using Siemens dual-source scanners (Somatom Definition, Siemens Healthcare, Germany) or GE spectral CT scanners (Discovery CT 750 HD, GE Healthcare, USA). Imaging included non-contrast (NP), arterial (AP), portal venous (PVP), and delayed (DP) phases. After NP scanning, iodinated contrast (350 mgI/mL, GE Healthcare, China) was administered intravenously at 1.5 mL/kg and 3.5 mL/s, followed by a 40-mL saline flush. Detailed scanning parameters are provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eScanning parameters and acquisition protocols of two center\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCenter 1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCenter 2\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCT scanner\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSomatom Definition CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDiscovery CT 750 HD\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eManufacture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSiemens Healthcare\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGE Healthcare\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTube voltage\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e120 kVp\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e120 kVp\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTube current\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAutomated Tube Current Modulation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAutomated Tube Current Modulation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDetector configuration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e64 \u0026times; 0.600 mm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e64 \u0026times; 0.625 mm\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePitch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.375\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRotation time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.33 s/r\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.5 s/r\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImage reconstruction thickness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1 mm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.375 mm\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAP initiation time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e20\u0026ndash;25 s\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25\u0026ndash;30 s\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePVP initiation time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e40\u0026ndash;60 s\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60\u0026ndash;70 s\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDP initiation time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e90\u0026ndash;120 s\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e100\u0026ndash;120 s\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eDefinition of radiological CR\u003c/h3\u003e\n\u003cp\u003eFollow-up CECT or contrast-enhanced MRI was performed 4\u0026ndash;6 weeks post-TACE. According to mRECIST criteria, CR was defined as complete disappearance of intratumoral arterial enhancement in all target lesions[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Additional embolization was performed when residual enhancement or incomplete lipiodol deposition was present, repeated until CR was achieved. CR assessments were conducted by a radiologist with over 20 years\u0026rsquo; experience in liver tumor imaging.\u003c/p\u003e\n\u003ch3\u003eDate collection and image analysis\u003c/h3\u003e\n\u003cp\u003eClinical and laboratory parameters included age, sex, hepatitis B (HBV) and C (HCV) infection, ECOG PS, Barcelona Clinic Liver Cancer (BCLC) and China Liver Cancer Staging (CNLC) grades, Child-Pugh grade, albumin-bilirubin (ALBI) score, serum AFP, fibrosis-4 (FIB-4) index, aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin (ALB), total bilirubin (TBIL), C-reactive protein (CRP), prothrombin time (PT), platelet count, neutrophil count, lymphocyte count, neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), platelet-to-neutrophil ratio (PNR), and hematocrit (HCT). The ALBI score was calculated as: (log10 TBIL [\u0026micro;mol/L] \u0026times; 0.66) + (ALB [g/L] \u0026times; -0.085) and classified into three grades: Grade 1 (\u0026le;-2.60), Grade 2 (-2.60 to -1.39), and Grade 3 (\u0026gt;-1.39). The FIB-4 index was computed as: (age \u0026times; AST [U/L]) / (platelet count [\u0026times;10⁹/L] \u0026times; ALT [U/L]\u003csup\u003e1/2\u003c/sup\u003e).\u003c/p\u003e\u003cp\u003eCECT image analysis was performed independently by two radiologists with 3 and 5 years of liver tumor imaging experience. Discrepancies were resolved by a senior radiologist with 12 years of experience. In cases with multiple lesions, the largest tumor was evaluated. Qualitative features assessed included tumor number, location, cirrhosis, ascites, tumor margin, intratumoral necrosis, peritumoral enhancement, capsule appearance, and intratumoral arteries.\u003c/p\u003e\u003cp\u003eFor quantitative analysis, the radiologist with 3 years of experience selected the largest cross-section of the tumor, avoiding the area of vessels, necrosis, and heterogeneous enhancement. A circular region of interest (ROI) measuring 20\u0026ndash;40 mm\u0026sup2; was manually delineated to obtain CT values for each tumor phase. Additionally, aortic CT values were measured by placing an ROI in the abdominal aorta during NP and DP. A schematic of ROI delineation is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Tumor CT values in NP、AP、PVP、DP (T0, Ta, Tp, Td) and aortic CT values in NP and DP (A0, Ad) were recorded. Four CECT-derived quantitative parameters were calculated:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eArterial enhancement ratio (AER) = (Ta - T0) / T0\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003ePortal venous enhancement ratio (PER) = (Tp - T0) / T0\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eArterial enhancement fraction (AEF) = [(Ta - T0) / (Tp - T0)] \u0026times; 100%\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eExtracellular volume fraction (ECV) = (1 - HCT) \u0026times; [(Td - T0) / (Ad - A0)] \u0026times; 100%\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003cp\u003eTumor size was defined as the longest axial measurement on the largest cross-section and recorded in millimeters. To assess reproducibility, a randomly selected subset of 30 cases was reanalyzed by the radiologist with 5 years of experience after a 2-week interval. Interobserver consistency was evaluated using the intraclass correlation coefficient (ICC), with ICC\u0026thinsp;\u0026gt;\u0026thinsp;0.75 indicating excellent consistency.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eFollow-up\u003c/h3\u003e\n\u003cp\u003ePost-treatment surveillance with CECT or contrast-enhanced MRI was performed every 3\u0026ndash;6 months. RFS was defined as the interval from the final TACE achieving radiological CR to imaging-confirmed local recurrence, intrahepatic recurrence, or distant metastasis. The final follow-up date was August 1, 2024. Recurrence date and pattern were recorded, with RFS calculated in days. Patients were categorized by 1-year recurrence status: early recurrence (ER) (RFS\u0026thinsp;\u0026le;\u0026thinsp;365 days) and non-ER (late recurrence (RFS\u0026thinsp;\u0026gt;\u0026thinsp;365 days) or recurrence-free at last follow-up).\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eML modeling workflow\u003c/h2\u003e\u003cp\u003eML modeling was conducted on the \u0026times;\u0026times;Platform following these steps:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eFeature selection: Variables with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.1 in univariate analysis were assessed for multicollinearity via Spearman correlation. For correlated pairs (\u003cem\u003er\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.75), the variable less associated with the outcome was excluded. Remaining features were refined using least absolute shrinkage and selection operator (LASSO) regression.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eModel construction: Using Center 1 data, six ML algorithms\u0026mdash;logistic regression (LR), random forest (RF), support vector machine (SVM), decision tree (DT), extreme gradient boosting (XGBoost), and Gaussian naive Bayes (GNB)\u0026mdash;were trained with 10-fold cross-validation to predict ER after TACE achieving CR. Center 2 served for external validation.