{"paper_id":"08b064a5-3e73-44ad-923b-9ec02c6899fd","body_text":"Tumor-to-liver volume ratio (TLVR)-integrated Radiomics-clinicopathological Fusion Model for Prognosis Prediction in Colorectal Cancer Liver Metastases | 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 Tumor-to-liver volume ratio (TLVR)-integrated Radiomics-clinicopathological Fusion Model for Prognosis Prediction in Colorectal Cancer Liver Metastases Haichao Zou, Haiyang Xie, Guoxin Liang, Shanshan Luo, Yumeng Guo, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9008242/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 Colorectal cancer (CRC) ranks as the third most common malignancy worldwide, with colorectal liver metastasis (CRLM) being the leading cause of CRC-related mortality. In this study, we propose the tumor-to-liver volume ratio (TLVR) as a standardized physiological biomarker and develop a multimodal fusion model integrating TLVR, CT radiomics and clinicopathological factors to accurately predict overall survival (OS) in CRLM patients. Methods In this retrospective study, 218 CRLM patients were enrolled and randomly divided into a training cohort (n = 152) and a validation cohort (n = 66). Radiomic features and clinical data were extracted from treatment-naive CT scans and medical records. The cut-off value for TLVR was determined by receiver operating characteristic (ROC) curve analyses. The random survival forest algorithm was used to construct the clinical, radiomics, and fusion models. The model performance was assessed with C-index, time-dependent area under the curve (AUC), and decision curve analysis (DCA). Results TLVR ≥ 0.015 (1.5%) was significantly correlated with poorer OS for CRLM patients. Fifteen radiomic features and five clinical variables were incorporated for model construction. The fusion model demonstrated superior prognosic accuracy in the validation cohort with AUC of 0.855, compared to the clinical model (AUC = 0.831) and the radiomics model (AUC = 0.828). Conclusion TLVR is a potent prognostic biomarker reflecting tumor-host volumetric equilibrium. Its integration into a CT radiomics-clinical fusion model significantly enhances OS prediction accuracy for CRLM patients. This non-invasive tool enables personalized therapeutic strategies, including TLVR-guided adjuvant therapy allocation and avoidance of futile conversion surgery. Radiomics Colorectal cancer Liver metastases TLVR Multimodal fusion Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 Introduction Colorectal cancer (CRC) remains a significant global public health challenge, ranking as the third most common malignancy and the second leading cause of cancer-related deaths. Distant metastasis is the primary cause of treatment failure and mortality among CRC patients [ 1 – 3 ]. Colorectal cancer liver metastasis (CRLM) is the most prevalent form, affecting approximately 50% of patients during their disease progression [ 4 ]. Despite advancements in surgery and systemic therapies, the prognosis for CRLM patients varies widely, underscoring the urgent need for personalized precision medicine strategies to improve survival rates and quality of life [ 5 ]. Recent studies indicate that total tumor volume (TTV) serves as an independent prognostic factor for CRLM [ 6 ]. However, TTV alone fails to account for inter-patient variability in liver volume. Consequently, identical TTV values may carry different clinical implications depending on individual liver size. The tumor-to-liver volume ratio (TLVR) addresses this limitation by integrating tumor burden with functional liver reserve, potentially offering a more accurate reflection of the dynamic balance between tumor aggressiveness and host compensatory capacity. Additionally, radiomics and machine learning have gained significant traction in radiology in recent years [ 7 – 9 ], demonstrating utility in disease prediction and survival analysis [ 10 – 12 ]. These approaches can evaluate radiological images using quantitative texture information known as imaging features [ 13 , 14 ]. Radiological texture features can assess prognostic indicators related to tumor heterogeneity and metastasis within the tumor microenvironment. Specifically, CRLM radiomics has emerged as a valuable tool for pre-operative diagnosis and post-treatment survival prediction [ 15 , 16 ]. In this study, we developed an integrated predictive model combining CT radiomics features with clinical characteristics based on TLVR. Utilizing the random forest algorithm, this model accurately forecasts OS in CRLM patients. Moreover, it significantly enhances risk stratification capabilities and facilitates personalized treatment planning and clinical management. 2 Methods 2.1 Ethical approval and cohort details This retrospective cohort study consecutively enrolled all the patients diagnosed with colorectal liver metastases (CRLM) and received primary therapy (hepatectomy or chemotherapy) at the Affiliated Tumor Hospital of Guangxi Medical University from April 2013 to November 2019. Key inclusion criteria were the absence of extrahepatic metastatic diseases at the time of CRLM diagnosis. The exclusion criteria comprised: (1) documented extrahepatic metastases prior to CRLM diagnosis; (2) other malignant disease present during the study period; (3) insufficient clinicopathological data, baseline CT imaging, or follow-up information. As this study involve retrospective analysis of anonymized patient data, the requirement for written informed consent was waived. The study protocol was strictly in line with the principles of the Declaration of Helsinki and obtained the institutional ethics committee approval of the school of oncology, Guangxi Medical University in 2025. 2.2 Collection of clinical data Baseline demographic and clinicopathologic variables were extracted from electronic medical records. Serological tumor markers, including carcinoembryonic antigen (CEA) and carbohydrate antigen 19 − 9 (CA19-9), were documented. To comprehensively evaluate the systemic inflammatory and nutritional status, composite indices including the albumin-bilirubin (ALBI), neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR) and platelet-to-lymphocyte ratio (PLR) were calculated using results from preoperative blood tests. 2.3 Image acquisition and segmentation Abdominal CT examinations were performed on a multi-detector row scanner (SOMATOM Definition Flash, Siemens Medical Systems). Images were reconstructed with a standard reconstruction kernel and stored in DICOM (Digital Imaging and Communications in Medicine) format. Regions-of-interest (ROI) encompassing the tumors and the entire liver were manually delineated slice by slice on portal venous phase images using 3D Slicer (version 5.8.1, https://www.slicer.org/ ). Total tumor volume (TTV) and total liver volume (TLV) were subsequently computed from these segmentations. Two experienced radiologists independently performed the volumetric segmentation. Inter-observer variability was minimized by resolving any disagreements through open discussion to reach a final consensus. 2.4 Extraction of radiomics features Following image acquisition and segmentation, cubic B-spline interpolation was applied to resample both the CT images and their ROI masks to an isotropic voxel size of 1 x 1 x 1 mm 3 to ensure a consistent spatial resolution. Radiomics feature extraction was performed using the pyradiomics library, a widely validated open-source tool for quantitative feature analysis of medical images. Feature selection proceeded in two stages. First, univariate Cox proportional hazards regression analysis was performed to screen out features significantly associated with OS (p < 0.05). Subsequently, these identified variables were subjected to the Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression to mitigate overfitting and enhance model generalizability. The optimal regularization parameter for the LASSO model was determined via ten-fold cross-validation to select the most informative features with non-zero coefficients. 2.5 Definition of TLVR cut-off value The TLVR cut-off values that would best predict OS were determined using receiver operating characteristic (ROC) curve analysis. The 3-year OS rates were used as objective variables. 2.6 Selection of clinical features Patients were categorized into high- and low-TLVR groups according to the predefined OS cutoff value. Clinical variables potentially associated with prognostic were first evaluated using univariate Cox proportional-hazards regression with a significance threshold (P < 0.05). Variables demonstrating significant association in univariate analysis were subsequently entered into a multivariate Cox regression model. Stepwise bidirectional regression based on the akaike information criterion was used to identify the independent prognostic factors. 2.7 Construction of prediction models All features were standardized according to the distribution of the training set. A random survival forest classifier was then used to build two independent models: a radiomics model utilizing CT radiomic features and a clinical model utilizing clinicopathological features. The prediction scores generated by these two models were integrated via score-level fusion. Three fusion strategies were evaluated including minimum score fusion, maximum score fusion, and weighted score fusion. For the minimum and maximum strategies, the two model scores were compared case-by-case and the smaller or larger score, respectively, was selected as the fused result. For the weighted strategy, the fused score was the weighted sum of the two model scores, with weights summing to one. 2.8 Statistical analysis The primary endpoint was OS, defined as the interval from the date of initial treatment to the date of death from any cause. Data for patients who remained alive or were lost to follow-up were censored at the date of their last known contact. All statistical analyses were performed using Python (version 3.9) and R software (version 4.5.1). Continuous variables are presented as the median (i.q.r.) values and were analyzed using the Mann–Whitney U-test or student’s t-test. Categorical variables are summarized as frequencies and percentages and were analyzed using the chi-squared test or Fisher’s exact test. The survival rate was estimated using the Kaplan–Meier method, and the survival rate estimates were compared using the log-rank test. A two-sided p-value less than 0.05 was considered statistically significant. 