Exploring the Value of a Radiomics Model Based on Hepatobiliary Phase Magnetic Resonance Imaging in Predicting Glypican-3-Positive Hepatocellular Carcinoma

Exploring the Value of a Radiomics Model Based on Hepatobiliary Phase Magnetic Resonance Imaging in Predicting Glypican-3-Positive Hepatocellular Carcinoma | 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 Exploring the Value of a Radiomics Model Based on Hepatobiliary Phase Magnetic Resonance Imaging in Predicting Glypican-3-Positive Hepatocellular Carcinoma Peipei Ye, Li Guo, Guangwen Tong, Youwei Wang, Shua Cui, Sirui Yang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8992745/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background Hepatocellular carcinoma (HCC) is the most common primary liver cancer (70%-80% of cases) with a heavy global disease burden and high postoperative recurrence rate, highlighting the need for improved preoperative diagnosis. Glypican-3 (GPC3) is a key diagnostic/therapeutic biomarker for HCC, but current detection relies on invasive methods (postoperative immunohistochemistry, preoperative needle biopsy) with limitations. Hepatobiliary phase contrast-enhanced MRI (HBP-MRI) and radiomics offer promising noninvasive alternatives. Purpose To explore the value of an HBP-MRI-based radiomics model for noninvasive preoperative prediction of GPC3-positive HCC, and develop an optimized model integrating imaging and clinical features. Methods A total of 151 HCC patients who underwent Gd-EOB-DTPA-enhanced MRI and hepatic resection/needle biopsy (2020–2024) were retrospectively divided into training (n = 107) and validation (n = 44) cohorts (7:3 ratio). Eight hundred fifty radiomic features were extracted via 3D Slicer, filtered by ICC (≥ 0.75), independent sample t-tests (P < 0.05), and LASSO (11 key features retained). Eight machine learning models were constructed; the optimal model was combined with clinical factors (AFP, liver cirrhosis) to form a combined model. Performance was evaluated by ROC curves, calibration curves, and DCA. Results The random forest model performed best (training AUC = 0.918; validation AUC = 0.903). The combined model achieved superior efficacy (training AUC = 0.936; validation AUC = 0.940), outperforming the clinical model (AUC = 0.762) and standalone radiomics model, with the highest net clinical benefit confirmed by DCA. Conclusions The HBP-MRI-based radiomics model enables noninvasive prediction of GPC3-positive HCC. The nomogram integrating radiomic score and clinical features exhibits higher diagnostic efficacy, providing a novel noninvasive tool for personalized HCC management. Hepatocellular carcinoma(HCC) Glypican-3 (GPC3) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Primary liver cancer is the sixth most common cancer globally[ 1 ] and the third leading cause of cancer-related death[ 2 ] in recent years. In China, this malignancy is ranked fourth in incidence and second in mortality, representing a major public health challenge due to its considerable disease burden[ 3 ].Hepatocellular carcinoma (HCC) is the most common form of liver cancer, making up around 70%-80% of all cases[ 4 ]. Consequently, it serves as the primary driver of both disease progression and prognosis in this malignancy[ 5 ]. The clinical outcomes for patients with HCC remain poor, despite ongoing progress in diagnosis and treatment, largely due to a high rate of postoperative recurrence[ 6 ]. Thus, the development of enhanced preoperative diagnostic and prognostic evaluation methods is essential for optimizing treatment outcomes in HCC patients. Recent advances in medical imaging have paved the way for novel approaches to non-invasive HCC assessment. One notable example is hepatobiliary phase contrast-enhanced magnetic resonance imaging (HBP-MRI), which stands out due to its unique capabilities[ 7 ]. The introduction of gadoxetic acid disodium (Gd-EOB-DTPA) as a contrast agent delineates a pronounced imaging disparity between normal hepatic parenchyma and malignant tumors, owing to their differing contrast uptake[ 8 ]. This improves our capacity for the non-invasive diagnosis and prognostic evaluation of HCC. As an emerging diagnostic imaging technology, radiomics capitalizes on advanced computational algorithms to extract and quantify sub-visual biological features from medical images, which underlies its capacity for constructing robust predictive models[ 9 ]. This methodology has shown significant potential in diagnosis, prognostic forecasting, and therapy planning, and has been extensively utilized in research across a spectrum of cancer types. Radiomic models derived from Gd-EOB-DTPA-enhanced MRI offer the capability for early, non-invasive assessment of HCC pathological features, complemented by their utility in qualitative diagnosis, treatment response evaluation, and forecasting postoperative recurrence[ 10 ]. Glypican-3 (GPC3), which is typically undetectable in normal liver tissue, exhibits a highly restricted expression profile characterized by significant overexpression in a majority of HCC cases. This distinct pattern not only underscores its promise as an immunotherapeutic target but also confirms its critical role in the diagnosis, therapeutic management, and prognosis of HCC. Nevertheless, the current clinical detection of GPC3 primarily depends on immunohistochemical analysis of resected postoperative specimens. Furthermore, while preoperative needle biopsy can assess GPC3 status, it is an invasive procedure that fails to comprehensively represent the tumor's inherent heterogeneity[ 11 ]. Consequently, a pressing clinical need exists for a non-invasive technique capable of predicting GPC3 status prior to treatment. This study seeks to assess the efficacy of an HBP-MRI-based radiomics model in the preoperative prediction of GPC3-positive HCC and to explore its clinical utility, thereby aiming to inform novel strategies for HCC diagnosis and precision medicine[ 12 ]. 2. Materials and Methods 2.1 Study Participants This retrospective study has been approved by the Institutional Review Board, which has also granted a waiver of informed consent. A patient cohort was identified from those with imaging features suggestive of primary liver cancer during the period from January 2020 to December 2024 for preliminary inclusion. 2.1.1 Inclusion criteria: (1)Patients with preoperative imaging highly suggestive of HCC; (2)Patients who underwent Gd-EOB-DTPA-enhanced MRI and subsequently received hepatic resection or needle biopsy within two weeks. 2.1.2 Exclusion criteria: (1) Pathological diagnosis other than HCC; (2) Pathology reports lacking essential pathological indicators for inclusion; (3) Tumor diameter less than 1 cm; (4) Images compromised by metal or motion artifacts; (5)History of prior treatment including chemotherapy, radiotherapy, transarterial chemoembolization (TACE), or radiofrequency ablation. Patients were classified into HCC groups based on the pathological records of GPC3 expression status. A total of 151 eligible HCC patients were included in the study. The cohort was divided into a training set (n = 107) and a validation set (n = 44) using stratified randomisation, as illustrated in Fig. 1 . 2.2 Methods 2.2.1 MRI Acquisition All patients underwent scanning using a GE Signa Premier 3.0T MRI scanner, with coverage extending from the diaphragm to the inferior poles of the kidneys. A bolus injection of the Gd-EOB-DTPA contrast agent was administered via a peripheral vein at a dose of 0.025 mmol/kg. The imaging protocol included T2-weighted imaging (T2WI), T2WI with fat suppression, and in-phase and out-of-phase T1-weighted imaging (T1WI). Gd-EOB-DTPA-enhanced multiphase sequences were acquired in the arterial, portal venous, transitional and HBPs. Images for the arterial, portal venous, transitional and HBPs were acquired at 15 seconds, 45 seconds, 50 seconds and 20 minutes after contrast injection, respectively. The axial T1WI parameters before and after enhancement were as follows: repetition time (TR) 3.11 ms, echo time (TE) 1.44 ms, field of view (FOV) 34.0 cm × 34.0 cm, matrix 192 × 384, slice thickness 5.0 mm, flip angle 15°, and acquisition time 12 s. 2.2.2 Clinical and Laboratory Data The following information was obtained from the hospital's information system: demographic characteristics (age, sex), preoperative blood test results (hepatitis B serological status, presence of cirrhosis), and biochemical indicators including alpha-fetoprotein (AFP) level, activated partial thromboplastin time (APTT), total bilirubin (TBIL), serum albumin (ALB), alanine aminotransferase(ALT) and aspartate aminotransferase (AST). 2.2.3 Histopathological Analysis Tissue samples were obtained via surgical resection or needle biopsy and subjected to histopathological evaluation. Cases were classified as GPC3-positive if immunoreactive tumour cells were present in more than 10% of the total tumour area; otherwise, they were classified as GPC3-negative. If the interobserver agreement between two pathologists was below 90%, a consensus was reached through joint re-evaluation and, if necessary, arbitration by a third senior pathologist. 2.2.4 Radiomics Analysis The MR images were corrected for intensity non-uniformity using the N4 algorithm in the RMIT tool for medical imaging, developed by the Jilin Provincial Medical Imaging AI Laboratory. Two abdominal radiologists, each with over three years' experience of abdominal MRI, manually delineated regions of interest (ROI) on HBP images using 3D Slicer software. Although both radiologists were aware of the HCC diagnosis, they were blinded to the GPC3 expression status and other clinical data. In cases of disagreement during ROI delineation, consensus was reached through discussion to ensure accurate and objective image interpretation. 2.2.4.1 Feature Extraction: A total of 850 radiomic features were automatically computed from the segmented tumour volumes using the Radiomics extension in 3D Slicer. 