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eModel evaluation: Performance was assessed using receiver operating characteristic (ROC) curves and metrics including area under the curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score. Decision curve analysis (DCA) evaluated clinical utility. The best-performing model was selected based on overall performance and calibration.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eInterpretability analysis: Shapley Additive exPlanations (SHAP) analysis identified the most influential predictors in the optimal model.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eCox regression nomogram: A multivariate Cox proportional hazards model incorporating key SHAP-identified parameters was used to develop a nomogram predicting 6, 12, and 18-month recurrence.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eRisk stratification: Patients were classified into low-, middle-, and high-risk groups by nomogram-derived scores, with stratification assessed by Kaplan-Meier analysis and log-rank test.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eAnalyses were performed using SPSS 26.0 (IBM, Armonk, NY, USA) and R 4.4.2 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.r-project.org\u003c/span\u003e\u003cspan address=\"https://www.r-project.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Parameters with \u0026gt;\u0026thinsp;15% missing data were excluded; those with \u0026le;\u0026thinsp;15% missingness were imputed using the RF algorithm. In Center 1, group comparisons between ER and non-ER were as follows: normally distributed continuous variables were expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation and compared with independent \u003cem\u003et\u003c/em\u003e-tests; non-normally distributed data were expressed as median (interquartile range) and compared using Mann-Whitney \u003cem\u003eU\u003c/em\u003e tests; categorical variables were reported as counts (%) and compared using chi-square or Fisher\u0026rsquo;s exact tests. A two-sided \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003ePatient recurrence and baseline characteristics\u003c/h2\u003e\u003cp\u003eAt final follow-up, among 100 patients from Center 1, 54 (54.0%) had ER, 40 (40.0%) late recurrence, and 6 (6.0%) remained recurrence-free. Median RFS was 304 days (95% CI: 248.14-359.86). In the ER group, 34 (62.96%) developed local recurrence and 20 (37.04%) intrahepatic or distant metastases. In late recurrence, 10 (25.0%) were local and 30 (75.0%) intrahepatic or distant metastases. In Center 2, 13 (59.09%) had ER, 2 (9.09%) late recurrence, and 7 (31.82%) no recurrence; median RFS was 326 days (95% CI: 275.44\u0026ndash;376.56). ER in this cohort comprised 7 (53.85%) local and 6 (46.15%) intrahepatic or distant metastases; all late recurrences were intrahepatic or distant metastases.\u003c/p\u003e\u003cp\u003eBaseline characteristics in Center 1 included 76 (76.0%) males and 24 (24.0%) females, aged 30\u0026ndash;78 (57.96\u0026thinsp;\u0026plusmn;\u0026thinsp;10.41) years. Center 2 had 19 (86.36%) males and 3 (13.64%) females, aged 43\u0026ndash;74 years (62.05\u0026thinsp;\u0026plusmn;\u0026thinsp;9.35). No significant intercenter differences were observed for age, sex, HBV or HCV infection, BCLC grade, and CNLC grade (all \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eInterobserver consistency\u003c/h2\u003e\u003cp\u003eICCs for CECT quantitative parameters (T0, Ta, Tp, Td, A0, Ad, AER, PER, AEF, ECV) ranged from 0.945 to 0.998, all \u0026gt;\u0026thinsp;0.75, indicating excellent agreement and measurement reliability.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eUnivariate analysis in the development dataset\u003c/h2\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, ER was associated with advanced CNLC grade (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.032), higher FIB-4 index (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.025), elevated CRP (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), higher neutrophil count (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.049), higher NLR (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003), and lower PNR (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.004). Age also differed significantly (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). No significant differences were found for sex, HBV or HCV infection, Child-Pugh grade, BCLC grade, ALBI score, AFP, AST, ALT, ALB, TBIL, PT, platelet count, lymphocyte count, or PLR (all \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, significant CECT differences included lower Td (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.021), reduced ECV (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), larger tumor size (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), more frequent intratumoral necrosis (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and intratumoral arteries (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.012), and fewer complete capsules (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007) in ER. No differences were noted for T0, Ta, Tp, AER, PER, AEF, tumor number, location, cirrhosis, ascites, tumor margin, or arterial peritumoral enhancement (all \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eUnivariate analysis of clinical and laboratory parameters between ER and non-ER groups in the development dataset\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParameters\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eER (n\u0026thinsp;=\u0026thinsp;54)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNon-ER (n\u0026thinsp;=\u0026thinsp;46)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003et/Z/χ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e60.28\u0026thinsp;\u0026plusmn;\u0026thinsp;10.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e55.24\u0026thinsp;\u0026plusmn;\u0026thinsp;9.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2.475\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.015\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.000\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.985\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e41 (75.93%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35 (76.09%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13 (24.07%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11 (23.91%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHBV infection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e47 (87.04%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e38 (82.61%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.382\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.537\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHCV infection\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 (9.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5 (10.87%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026gt;\u0026thinsp;0.999\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBCLC grade\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3.382\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.066\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42 (77.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e42 (91.30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12 (22.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4 (8.70%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCNLC grade\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e8.420\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.032\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e28 (51.85%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35 (76.09%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIb\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e17 (31.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (17.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIIa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4 (7.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (6.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIIb\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 (9.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChild-Pugh grade\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.756\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.385\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e27 (50.