3 Results 3.1 Patient enrollment and patient baseline characteristics Between April 2013 to November 2019, 218 out of 296 CRLM patients were retrospectively enrolled in this study according to the inclusion and exclusion criteria (Fig. 1 ). Then, this cohort was randomly partitioned into a training set (n = 152) and a validation set (n = 66). The baseline demographic and clinicopathological characteristics of the entire cohort are detailed in Table 1 . Notably, the overall 3-year OS was 41.7%, with a median follow-up duration of 26 months. Table 1 Clinical characteristics of patients Clinicopathological features Total (n = 218) Training set (n = 152) Validation set (n = 66) p Gender (male) 145 (67) 106 (70) 39 (59) 0.169 Age * 60 (50.25, 66) 59 (50.75, 66) 61 (50.25, 66.75) 0.485 BMI * 22.49 (20.31, 24.77) 22.34 (20.31, 24.55) 22.77 (20.23, 25.63) 0.453 Primary site Location (right) 60 (28) 44 (29) 16 (24) 0.583 Lymphatic invasion (+) 122 (56) 88 (58) 34 (52) 0.473 Lymph node metastasis (+) 153 (70) 108 (71) 45 (68) 0.791 Metastases Maximal tumor diameter(cm)* 3.2 (2, 5.38) 3.2 (2, 4.95) 3.5 (2.15, 5.88) 0.292 Multiple tumour number 128 (59) 87 (57) 41 (62) 0.601 Tumor distribution (bilobar) 123 (56) 87 (57) 36 (55) 0.826 Metachronous metastases 82 (38) 59 (39) 23 (35) 0.687 TLVR* 1.86 (0.43, 9.21) 1.91 (0.42, 8.8) 1.75 (0.46, 9.48) 0.881 Laboratory tests ALBI * -2.5 (-2.73, -2.24) -2.5 (-2.73, -2.25) -2.46 (-2.69, -2.2) 0.384 NLR* 2.57 (1.79, 3.85) 2.53 (1.73, 3.8) 2.69 (1.94, 3.9) 0.415 LMR* 3.35 (2.33, 4.3) 3.37 (2.39, 4.37) 3.21 (2.3, 4.16) 0.381 PLR* 166.91 (123.4, 234.13) 164.71 (125.21, 232.61) 171.41 (122.65, 241.23) 0.788 CEA* 19.56 (5.59, 99.55) 22.26 (6.71, 99.94) 12.48 (4.46, 96.4) 0.351 CA19-9* 29.08 (7.94, 251.73) 32.05 (9.57, 259.35) 18 (4.56, 203.73) 0.181 Values in parentheses are percentages unless indicated otherwise; *values are median (i.q.r.). BMI, body mass index; TLVR, tumor-to-liver volume ratio; ALBI, albumin–bilirubin; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; CEA, carcinoembryonic antigen; CA19-9,carbohydrate antigen 19 − 9. 3.2 Determination of TLVR cut-off values and survival analysis A 3-year ROC analysis was utilized to assess the association between TLVR and OS within the training cohort. The analysis yielded an AUC of 0.677, determining an optimal discriminative cutoff of 0.015 for prognostic stratification (sensitivity 67%, specificity 63%; Fig. 2 a). In the validation cohort, patients with TLVR ≥ 0.015 was significantly correlated with poorer OS (P < 0.01) (Fig. 2 b). To account for treatment heterogeneity, subgroup analyses was performed based on therapeutic modality of either surgical resection or chemotherapy. The established TLVR cutoff (≥ 0.015) remained a robust predictor of reduced OS in both the surgical resection (Log-rank, P = 0.076, Fig. 2 c) and the chemotherapy subgroup (Log-rank, P = 0.012, Fig. 2 d), confirming its prognostic independence from therapeutic approach. Table 2 Univariable and multivariable Cox proportional hazards analysis of factors associated with overall survival Characteristics Univariable analysis Multivariable analysis Hazard ratio (95% CI) p_value Hazard ratio (95% CI) p_value Gender (male) 1.37(0.95–1.98) 0.089 Age 1.03(0.87–1.24) 0.705 BMI 0.97(0.83–1.16) 0.801 Primary site Location (right) 1.25(0.86–1.82) 0.241 Lymphatic invasion 1.44(1.02–2.06) 0.282 Lymph node metastasis 2.44(1.61–3.69) < 0.001 2.30(1.51–3.53) < 0.001 Metastases Maximal tumor diameter(cm) 1.2(1.02–1.42) 0.033 Multiple tumour number 1.23(0.87–1.34) 0.248 Tumor distribution (bilobar) 1.84(1.29–2.63) < 0.001 1.80(1.25–2.58) 0.002 Metachronous metastases 0.78(0.52–1.11) 0.171 TLVR ≥ 0.015 2.28(1.58–3.28) < 0.001 1.86(1.26–2.74) 0.002 Laboratory tests ALBI 1.23(1.03–1.47) 0.024 NLR 1.09(0.94–1.25) 0.265 LMR 0.76(0.61–0.94) 0.011 0.86(0.67–1.06) 0.159 PLR 1.16(0.97–1.31) 0.120 CEA 1.33(1.14–1.56) < 0.001 1.15(0.97–1.36) 0.099 CA19-9 1.29(1.08–1.57) 0.031 BMI, body mass index; TLVR, tumor-to-liver volume ratio; ALBI, albumin–bilirubin; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; CEA, carcinoembryonic antigen; CA19-9, carbohydrate antigen 19 − 9. 3.3 Feature selection for clinical models The associations between clinicopathological variables and OS in the CRLM patients are summarized in Table 2 . Following Z-score normalization using training set parameters, univariate Cox regression identified eight clinical features significantly associated with OS ( P < 0.05). Subsequent multivariate stepwise Cox regression refined these to five independent prognostic factors. Lymph node metastasis, bilobar tumor distribution, TLVR ≥ 0.015, and elevated CEA were independently associated with worse prognosis. Conversely, higher lymphocyte-to-monocyte ratio (LMR) demonstrated a protective effect. 3.4 Features selected for the CT radiomics model A total of 1, 316 radiomic texture features were extracted from the raw CT images using pyradiomics, encompassing features from both the original images as well as transformed images processed with exponential, gradient, logarithmic, square, square-root, and wavelet filters. After Z-score normalization, univariate Cox analysis was performed to identify 667 candidate features significantly associated with OS ( P < 0.05). To mitigate overfitting, L1-regularized LASSO regression was applied, retaining 15 radiomic features with non-zero coefficients. The selected features, their coefficients and relative importance rankings are visualized (Fig. 3 a-c). 3.5 Model development and validation Three prognostic models were thereby developed for CRLM, including a radiomics feature-based model, a clinical feature-based model, and a fusion model. In the validation cohort, the AUC of the 3-year ROC curves was 0.828 (95% Cl: 0.722–0.934) for the radiomics model and 0.831 (95% Cl: 0.728–0.934) for the clinical model, respectively. To optimize predictive performance, a fusion model was constructed based on the multiple weighed fusion strategies combining radiomics and clinical prediction scores (Table 3 ). Generally, the fusion model demonstrated superior discriminative ability AUC and C-index as compared to the radiomics or clinical model alone. Notably, the fusion model employing a 0.4 x Radiation score (Rad) + 0.6 x Clinical score (Cli) strategy achieved the highest validation AUC of 0.855 (95% CI: 0.759–0.950; p < 0.05 versus individual models). As visualized in the 3-year ROC curves (Figs. 4 a and 4 b), the fusion approach exhibited superior discriminative capacity, significantly outperforming both the radiomics and clinical models. Calibration analysis further corroborated the model's reliability, demonstrating excellent agreement between predicted probabilities and observed survival outcomes (Fig. 4 c). Moreover, Decision Curve Analysis (DCA) highlighted the clinical relevance of our approach: the fusion model conferred a robust net clinical benefit that matched or exceeded that of single-modality models across a broad range of decision thresholds (Fig. 4 d). The model also demonstrated robust predictive performance at 1 and 5 years (Supplementary Figures S1 a-f). Taken together, the fusion model showed robust prognostic performance for CRLM and demonstrate potential value for precise stratification in clinical practice. To better illustrate our finding, the workflow of the construction of the CRLM prognostic model was presented (Fig. 5 ). Patients were randomly assigned to training and validation cohorts in a 7:3 ratio. After manual segmentation of tumor ROIs on CT images, radiomic features and clinical characteristics were selected from the training cohort to develop the prediction model. The model’s efficacy was then validated in the independent validation cohort. Table 3 The predictive performance of different combinations of radiomics and clinical features Model AUC 95%CI C_index Radiomics 0.828 (0.722–0.934) 0.726 Clinical 0.831 (0.728–0.934) 0.754 Minimum 0.823 (0.719–0.927) 0.735 Maximum 0.864 (0.772–0.957) 0.761 0.1 Rad + 0.9 Cli 0.835 (0.732–0.937) 0.759 0.2 Rad + 0.8 Cli 0.848 (0.749–0.947) 0.766 0.3 Rad + 0.7 Cli 0.850 (0.753–0.947) 0.766 0.4 Rad + 0.6 Cli 0.855 (0.759–0.950) 0.769 0.5 Rad + 0.5 Cli 0.846 (0.747–0.944) 0.764 0.6 Rad + 0.4 Cli 0.839 (0.740–0.938) 0.756 0.7 Rad + 0.3 Cli 0.839 (0.739–0.939) 0.746 0.8 Rad + 0.2 Cli 0.840 (0.740–0.939) 0.739 0.9 Rad + 0.1 Cli 0.835 (0.733–0.938) 0.730 4 Discussion In this study, we established 0.015 as the critical TLVR threshold for survival stratification analysis of CRLM patients, where there was an association between elevated TLVR and significantly shortened OS under different treatment modalities of either surgery or systemic therapy. We further developed a fusion model by combining CT features and clinical data. The model demonstrated satisfactory performance in both the training and validation set (with respective AUCs of 0.847 and 0.855), indicating that the fusion model reached potent generalization ability in different cohorts. Unlike the traditional clinical risk scores including the Fong score relying on static integer counts [ 17 ] or the absolute measurement of TTV [ 6 , 18 ], TLVR incorporates the physiological context of the future liver remnant. Previous studies have demonstrated that TTV is a more accurate surrogate for tumor burden than maximal tumor diameter because it takes non-spherical tumor shapes into account. However, TTV alone does not reflect the functional reserve of the liver. In our working model, a TLVR ≥ 0.015 was associated with significantly poorer clinical outcomes (P < 0.05), suggesting that this parameter captures not only the spatial occupation of the tumor but also the relative compromise of hepatic reserve. This is consistent with the hypothesis that a high relative tumor volume correlates with \"geometric metastatic spread\" and a higher likelihood of micrometastases that are undetectable by conventional imaging [ 19 ]. Therefore, TLVR may serve as a more refined biological indicator than absolute volume alone. The volumetric stratification provided by TLVR offers potential solutions to the clinical dilemma highlighted by the JCOG0603 trial regarding adjuvant chemotherapy [ 20 ]. Our results suggest that patients with a low volumetric burden (TLVR < 0.015) may represent a subgroup with a favorable prognosis, for whom the toxicity of adjuvant therapy, such as sinusoidal obstruction syndrome or chemotherapy-associated steatohepatitis [ 21 ], might outweigh the oncological benefits. Conversely, a suprathreshold TLVR indicates a high risk of occult disease, warranting aggressive systemic therapy. A persistently high ratio may imply that the rate of parenchymal injury from cytotoxicity exceeds the rate of tumor regression. In such cases of functional inoperability, alternative non-surgical modalities might be prioritized to avoid futile resections that compromise the functional liver reserve. Beyond macroscopic volume, this study utilized CT-based radiomics to non-invasively characterize tumor heterogeneity. It was reported that radiomics features can decode the tumor microenvironment and provide prognostic information that is not visible to the naked eye [ 14 ]. By employing LASSO regression, we identified 15 key radiomic features, primarily texture and wavelet indices, which reflect voxel-level tissue complexity. Consistent with the previous findings [ 22 , 23 ], our radiomics signature demonstrated strong predictive capability (AUC = 0.828). The inclusion of wavelet transformation features was particularly crucial for revealing sub-visual spatial patterns. Methodologically, the use of the random survival forest algorithm allowed for the modeling of complex, non-linear survival associations that traditional Cox proportional hazards models may miss [ 24 ]. The finding that the optimal fusion strategy assigned a slightly higher weight to clinical features (0.6) than to radiomics (0.4) implies that systemic host status remains the dominant driver of prognosis. However, the radiomics signature contributes critical supplementary biological information that standard clinical metrics cannot capture, thereby refining the prediction accuracy. Collectively, this study shows that combining anatomical and textural data better characterizes the tumor phenotype than either method alone. LMR was identified as the sole independent protective factor in our analysis. This metric likely reflects the balance of the host-tumor immune interface. Lymphocytes mediate cytotoxic immune surveillance, while monocytes can differentiate into Tumor-Associated Macrophages (TAMs), promoting angiogenesis and immune evasion [ 25 ]. A lower LMR has been widely reported to correlate with poor prognosis in CRLM [ 26 – 28 ], and our findings reinforce its value as a systemic immunological marker complementing the local anatomical information provided by TLVR. Additionally, bilobar tumor distribution remained a significant risk factor. As a core component of the classic Fong score [ 17 ], bilobar distribution indicates extensive spatial dispersion, representing both a technical challenge for R0 resection and a marker of biologically advanced disease, independent of TTV. Several limitations of this study should be acknowledged. First, the retrospective, single-center design introduces inherent selection bias and potential unknown confounders. Specifically, the patient population was derived from a single institution, which may limit the generalizability of the model to broader demographics or different healthcare settings. Second, the potential of the fusion model to guide specific therapeutic decisions remains to be tested. Although our subgroup analyses suggest that the prognostic value of TLVR is robust across different treatment cohorts, it is unclear whether the model can predict chemosensitivity or determine the optimal timing for surgery. Future studies should evaluate whether this model can identify patients who would benefit most from neoadjuvant chemotherapy versus immediate surgery. 