2.2.4.2 Feature Selection: Radiomic features were first screened for reliability based on the intraclass correlation coefficient (ICC) for both inter- and intra-observer variability. Those with ICC values greater than 0.75 were deemed reproducible, and their averaged values across observers were adopted for subsequent modeling. In order to reduce dimensionality and enhance the robustness of predictions, a two-stage feature selection strategy was implemented. The study cohort was then randomly partitioned into a training subset and a validation subset, with a ratio of 7:3. Independent-sample t-tests (p < 0.05) were performed on the training set inorder to filter out preliminary features. The least absolute shrinkage and selection operator (LASSO) algorithm was then used to refine the features further. A multi-stage predictive modeling framework was constructed: Initially, eight machine learning models were built based on the selected features to predict GPC3 status, including random forest (RF), support vector machine (SVM), logistic regression (LR), gradient boosting machine (GBM), eXtreme Gradient Boosting (XGB), Naive Bayes (NB), K-Nearest Neighbors (KNN), and decision tree (DT). Subsequently, through model performance comparison (with the area under the curve (AUC) as the core evaluation index), the best-performing model (random forest model) was selected, and the radiomic score (RadScore) was calculated based on this model. Finally, a combined model integrating the RadScore derived from the random forest model and clinical predictors (AFP, liver cirrhosis) was constructed, and validation was completed in the independent validation cohort (n = 44). The detailed process is shown in Fig. 2 . 2.2.5 Statistical Analysis All statistical analyses were performed using SPSS software (version 24) and R software (version 4.4.3). The Kolmogorov–Smirnov test was employed to evaluate the normality of continuous variables, while Levene's test was utilised to assess the homogeneity of variance. The statistical analysis of continuous variables between groups was conducted using either independent samples t-tests or Mann-Whitney U tests, depending on the fulfilment of the assumption of normality. The assessment of associations between categorical variables was conducted through the utilisation of the chi-square test. Clinical variables that demonstrated statistically significant differences (P < 0.05) in the univariate analyses were directly incorporated into multivariate logistic regression models for further analysis to identify factors independently associated with GPC3 expression. Throughout the analysis, the statistical significance threshold was set at P < 0.05. The predictive performance of the developed models was assessed by receiver operating characteristic (ROC) analysis, with model efficacy further detailed by calculating a suite of metrics encompassing the AUC, accuracy, sensitivity, specificity, and F1 score. 3. Results 3.1 Baseline Patient Characteristics The preoperative clinicopathological characteristics of GPC3-positive and GPC3-negative patients in the training and validation cohorts are summarized in Table 1 . Univariate logistic regression analysis in the training cohort revealed that AFP level and cirrhosis status were significantly associated with GPC3 expression (P < 0.05). All relevant factors were subsequently included in a multivariate logistic regression model. The results identified AFP (P < 0.01) and cirrhosis (P = 0.001) as independent predictors of GPC3 expression, which were subsequently incorporated into the clinical-radiomic model. Table 1 Clinical Characteristics of Patients with Hepatocellular Carcinoma (HCC) Category Training Cohort (n = 107) Validation Cohort (n = 44) gpc3+ gpc3- p gpc3+ gpc3- p n = 85 n = 22 n = 35 n = 9 Age 55.91 ± 10.289 60.64 ± 9.384 0.4 58.91 ± 8.329 54.44 ± 5.199 0.192 Gender 0.78 38,6 0.424 Male 68 17 31 7 Female 17 5 4 2 AFP > 0.01 0.066 AFP < 20 39 20 18 6 20 < AFP 400 18 0 8 0 KI67 30(20,75) 30(10,52) 0.23 20(10,30) 50(8.75,60) 0.262 APTT 30.1(27.9,34.8) 30.65(27.675,33.725) 0.96 30.2(28,35) 33.1(30.45,34.9) 0.147 ALB 36.304 ± 7.6352 39.25 ± 5.6085 0.121 35.7 ± 6.6662 34.767 ± 9.1789 0.399 Total Bilirubin 39.1(20,77) 26.15(14.125,45.5) 0.0502 44(23,150) 35(15.05,66) 0.21 ALT 52(30,108) 36(27.25,57.25) 0.127 61(33,184) 56(22,118.5) 0.33 AST 19(14.95,34.05)) 21.2(14.725,37.1) 0.0997 21.5(16,36.8) 32(18.8,58.5) 0.301 HBV Status 0.754 0.968 Positive 18 4 10 2 Negative 67 18 24 7 Cirrhosis 0.001 0.027 Positive 48 4 24 2 Negative 37 18 20 7 Differentiation 0.053 0.588 Well-differentiated 4 4 3 1 Moderately-Poorly Differentiated 81 18 41 8 Satellite Nodules 0.172 0.34 Positive 12 1 2 1 Negative 73 21 42 8 MVI 0.074 0.626 MVI0 53 19 33 8 MVI1 19 2 9 1 MVI2 13 1 2 0 3.2 Feature Selection and Construction of the MRI Radiomics Model A total of 850 radiomic features were extracted from the hepatobiliary phase (HBP) images of each patient in the training cohort. After excluding features with ICC below 0.75 due to poor stability, 796 features were retained as robust features. These were further screened using independent sample t-tests, resulting in 96 statistically significant features. Finally, 11 key radiomic features were selected through LASSO regression analysis (see Fig. 3). 3.3 Model Performance Evaluation Based on the selected radiomic features, eight machine learning models were constructed to predict GPC3 status. Among the above eight models, the RF model performed the best, with an AUC of 0.918 in the training cohort and 0.903 in the validation cohort; for the specific performance indicators of the remaining seven models and the performance comparison between all models (including the RF model), please refer to the performance radar chart of each model (see Fig. 4 ).In the training cohort, the AUC of the standalone clinical model was 0.758, while the combined model integrating the RadScore calculated by the RF model and clinical factors (AFP, liver cirrhosis) achieved an AUC of 0.936. In the validation cohort, the AUC of the standalone clinical model was 0.762, and the AUC of the combined model was 0.940. The ROC curves of each model are shown in Fig. 5 .A nomogram was developed based on this combined model (Fig. 6), which integrates AFP level, liver cirrhosis status, and RadScore, providing a visual tool for the individualized prediction of GPC3-positive HCC. Decision curve analysis (DCA) was used to evaluate the clinical utility of the clinical model, the RF model (standalone radiomics model), and the combined model (Fig. 7 ). The results showed that within a wide range of threshold probabilities, the combined model provided significantly higher net clinical benefit than the standalone radiomics model (RF model) or the standalone clinical model, confirming its high clinical application value in predicting GPC3-positive HCC. 4. Discussion This study aimed to preoperatively predict GPC3 expression in hepatocellular carcinoma (HCC) by constructing and comparing a clinical model, radiomics models, and a combined model, while systematically evaluating the predictive performance of different modeling architectures and the value of model integration strategies. Notably, the dataset of this study exhibited a significant class imbalance, with the ratio of GPC3-positive to negative samples being 3.72:1. A review of relevant literature revealed that this proportional imbalance is prevalent in GPC3-related studies: Han et al. (2023)[ 12 ] reported a ratio of 3.07:1 for GPC3-positive to negative samples and addressed the issue using Synthetic Minority Over-sampling Technique (SMOTE), whereas Geng et al. (2021)[ 13 ], Zhang et al. (2023)[ 14 ], and Li et al. (2024)[ 15 ]documented imbalanced ratios of 3.08:1, 3.43:1, and 3.78:1, respectively, without implementing any correction techniques for this distribution characteristic. Although this study initially attempted to improve the data distribution imbalance by applying SMOTE to the training set, this approach did not yield the expected results, and thus no synthetic over-sampling techniques were ultimately used to intervene in the data distribution. Analysis yielded 11 critical radiomic features spanning shape, GLDM, GLRLM, GLCM, and NGTDM categories. These features offer multi-perspective quantifications of a tumor's intensity distribution, textural patterns, and spatial heterogeneity, enabling the non-invasive preoperative prediction of GPC3-positive HCC [ 16 ]. The wavelet transform, a method utilized in radiomics analysis, excels at resolving fine local details and high-frequency components within medical images, thereby refining the characterization of tumor heterogeneity [ 17 ]. In alignment with our results, the wavelet transform separates medical images into low-frequency (LF) and high-frequency (HF) constituents. The LF elements primarily encapsulate global semantic content, such as the gross morphology and contour of a tumor, whereas the HF elements depict fine-grained local details, including edges and textural patterns[ 18 ]. The ensemble of eight selected wavelet features was dominated by textural descriptors from GLCM, GLRLM, and GLDM.GLCM reflects the frequency (consistent with the statistical logic of probability) of pixel pairs with the same gray-level value occurring at a certain distance and direction within a ROI, which can indirectly represent tumor phenotype; GLRLM quantifies the distribution of gray-level runs[ 19 ]; GLDM quantifies gray-level dependencies in image[ 19 ][ 20 ], A grey-level dependency is defined as the number of connected voxels within a distance of δ that depend on the central voxel[ 21 ]. In terms of model performance, the random forest model exhibited the best predictive ability among the eight constructed machine learning models, with an AUC of 0.918 in the training cohort and 0.903 in the validation cohort. This outperformed the other seven radiomics models. Notably, the