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e27 (58.70%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e27 (50.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19 (41.30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALBI score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.090\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.956\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11 (20.37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9 (19.57%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e37 (68.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31 (67.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6 (11.11%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6 (13.04%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSerum AFP (ng/mL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.418\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.234\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026le;400\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42 (77.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e40 (86.96%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026gt;400\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12 (22.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6 (13.04%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFIB-4 index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.29 (3.15, 9.56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.97 (2.55, 6.25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-2.234\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.025\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAST (U/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44.35 (28.10, 71.06)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e42.10 (30.78, 53.17)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-1.477\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.140\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALT (U/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e39.20 (24.48, 62.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37.11 (24.10, 53.38)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.916\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.359\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eALB (g/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.95 (31.50, 38.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e36.66 (33.50, 38.95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.896\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.370\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTBIL (\u0026micro;mol/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21.40 (15.18, 35.93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25.35 (15.30, 33.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.048\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRP (ng/ml)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.40 (2.23, 24.61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.85 (1.06, 6.69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-3.320\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePT (s)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12.60 (12.00, 13.90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12.78 (11.85, 13.70)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.246\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.806\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlatelet count (\u0026times;10\u003csup\u003e9\u003c/sup\u003e/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e75.00 (56.00, 114.25)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e87.50 (61.00, 119.75)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.955\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.340\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeutrophil count (\u0026times;10\u003csup\u003e9\u003c/sup\u003e/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.12 (2.01, 3.79)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.12 (1.59, 2.15)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-1.967\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.049\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymphocyte count (\u0026times;10\u003csup\u003e9\u003c/sup\u003e/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.85 (0.62, 1.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.85 (0.81, 1.10)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-1.589\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.112\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.87 (2.29, 4.46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.38 (1.56, 2.70)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-2.988\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e95.63 (71.03, 137.37)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e93.28 (62.42, 136.27)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.660\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.509\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePNR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e31.57 (21.72, 44.91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e43.11 (27.99, 61.72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-2.877\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.004\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHBV hepatitis B virus; HCV hepatitis C virus; BCLC Barcelona Clinic Liver Cancer; CNLC China Liver Cancer Staging; ALBI albumin-bilirubin; AFP alpha-fetoprotein; AST aspartate aminotransferase; ALT alanine aminotransferase; ALB albumin; TBIL total bilirubin; CRP C-reactive protein; PT prothrombin time; NLR neutrophil-to-lymphocyte ratio; PLR platelet-to-lymphocyte ratio; PNR platelet-to-neutrophil ratio\u003c/p\u003e\u003cp\u003e\u003csup\u003ea\u003c/sup\u003e \u003cem\u003et\u003c/em\u003e value, \u003csup\u003eb\u003c/sup\u003e \u003cem\u003eZ\u003c/em\u003e value, \u003csup\u003ec\u003c/sup\u003e \u003cem\u003eχ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e value by chi-square test, \u003csup\u003ed\u003c/sup\u003e χ2 value by Fisher's exact test, - indicates no statistical value\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eUnivariate analysis of CECT parameters between ER and non-ER groups in the development dataset\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParameters\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eER (n\u0026thinsp;=\u0026thinsp;54)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNon-ER (n\u0026thinsp;=\u0026thinsp;46)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003et/Z/χ\u003c/em\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eT0 (Hu)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44.19\u0026thinsp;\u0026plusmn;\u0026thinsp;5.61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e44.45\u0026thinsp;\u0026plusmn;\u0026thinsp;6.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.215\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.830\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTa (Hu)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e86.92\u0026thinsp;\u0026plusmn;\u0026thinsp;19.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e87.78\u0026thinsp;\u0026plusmn;\u0026thinsp;22.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.207\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.836\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTp (Hu)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e89.98\u0026thinsp;\u0026plusmn;\u0026thinsp;18.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e89.78\u0026thinsp;\u0026plusmn;\u0026thinsp;20.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.049\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTd (Hu)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e75.57\u0026thinsp;\u0026plusmn;\u0026thinsp;10.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e81.13\u0026thinsp;\u0026plusmn;\u0026thinsp;12.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-2.340\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.021\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAER\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.98\u0026thinsp;\u0026plusmn;\u0026thinsp;0.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.219\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.827\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePER\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.05\u0026thinsp;\u0026plusmn;\u0026thinsp;0.