5 Conclusion This study proposes the Tumor-to-Liver Volume Ratio (TLVR) as a novel physiological volumetric index and demonstrates its independent prognostic significance for colorectal cancer liver metastases (CRLM). We have established an efficient, non-invasive and precise prognostic model of TLVR combined with indicators of tumor burden, host immunity and microscopic heterogeneity. The model quantifies their interactions and offers a basis for individualized, precision treatment of CRLM patients, supporting broad clinical applicability. Abbreviations ALBI Albumin-bilirubin AUC Area under the curve BMI Body mass index CA19-9 Carbohydrate antigen 19-9 CEA Carcinoembryonic antigen CT Computed tomography CRC Colorectal cancer CRLM Colorectal liver metastasis DCA Decision curve analysis LASSO Least absolute shrinkage and selection operator LMR Lymphocyte-to-monocyte ratio NLR Neutrophil-to-lymphocyte ratio OS Overall survival PLR Platelet-to-lymphocyte ratio ROC Receiver-operating characteristic ROI Region of interest TLV Total liver volume TLVR Tumor-to-liver volume ratio TTV Total tumor volume Declarations Competing interests All authors have no conflicts of interest to declare. Ethics approval and consent to participate Ethics approval for this study was granted from the local ethical committee of The Affiliated Tumor Hospital of Guangxi Medical University (NO. KY2024230),and the study was performed in accordance with the principles of the Declaration of Helsinki. Consent for publication Not applicable. Funding This work was supported by the Natural Science Foundation of Guangxi Province (Grant number: 2018GXNSFAA294013), Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation under (Grant number: 2024GXNSFAA010011) and Guangxi Medical and Health Appropriate Technology Development and Promotion and Application Project (Grant number: S2023089). Author Contribution XH and HZ conceived the study. HX and GL carried out the research. SL, HZ and YG analysed the data.HZ and YG wrote the paper. All authors read and approved the final manuscript. Acknowledgements. Not applicable. Data Availability The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request. References World Health Organization (WHO). Colorectal Cancer. Available online: https://www.who.int/news-room/fact-sheets/detail/colorectal-cancer (accessed on 10 December 2025). Siegel RL et al. Colorectal cancer statistics, 2023. CA: a cancer journal for clinicians vol. 73,3 (2023): 233–54. 10.3322/caac.21772 Van Cutsem E, Cervantes A, Adam R, et al. ESMO consensus guidelines for the management of patients with metastatic colorectal cancer. Ann Oncol. 2016;27(8):1386–422. 10.1093/annonc/mdw235 . Zeineddine FA et al. Feb. Survival improvement for patients with metastatic colorectal cancer over twenty years. NPJ precision oncology vol. 7,1 16. 13 2023, 10.1038/s41698-023-00353-4 Meng Q, et al. Preoperative nomogram to predict survival following colorectal cancer liver metastasis simultaneous resection. J Gastrointest Oncol vol. 2021;12(2):556–67. 10.21037/jgo-20-329 . Tai K, et al. Total tumour volume as a prognostic factor in patients with resectable colorectal cancer liver metastases. BJS open vol. 2020;4(3):456–66. 10.1002/bjs5.50280 . Zhou SK, Greenspan H, Davatzikos C et al. A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises. Proc IEEE Inst Electr Electron Eng. 2021;109(5):820–838. 10.1109/JPROC.2021.3054390 Litjens G, et al. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60–88. 10.1016/j.media.2017.07.005 . Choy G, et al. Current Applications and Future Impact of Machine Learning in Radiology. Radiol vol. 2018;288(2):318–28. 10.1148/radiol.2018171820 . Cavallo AU, et al. CT and MRI radiomics in cardiovascular risk prediction: a systematic review and meta-analysis by the EuSoMII Radiomics Auditing Group. European radiology, 10.1007/s00330-025-12236-2. 24 Dec. 2025. 10.1007/s00330-025-12236-2 . Li Z, et al. Multi-sequence MRI deep learning and habitat radiomics for predicting mismatch repair status and prognosis in colorectal liver metastasis: a multicenter study. La Radiologia medica, 10.1007/s11547-025-02155-z. 13 Nov. 2025. 10.1007/s11547-025-02155-z . Li Z-F, et al. Dual-Time-Point Radiomics for Prognosis Prediction in Colorectal Liver Metastasis Treated with Neoadjuvant Therapy Before Radical Resection: A Two-Center Study. Annals Surg Oncol vol. 2025;32(5):3516–25. 10.1245/s10434-025-16941-6 . Aerts HJWL et al. Jun. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nature communications vol. 5 4006. 3 2014, 10.1038/ncomms5006 Gillies RJ, et al. Radiomics: Images Are More than Pictures. They Are Data Radiol vol. 2016;278(2):563–77. 10.1148/radiol.2015151169 . Inchingolo R, et al. Radiomics in colorectal cancer patients. World J Gastroenterol vol. 2023;29:2888–904. 10.3748/wjg.v29.i19.2888 . Rompianesi G, et al. Artificial intelligence in the diagnosis and management of colorectal cancer liver metastases. World J Gastroenterol vol. 2022;28(1):108–22. 10.3748/wjg.v28.i1.108 . Fong Y, Fortner J, Sun RL, et al. Clinical score for predicting recurrence after hepatic resection for metastatic colorectal cancer: analysis of 1001 consecutive cases. Ann Surg. 1999;230:309–18. Wesdorp NJ, et al. The Prognostic Value of Total Tumor Volume Response Compared With RECIST1.1 in Patients With Initially Unresectable Colorectal Liver Metastases Undergoing Systemic Treatment. Annals of surgery open: perspectives of surgical history, education, and clinical approaches 2,4 e103. 28 Oct. 2021. 10.1097/AS9.0000000000000103 . Mühlberg A, et al. The relevance of CT-based geometric and radiomics analysis of whole liver tumor burden to predict survival of patients with metastatic colorectal cancer. Eur Radiol vol. 2021;31(2):834–46. 10.1007/s00330-020-07192-y . Kanemitsu Y, et al. Hepatectomy Followed by mFOLFOX6 Versus Hepatectomy Alone for Liver-Only Metastatic Colorectal Cancer (JCOG0603): A Phase II or III Randomized Controlled Trial. J Clin oncology: official J Am Soc Clin Oncol vol. 2021;39(34):3789–99. 10.1200/JCO.21.01032 . Granata V, et al. Liver radiologic findings of chemotherapy-induced toxicity in liver colorectal metastases patients. Eur Rev Med Pharmacol Sci vol. 2019;23(22):9697–706. 10.26355/eurrev_201911_19531 . Li Z-F, et al. Dual-Time-Point Radiomics for Prognosis Prediction in Colorectal Liver Metastasis Treated with Neoadjuvant Therapy Before Radical Resection: A Two-Center Study. Annals Surg Oncol vol. 2025;32(5):3516–25. 10.1245/s10434-025-16941-6 . Granata V, et al. Radiomics and machine learning analysis by computed tomography and magnetic resonance imaging in colorectal liver metastases prognostic assessment. La Radiologia Med vol. 2023;128:1310–32. 10.1007/s11547-023-01710-w . Ma W et al. Dec. Machine learning models and restricted cubic spline were employed to analyze and predict postoperative ischemic stroke in type A aortic dissection patients. BMC cardiovascular disordersvol. 26,1 34. 10 2025, 10.1186/s12872-025-05375-3 Kwiecień I et al. Sep. Blood Monocyte Subsets with Activation Markers in Relation with Macrophages in Non-Small Cell Lung Cancer. Cancers vol. 12,9 2513. 4 2020, 10.3390/cancers12092513 Peng J et al. Jul. Preoperative lymphocyte-to-monocyte ratio represents a superior predictor compared with neutrophil-to-lymphocyte and platelet-to-lymphocyte ratios for colorectal liver-only metastases survival. OncoTargets and therapy vol. 10 3789–3799. 27 2017, 10.2147/OTT.S140872 McCluney SJ, et al. Predicting complications in hepatic resection for colorectal liver metastasis: the lymphocyte-to-monocyte ratio. ANZ J Surg vol. 2018;88:E782–6. 10.1111/ans.14725 . Facciorusso A et al. Lymphocyte-to-monocyte ratio predicts survival after radiofrequency ablation for colorectal liver metastases. World journal of gastroenterology vol. 22,16 (2016): 4211-8. 10.3748/wjg.v22.i16.4211 Additional Declarations No competing interests reported. Supplementary Files Additionalfile.docx Supplementary information File name：Additional file 1 File format：docx Title of data：Multimodal Fusion for Predicting Prognosis in Colorectal Cancer Liver Metastases SUPPLEMENTARY MATERIAL Description of data: Performance evaluation and comparison of the prognostic models. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 05 May, 2026 Reviewers agreed at journal 01 Apr, 2026 Reviewers agreed at journal 29 Mar, 2026 Reviewers invited by journal 25 Mar, 2026 Editor invited by journal 04 Mar, 2026 Editor assigned by journal 03 Mar, 2026 Submission checks completed at journal 03 Mar, 2026 First submitted to journal 02 Mar, 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-9008242\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":611782071,\"identity\":\"f329921f-fbb5-4e65-9e4a-f49383e96ec5\",\"order_by\":0,\"name\":\"Haichao Zou\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Guangxi Medical University Cancer Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Haichao\",\"middleName\":\"\",\"lastName\":\"Zou\",\"suffix\":\"\"},{\"id\":611782072,\"identity\":\"d099cdc3-7f8f-4ee5-9752-3ab90ec2e69d\",\"order_by\":1,\"name\":\"Haiyang Xie\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Guangxi Medical University Cancer Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Haiyang\",\"middleName\":\"\",\"lastName\":\"Xie\",\"suffix\":\"\"},{\"id\":611782073,\"identity\":\"678c8dad-c9df-4f4f-a3a0-fd0a3fac8b6d\",\"order_by\":2,\"name\":\"Guoxin Liang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Guangxi Medical University Cancer Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Guoxin\",\"middleName\":\"\",\"lastName\":\"Liang\",\"suffix\":\"\"},{\"id\":611782075,\"identity\":\"3beec42c-94e2-4528-9467-dbb30106ad99\",\"order_by\":3,\"name\":\"Shanshan Luo\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Guangxi Medical University Cancer Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Shanshan\",\"middleName\":\"\",\"lastName\":\"Luo\",\"suffix\":\"\"},{\"id\":611782078,\"identity\":\"733234bd-1519-43ee-9ec6-10a3dc97e67e\",\"order_by\":4,\"name\":\"Yumeng Guo\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Huashan Hospital, Fudan University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yumeng\",\"middleName\":\"\",\"lastName\":\"Guo\",\"suffix\":\"\"},{\"id\":611782079,\"identity\":\"acd141b9-6428-434b-b45a-2c2073b4998b\",\"order_by\":5,\"name\":\"Xinxin He\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYBACAyA+wMAgwcAPFWBsIFqLZAMpWiCMA8RqMWfvMTxc2GaRZ3x+jelmHgYb2Q0HmJ89wKfFsudYwuEZZySKzW48S7vNw5BmvOEAm7kBPi0GN5IPHOapkEjcduPwMaCWw4kbDvCwSeDVcv9hw2EeA4nEzTMOtgG1/CdCyw1miC0b+JtBthwgQsuZtITDPGckEmfcYEu7Occg2XjmYTYz/FqOnzH+zNtWl9jff8bsxpsKO9m+483P8GpBAIkEBkg0MROnHgj4DxCtdBSMglEwCkYYAAAEX051gG8G6gAAAABJRU5ErkJggg==\",\"orcid\":\"\",\"institution\":\"Guangxi Medical University Cancer Hospital\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Xinxin\",\"middleName\":\"\",\"lastName\":\"He\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2026-03-02 09:25:22\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-9008242/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-9008242/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":105571849,\"identity\":\"535f5277-7ed4-4de4-b011-784c9d1db302\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:24:47\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":622445,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eThe flowchart of patient inclusion.