amalgamation of the random forest-calculated RadScore (the optimal radiomics model) with clinical independent risk factors (AFP and liver cirrhosis) translated into a model with refined predictive performance, evidenced by AUCs of 0.936 (training) and 0.940 (validation). The integrated model demonstrated a marked superiority over both the clinical-only model (training AUC: 0.758; validation AUC: 0.762) and the standalone radiomics model. This multimodal integration strategy holds substantial scientific significance: while radiomic features quantify tumor heterogeneity at a sub-visual level, thereby overcoming the limitations of conventional clinical parameters that reflect macroscopic pathological states, clinical variables provide essential biological context, creating a synergistic diagnostic framework. The DCA subsequently validated the superior clinical value of the integrated modeling approach. The results demonstrated that across a broad spectrum of threshold probabilities, the combined model yielded a consistently superior net clinical benefit when benchmarked against both the clinical-only and radiomics-only models[ 22 ]. The findings demonstrate the model's viability for clinical decision support, providing substantial basis for precision intervention selection before surgery and risk-adapted management after resection. The proposed modeling methodology not only strengthens forecasting accuracy but also provides a clinically viable and intelligible guidance instrument, representing a stride forward in precision hepatology. The study's limitations should be considered alongside its contributions. The single-center derivation cohort may limit generalizability due to potential selection bias, an issue that must be addressed through future multi-center, large-scale validation to confirm the nomogram's clinical utility (Fig. 6). This investigation was additionally constrained by its exclusive focus on HBP features, without incorporating complementary multiparametric MRI sequences such as diffusion-weighted imaging (DWI) and T2WI. This omission is particularly relevant given that the apparent diffusion coefficient (ADC) derived from DWI has established correlations with tumor cellularity and could potentially enhance the prediction of GPC3 expression[ 23 ][ 24 ]. 5. Conclusion This research establishes the combined model as a potential preoperative screening tool for biologically aggressive GPC3-positive HCC. Its additional capacity to guide non-surgical therapeutic decisions—particularly regarding targeted or combined immunotherapy—supports the evolution of an integrated precision medicine paradigm that bridges imaging biomarkers, molecular pathology, and standard clinical parameters. Abbreviations ADC Apparent Diffusion Coefficient AFP Alpha-Fetoprotein ALB Serum Albumin ALT Alanine Aminotransferase AUC Area Under the Curve APTT Activated Partial Thromboplastin Time AST Aspartate Aminotransferase DCA Decision Curve Analysis DT Decision Tree DWI Diffusion-Weighted Imaging FOV Field of View GBM Gradient Boosting Machine GLCM Gray Level Co-occurrence Matrix GLDM Gray Level Dependence Matrix GLRLM Gray Level Run Length Matrix GPC3 Glypican-3 Gd-EOB-DTPA Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid HBP-MRI Hepatobiliary Phase contrast-enhanced Magnetic Resonance Imaging HCC Hepatocellular Carcinoma HBV Hepatitis B Virus HF High Frequency ICC Intraclass Correlation Coefficient IRB Institutional Review Board KNN K-Nearest Neighbors LASSO Least Absolute Shrinkage and Selection Operator LF Low Frequency LR Logistic Regression MVI Microvascular Invasion NB Naive Bayes NGTDM Neighborhood Gray Tone Difference Matrix ROI Region of Interest RadScore Radiomic Score RF Random Forest ROC Receiver Operating Characteristic SMOTE Synthetic Minority Over-sampling Technique SPSS Statistical Product and Service Solutions SVM Support Vector Machine TACE Transarterial Chemoembolization TBIL Total Bilirubin TE Echo Time T1WI T1-Weighted Imaging T2WI T2-Weighted Imaging TR Repetition Time XGB eXtreme Gradient Boosting Declarations Ethics approval and consent to participate： This study was approved by the Ethics Committee of The Second Affiliated Hospital of Kunming Medical University. Written informed consent was obtained from all individual participants included in the study. Consent for publication： Not applicable Availability of data and materials： The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Competing Interests ： The authors declare that they have no competing interests. Funding ： Not applicable Authors' contributions ： Peipei Ye : Writing – original draft ,Data curation, formal analysis, table revision Li Guo: Writing – review and editing. Guangwen tong,Youwei wang: Data curation, formal analysis, table revision. Shuai Cui, Sirui Yang: Investigation, experiment assistance. All authors have read and approved the final manuscript. Acknowledgements ： The authors thank the patients and staff of The Second Affiliated Hospital of Kunming Medical University for their support in this study. References Cao M, Xia C, Cao M, Yang F, Yan X, He S, Chen W. Attributable liver cancer deaths and disability-adjusted life years in China and worldwide: profiles and changing trends. Cancer biology Med. 2024;21(8):679–91. Cao G, Liu J, Liu M. Global, regional, and national trends in incidence and mortality of primary liver cancer and its underlying etiologies from 1990 to 2019: results from the global burden of disease study 2019. J Epidemiol Global Health. 2023;13(2):344–60. Cao M, Li H, Sun D, He S, Yan X, Yang F, Chen W. Current cancer burden in China: epidemiology, etiology, and prevention. Cancer biology Med. 2022;19(8):1121–38. McGlynn KA, Petrick JL, El-Serag HB. Epidemiology of hepatocellular carcinoma. Hepatology. 2021;73:4–13. Ke J, Liu Y, Liu F, Cai H, Li X, Zhang Z, Ji B. In-situ-formed immunotherapeutic and hemostatic dual drug-loaded nanohydrogel for preventing postoperative recurrence of hepatocellular carcinoma. J Controlled Release. 2024;372:141–54. Park HJ, Seo N, Kim SY. Current landscape and future perspectives of abbreviated MRI for hepatocellular carcinoma surveillance. Korean J Radiol. 2022;23(6):598. Li XQ, Wang X, Zhao DW, Sun J, Liu JJ, Lin DD, Li HJ. Application of Gd-EOB-DTPA-enhanced magnetic resonance imaging (MRI) in hepatocellular carcinoma. World J Surg Oncol. 2020;18(1):219. Scapicchio C, Gabelloni M, Barucci A, Cioni D, Saba L, Neri E. A deep look into radiomics. Radiol Med. 2021;126(10):1296–311. Chen S, Feng S, Wei J, Liu F, Li B, Li X, Kuang M. Pretreatment prediction of immunoscore in hepatocellular cancer: a radiomics-based clinical model based on Gd-EOB-DTPA-enhanced MRI imaging. Eur Radiol. 2019;29(8):4177–87. Zhou F, Shang W, Yu X, Tian J. Glypican-3: a promising biomarker for hepatocellular carcinoma diagnosis and treatment. Med Res Rev. 2020;38(2):741–67. Gu D, Xie Y, Wei J, Li W, Ye Z, Zhu Z, Li X. MRI-based radiomics signature: a potential biomarker for identifying glypican 3‐positive hepatocellular carcinoma. J Magn Reson Imaging. 2020;52(6):1679–87. Han Z, Dai H, Chen X, Gao L, Chen X, Yan C, Li Y. Delta-radiomics models based on multi-phase contrast-enhanced magnetic resonance imaging can preoperatively predict glypican-3-positive hepatocellular carcinoma. Front Physiol. 2023;14:1138239. Geng Z, Zhang Y, Wang S, Li H, Zhang C, Yin S, Dai Y. Radiomics analysis of susceptibility weighted imaging for hepatocellular carcinoma: exploring the correlation between histopathology and radiomics features. Magn Reson Med Sci. 2021;20(3):253–63. Zhang N, Wu M, Zhou Y, Yu C, Shi D, Wang C, Zhu S. Radiomics nomogram for prediction of glypican-3 positive hepatocellular carcinoma based on hepatobiliary phase imaging. Front Oncol. 2023;13:1209814. Li SQ, Yang CX, Wu CM, Cui JJ, Wang JN, Yin XP. Prediction of glypican-3 expression in hepatocellular carcinoma using multisequence magnetic resonance imaging-based histology nomograms. Quant Imaging Med Surg. 2024;14(7):4436. Masokano IB, Liu W, Xie S, Marcellin DFH, Pei Y, Li W. The application of texture quantification in hepatocellular carcinoma using CT and MRI: a review of perspectives and challenges. Cancer Imaging. 2020;20(1):67. Zhang M, Wang L, Gu L, Ling T, Tao X. (2025). WT-BCP: Wavelet Transform based Bidirectional Copy-Paste for Semi-Supervised Medical Image Segmentation. arXiv preprint arXiv:2504.14445 . Varghese BA, Fields BK, Hwang DH, Duddalwar VA, Matcuk GR Jr, Cen SY. Spatial assessments in texture analysis: what the radiologist needs to know. Front Radiol. 2023;3:1240544. Kim YJ. Machine learning models for sarcopenia identification based on radiomic features of muscles in computed tomography. Int J Environ Res Public Health. 2021;18(16):8710. Varriano G. (2024). Radiomics and Formal Methods: a dynamic duo for medicine in pixels. Piovani D, Sokou R, Tsantes AG, Vitello AS, Bonovas S. August). Optimizing clinical decision making with decision curve analysis: insights for clinical investigators. Healthcare. Volume 11. MDPI; 2023. p. 2244. 16. Zhao, J., Gao, S., Sun, W., Grimm, R., Fu, C., Han, J., … Zeng, M. (2021). Magnetic resonance imaging and diffusion-weighted imaging-based histogram analyses in predicting glypican 3-positive hepatocellular carcinoma. European Journal of Radiology, 139, 109732.. Liu, P., Li, W., Qiu, G., Chen, J., Liu, Y., Wen, Z., … Zhao, Y. (2023). Multiparametric MRI combined with clinical factors to predict glypican-3 expression of hepatocellular carcinoma. Frontiers in Oncology, 13, 1142916.. Rastogi A, A. Changing role of histopathology in the diagnosis and management of hepatocellular carcinoma. World J Gastroenterol. 2020;24(35):4000. Additional Declarations No competing interests reported. Supplementary Files date.