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.02\u0026thinsp;\u0026plusmn;\u0026thinsp;0.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.264\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.792\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAEF (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e99.06\u0026thinsp;\u0026plusmn;\u0026thinsp;38.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e98.90\u0026thinsp;\u0026plusmn;\u0026thinsp;39.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.021\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.983\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eECV (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e21.45\u0026thinsp;\u0026plusmn;\u0026thinsp;5.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e29.33\u0026thinsp;\u0026plusmn;\u0026thinsp;6.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-6.836\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor size (mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e40.00 (28.50, 55.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e22.00 (18.25, 33.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-4.432\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor number\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.513\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.522\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42 (77.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39 (84.78%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6 (11.11%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5 (10.87%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u0026gt;2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6 (11.11%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2 (4.35%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor location\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e4.833\u003csup\u003ed\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.145\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLeft lobe\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12 (22.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (17.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRight lobe\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e38 (70.37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37 (80.43%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCaudate lobe\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0 (0.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (2.17%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJunction area\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4 (7.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCirrhosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e49 (90.74%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e42 (91.30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026gt;\u0026thinsp;0.999\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAscites\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e19 (35.19%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11 (23.91%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.503\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.220\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor margin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1.307\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.253\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22 (40.74%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24 (52.17%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNon-smooth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32 (59.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e22 (47.83%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntratumoral necrosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e26.235\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbsence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e17 (31.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e38 (82.61%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePresence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e37 (68.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (17.39%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eArterial peritumoral enhancement\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.723\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbsence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e49 (90.74%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e43 (93.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePresence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 (9.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (6.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCapsule appearance\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e7.369\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eComplete\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e10 (18.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20 (43.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIncomplete\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44 (81.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26 (56.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntratumoral arteries\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6.355\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.012\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAbsence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e17 (31.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26 (56.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePresence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e37 (68.52%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20 (43.48%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eT0, Ta, TP, Td Tumor CT values of non-contrast phase, arterial phase, portal venous phase and delayed phase; AER arterial enhancement ratio; PER portal venous enhancement ratio; AEF arterial enhancement fraction; ECV extracellular volume fraction\u003c/p\u003e\u003cp\u003eThe note is the same as in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eML feature selection\u003c/h2\u003e\u003cp\u003eA total of 40 clinical, laboratory, and CECT parameters were analyzed using a stepwise selection strategy. Parameters with \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.1 in univariate analysis were subjected to Spearman correlation analysis, followed by the LASSO regression (Supplementary Fig.\u0026nbsp;1). Thirteen parameters were retained for model development: age, BCLC grade, CNLC grade, FIB-4 index, PNR, NLR, CRP, Td, ECV, tumor size, capsule appearance, intratumoral necrosis, and intratumoral arteries. A correlation heatmap was generated to illustrate associations between these variables and ER (Supplementary Fig.\u0026nbsp;2).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eModel construction and evaluation\u003c/h2\u003e\u003cp\u003eSix ML algorithms\u0026mdash;LR, RF, SVM, DT, XGBoost, and GNB\u0026mdash;were trained with 10-fold cross-validation on the Center 1 dataset, with external validation in Center 2. Internal validation AUCs were: LR 0.934, RF 0.815, SVM 0.912, DT 0.815, XGBoost 0.913, GNB 0.865; external validation AUCs were 0.769, 0.791, 0.744, 0.650, 0.812, and 0.769, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA-B; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). DCA confirmed clinical utility across threshold probabilities (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC-D).\u003c/p\u003e\u003cp\u003eAlthough LR yielded the highest internal AUC, its external performance was inferior. XGBoost, with the highest external AUC (0.812) and strong internal performance, demonstrated greater robustness and was selected as the optimal model. Calibration analysis showed excellent agreement between predicted and observed outcomes (internal: χ\u0026sup2; = 22.867, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.196; external: χ\u0026sup2; = 22.060, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.229) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE-F).