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"image1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/39d7acf56af909a4f55e2110.png\"},{\"id\":105571520,\"identity\":\"58f96b46-df99-41c6-a68a-e4c4ad943a82\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:23:14\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":8235009,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eDetermination of the optimal TLVR cutoff value and survival analysis. (a) Time-dependent ROC curve analysis in the training set for predicting 3-year OS. (b) Kaplan-Meier survival curves for patients in the validation set stratified by the TLVR cutoff (High vs. Low). (c, d) Subgroup survival analyses based on treatment modality in the validation set. Kaplan-Meier curves compare OS between high- and low-TLVR groups in (c) the surgery cohort and (d) the chemotherapy cohort.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"image2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/91bd55a7533d6b47eb103091.png\"},{\"id\":105571348,\"identity\":\"c1397ca4-82fc-460e-90fa-85f0d4a57740\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:22:51\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":14633986,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) Binomial deviance calculation for radiomic features; (b) Changes in coefficients during LASSO regularization; (c) Contribution of radiomic features to model performance.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"image3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/c8efb60cb76449a9d058bf41.png\"},{\"id\":105571850,\"identity\":\"105746f7-9165-4a25-acf9-3d389a81ab57\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:24:47\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":9889599,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003ePerformance evaluation and comparison of the prognostic models. (a, b) 3-year time-dependent ROC curves of the Clinical model, Radiomics model, and Fusion model in the (a) training set and (b) validation set. (c) Calibration plots of the three models in the validation set. (d) DCA in using radiomics model, clinic model, and fusion model.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"image4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/59ee0f0ff8b4164076dea910.png\"},{\"id\":105571524,\"identity\":\"b30340db-861e-4e03-b532-c86f541c44ee\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:23:15\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":348382,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eThe flowchart of our proposed prediction model.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"image5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/e17fc72706535cca68079d11.png\"},{\"id\":105728138,\"identity\":\"5c45d41b-79c7-4481-969b-d71a55df3711\",\"added_by\":\"auto\",\"created_at\":\"2026-03-30 11:10:05\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":22021477,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/9bcbc5a5-7bbb-406d-a843-02c8ccc1e88d.pdf\"},{\"id\":105571418,\"identity\":\"91c97551-4bd8-4a42-8ead-01d3b2136b1a\",\"added_by\":\"auto\",\"created_at\":\"2026-03-27 13:23:10\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":1532149,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eSupplementary information\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eFile name：Additional file 1\\u003c/p\\u003e\\n\\u003cp\\u003eFile format：docx\\u003c/p\\u003e\\n\\u003cp\\u003eTitle of data：Multimodal Fusion for Predicting Prognosis in Colorectal Cancer Liver Metastases\\u003c/p\\u003e\\n\\u003cp\\u003eSUPPLEMENTARY MATERIAL\\u003c/p\\u003e\\n\\u003cp\\u003eDescription of data: Performance evaluation and comparison of the prognostic models.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Additionalfile.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-9008242/v1/dba5919f74d1d9b448c0d2ea.docx\"}],\"financialInterests\":\"No competing interests reported.\",\"formattedTitle\":\"Tumor-to-liver volume ratio (TLVR)-integrated Radiomics-clinicopathological Fusion Model for Prognosis Prediction in Colorectal Cancer Liver Metastases\",\"fulltext\":[{\"header\":\"1 Introduction\",\"content\":\"\\u003cp\\u003eColorectal cancer (CRC) remains a significant global public health challenge, ranking as the third most common malignancy and the second leading cause of cancer-related deaths. Distant metastasis is the primary cause of treatment failure and mortality among CRC patients [\\u003cspan additionalcitationids=\\\"CR2\\\" citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e]. Colorectal cancer liver metastasis (CRLM) is the most prevalent form, affecting approximately 50% of patients during their disease progression [\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e]. Despite advancements in surgery and systemic therapies, the prognosis for CRLM patients varies widely, underscoring the urgent need for personalized precision medicine strategies to improve survival rates and quality of life [\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003eRecent studies indicate that total tumor volume (TTV) serves as an independent prognostic factor for CRLM [\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e]. However, TTV alone fails to account for inter-patient variability in liver volume. Consequently, identical TTV values may carry different clinical implications depending on individual liver size. The tumor-to-liver volume ratio (TLVR) addresses this limitation by integrating tumor burden with functional liver reserve, potentially offering a more accurate reflection of the dynamic balance between tumor aggressiveness and host compensatory capacity. Additionally, radiomics and machine learning have gained significant traction in radiology in recent years [\\u003cspan additionalcitationids=\\\"CR8\\\" citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e], demonstrating utility in disease prediction and survival analysis [\\u003cspan additionalcitationids=\\\"CR11\\\" citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e]. These approaches can evaluate radiological images using quantitative texture information known as imaging features [\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e]. Radiological texture features can assess prognostic indicators related to tumor heterogeneity and metastasis within the tumor microenvironment. Specifically, CRLM radiomics has emerged as a valuable tool for pre-operative diagnosis and post-treatment survival prediction [\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e16\\u003c/span\\u003e].\\u003c/p\\u003e \\u003cp\\u003eIn this study, we developed an integrated predictive model combining CT radiomics features with clinical characteristics based on TLVR. Utilizing the random forest algorithm, this model accurately forecasts OS in CRLM patients. Moreover, it significantly enhances risk stratification capabilities and facilitates personalized treatment planning and clinical management.\\u003c/p\\u003e\"},{\"header\":\"2 Methods\",\"content\":\"\\u003ch2\\u003e2.1 Ethical approval and cohort details\\u003c/p\\u003e \\u003c/h2\\u003e \\u003cp\\u003eThis retrospective cohort study consecutively enrolled all the patients diagnosed with colorectal liver metastases (CRLM) and received primary therapy (hepatectomy or chemotherapy) at the Affiliated Tumor Hospital of Guangxi Medical University from April 2013 to November 2019. Key inclusion criteria were the absence of extrahepatic metastatic diseases at the time of CRLM diagnosis. The exclusion criteria comprised: (1) documented extrahepatic metastases prior to CRLM diagnosis; (2) other malignant disease present during the study period; (3) insufficient clinicopathological data, baseline CT imaging, or follow-up information. As this study involve retrospective analysis of anonymized patient data, the requirement for written informed consent was waived. The study protocol was strictly in line with the principles of the Declaration of Helsinki and obtained the institutional ethics committee approval of the school of oncology, Guangxi Medical University in 2025.\\u003c/p\\u003e \\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.2 Collection of clinical data\\u003c/h2\\u003e \\u003cp\\u003eBaseline demographic and clinicopathologic variables were extracted from electronic medical records. Serological tumor markers, including carcinoembryonic antigen (CEA) and carbohydrate antigen 19\\u0026thinsp;\\u0026minus;\\u0026thinsp;9 (CA19-9), were documented. To comprehensively evaluate the systemic inflammatory and nutritional status, composite indices including the albumin-bilirubin (ALBI), neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR) and platelet-to-lymphocyte ratio (PLR) were calculated using results from preoperative blood tests.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec4\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.3 Image acquisition and segmentation\\u003c/h2\\u003e \\u003cp\\u003eAbdominal CT examinations were performed on a multi-detector row scanner (SOMATOM Definition Flash, Siemens Medical Systems). Images were reconstructed with a standard reconstruction kernel and stored in DICOM (Digital Imaging and Communications in Medicine) format. Regions-of-interest (ROI) encompassing the tumors and the entire liver were manually delineated slice by slice on portal venous phase images using 3D Slicer (version 5.8.1, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003ehttps://www.slicer.org/\\u003c/span\\u003e\\u003cspan address=\\\"https://www.slicer.org/\\\" targettype=\\\"URL\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e). Total tumor volume (TTV) and total liver volume (TLV) were subsequently computed from these segmentations. Two experienced radiologists independently performed the volumetric segmentation. Inter-observer variability was minimized by resolving any disagreements through open discussion to reach a final consensus.