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 30 Apr, 2026 Reviews received at journal 27 Apr, 2026 Reviewers agreed at journal 01 Apr, 2026 Reviewers agreed at journal 01 Apr, 2026 Reviewers invited by journal 01 Apr, 2026 Editor assigned by journal 05 Mar, 2026 Submission checks completed at journal 05 Mar, 2026 First submitted to journal 28 Feb, 2026 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-8992745","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":616011695,"identity":"4e69a7a2-d97d-4dfd-bcf0-745ac9e7e2f1","order_by":0,"name":"Peipei Ye","email":"","orcid":"","institution":"Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Peipei","middleName":"","lastName":"Ye","suffix":""},{"id":616011696,"identity":"43b79e59-db10-4e5a-afca-8e7a4b23bb30","order_by":1,"name":"Li Guo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAs0lEQVRIiWNgGAWjYHACxgMJbAwM/MzMhx8QrQesRbKdLc2AeC0MQC0G53kUJIhSLh+RfODAgzIbeePDPAwGDDU20QS1GN5ISziQcC7NcNth3gMPGI6l5TYQ1DIjx+BAYtvhBLPDfAkGjA2HSdBi3MxjIEGUFnkJqBYDZmK1GPA8g/hlxmFgICcQ4xf59uSDD38AQ4y///DhBx9qbIiw5QAyL4GQcrAtBA0dBaNgFIyCUQAANd1BV8ijSW0AAAAASUVORK5CYII=","orcid":"","institution":"Kunming Medical University","correspondingAuthor":true,"prefix":"","firstName":"Li","middleName":"","lastName":"Guo","suffix":""},{"id":616011697,"identity":"70f786f8-666c-44f1-9036-9cdde3a89974","order_by":2,"name":"Guangwen Tong","email":"","orcid":"","institution":"Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Guangwen","middleName":"","lastName":"Tong","suffix":""},{"id":616011698,"identity":"2330bd06-9d5d-4ef8-89dd-38605ef01818","order_by":3,"name":"Youwei Wang","email":"","orcid":"","institution":"Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Youwei","middleName":"","lastName":"Wang","suffix":""},{"id":616011699,"identity":"9dbe1147-a5f6-4deb-b2ff-ed2d00e21e10","order_by":4,"name":"Shua Cui","email":"","orcid":"","institution":"Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shua","middleName":"","lastName":"Cui","suffix":""},{"id":616011700,"identity":"2ad4cb59-4b68-43db-b2fd-5b561803271f","order_by":5,"name":"Sirui Yang","email":"","orcid":"","institution":"Kunming Medical University","correspondingAuthor":false,"prefix":"","firstName":"Sirui","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2026-02-28 07:40:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8992745/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8992745/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106403995,"identity":"6b020706-dfd2-405c-93b0-e8db0aa34353","added_by":"auto","created_at":"2026-04-08 09:15:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":65908,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of Study Population Inclusion and Exclusion\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/30e6193ae53144dd0bc0be1c.png"},{"id":106403836,"identity":"24dc7ded-c829-4baf-a84e-e92f79056203","added_by":"auto","created_at":"2026-04-08 09:15:02","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":76775,"visible":true,"origin":"","legend":"\u003cp\u003eOverall Research Flowchart\u003c/p\u003e\n\u003cp\u003e(A) Acquisition of hepatobiliary phase contrast-enhanced magnetic resonance imaging (HBP-MRI) images and volume delineation of regions of interest (ROIs).(B) Types of extracted features and use of filters.(C) Selection of extracted features.(D) Creation of the radiomics nomogram model and model evaluation.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/61b2a4be1d133aff9efdd091.png"},{"id":106404116,"identity":"e4d46ebe-2ea8-40a4-8cf5-4bbccf7fa6c2","added_by":"auto","created_at":"2026-04-08 09:15:31","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":169794,"visible":true,"origin":"","legend":"\u003cp\u003eRadiomic Feature Extraction and Selection Process (A) LASSO Feature Selection Graph; (B) Cross-Validation Graph\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/e1fa80ce83f3e1aab857ab92.png"},{"id":106403456,"identity":"9fff0a62-de02-43da-ad77-38f17d08ee24","added_by":"auto","created_at":"2026-04-08 09:14:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":99011,"visible":true,"origin":"","legend":"\u003cp\u003ePerformance Radar Chart of Eight Machine Learning Models\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/d5455111de40bbbf833b917b.png"},{"id":106309324,"identity":"d75b1043-290d-4d98-97c0-c83373d07aba","added_by":"auto","created_at":"2026-04-07 10:16:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":542416,"visible":true,"origin":"","legend":"\u003cp\u003eROC Curves of Models Upper part: ROC curves of the eight machine learning models;\u003c/p\u003e\n\u003cp\u003eLower part: ROC curves of the optimal machine learning model (RFmodel), the clinical model, and the combined model\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/a495f9ac8fbf5757d2c4280d.png"},{"id":106309326,"identity":"9ec6d471-8e8e-49fd-a44a-ab1c45fcd338","added_by":"auto","created_at":"2026-04-07 10:16:52","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":24963,"visible":true,"origin":"","legend":"\u003cp\u003eNomogram for the Individualized Prediction of GPC3-Positive HCC\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/d326c09ee79872892a2b6d1c.png"},{"id":106403971,"identity":"f7c18330-285f-45a2-8ff3-75a011945d29","added_by":"auto","created_at":"2026-04-08 09:15:18","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":52003,"visible":true,"origin":"","legend":"\u003cp\u003eDCA of the Combined Model, the Optimal Machine Learning Model (RF Model) and the Clinical Model\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/cd9580ecb210eb34b5108b64.png"},{"id":106405687,"identity":"3e5dfa1e-19bd-4773-8c3c-9dcc7948b0de","added_by":"auto","created_at":"2026-04-08 09:28:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1789732,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/f92bc5b1-e877-47ee-aeff-43cd5ce8de1a.pdf"},{"id":106309321,"identity":"37efdd51-1eb1-488f-95df-3fad6536a766","added_by":"auto","created_at":"2026-04-07 10:16:52","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":3601893,"visible":true,"origin":"","legend":"","description":"","filename":"date.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8992745/v1/2477ffa6f562ef9c4aac3796.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Exploring the Value of a Radiomics Model Based on Hepatobiliary Phase Magnetic Resonance Imaging in Predicting Glypican-3-Positive Hepatocellular Carcinoma","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003ePrimary liver cancer is the sixth most common cancer globally[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e] and the third leading cause of cancer-related death[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] in recent years. In China, this malignancy is ranked fourth in incidence and second in mortality, representing a major public health challenge due to its considerable disease burden[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].Hepatocellular carcinoma (HCC) is the most common form of liver cancer, making up around 70%-80% of all cases[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Consequently, it serves as the primary driver of both disease progression and prognosis in this malignancy[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. The clinical outcomes for patients with HCC remain poor, despite ongoing progress in diagnosis and treatment, largely due to a high rate of postoperative recurrence[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Thus, the development of enhanced preoperative diagnostic and prognostic evaluation methods is essential for optimizing treatment outcomes in HCC patients.\u003c/p\u003e \u003cp\u003eRecent advances in medical imaging have paved the way for novel approaches to non-invasive HCC assessment. One notable example is hepatobiliary phase contrast-enhanced magnetic resonance imaging (HBP-MRI), which stands out due to its unique capabilities[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. The introduction of gadoxetic acid disodium (Gd-EOB-DTPA) as a contrast agent delineates a pronounced imaging disparity between normal hepatic parenchyma and malignant tumors, owing to their differing contrast uptake[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. This improves our capacity for the non-invasive diagnosis and prognostic evaluation of HCC.\u003c/p\u003e \u003cp\u003eAs an emerging diagnostic imaging technology, radiomics capitalizes on advanced computational algorithms to extract and quantify sub-visual biological features from medical images, which underlies its capacity for constructing robust predictive models[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. This methodology has shown significant potential in diagnosis, prognostic forecasting, and therapy planning, and has been extensively utilized in research across a spectrum of cancer types. Radiomic models derived from Gd-EOB-DTPA-enhanced MRI offer the capability for early, non-invasive assessment of HCC pathological features, complemented by their utility in qualitative diagnosis, treatment response evaluation, and forecasting postoperative recurrence[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eGlypican-3 (GPC3), which is typically undetectable in normal liver tissue, exhibits a highly restricted expression profile characterized by significant overexpression in a majority of HCC cases. This distinct pattern not only underscores its promise as an immunotherapeutic target but also confirms its critical role in the diagnosis, therapeutic management, and prognosis of HCC. Nevertheless, the current clinical detection of GPC3 primarily depends on immunohistochemical analysis of resected postoperative specimens. Furthermore, while preoperative needle biopsy can assess GPC3 status, it is an invasive procedure that fails to comprehensively represent the tumor's inherent heterogeneity[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Consequently, a pressing clinical need exists for a non-invasive technique capable of predicting GPC3 status prior to treatment. This study seeks to assess the efficacy of an HBP-MRI-based radiomics model in the preoperative prediction of GPC3-positive HCC and to explore its clinical utility, thereby aiming to inform novel strategies for HCC diagnosis and precision medicine[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study Participants\u003c/h2\u003e \u003cp\u003e This retrospective study has been approved by the Institutional Review Board, which has also granted a waiver of informed consent. A patient cohort was identified from those with imaging features suggestive of primary liver cancer during the period from January 2020 to December 2024 for preliminary inclusion.