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance of six ML algorithms in the internal validation dataset and external validation dataset\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"9\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDataset\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAUC (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eF1 score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.934 (0.877\u0026ndash;0.991)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.900\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.891\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.891\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.907\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.769 (0.517-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.773\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.846\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.786\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.750\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.815\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.815 (0.726\u0026ndash;0.904)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.810\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.889\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.717\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.787\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.846\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.835\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.791 (0.587\u0026ndash;0.994)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.556\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.765\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e1.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.912 (0.857\u0026ndash;0.968)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.810\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.852\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.761\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.807\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.814\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.829\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.744 (0.498\u0026ndash;0.989)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.857\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.857\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.815 (0.726\u0026ndash;0.904)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.790\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.778\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.804\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.824\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.755\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.650 (0.402\u0026ndash;0.897)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.545\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.462\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.462\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.546\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.913 (0.861\u0026ndash;0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.830\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.870\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.783\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.825\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.837\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.847\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.812 (0.624-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.923\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.667\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.857\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.857\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGNB\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.865 (0.789\u0026ndash;0.941)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.222\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.935\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.800\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.506\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.348\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExternal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.769 (0.517-1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.364\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.889\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.381\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"9\"\u003e- indicates no statistical value\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eModel interpretability\u003c/h2\u003e\u003cp\u003eSHAP analysis of the XGBoost model (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) identified the top 10 contributors to ER prediction: ECV, complete capsule, CRP, intratumoral necrosis, FIB-4 index, tumor size, age, PNR, Td, and NLR.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eMultivariate Cox regression analysis\u003c/h2\u003e\u003cp\u003eTen SHAP-derived parameters were entered into multivariate Cox analysis, yielding six independent RFS predictors: ECV (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.845, 95% \u003cem\u003eCI\u003c/em\u003e 0.803\u0026ndash;0.890, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), complete capsule (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.273, 95% \u003cem\u003eCI\u003c/em\u003e 0.151\u0026ndash;0.492, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), FIB-4 index (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.048, 95% \u003cem\u003eCI\u003c/em\u003e 1.004\u0026ndash;1.094, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.032), tumor size (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.020, 95% \u003cem\u003eCI\u003c/em\u003e 1.003\u0026ndash;1.038, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.020), PNR (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.984, 95% \u003cem\u003eCI\u003c/em\u003e 0.972\u0026ndash;0.996, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.011), and Td (\u003cem\u003eHR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.033, 95% \u003cem\u003eCI\u003c/em\u003e 1.008\u0026ndash;1.057, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.008) (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eMultivariate Cox regression analysis of RFS\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eβ value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHR (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eECV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.146\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.864 (0.823\u0026ndash;0.907)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eComplete capsule\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-1.148\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.317 (0.177\u0026ndash;0.568)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.007 (0.997\u0026ndash;1.016)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.187\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntratumoral necrosis\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.583\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.792 (0.883\u0026ndash;3.633)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.106\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFIB-4 index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.052\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.054 (1.008\u0026ndash;1.102)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.021\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor size\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.018 (1.002\u0026ndash;1.036)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.032\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.012\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.012 (0.990\u0026ndash;1.034)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.296\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePNR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.013\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.987 (0.975\u0026ndash;0.998)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.025\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTd\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.031\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.032 (1.008\u0026ndash;1.057)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.010\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNLR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.035\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.965 (0.917\u0026ndash;1.017)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.183\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eCox regression nomogram and risk stratification\u003c/h2\u003e\u003cp\u003eA nomogram incorporating these six predictors was developed to estimate 6, 12, and 18-month recurrence risk after TACE with radiological CR (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Risk scores stratified patients into low- (0-33rd percentile), middle- (34th-67th percentile), and high-risk (68th-100th percentile) groups: 34, 33, and 33 patients in Center 1; 8, 7, and 7 patients in Center 2, respectively. Kaplan\u0026ndash;Meier curves with log-rank test demonstrated significant RFS differences among subgroups (Center 1: \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; Center 2: \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.0026) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB-C).