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec5\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.4 Extraction of radiomics features\\u003c/h2\\u003e \\u003cp\\u003eFollowing image acquisition and segmentation, cubic B-spline interpolation was applied to resample both the CT images and their ROI masks to an isotropic voxel size of 1 x 1 x 1 mm\\u003csup\\u003e3\\u003c/sup\\u003e to ensure a consistent spatial resolution. Radiomics feature extraction was performed using the pyradiomics library, a widely validated open-source tool for quantitative feature analysis of medical images.\\u003c/p\\u003e \\u003cp\\u003eFeature selection proceeded in two stages. First, univariate Cox proportional hazards regression analysis was performed to screen out features significantly associated with OS (p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05). Subsequently, these identified variables were subjected to the Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression to mitigate overfitting and enhance model generalizability. The optimal regularization parameter for the LASSO model was determined via ten-fold cross-validation to select the most informative features with non-zero coefficients.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec6\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.5 Definition of TLVR cut-off value\\u003c/h2\\u003e \\u003cp\\u003eThe TLVR cut-off values that would best predict OS were determined using receiver operating characteristic (ROC) curve analysis. The 3-year OS rates were used as objective variables.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec7\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.6 Selection of clinical features\\u003c/h2\\u003e \\u003cp\\u003ePatients were categorized into high- and low-TLVR groups according to the predefined OS cutoff value. Clinical variables potentially associated with prognostic were first evaluated using univariate Cox proportional-hazards regression with a significance threshold (P\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05). Variables demonstrating significant association in univariate analysis were subsequently entered into a multivariate Cox regression model. Stepwise bidirectional regression based on the akaike information criterion was used to identify the independent prognostic factors.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec8\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.7 Construction of prediction models\\u003c/h2\\u003e \\u003cp\\u003eAll features were standardized according to the distribution of the training set. A random survival forest classifier was then used to build two independent models: a radiomics model utilizing CT radiomic features and a clinical model utilizing clinicopathological features. The prediction scores generated by these two models were integrated via score-level fusion. Three fusion strategies were evaluated including minimum score fusion, maximum score fusion, and weighted score fusion. For the minimum and maximum strategies, the two model scores were compared case-by-case and the smaller or larger score, respectively, was selected as the fused result. For the weighted strategy, the fused score was the weighted sum of the two model scores, with weights summing to one.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec9\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.8 Statistical analysis\\u003c/h2\\u003e \\u003cp\\u003eThe primary endpoint was OS, defined as the interval from the date of initial treatment to the date of death from any cause. Data for patients who remained alive or were lost to follow-up were censored at the date of their last known contact. All statistical analyses were performed using Python (version 3.9) and R software (version 4.5.1). Continuous variables are presented as the median (i.q.r.) values and were analyzed using the Mann\\u0026ndash;Whitney U-test or student\\u0026rsquo;s t-test. Categorical variables are summarized as frequencies and percentages and were analyzed using the chi-squared test or Fisher\\u0026rsquo;s exact test. The survival rate was estimated using the Kaplan\\u0026ndash;Meier method, and the survival rate estimates were compared using the log-rank test. A two-sided p-value less than 0.05 was considered statistically significant.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"3 Results\",\"content\":\"\\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.1 Patient enrollment and patient baseline characteristics\\u003c/h2\\u003e \\u003cp\\u003eBetween April 2013 to November 2019, 218 out of 296 CRLM patients were retrospectively enrolled in this study according to the inclusion and exclusion criteria (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e). Then, this cohort was randomly partitioned into a training set (n\\u0026thinsp;=\\u0026thinsp;152) and a validation set (n\\u0026thinsp;=\\u0026thinsp;66). The baseline demographic and clinicopathological characteristics of the entire cohort are detailed in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e. Notably, the overall 3-year OS was 41.7%, with a median follow-up duration of 26 months.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab1\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 1\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eClinical characteristics of patients\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"5\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eClinicopathological features\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eTotal (n\\u0026thinsp;=\\u0026thinsp;218)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eTraining set (n\\u0026thinsp;=\\u0026thinsp;152)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eValidation set (n\\u0026thinsp;=\\u0026thinsp;66)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\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 \\u003cp\\u003eGender (male)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e145 (67)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e106 (70)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e39 (59)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.169\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eAge *\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e60 (50.25, 66)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e59 (50.75, 66)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e61 (50.25, 66.75)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.485\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eBMI *\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e22.49 (20.31, 24.77)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e22.34 (20.31, 24.55)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e22.77 (20.23, 25.63)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.453\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003ePrimary site\\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=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLocation (right)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e60 (28)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e44 (29)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e16 (24)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.583\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLymphatic invasion (+)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e122 (56)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e88 (58)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e34 (52)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.473\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLymph node metastasis (+)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e153 (70)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e108 (71)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e45 (68)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.791\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eMetastases\\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=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximal tumor diameter(cm)*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e3.2 (2, 5.38)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e3.2 (2, 4.95)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e3.5 (2.15, 5.88)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.292\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMultiple tumour number\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e128 (59)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e87 (57)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e41 (62)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.601\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTumor distribution (bilobar)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e123 (56)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e87 (57)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e36 (55)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.826\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMetachronous metastases\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e82 (38)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e59 (39)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e23 (35)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.687\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTLVR*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.86 (0.43, 9.21)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e1.91 (0.42, 8.8)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.75 (0.46, 9.48)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.881\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eLaboratory tests\\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=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eALBI *\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e-2.5 (-2.73, -2.24)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e-2.5 (-2.73, -2.25)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e-2.46 (-2.69, -2.2)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.384\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eNLR*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e2.57 (1.79, 3.85)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e2.53 (1.73, 3.8)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e2.69 (1.94, 3.9)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.415\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLMR*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e3.35 (2.33, 4.3)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e3.37 (2.39, 4.37)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e3.21 (2.3, 4.16)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.381\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePLR*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e166.91 (123.4, 234.13)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e164.71 (125.21, 232.61)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e171.41 (122.65, 241.23)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.788\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCEA*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e19.56 (5.59, 99.55)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e22.26 (6.71, 99.94)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e12.48 (4.46, 96.4)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.351\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCA19-9*\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e29.08 (7.94, 251.73)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e32.05 (9.57, 259.35)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e18 (4.56, 203.73)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.181\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003eValues in parentheses are percentages unless indicated otherwise; *values are median (i.q.r.). BMI, body mass index; TLVR, tumor-to-liver volume ratio; ALBI, albumin\\u0026ndash;bilirubin; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; CEA, carcinoembryonic antigen; CA19-9,carbohydrate antigen 19\\u0026thinsp;\\u0026minus;\\u0026thinsp;9.