\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section3\"\u003e \u003ch2\u003e2.1.1 Inclusion criteria:\u003c/h2\u003e \u003cp\u003e(1)Patients with preoperative imaging highly suggestive of HCC;\u003c/p\u003e \u003cp\u003e(2)Patients who underwent Gd-EOB-DTPA-enhanced MRI and subsequently received hepatic resection or needle biopsy within two weeks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.1.2 Exclusion criteria:\u003c/h2\u003e \u003cp\u003e(1) Pathological diagnosis other than HCC;\u003c/p\u003e \u003cp\u003e(2) Pathology reports lacking essential pathological indicators for inclusion;\u003c/p\u003e \u003cp\u003e(3) Tumor diameter less than 1 cm;\u003c/p\u003e \u003cp\u003e(4) Images compromised by metal or motion artifacts;\u003c/p\u003e \u003cp\u003e(5)History of prior treatment including chemotherapy, radiotherapy, transarterial chemoembolization (TACE), or radiofrequency ablation.\u003c/p\u003e \u003cp\u003ePatients were classified into HCC groups based on the pathological records of GPC3 expression status. A total of 151 eligible HCC patients were included in the study. The cohort was divided into a training set (n\u0026thinsp;=\u0026thinsp;107) and a validation set (n\u0026thinsp;=\u0026thinsp;44) using stratified randomisation, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Methods\u003c/h2\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 MRI Acquisition\u003c/h2\u003e \u003cp\u003eAll patients underwent scanning using a GE Signa Premier 3.0T MRI scanner, with coverage extending from the diaphragm to the inferior poles of the kidneys. A bolus injection of the Gd-EOB-DTPA contrast agent was administered via a peripheral vein at a dose of 0.025 mmol/kg. The imaging protocol included T2-weighted imaging (T2WI), T2WI with fat suppression, and in-phase and out-of-phase T1-weighted imaging (T1WI). Gd-EOB-DTPA-enhanced multiphase sequences were acquired in the arterial, portal venous, transitional and HBPs. Images for the arterial, portal venous, transitional and HBPs were acquired at 15 seconds, 45 seconds, 50 seconds and 20 minutes after contrast injection, respectively. The axial T1WI parameters before and after enhancement were as follows: repetition time (TR) 3.11 ms, echo time (TE) 1.44 ms, field of view (FOV) 34.0 cm \u0026times; 34.0 cm, matrix 192 \u0026times; 384, slice thickness 5.0 mm, flip angle 15\u0026deg;, and acquisition time 12 s.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Clinical and Laboratory Data\u003c/h2\u003e \u003cp\u003eThe following information was obtained from the hospital's information system: demographic characteristics (age, sex), preoperative blood test results (hepatitis B serological status, presence of cirrhosis), and biochemical indicators including alpha-fetoprotein (AFP) level, activated partial thromboplastin time (APTT), total bilirubin (TBIL), serum albumin (ALB), alanine aminotransferase(ALT) and aspartate aminotransferase (AST).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3 Histopathological Analysis\u003c/h2\u003e \u003cp\u003eTissue samples were obtained via surgical resection or needle biopsy and subjected to histopathological evaluation. Cases were classified as GPC3-positive if immunoreactive tumour cells were present in more than 10% of the total tumour area; otherwise, they were classified as GPC3-negative. If the interobserver agreement between two pathologists was below 90%, a consensus was reached through joint re-evaluation and, if necessary, arbitration by a third senior pathologist.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.2.4 Radiomics Analysis\u003c/h2\u003e \u003cp\u003eThe MR images were corrected for intensity non-uniformity using the N4 algorithm in the RMIT tool for medical imaging, developed by the Jilin Provincial Medical Imaging AI Laboratory. Two abdominal radiologists, each with over three years' experience of abdominal MRI, manually delineated regions of interest (ROI) on HBP images using 3D Slicer software. Although both radiologists were aware of the HCC diagnosis, they were blinded to the GPC3 expression status and other clinical data. In cases of disagreement during ROI delineation, consensus was reached through discussion to ensure accurate and objective image interpretation.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section4\"\u003e \u003ch2\u003e2.2.4.1 Feature Extraction:\u003c/h2\u003e \u003cp\u003eA total of 850 radiomic features were automatically computed from the segmented tumour volumes using the Radiomics extension in 3D Slicer.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section4\"\u003e \u003ch2\u003e2.2.4.2 Feature Selection:\u003c/h2\u003e \u003cp\u003eRadiomic features were first screened for reliability based on the intraclass correlation coefficient (ICC) for both inter- and intra-observer variability. Those with ICC values greater than 0.75 were deemed reproducible, and their averaged values across observers were adopted for subsequent modeling.\u003c/p\u003e \u003cp\u003eIn order to reduce dimensionality and enhance the robustness of predictions, a two-stage feature selection strategy was implemented. The study cohort was then randomly partitioned into a training subset and a validation subset, with a ratio of 7:3. Independent-sample t-tests (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) were performed on the training set inorder to filter out preliminary features. The least absolute shrinkage and selection operator (LASSO) algorithm was then used to refine the features further.\u003c/p\u003e \u003cp\u003eA multi-stage predictive modeling framework was constructed: Initially, eight machine learning models were built based on the selected features to predict GPC3 status, including random forest (RF), support vector machine (SVM), logistic regression (LR), gradient boosting machine (GBM), eXtreme Gradient Boosting (XGB), Naive Bayes (NB), K-Nearest Neighbors (KNN), and decision tree (DT). Subsequently, through model performance comparison (with the area under the curve (AUC) as the core evaluation index), the best-performing model (random forest model) was selected, and the radiomic score (RadScore) was calculated based on this model. Finally, a combined model integrating the RadScore derived from the random forest model and clinical predictors (AFP, liver cirrhosis) was constructed, and validation was completed in the independent validation cohort (n\u0026thinsp;=\u0026thinsp;44). The detailed process is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e2.2.5 Statistical Analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed using SPSS software (version 24) and R software (version 4.4.3). The Kolmogorov\u0026ndash;Smirnov test was employed to evaluate the normality of continuous variables, while Levene's test was utilised to assess the homogeneity of variance. The statistical analysis of continuous variables between groups was conducted using either independent samples t-tests or Mann-Whitney U tests, depending on the fulfilment of the assumption of normality. The assessment of associations between categorical variables was conducted through the utilisation of the chi-square test. Clinical variables that demonstrated statistically significant differences (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the univariate analyses were directly incorporated into multivariate logistic regression models for further analysis to identify factors independently associated with GPC3 expression. Throughout the analysis, the statistical significance threshold was set at P\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003cp\u003eThe predictive performance of the developed models was assessed by receiver operating characteristic (ROC) analysis, with model efficacy further detailed by calculating a suite of metrics encompassing the AUC, accuracy, sensitivity, specificity, and F1 score.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Baseline Patient Characteristics\u003c/h2\u003e \u003cp\u003eThe preoperative clinicopathological characteristics of GPC3-positive and GPC3-negative patients in the training and validation cohorts are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Univariate logistic regression analysis in the training cohort revealed that AFP level and cirrhosis status were significantly associated with GPC3 expression (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). All relevant factors were subsequently included in a multivariate logistic regression model. The results identified AFP (P\u0026thinsp;\u0026lt;\u0026thinsp;0.01) and cirrhosis (P\u0026thinsp;=\u0026thinsp;0.001) as independent predictors of GPC3 expression, which were subsequently incorporated into the clinical-radiomic model.