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eTACE is a widely used minimally invasive treatment for unresectable HCC, offering tumor control and survival benefits. While ML is often applied to predict TACE treatment response, post-TACE recurrence risk assessment remains underexplored[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. To address this, we decided to develop the recurrence assessment model to guide clinical decision-making. Specifically, we created and validated an ER prediction model using the XGBoost algorithm with 13 parameters, including CECT-derived ECV, complete capsule, CRP, et al. The model showed strong performance, with AUC values of 0.913 and 0.812 in internal and external validation datasets. Additionally, we constructed a Cox regression nomogram to predict post-TACE recurrence risk, successfully stratifying patients into three subgroups, demonstrating its potential for clinical implementation.\u003c/p\u003e\u003cp\u003ePrevious studies reported 1-year recurrence or progression rates post-TACE ranging from 23% to 76%, while our study observed rates of 54%-59%, with a median recurrence time of 304\u0026ndash;326 days[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan additionalcitationids=\"CR19\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Furthermore, we found that over 50% of ER cases involved local recurrence, whereas more than 75% of late recurrence cases developed intrahepatic or distant metastases. ER likely arises from undetectable residual or satellite tumors, while late recurrence may result from de novo tumors in cirrhotic livers or underlying liver disease. This supports Ilagan's view that radiologic CR does not equate to pathologic CR, with local progression often preceding distant progression[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eWe identified 13 recurrence-associated predictors in the XGBoost model and used SHAP analysis to interpret their relationship with ER. Among these, CECT-derived ECV had the highest contribution, underscoring its critical role in post-TACE recurrence assessment. Our findings showed that low ECV values were associated with higher ER risk, and ECV was an independent predictor of RFS. Previous studies support ECV as a predictive imaging biomarker for HCC. Fu et al reported longer PFS and OS in patients with higher tumor ECV receiving immune checkpoint inhibitors[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Li et al integrated ECV into a LightGBM model for non-invasive pathological grading[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. ECV has also been used to stage liver fibrosis and predict liver-related events such as post-hepatectomy liver failure[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. By quantifying extracellular matrix, ECV reflects key tumor microenvironment features\u0026mdash;including angiogenesis, fibroblast activation, immune infiltration, and tumor-stromal interactions\u0026mdash;that influence treatment response and prognosis[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], though the underlying mechanisms require further clarification. Complete capsule and intratumoral necrosis were also pivotal qualitative CT features. The HCC capsule, composed of an outer collagen fiber layer and inner neovascularization, may limit tumor invasion and enhance embolization efficacy, explaining its association with lower ER risk, consistent with Lan et al[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. In contrast, tumors with necrotic components exhibited higher ER risk, likely reflecting aggressive biology, ischemia and hypoxia-induced necrosis, and impaired anti-tumor immunity[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Tumor size was another significant predictor, as larger tumors correlate with higher tumor burden, microvascular invasion, and advanced stage[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Notably, tumor number was not considered a predictive factor in this study, whereas Cerban et al identified multiple tumors as a risk factor for recurrence after TACE treatment[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The reason may be that our study mainly included early-stage single-nodule HCC, which did not reflect the impact of multinodular lesions on recurrence.\u003c/p\u003e\u003cp\u003eClinical and laboratory markers are key in predictive model development. CRP, an acute-phase protein produced by hepatocytes, reflects tumor-induced inflammation and correlates with tumor progression[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. We found higher CRP levels linked to increased ER risk. Ota et al. also identified CRP as an independent risk factor for overall and recurrence-free survival post-hepatectomy, showing that CRP-based ML models aid in effective risk stratification[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Low PNR and high NLR, indicative of chronic inflammation, were associated with higher ER risk, as elevated neutrophil levels promote a pro-tumor microenvironment[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. The FIB-4 index, which gauges liver fibrosis, predicts ER and RFS post-TACE. Recent studies have shown FIB-4 can predict recurrence in HBV-related HCC after stereotactic body radiation therapy[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Although serum AFP is a key HCC marker, we found no significant difference in AFP levels between ER and non-ER groups, likely due to the inclusion of early-stage patients with smaller tumors. Previous studies have suggested serum AFP\u0026thinsp;\u0026gt;\u0026thinsp;20 ng/mL predicts recurrence after TACE, warranting further research[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eAlthough ML models achieve strong predictive performance, their reliance on specialized computational tools. In contrast, nomograms\u0026mdash;visual tools derived from multivariate statistical analysis\u0026mdash;are more accessible and practical. In this study, we performed multivariate Cox regression on key variables identified by SHAP analysis of the XGBoost model to determine independent predictors of RFS. A Cox regression-based nomogram was then developed to estimate recurrence risk at 6, 12, and 18 months after TACE achieving radiological CR, offering a more comprehensive survival prediction than the ER model. The nomogram stratified patients into low-, middle-, and high-risk groups, enabling tailored clinical management. High-risk patients may benefit from early, intensive interventions such as adjuvant resection, combination ablation, or more frequent surveillance with shorter follow-up intervals. In contrast, follow-up for low- and middle-risk patients may be extended, with greater emphasis on quality of life. However, this does not mean the lesions will remain in CR status, as late recurrence is more likely to involve new lesions in the liver or distant organs like the lungs or bones. Clinicians should therefore closely evaluate liver regions beyond 1 cm from the primary lesion and screen organs susceptible to distant spread[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThis dual-center study, despite differences in equipment and scanning parameters, demonstrated good performance and effective risk stratification in the external validation set, confirming the model's generalizability. However, there are several limitations. First, the small sample size due to the limited number of high-quality cases achieving radiological CR after TACE warrants further validation. Second, CR determination based on CECT may be suboptimal due to dense lipiodol interference. Third, the study focused only on TACE monotherapy, excluding patients who received combination therapies like ablation, radiotherapy, targeted therapy, or immunotherapy, which may help reduce recurrence risk. Fourth, manual ROI measurement for CT quantitative parameters may affect robustness. Although senior radiologists validated measurements with ICC values above 0.9, future studies should adopt automated segmentation or advanced techniques like dual-energy CT iodine density or MRI T1 mapping for more stable tumor parameter values[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn conclusion, we developed and externally validated an ER prediction model based on the XGBoost algorithm and a Cox regression nomogram to predict recurrence after TACE achieving radiological CR in unresectable HCC patients. These easily obtainable clinical and imaging parameters, along with the visual model, offer a reliable, practical solution for identifying high-risk recurrence patients and guiding personalized management.