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec12\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.2 Determination of TLVR cut-off values and survival analysis\\u003c/h2\\u003e \\u003cp\\u003eA 3-year ROC analysis was utilized to assess the association between TLVR and OS within the training cohort. The analysis yielded an AUC of 0.677, determining an optimal discriminative cutoff of 0.015 for prognostic stratification (sensitivity 67%, specificity 63%; Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ea). In the validation cohort, patients with TLVR\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015 was significantly correlated with poorer OS (P\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.01) (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003eb). To account for treatment heterogeneity, subgroup analyses was performed based on therapeutic modality of either surgical resection or chemotherapy. The established TLVR cutoff (\\u0026ge;\\u0026thinsp;0.015) remained a robust predictor of reduced OS in both the surgical resection (Log-rank, P\\u0026thinsp;=\\u0026thinsp;0.076, Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ec) and the chemotherapy subgroup (Log-rank, P\\u0026thinsp;=\\u0026thinsp;0.012, Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ed), confirming its prognostic independence from therapeutic approach.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab2\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 2\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eUnivariable and multivariable Cox proportional hazards analysis of factors associated with overall survival\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"5\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\" morerows=\\\"1\\\" rowspan=\\\"2\\\"\\u003e \\u003cp\\u003eCharacteristics\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colspan=\\\"2\\\" nameend=\\\"c3\\\" namest=\\\"c2\\\"\\u003e \\u003cp\\u003eUnivariable analysis\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colspan=\\\"2\\\" nameend=\\\"c5\\\" namest=\\\"c4\\\"\\u003e \\u003cp\\u003eMultivariable analysis\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eHazard ratio (95% CI)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003ep_value\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eHazard ratio (95% CI)\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003ep_value\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eGender (male)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.37(0.95\\u0026ndash;1.98)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.089\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eAge\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.03(0.87\\u0026ndash;1.24)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.705\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eBMI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.97(0.83\\u0026ndash;1.16)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.801\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePrimary site\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLocation (right)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.25(0.86\\u0026ndash;1.82)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.241\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLymphatic invasion\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.44(1.02\\u0026ndash;2.06)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.282\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLymph node metastasis\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e2.44(1.61\\u0026ndash;3.69)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e2.30(1.51\\u0026ndash;3.53)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMetastases\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximal tumor diameter(cm)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.2(1.02\\u0026ndash;1.42)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.033\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMultiple tumour number\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.23(0.87\\u0026ndash;1.34)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.248\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTumor distribution (bilobar)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.84(1.29\\u0026ndash;2.63)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.80(1.25\\u0026ndash;2.58)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.002\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMetachronous metastases\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.78(0.52\\u0026ndash;1.11)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.171\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTLVR\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e2.28(1.58\\u0026ndash;3.28)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.86(1.26\\u0026ndash;2.74)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.002\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLaboratory tests\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eALBI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.23(1.03\\u0026ndash;1.47)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.024\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eNLR\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.09(0.94\\u0026ndash;1.25)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.265\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eLMR\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.76(0.61\\u0026ndash;0.94)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.011\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e0.86(0.67\\u0026ndash;1.06)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.159\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003ePLR\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.16(0.97\\u0026ndash;1.31)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.120\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCEA\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.33(1.14\\u0026ndash;1.56)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e\\u0026lt;\\u0026thinsp;0.001\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.15(0.97\\u0026ndash;1.36)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e0.099\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eCA19-9\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e1.29(1.08\\u0026ndash;1.57)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.031\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e\\u0026nbsp;\\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003eBMI, body mass index; TLVR, tumor-to-liver volume ratio; ALBI, albumin\\u0026ndash;bilirubin; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; CEA, carcinoembryonic antigen; CA19-9, carbohydrate antigen 19\\u0026thinsp;\\u0026minus;\\u0026thinsp;9.\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec13\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.3 Feature selection for clinical models\\u003c/h2\\u003e \\u003cp\\u003eThe associations between clinicopathological variables and OS in the CRLM patients are summarized in Table\\u0026nbsp;\\u003cspan refid=\\\"Tab2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e. Following Z-score normalization using training set parameters, univariate Cox regression identified eight clinical features significantly associated with OS (\\u003cem\\u003eP\\u003c/em\\u003e\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05). Subsequent multivariate stepwise Cox regression refined these to five independent prognostic factors. Lymph node metastasis, bilobar tumor distribution, TLVR\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015, and elevated CEA were independently associated with worse prognosis. Conversely, higher lymphocyte-to-monocyte ratio (LMR) demonstrated a protective effect.\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.4 Features selected for the CT radiomics model\\u003c/h2\\u003e \\u003cp\\u003eA total of 1, 316 radiomic texture features were extracted from the raw CT images using pyradiomics, encompassing features from both the original images as well as transformed images processed with exponential, gradient, logarithmic, square, square-root, and wavelet filters. After Z-score normalization, univariate Cox analysis was performed to identify 667 candidate features significantly associated with OS (\\u003cem\\u003eP\\u003c/em\\u003e\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05). To mitigate overfitting, L1-regularized LASSO regression was applied, retaining 15 radiomic features with non-zero coefficients. The selected features, their coefficients and relative importance rankings are visualized (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003ea-c).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.5 Model development and validation\\u003c/h2\\u003e \\u003cp\\u003eThree prognostic models were thereby developed for CRLM, including a radiomics feature-based model, a clinical feature-based model, and a fusion model. In the validation cohort, the AUC of the 3-year ROC curves was 0.828 (95% Cl: 0.722\\u0026ndash;0.934) for the radiomics model and 0.831 (95% Cl: 0.728\\u0026ndash;0.934) for the clinical model, respectively. To optimize predictive performance, a fusion model was constructed based on the multiple weighed fusion strategies combining radiomics and clinical prediction scores (Table\\u0026nbsp;\\u003cspan refid=\\\"Tab3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e). Generally, the fusion model demonstrated superior discriminative ability AUC and C-index as compared to the radiomics or clinical model alone. Notably, the fusion model employing a 0.4 x Radiation score (Rad)\\u0026thinsp;+\\u0026thinsp;0.6 x Clinical score (Cli) strategy achieved the highest validation AUC of 0.855 (95% CI: 0.759\\u0026ndash;0.950; p\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05 versus individual models). As visualized in the 3-year ROC curves (Figs.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ea and \\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003eb), the fusion approach exhibited superior discriminative capacity, significantly outperforming both the radiomics and clinical models. Calibration analysis further corroborated the model's reliability, demonstrating excellent agreement between predicted probabilities and observed survival outcomes (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ec). Moreover, Decision Curve Analysis (DCA) highlighted the clinical relevance of our approach: the fusion model conferred a robust net clinical benefit that matched or exceeded that of single-modality models across a broad range of decision thresholds (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ed). The model also demonstrated robust predictive performance at 1 and 5 years (Supplementary Figures \\u003cspan refid=\\\"MOESM1\\\" class=\\\"InternalRef\\\"\\u003eS1\\u003c/span\\u003ea-f). Taken together, the fusion model showed robust prognostic performance for CRLM and demonstrate potential value for precise stratification in clinical practice.