\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\u003eClinical Characteristics of Patients with Hepatocellular Carcinoma (HCC)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eTraining Cohort (n\u0026thinsp;=\u0026thinsp;107)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003eValidation Cohort (n\u0026thinsp;=\u0026thinsp;44)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003egpc3+\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003egpc3-\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003egpc3+\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003egpc3-\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e55.91\u0026thinsp;\u0026plusmn;\u0026thinsp;10.289\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e60.64\u0026thinsp;\u0026plusmn;\u0026thinsp;9.384\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e58.91\u0026thinsp;\u0026plusmn;\u0026thinsp;8.329\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e54.44\u0026thinsp;\u0026plusmn;\u0026thinsp;5.199\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.192\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e38,6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.424\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\u003e68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\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\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAFP\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;0.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.066\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAFP\u0026thinsp;\u0026lt;\u0026thinsp;20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e20\u0026thinsp;\u0026lt;\u0026thinsp;AFP\u0026thinsp;\u0026lt;\u0026thinsp;400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAFP\u0026thinsp;\u0026gt;\u0026thinsp;400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKI67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30(20,75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30(10,52)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20(10,30)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e50(8.75,60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.262\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAPTT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30.1(27.9,34.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30.65(27.675,33.725)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e30.2(28,35)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e33.1(30.45,34.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.147\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36.304\u0026thinsp;\u0026plusmn;\u0026thinsp;7.6352\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e39.25\u0026thinsp;\u0026plusmn;\u0026thinsp;5.6085\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.121\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35.7\u0026thinsp;\u0026plusmn;\u0026thinsp;6.6662\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e34.767\u0026thinsp;\u0026plusmn;\u0026thinsp;9.1789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.399\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTotal Bilirubin\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39.1(20,77)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.15(14.125,45.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0502\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e44(23,150)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e35(15.05,66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52(30,108)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36(27.25,57.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e61(33,184)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e56(22,118.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.33\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAST\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19(14.95,34.05))\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.2(14.725,37.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0997\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e21.5(16,36.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e32(18.8,58.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.301\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHBV Status\u003c/b\u003e\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.754\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.968\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCirrhosis\u003c/b\u003e\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.027\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDifferentiation\u003c/b\u003e\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.053\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.588\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWell-differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerately-Poorly Differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSatellite Nodules\u003c/b\u003e\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.172\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMVI\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.074\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.626\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMVI0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMVI1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMVI2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\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=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Feature Selection and Construction of the MRI Radiomics Model\u003c/h2\u003e \u003cp\u003eA total of 850 radiomic features were extracted from the hepatobiliary phase (HBP) images of each patient in the training cohort. After excluding features with ICC below 0.75 due to poor stability, 796 features were retained as robust features. These were further screened using independent sample t-tests, resulting in 96 statistically significant features. Finally, 11 key radiomic features were selected through LASSO regression analysis (see Fig.\u0026nbsp;3).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Model Performance Evaluation\u003c/h2\u003e \u003cp\u003eBased on the selected radiomic features, eight machine learning models were constructed to predict GPC3 status. Among the above eight models, the RF model performed the best, with an AUC of 0.918 in the training cohort and 0.903 in the validation cohort; for the specific performance indicators of the remaining seven models and the performance comparison between all models (including the RF model), please refer to the performance radar chart of each model (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e).In the training cohort, the AUC of the standalone clinical model was 0.758, while the combined model integrating the RadScore calculated by the RF model and clinical factors (AFP, liver cirrhosis) achieved an AUC of 0.936. In the validation cohort, the AUC of the standalone clinical model was 0.762, and the AUC of the combined model was 0.940. The ROC curves of each model are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e5\u003c/span\u003e.A nomogram was developed based on this combined model (Fig.\u0026nbsp;6), which integrates AFP level, liver cirrhosis status, and RadScore, providing a visual tool for the individualized prediction of GPC3-positive HCC. Decision curve analysis (DCA) was used to evaluate the clinical utility of the clinical model, the RF model (standalone radiomics model), and the combined model (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The results showed that within a wide range of threshold probabilities, the combined model provided significantly higher net clinical benefit than the standalone radiomics model (RF model) or the standalone clinical model, confirming its high clinical application value in predicting GPC3-positive HCC.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study aimed to preoperatively predict GPC3 expression in hepatocellular carcinoma (HCC) by constructing and comparing a clinical model, radiomics models, and a combined model, while systematically evaluating the predictive performance of different modeling architectures and the value of model integration strategies. Notably, the dataset of this study exhibited a significant class imbalance, with the ratio of GPC3-positive to negative samples being 3.72:1. A review of relevant literature revealed that this proportional imbalance is prevalent in GPC3-related studies: Han et al. (2023)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] reported a ratio of 3.07:1 for GPC3-positive to negative samples and addressed the issue using Synthetic Minority Over-sampling Technique (SMOTE), whereas Geng et al. (2021)[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], Zhang et al. (2023)[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], and Li et al. (2024)[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]documented imbalanced ratios of 3.08:1, 3.43:1, and 3.78:1, respectively, without implementing any correction techniques for this distribution characteristic. Although this study initially attempted to improve the data distribution imbalance by applying SMOTE to the training set, this approach did not yield the expected results, and thus no synthetic over-sampling techniques were ultimately used to intervene in the data distribution.