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAFP Alpha-fetoprotein\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;ALBI Albumin-bilirubin\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;AUC Area under the curve\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;BCLC Barcelona Clinic Liver Cancer\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;CECT Contrast-enhanced CT\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;CNLC China Liver Cancer Staging\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;CR Complete response\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;ECV Extracellular volume fraction\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;ER Early recurrence\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;GNB Gaussian naive Bayes\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;HCC Hepatocellular carcinoma\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;ICC Intraclass correlation coefficient\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;LASSO Least absolute shrinkage and selection operator\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;LR Logistic regression\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;RF Random forest\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;RFS Recurrence free survival\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;ROI Region of interest\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;SHAP Shapley Additive Explanations\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;SVM Support vector machine\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;TACE Transarterial chemoembolization\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;XGBoost Extreme gradient boosting\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eLHL, YYL, DXL, KM, and XMY were involved in the design of the article study. LHL, DXL, JLZ, and XRN contributed to the collection, processing, analysis, or interpretation of the data. LHL, CLY, and ZZ contributed to the drafting and revision of the manuscript. All authors read and approved the final version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors acknowledge the expert team of Deepwise Intelligent Technology Co. for their technical support.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets uesd or analysed during this study are available from the corresonding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBray F, Laversanne M, Sung H et al (2024) Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 74:229\u0026ndash;263\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHwang SY, Danpanichkul P, Agopian V et al (2025) Hepatocellular carcinoma: updates on epidemiology, surveillance, diagnosis and treatment. Clin Mol Hepatol 31:S228\u0026ndash;S254\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTorimura T, Iwamoto H (2022) Treatment and the prognosis of hepatocellular carcinoma in Asia. Liver Int 42:2042\u0026ndash;2054\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLlovet JM, Pinyol R, Yarchoan M et al (2024) Adjuvant and neoadjuvant immunotherapies in hepatocellular carcinoma. Nat Rev Clin Oncol 21:294\u0026ndash;311\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhong BY, Jin ZC, Chen JJ, Zhu HD, Zhu XL (2023) Role of transarterial chemoembolization in the treatment of hepatocellular carcinoma. J Clin Transl Hepatol 11:480\u0026ndash;489\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRou WS, Lee BS, Moon HS, Lee ES, Kim SH, Lee HY (2014) Risk factors and therapeutic results of early local recurrence after transcatheter arterial chemoembolization. World J Gastroenterol 20:6995\u0026ndash;7004\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiang H, Yang C, Chen Y et al (2023) Development of a model including MRI features for predicting advanced-stage recurrence of hepatocellular carcinoma after liver resection. Radiology 309:e230527\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKraglund F, Skou N, Villadsen GE, Jepsen P (2024) Landmark analysis of the risk of recurrence after resection or ablation for HCC: A nationwide study. Hepatol Commun 8:e0472\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQin S, Chen M, Cheng AL et al (2023) Atezolizumab plus bevacizumab versus active surveillance in patients with resected or ablated high-risk hepatocellular carcinoma (IMbrave050): a randomised, open-label, multicentre, phase 3 trial. Lancet 402:1835\u0026ndash;1847\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJeong SO, Kim EB, Jeong SW et al (2017) Predictive Factors for Complete Response and Recurrence after Transarterial Chemoembolization in Hepatocellular Carcinoma. Gut Liver 11:409\u0026ndash;416\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCerban R, Ester C, Iacob S et al (2018) Predictive Factors of Tumor Recurrence and Survival in Patients with Hepatocellular Carcinoma treated with Transarterial Chemoembolization. J Gastrointestin Liver Dis 27:409\u0026ndash;417\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAsano K, Kageyama K, Yamamoto A et al (2023) Transcatheter arterial chemoembolization for treatment-naive hepatocellular carcinoma has different treatment effects depending on central or peripheral tumor location. Liver Cancer 12:576\u0026ndash;589\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSheng Y, Wang Q, Liu HF, Chen WH, He ZM, Wang Q (2023) Preoperative nomogram incorporating clinical factors, serological markers and LI-RADS MRI features to predict early recurrence of hepatocellular carcinoma treated with transarterial chemoembolization. Acad Radiol 30:1288\u0026ndash;1297\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu W, Wei R, Chen J et al (2024) Prognosis prediction and risk stratification of transarterial chemoembolization or intraarterial chemotherapy for unresectable hepatocellular carcinoma based on machine learning. Eur Radiol 34:5094\u0026ndash;5107\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSingal AG, Llovet JM, Yarchoan M et al (2023) AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology 78:1922\u0026ndash;1965\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLlovet JM, Lencioni R (2020) mRECIST for HCC: Performance and novel refinements. J Hepatol 72:288\u0026ndash;306\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSoni L, Soopramanien J, Acharya A, Ashrafian H, Giannarou S, Fotiadis N, Darzi A (2025) The use of machine learning in transarterial chemoembolisation/transarterial embolisation for patients with intermediate-stage hepatocellular carcinoma: a systematic review. Radiol Med 130:1124\u0026ndash;1138\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYoung S, Sanghvi T, Sharma S et al (2022) Local recurrence following complete radiologic response in patients treated with transarterial chemoembolization for hepatocellular carcinoma. Diagn Interv Imaging 103:143\u0026ndash;149\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIlagan CH, Goldman DA, G\u0026ouml;nen M et al (2022) Recurrence of hepatocellular carcinoma after complete radiologic response to trans-arterial embolization: A retrospective study on patterns, treatments, and prognoses. Ann Surg Oncol 29:6815\u0026ndash;6826\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKinugasa H, Nouso K, Takeuchi Y et al (2012) Risk factors for recurrence after transarterial chemoembolization for early-stage hepatocellular carcinoma. J Gastroenterol 47:421\u0026ndash;426\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFu X, Guo Y, Zhang K et al (2025) Prognostic impact of extracellular volume fraction derived from equilibrium contrast-enhanced CT in HCC patients receiving immune checkpoint inhibitors. Sci Rep 15:13643\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi J, Zou L, Ma H et al (2024) Interpretable machine learning based on CT-derived extracellular volume fraction to predict pathological grading of hepatocellular carcinoma. Abdom Radiol (NY) 49:3383\u0026ndash;3396\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeng Y, Shen H, Tang H et al (2022) Nomogram based on CT-derived extracellular volume for the prediction of post-hepatectomy liver failure in patients with resectable hepatocellular carcinoma. Eur Radiol 32:8529\u0026ndash;8539\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeng Y, Tang H, Huang Y et al (2023) CT-derived extracellular volume and liver volumetry can predict posthepatectomy liver failure in hepatocellular carcinoma. Insights Imaging 14:145\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaehara J, Masugi Y, Abe T et al (2020) Quantification of intratumoral collagen and elastin fibers within hepatocellular carcinoma tissues finds correlations with clinico-patho-radiological features. Hepatol Res 50:607\u0026ndash;619\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu Q, Zhu Y, Mei J, Liu Y, Zhou G (2025) Extracellular matrix dynamics in tumor immunoregulation: from tumor microenvironment to immunotherapy. J Hematol Oncol 18:65\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLan YF,Chen LX,Tang LL,Pan X,Wu SJ,Fang ZT (2025) MRI imaging features for evaluating the efficacy and prognosis of TACE combined with targeted immunotherapy in unresectable hepatocellular carcinoma. Eur J Radiol 191:112296\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYen YH, Kuo FY, Eng HL et al (2023) Tumor necrosis as a predictor of early tumor recurrence after resection in patients with hepatoma. PLoS One 18:e0292144\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHou J, Zhao R, Xia W et al (2020) PD-L1-mediated gasdermin C expression switches apoptosis to pyroptosis in cancer cells and facilitates tumour necrosis. Nat Cell Biol 22:1264\u0026ndash;1275\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShinkawa H, Tanaka S, Kabata D et al (2021) The prognostic impact of tumor differentiation on recurrence and survival after resection of hepatocellular carcinoma is dependent on tumor size. Liver Cancer 10:461\u0026ndash;472\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOkumura T, Kimura T, Iwadare T et al (2023) Prognostic significance of C-reactive protein in Lenvatinib-treated unresectable hepatocellular carcinoma: A multi-institutional study. Cancers (Basel) 15:5343\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOta M, Komeda K, Iida H et al (2023) The prognostic value of preoperative serum markers and risk classification in patients with hepatocellular carcinoma. Ann Surg Oncol 30:2807\u0026ndash;2815\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGeh D, Leslie J, Rumney R, Reeves HL, Bird TG, Mann DA (2022) Neutrophils as potential therapeutic targets in hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol 19:257\u0026ndash;273\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng Z, Han Y, Li W, Chen H, Lin N, Yu Y, Xu X (2025) Baseline FIB-4 may be a risk factor of recurrence after SBRT in patients with HBV-related small HCC. Cancer Med 14:e70535\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHsiao CY, Ho CM, Ho MC, Cheng HY, Wu YM, Lee PH, Hu RH (2024) Risk factors, patterns, and outcome predictors of late recurrence in patients with hepatocellular carcinoma after curative resection: A large cohort study with long-term follow-up results. Surgery 176:2\u0026ndash;10\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi S, Zhou D, Sirajuddin A et al (2022) T1 mapping and extracellular volume fraction in dilated cardiomyopathy: A prognosis study. JACC Cardiovasc Imaging 15:578\u0026ndash;590\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Hepatocellular carcinoma, Transarterial chemoembolization, X-Ray Computed, Machine learning, Recurrence free survival","lastPublishedDoi":"10.21203/rs.3.rs-7611742/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7611742/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e\u003cp\u003eTo develop and validate machine learning (ML) models using clinical and contrast-enhanced CT (CECT) parameters to assess recurrence risk in hepatocellular carcinoma (HCC) after transarterial chemoembolization (TACE) achieving radiological complete response (CR).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003e122 HCC patients who underwent TACE and achieved radiological CR from two centers were divided into the development (n\u0026thinsp;=\u0026thinsp;100) and external validation dataset (n\u0026thinsp;=\u0026thinsp;22). Recurrence free survival (RFS) was tracked, and patients were categorized into early recurrence (ER) and non-ER groups based on a 1-year cutoff. Forty clinical and CECT parameters were collected and screened. Six ML models were constructed and compared using the area under the curve (AUC) and decision curve analysis (DCA). Key parameters were used to construct a Cox regression nomogram and stratify recurrence risk using log-rank test.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThe extreme gradient boosting (XGBoost) model demonstrated the best predictive performance based on 13 parameters, with AUCs of 0.913 and 0.812 for the internal and external validation datasets. SHapley Additive exPlanations (SHAP) analysis identified the top 10 parameters. The Cox regression nomogram was constructed with ECV, complete capsule, FIB-4 index, tumor size, platelet-to-neutrophil ratio, and delayed phase tumor CT value. Log-rank test demonstrated significant risk stratification between the two datasets (both \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThe XGBoost-based ER prediction model identifies 1-year recurrence following TACE with radiological CR. The Cox regression nomogram enables risk stratification, dividing patients into three subgroups.\u003c/p\u003e","manuscriptTitle":"Early recurrence prediction and risk stratification of hepatocellular carcinoma after transarterial chemoembolization achieving radiological complete response based on contrast-enhanced CT machine learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-05 11:32:36","doi":"10.21203/rs.3.rs-7611742/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-15T17:07:36+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-08T06:59:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"242603005694361682470207084817695892894","date":"2026-01-08T04:39:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211042074080837387645127335523722160946","date":"2026-01-04T23:05:08+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-03T14:14:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"80975356957220111741633818575295438505","date":"2025-09-23T13:35:31+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-22T01:50:21+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-19T12:30:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-19T09:10:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"Abdominal Radiology","date":"2025-09-14T09:32:36+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"abdominal-radiology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aima","sideBox":"Learn more about [Abdominal Radiology](http://link.springer.com/journal/261)","snPcode":"261","submissionUrl":"https://submission.springernature.com/new-submission/261/3","title":"Abdominal Radiology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"21105c02-5b95-48f8-a0b0-882662034f6b","owner":[],"postedDate":"October 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-09T16:16:55+00:00","versionOfRecord":{"articleIdentity":"rs-7611742","link":"https://doi.org/10.1007/s00261-026-05436-x","journal":{"identity":"abdominal-radiology","isVorOnly":false,"title":"Abdominal Radiology"},"publishedOn":"2026-03-03 15:57:45","publishedOnDateReadable":"March 3rd, 2026"},"versionCreatedAt":"2025-10-05 11:32:36","video":"","vorDoi":"10.1007/s00261-026-05436-x","vorDoiUrl":"https://doi.org/10.1007/s00261-026-05436-x","workflowStages":[]},"version":"v1","identity":"rs-7611742","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7611742","identity":"rs-7611742","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}