\\u003c/p\\u003e \\u003cp\\u003eTo better illustrate our finding, the workflow of the construction of the CRLM prognostic model was presented (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e). Patients were randomly assigned to training and validation cohorts in a 7:3 ratio. After manual segmentation of tumor ROIs on CT images, radiomic features and clinical characteristics were selected from the training cohort to develop the prediction model. The model\\u0026rsquo;s efficacy was then validated in the independent validation cohort.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab3\\\" border=\\\"1\\\"\\u003e \\u003ccaption language=\\\"En\\\"\\u003e \\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 3\\u003c/div\\u003e \\u003cdiv class=\\\"CaptionContent\\\"\\u003e \\u003cp\\u003eThe predictive performance of different combinations of radiomics and clinical features\\u003c/p\\u003e \\u003c/div\\u003e \\u003c/caption\\u003e \\u003ccolgroup cols=\\\"3\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"char\\\" char=\\\".\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eModel\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eAUC 95%CI\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eC_index\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eRadiomics\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.828 (0.722\\u0026ndash;0.934)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.726\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eClinical\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.831 (0.728\\u0026ndash;0.934)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.754\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMinimum\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.823 (0.719\\u0026ndash;0.927)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.735\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximum\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.864 (0.772\\u0026ndash;0.957)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.761\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.1 Rad\\u0026thinsp;+\\u0026thinsp;0.9 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.835 (0.732\\u0026ndash;0.937)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.759\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.2 Rad\\u0026thinsp;+\\u0026thinsp;0.8 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.848 (0.749\\u0026ndash;0.947)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.766\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.3 Rad\\u0026thinsp;+\\u0026thinsp;0.7 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.850 (0.753\\u0026ndash;0.947)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.766\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.4 Rad\\u0026thinsp;+\\u0026thinsp;0.6 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.855 (0.759\\u0026ndash;0.950)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.769\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.5 Rad\\u0026thinsp;+\\u0026thinsp;0.5 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.846 (0.747\\u0026ndash;0.944)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.764\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.6 Rad\\u0026thinsp;+\\u0026thinsp;0.4 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.839 (0.740\\u0026ndash;0.938)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.756\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.7 Rad\\u0026thinsp;+\\u0026thinsp;0.3 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.839 (0.739\\u0026ndash;0.939)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.746\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.8 Rad\\u0026thinsp;+\\u0026thinsp;0.2 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.840 (0.740\\u0026ndash;0.939)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.739\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e0.9 Rad\\u0026thinsp;+\\u0026thinsp;0.1 Cli\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e0.835 (0.733\\u0026ndash;0.938)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"char\\\" char=\\\".\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e0.730\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"4 Discussion\",\"content\":\"\\u003cp\\u003eIn this study, we established 0.015 as the critical TLVR threshold for survival stratification analysis of CRLM patients, where there was an association between elevated TLVR and significantly shortened OS under different treatment modalities of either surgery or systemic therapy. We further developed a fusion model by combining CT features and clinical data. The model demonstrated satisfactory performance in both the training and validation set (with respective AUCs of 0.847 and 0.855), indicating that the fusion model reached potent generalization ability in different cohorts.\\u003c/p\\u003e \\u003cp\\u003eUnlike the traditional clinical risk scores including the Fong score relying on static integer counts [\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e] or the absolute measurement of TTV [\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e], TLVR incorporates the physiological context of the future liver remnant. Previous studies have demonstrated that TTV is a more accurate surrogate for tumor burden than maximal tumor diameter because it takes non-spherical tumor shapes into account. However, TTV alone does not reflect the functional reserve of the liver. In our working model, a TLVR\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015 was associated with significantly poorer clinical outcomes (P\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.05), suggesting that this parameter captures not only the spatial occupation of the tumor but also the relative compromise of hepatic reserve. This is consistent with the hypothesis that a high relative tumor volume correlates with \\\"geometric metastatic spread\\\" and a higher likelihood of micrometastases that are undetectable by conventional imaging [\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e]. Therefore, TLVR may serve as a more refined biological indicator than absolute volume alone. The volumetric stratification provided by TLVR offers potential solutions to the clinical dilemma highlighted by the JCOG0603 trial regarding adjuvant chemotherapy [\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e]. Our results suggest that patients with a low volumetric burden (TLVR\\u0026thinsp;\\u0026lt;\\u0026thinsp;0.015) may represent a subgroup with a favorable prognosis, for whom the toxicity of adjuvant therapy, such as sinusoidal obstruction syndrome or chemotherapy-associated steatohepatitis [\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e], might outweigh the oncological benefits. Conversely, a suprathreshold TLVR indicates a high risk of occult disease, warranting aggressive systemic therapy. A persistently high ratio may imply that the rate of parenchymal injury from cytotoxicity exceeds the rate of tumor regression. In such cases of functional inoperability, alternative non-surgical modalities might be prioritized to avoid futile resections that compromise the functional liver reserve.\\u003c/p\\u003e \\u003cp\\u003eBeyond macroscopic volume, this study utilized CT-based radiomics to non-invasively characterize tumor heterogeneity. It was reported that radiomics features can decode the tumor microenvironment and provide prognostic information that is not visible to the naked eye [\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e]. By employing LASSO regression, we identified 15 key radiomic features, primarily texture and wavelet indices, which reflect voxel-level tissue complexity. Consistent with the previous findings [\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e], our radiomics signature demonstrated strong predictive capability (AUC\\u0026thinsp;=\\u0026thinsp;0.828). The inclusion of wavelet transformation features was particularly crucial for revealing sub-visual spatial patterns. Methodologically, the use of the random survival forest algorithm allowed for the modeling of complex, non-linear survival associations that traditional Cox proportional hazards models may miss [\\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e]. The finding that the optimal fusion strategy assigned a slightly higher weight to clinical features (0.6) than to radiomics (0.4) implies that systemic host status remains the dominant driver of prognosis. However, the radiomics signature contributes critical supplementary biological information that standard clinical metrics cannot capture, thereby refining the prediction accuracy. Collectively, this study shows that combining anatomical and textural data better characterizes the tumor phenotype than either method alone.\\u003c/p\\u003e \\u003cp\\u003eLMR was identified as the sole independent protective factor in our analysis. This metric likely reflects the balance of the host-tumor immune interface. Lymphocytes mediate cytotoxic immune surveillance, while monocytes can differentiate into Tumor-Associated Macrophages (TAMs), promoting angiogenesis and immune evasion [\\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e]. A lower LMR has been widely reported to correlate with poor prognosis in CRLM [\\u003cspan additionalcitationids=\\\"CR27\\\" citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e28\\u003c/span\\u003e], and our findings reinforce its value as a systemic immunological marker complementing the local anatomical information provided by TLVR. Additionally, bilobar tumor distribution remained a significant risk factor. As a core component of the classic Fong score [\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e], bilobar distribution indicates extensive spatial dispersion, representing both a technical challenge for R0 resection and a marker of biologically advanced disease, independent of TTV.\\u003c/p\\u003e \\u003cp\\u003eSeveral limitations of this study should be acknowledged. First, the retrospective, single-center design introduces inherent selection bias and potential unknown confounders. Specifically, the patient population was derived from a single institution, which may limit the generalizability of the model to broader demographics or different healthcare settings. Second, the potential of the fusion model to guide specific therapeutic decisions remains to be tested. Although our subgroup analyses suggest that the prognostic value of TLVR is robust across different treatment cohorts, it is unclear whether the model can predict chemosensitivity or determine the optimal timing for surgery. Future studies should evaluate whether this model can identify patients who would benefit most from neoadjuvant chemotherapy versus immediate surgery.\\u003c/p\\u003e\"},{\"header\":\"5 Conclusion\",\"content\":\"\\u003cp\\u003eThis study proposes the Tumor-to-Liver Volume Ratio (TLVR) as a novel physiological volumetric index and demonstrates its independent prognostic significance for colorectal cancer liver metastases (CRLM). We have established an efficient, non-invasive and precise prognostic model of TLVR combined with indicators of tumor burden, host immunity and microscopic heterogeneity. The model quantifies their interactions and offers a basis for individualized, precision treatment of CRLM patients, supporting broad clinical applicability.