\u003c/p\u003e \u003cp\u003eAnalysis yielded 11 critical radiomic features spanning shape, GLDM, GLRLM, GLCM, and NGTDM categories. These features offer multi-perspective quantifications of a tumor's intensity distribution, textural patterns, and spatial heterogeneity, enabling the non-invasive preoperative prediction of GPC3-positive HCC [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The wavelet transform, a method utilized in radiomics analysis, excels at resolving fine local details and high-frequency components within medical images, thereby refining the characterization of tumor heterogeneity [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. In alignment with our results, the wavelet transform separates medical images into low-frequency (LF) and high-frequency (HF) constituents. The LF elements primarily encapsulate global semantic content, such as the gross morphology and contour of a tumor, whereas the HF elements depict fine-grained local details, including edges and textural patterns[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The ensemble of eight selected wavelet features was dominated by textural descriptors from GLCM, GLRLM, and GLDM.GLCM reflects the frequency (consistent with the statistical logic of probability) of pixel pairs with the same gray-level value occurring at a certain distance and direction within a ROI, which can indirectly represent tumor phenotype; GLRLM quantifies the distribution of gray-level runs[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]; GLDM quantifies gray-level dependencies in image[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e][\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], A grey-level dependency is defined as the number of connected voxels within a distance of δ that depend on the central voxel[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn terms of model performance, the random forest model exhibited the best predictive ability among the eight constructed machine learning models, with an AUC of 0.918 in the training cohort and 0.903 in the validation cohort. This outperformed the other seven radiomics models. Notably, the amalgamation of the random forest-calculated RadScore (the optimal radiomics model) with clinical independent risk factors (AFP and liver cirrhosis) translated into a model with refined predictive performance, evidenced by AUCs of 0.936 (training) and 0.940 (validation). The integrated model demonstrated a marked superiority over both the clinical-only model (training AUC: 0.758; validation AUC: 0.762) and the standalone radiomics model. This multimodal integration strategy holds substantial scientific significance: while radiomic features quantify tumor heterogeneity at a sub-visual level, thereby overcoming the limitations of conventional clinical parameters that reflect macroscopic pathological states, clinical variables provide essential biological context, creating a synergistic diagnostic framework.\u003c/p\u003e \u003cp\u003eThe DCA subsequently validated the superior clinical value of the integrated modeling approach. The results demonstrated that across a broad spectrum of threshold probabilities, the combined model yielded a consistently superior net clinical benefit when benchmarked against both the clinical-only and radiomics-only models[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The findings demonstrate the model's viability for clinical decision support, providing substantial basis for precision intervention selection before surgery and risk-adapted management after resection. The proposed modeling methodology not only strengthens forecasting accuracy but also provides a clinically viable and intelligible guidance instrument, representing a stride forward in precision hepatology.\u003c/p\u003e \u003cp\u003eThe study's limitations should be considered alongside its contributions. The single-center derivation cohort may limit generalizability due to potential selection bias, an issue that must be addressed through future multi-center, large-scale validation to confirm the nomogram's clinical utility (Fig.\u0026nbsp;6). This investigation was additionally constrained by its exclusive focus on HBP features, without incorporating complementary multiparametric MRI sequences such as diffusion-weighted imaging (DWI) and T2WI. This omission is particularly relevant given that the apparent diffusion coefficient (ADC) derived from DWI has established correlations with tumor cellularity and could potentially enhance the prediction of GPC3 expression[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e][\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis research establishes the combined model as a potential preoperative screening tool for biologically aggressive GPC3-positive HCC. Its additional capacity to guide non-surgical therapeutic decisions\u0026mdash;particularly regarding targeted or combined immunotherapy\u0026mdash;supports the evolution of an integrated precision medicine paradigm that bridges imaging biomarkers, molecular pathology, and standard clinical parameters.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eADC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eApparent Diffusion Coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAFP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAlpha-Fetoprotein\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eALB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSerum Albumin\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eALT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAlanine Aminotransferase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArea Under the Curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAPTT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eActivated Partial Thromboplastin Time\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAST\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAspartate Aminotransferase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDecision Curve Analysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDecision Tree\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDWI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDiffusion-Weighted Imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFOV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eField of View\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGBM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGradient Boosting Machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLCM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray Level Co-occurrence Matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLDM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray Level Dependence Matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLRLM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGray Level Run Length Matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGPC3\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGlypican-3\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGd-EOB-DTPA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGadolinium ethoxybenzyl diethylenetriamine pentaacetic acid\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHBP-MRI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHepatobiliary Phase contrast-enhanced Magnetic Resonance Imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHCC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHepatocellular Carcinoma\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHBV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHepatitis B Virus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHigh Frequency\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eICC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIntraclass Correlation Coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIRB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eInstitutional Review Board\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKNN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eK-Nearest Neighbors\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLeast Absolute Shrinkage and Selection Operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLow Frequency\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLogistic Regression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMVI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMicrovascular Invasion\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNaive Bayes\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNGTDM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNeighborhood Gray Tone Difference Matrix\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRegion of Interest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRadScore\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRadiomic Score\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom Forest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReceiver Operating Characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSMOTE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSynthetic Minority Over-sampling Technique\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSPSS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStatistical Product and Service Solutions\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSupport Vector Machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTACE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eTransarterial Chemoembolization\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTBIL\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eTotal Bilirubin\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eEcho Time\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT1WI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eT1-Weighted Imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT2WI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eT2-Weighted Imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRepetition Time\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eXGB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eeXtreme Gradient Boosting\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate：\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Ethics Committee of The Second Affiliated Hospital of Kunming Medical University. Written informed consent was obtained from all individual participants included in the study.\u003c/p\u003e\n\u003cp\u003eConsent for publication：\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials：\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePeipei Ye : Writing – original draft ,Data curation, formal analysis, table revision\u003c/p\u003e\n\u003cp\u003eLi Guo: Writing – review and editing.\u003c/p\u003e\n\u003cp\u003eGuangwen tong,Youwei wang: Data curation, formal analysis, table revision.\u003c/p\u003e\n\u003cp\u003eShuai Cui, Sirui Yang: Investigation, experiment assistance.\u003c/p\u003e\n\u003cp\u003eAll authors have read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors thank the patients and staff of The Second Affiliated Hospital of Kunming Medical University for their support in this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCao M, Xia C, Cao M, Yang F, Yan X, He S, Chen W. Attributable liver cancer deaths and disability-adjusted life years in China and worldwide: profiles and changing trends. Cancer biology Med. 2024;21(8):679\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao G, Liu J, Liu M. Global, regional, and national trends in incidence and mortality of primary liver cancer and its underlying etiologies from 1990 to 2019: results from the global burden of disease study 2019. J Epidemiol Global Health. 2023;13(2):344\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao M, Li H, Sun D, He S, Yan X, Yang F, Chen W. Current cancer burden in China: epidemiology, etiology, and prevention. Cancer biology Med. 2022;19(8):1121\u0026ndash;38.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGlynn KA, Petrick JL, El-Serag HB. Epidemiology of hepatocellular carcinoma. Hepatology. 2021;73:4\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKe J, Liu Y, Liu F, Cai H, Li X, Zhang Z, Ji B. In-situ-formed immunotherapeutic and hemostatic dual drug-loaded nanohydrogel for preventing postoperative recurrence of hepatocellular carcinoma. J Controlled Release. 2024;372:141\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark HJ, Seo N, Kim SY. Current landscape and future perspectives of abbreviated MRI for hepatocellular carcinoma surveillance. Korean J Radiol. 2022;23(6):598.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi XQ, Wang X, Zhao DW, Sun J, Liu JJ, Lin DD, Li HJ. Application of Gd-EOB-DTPA-enhanced magnetic resonance imaging (MRI) in hepatocellular carcinoma. World J Surg Oncol. 2020;18(1):219.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScapicchio C, Gabelloni M, Barucci A, Cioni D, Saba L, Neri E. A deep look into radiomics. Radiol Med. 2021;126(10):1296\u0026ndash;311.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen S, Feng S, Wei J, Liu F, Li B, Li X, Kuang M. Pretreatment prediction of immunoscore in hepatocellular cancer: a radiomics-based clinical model based on Gd-EOB-DTPA-enhanced MRI imaging. Eur Radiol. 2019;29(8):4177\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou F, Shang W, Yu X, Tian J. Glypican-3: a promising biomarker for hepatocellular carcinoma diagnosis and treatment. Med Res Rev. 2020;38(2):741\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGu D, Xie Y, Wei J, Li W, Ye Z, Zhu Z, Li X. MRI-based radiomics signature: a potential biomarker for identifying glypican 3‐positive hepatocellular carcinoma. J Magn Reson Imaging. 2020;52(6):1679\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan Z, Dai H, Chen X, Gao L, Chen X, Yan C, Li Y. Delta-radiomics models based on multi-phase contrast-enhanced magnetic resonance imaging can preoperatively predict glypican-3-positive hepatocellular carcinoma. Front Physiol. 2023;14:1138239.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeng Z, Zhang Y, Wang S, Li H, Zhang C, Yin S, Dai Y. Radiomics analysis of susceptibility weighted imaging for hepatocellular carcinoma: exploring the correlation between histopathology and radiomics features. Magn Reson Med Sci. 2021;20(3):253\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang N, Wu M, Zhou Y, Yu C, Shi D, Wang C, Zhu S. Radiomics nomogram for prediction of glypican-3 positive hepatocellular carcinoma based on hepatobiliary phase imaging. Front Oncol. 2023;13:1209814.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi SQ, Yang CX, Wu CM, Cui JJ, Wang JN, Yin XP. Prediction of glypican-3 expression in hepatocellular carcinoma using multisequence magnetic resonance imaging-based histology nomograms. Quant Imaging Med Surg. 2024;14(7):4436.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMasokano IB, Liu W, Xie S, Marcellin DFH, Pei Y, Li W. The application of texture quantification in hepatocellular carcinoma using CT and MRI: a review of perspectives and challenges. Cancer Imaging. 2020;20(1):67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang M, Wang L, Gu L, Ling T, Tao X. (2025). WT-BCP: Wavelet Transform based Bidirectional Copy-Paste for Semi-Supervised Medical Image Segmentation. \u003cem\u003earXiv preprint arXiv:2504.14445\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVarghese BA, Fields BK, Hwang DH, Duddalwar VA, Matcuk GR Jr, Cen SY. Spatial assessments in texture analysis: what the radiologist needs to know. Front Radiol. 2023;3:1240544.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim YJ. Machine learning models for sarcopenia identification based on radiomic features of muscles in computed tomography. Int J Environ Res Public Health. 2021;18(16):8710.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVarriano G. (2024). Radiomics and Formal Methods: a dynamic duo for medicine in pixels.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiovani D, Sokou R, Tsantes AG, Vitello AS, Bonovas S. August). Optimizing clinical decision making with decision curve analysis: insights for clinical investigators. Healthcare. Volume 11. MDPI; 2023. p. 2244. 16.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, J., Gao, S., Sun, W., Grimm, R., Fu, C., Han, J., \u0026hellip; Zeng, M. (2021). Magnetic resonance imaging and diffusion-weighted imaging-based histogram analyses in predicting glypican 3-positive hepatocellular carcinoma. European Journal of Radiology, 139, 109732..\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, P., Li, W., Qiu, G., Chen, J., Liu, Y., Wen, Z., \u0026hellip; Zhao, Y. (2023). Multiparametric MRI combined with clinical factors to predict glypican-3 expression of hepatocellular carcinoma. Frontiers in Oncology, 13, 1142916..\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRastogi A, A. Changing role of histopathology in the diagnosis and management of hepatocellular carcinoma. World J Gastroenterol. 2020;24(35):4000.\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":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Hepatocellular carcinoma(HCC), Glypican-3 (GPC3)","lastPublishedDoi":"10.21203/rs.3.rs-8992745/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8992745/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eHepatocellular carcinoma (HCC) is the most common primary liver cancer (70%-80% of cases) with a heavy global disease burden and high postoperative recurrence rate, highlighting the need for improved preoperative diagnosis. Glypican-3 (GPC3) is a key diagnostic/therapeutic biomarker for HCC, but current detection relies on invasive methods (postoperative immunohistochemistry, preoperative needle biopsy) with limitations. Hepatobiliary phase contrast-enhanced MRI (HBP-MRI) and radiomics offer promising noninvasive alternatives.\u003c/p\u003e\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo explore the value of an HBP-MRI-based radiomics model for noninvasive preoperative prediction of GPC3-positive HCC, and develop an optimized model integrating imaging and clinical features.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA total of 151 HCC patients who underwent Gd-EOB-DTPA-enhanced MRI and hepatic resection/needle biopsy (2020\u0026ndash;2024) were retrospectively divided into training (n\u0026thinsp;=\u0026thinsp;107) and validation (n\u0026thinsp;=\u0026thinsp;44) cohorts (7:3 ratio). Eight hundred fifty radiomic features were extracted via 3D Slicer, filtered by ICC (\u0026ge;\u0026thinsp;0.75), independent sample t-tests (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05), and LASSO (11 key features retained). Eight machine learning models were constructed; the optimal model was combined with clinical factors (AFP, liver cirrhosis) to form a combined model. Performance was evaluated by ROC curves, calibration curves, and DCA.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe random forest model performed best (training AUC\u0026thinsp;=\u0026thinsp;0.918; validation AUC\u0026thinsp;=\u0026thinsp;0.903). The combined model achieved superior efficacy (training AUC\u0026thinsp;=\u0026thinsp;0.936; validation AUC\u0026thinsp;=\u0026thinsp;0.940), outperforming the clinical model (AUC\u0026thinsp;=\u0026thinsp;0.762) and standalone radiomics model, with the highest net clinical benefit confirmed by DCA.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe HBP-MRI-based radiomics model enables noninvasive prediction of GPC3-positive HCC. The nomogram integrating radiomic score and clinical features exhibits higher diagnostic efficacy, providing a novel noninvasive tool for personalized HCC management.\u003c/p\u003e","manuscriptTitle":"Exploring the Value of a Radiomics Model Based on Hepatobiliary Phase Magnetic Resonance Imaging in Predicting Glypican-3-Positive Hepatocellular Carcinoma","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-07 10:16:47","doi":"10.21203/rs.3.rs-8992745/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-30T21:45:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-27T04:01:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"282753033868066629964275460570876958557","date":"2026-04-01T14:20:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"145698772158323316991904811180625041261","date":"2026-04-01T07:19:33+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-01T05:03:44+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-05T10:19:34+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-05T10:18:27+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2026-02-28T07:20:43+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"80736600-56c3-44af-9460-36c1bd7179e8","owner":[],"postedDate":"April 7th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-04-30T21:45:02+00:00","index":20,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-07T10:16:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-07 10:16:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8992745","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8992745","identity":"rs-8992745","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}