\\u003c/p\\u003e\"},{\"header\":\"Abbreviations\",\"content\":\"\\u003cp\\u003eALBI \\u0026nbsp; \\u0026nbsp; Albumin-bilirubin\\u003c/p\\u003e\\n\\u003cp\\u003eAUC \\u0026nbsp; \\u0026nbsp; Area under the curve\\u003c/p\\u003e\\n\\u003cp\\u003eBMI \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Body mass index\\u003c/p\\u003e\\n\\u003cp\\u003eCA19-9 \\u0026nbsp; \\u0026nbsp;Carbohydrate antigen 19-9\\u003c/p\\u003e\\n\\u003cp\\u003eCEA \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Carcinoembryonic antigen\\u003c/p\\u003e\\n\\u003cp\\u003eCT \\u0026nbsp; \\u0026nbsp; \\u0026nbsp; Computed tomography\\u003c/p\\u003e\\n\\u003cp\\u003eCRC \\u0026nbsp; \\u0026nbsp; Colorectal cancer\\u003c/p\\u003e\\n\\u003cp\\u003eCRLM \\u0026nbsp; Colorectal liver metastasis\\u003c/p\\u003e\\n\\u003cp\\u003eDCA \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Decision curve analysis\\u003c/p\\u003e\\n\\u003cp\\u003eLASSO \\u0026nbsp; Least absolute shrinkage and selection operator\\u003c/p\\u003e\\n\\u003cp\\u003eLMR \\u0026nbsp; \\u0026nbsp; Lymphocyte-to-monocyte ratio\\u003c/p\\u003e\\n\\u003cp\\u003eNLR \\u0026nbsp; \\u0026nbsp; Neutrophil-to-lymphocyte ratio\\u003c/p\\u003e\\n\\u003cp\\u003eOS \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Overall survival\\u003c/p\\u003e\\n\\u003cp\\u003ePLR \\u0026nbsp; \\u0026nbsp; Platelet-to-lymphocyte ratio\\u003c/p\\u003e\\n\\u003cp\\u003eROC \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Receiver-operating characteristic\\u003c/p\\u003e\\n\\u003cp\\u003eROI \\u0026nbsp; \\u0026nbsp; \\u0026nbsp; Region of interest\\u003c/p\\u003e\\n\\u003cp\\u003eTLV \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Total liver volume\\u003c/p\\u003e\\n\\u003cp\\u003eTLVR \\u0026nbsp; \\u0026nbsp;Tumor-to-liver volume ratio\\u003c/p\\u003e\\n\\u003cp\\u003eTTV \\u0026nbsp; \\u0026nbsp; \\u0026nbsp;Total tumor volume\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003ch2\\u003e\\u0026nbsp;\\u003c/h2\\u003e\\n\\u003ch2\\u003eCompeting interests\\u003c/h2\\u003e\\n\\u003cp\\u003eAll authors have no conflicts of interest to declare.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eEthics approval and consent to participate\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eEthics approval for this study was granted from the local ethical committee of The Affiliated Tumor Hospital of Guangxi Medical University (NO. KY2024230),and the study was performed in accordance with the principles of the Declaration of Helsinki.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eConsent for publication\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eNot applicable.\\u003c/p\\u003e\\n\\u003ch2\\u003eFunding\\u003c/h2\\u003e\\n\\u003cp\\u003eThis work was supported by the Natural Science Foundation of Guangxi Province\\u003c/p\\u003e\\n\\u003cp\\u003e(Grant number: 2018GXNSFAA294013), Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation under (Grant number: 2024GXNSFAA010011) and Guangxi Medical and Health Appropriate Technology Development and Promotion and Application Project (Grant number: S2023089).\\u003c/p\\u003e\\n\\u003ch2\\u003eAuthor Contribution\\u003c/h2\\u003e\\n\\u003cp\\u003eXH and HZ conceived the study. HX and GL carried out the research. SL, HZ and YG analysed the data.HZ and YG wrote the paper. All authors read and approved the final manuscript.\\u003c/p\\u003e\\n\\u003ch2\\u003eAcknowledgements.\\u003c/h2\\u003e\\n\\u003cp\\u003eNot applicable.\\u003c/p\\u003e\\n\\u003ch2\\u003eData Availability\\u003c/h2\\u003e\\n\\u003cp\\u003eThe datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\u003cli\\u003e\\u003cspan\\u003eWorld Health Organization (WHO). Colorectal Cancer. Available online: \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003ehttps://www.who.int/news-room/fact-sheets/detail/colorectal-cancer\\u003c/span\\u003e\\u003cspan address=\\\"https://www.who.int/news-room/fact-sheets/detail/colorectal-cancer\\\" targettype=\\\"URL\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e (accessed on 10 December 2025).\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSiegel RL et al. Colorectal cancer statistics, 2023. CA: a cancer journal for clinicians vol. 73,3 (2023): 233\\u0026ndash;54. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.3322/caac.21772\\u003c/span\\u003e\\u003cspan address=\\\"10.3322/caac.21772\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eVan Cutsem E, Cervantes A, Adam R, et al. ESMO consensus guidelines for the management of patients with metastatic colorectal cancer. Ann Oncol. 2016;27(8):1386\\u0026ndash;422. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1093/annonc/mdw235\\u003c/span\\u003e\\u003cspan address=\\\"10.1093/annonc/mdw235\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eZeineddine FA et al. Feb. Survival improvement for patients with metastatic colorectal cancer over twenty years. NPJ precision oncology vol. 7,1 16. 13 2023, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1038/s41698-023-00353-4\\u003c/span\\u003e\\u003cspan address=\\\"10.1038/s41698-023-00353-4\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMeng Q, et al. Preoperative nomogram to predict survival following colorectal cancer liver metastasis simultaneous resection. J Gastrointest Oncol vol. 2021;12(2):556\\u0026ndash;67. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.21037/jgo-20-329\\u003c/span\\u003e\\u003cspan address=\\\"10.21037/jgo-20-329\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eTai K, et al. Total tumour volume as a prognostic factor in patients with resectable colorectal cancer liver metastases. BJS open vol. 2020;4(3):456\\u0026ndash;66. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1002/bjs5.50280\\u003c/span\\u003e\\u003cspan address=\\\"10.1002/bjs5.50280\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eZhou SK, Greenspan H, Davatzikos C et al. A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises. Proc IEEE Inst Electr Electron Eng. 2021;109(5):820\\u0026ndash;838. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1109/JPROC.2021.3054390\\u003c/span\\u003e\\u003cspan address=\\\"10.1109/JPROC.2021.3054390\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLitjens G, et al. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60\\u0026ndash;88. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1016/j.media.2017.07.005\\u003c/span\\u003e\\u003cspan address=\\\"10.1016/j.media.2017.07.005\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChoy G, et al. Current Applications and Future Impact of Machine Learning in Radiology. Radiol vol. 2018;288(2):318\\u0026ndash;28. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1148/radiol.2018171820\\u003c/span\\u003e\\u003cspan address=\\\"10.1148/radiol.2018171820\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eCavallo AU, et al. CT and MRI radiomics in cardiovascular risk prediction: a systematic review and meta-analysis by the EuSoMII Radiomics Auditing Group. European radiology, 10.1007/s00330-025-12236-2. 24 Dec. 2025. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1007/s00330-025-12236-2\\u003c/span\\u003e\\u003cspan address=\\\"10.1007/s00330-025-12236-2\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLi Z, et al. Multi-sequence MRI deep learning and habitat radiomics for predicting mismatch repair status and prognosis in colorectal liver metastasis: a multicenter study. La Radiologia medica, 10.1007/s11547-025-02155-z. 13 Nov. 2025. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1007/s11547-025-02155-z\\u003c/span\\u003e\\u003cspan address=\\\"10.1007/s11547-025-02155-z\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLi Z-F, et al. Dual-Time-Point Radiomics for Prognosis Prediction in Colorectal Liver Metastasis Treated with Neoadjuvant Therapy Before Radical Resection: A Two-Center Study. Annals Surg Oncol vol. 2025;32(5):3516\\u0026ndash;25. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1245/s10434-025-16941-6\\u003c/span\\u003e\\u003cspan address=\\\"10.1245/s10434-025-16941-6\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eAerts HJWL et al. Jun. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nature communications vol. 5 4006. 3 2014, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1038/ncomms5006\\u003c/span\\u003e\\u003cspan address=\\\"10.1038/ncomms5006\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGillies RJ, et al. Radiomics: Images Are More than Pictures. They Are Data Radiol vol. 2016;278(2):563\\u0026ndash;77. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1148/radiol.2015151169\\u003c/span\\u003e\\u003cspan address=\\\"10.1148/radiol.2015151169\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eInchingolo R, et al. Radiomics in colorectal cancer patients. World J Gastroenterol vol. 2023;29:2888\\u0026ndash;904. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.3748/wjg.v29.i19.2888\\u003c/span\\u003e\\u003cspan address=\\\"10.3748/wjg.v29.i19.2888\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRompianesi G, et al. Artificial intelligence in the diagnosis and management of colorectal cancer liver metastases. World J Gastroenterol vol. 2022;28(1):108\\u0026ndash;22. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.3748/wjg.v28.i1.108\\u003c/span\\u003e\\u003cspan address=\\\"10.3748/wjg.v28.i1.108\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFong Y, Fortner J, Sun RL, et al. Clinical score for predicting recurrence after hepatic resection for metastatic colorectal cancer: analysis of 1001 consecutive cases. Ann Surg. 1999;230:309\\u0026ndash;18.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWesdorp NJ, et al. The Prognostic Value of Total Tumor Volume Response Compared With RECIST1.1 in Patients With Initially Unresectable Colorectal Liver Metastases Undergoing Systemic Treatment. Annals of surgery open: perspectives of surgical history, education, and clinical approaches 2,4 e103. 28 Oct. 2021. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1097/AS9.0000000000000103\\u003c/span\\u003e\\u003cspan address=\\\"10.1097/AS9.0000000000000103\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eM\\u0026uuml;hlberg A, et al. The relevance of CT-based geometric and radiomics analysis of whole liver tumor burden to predict survival of patients with metastatic colorectal cancer. Eur Radiol vol. 2021;31(2):834\\u0026ndash;46. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1007/s00330-020-07192-y\\u003c/span\\u003e\\u003cspan address=\\\"10.1007/s00330-020-07192-y\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKanemitsu Y, et al. Hepatectomy Followed by mFOLFOX6 Versus Hepatectomy Alone for Liver-Only Metastatic Colorectal Cancer (JCOG0603): A Phase II or III Randomized Controlled Trial. J Clin oncology: official J Am Soc Clin Oncol vol. 2021;39(34):3789\\u0026ndash;99. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1200/JCO.21.01032\\u003c/span\\u003e\\u003cspan address=\\\"10.1200/JCO.21.01032\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGranata V, et al. Liver radiologic findings of chemotherapy-induced toxicity in liver colorectal metastases patients. Eur Rev Med Pharmacol Sci vol. 2019;23(22):9697\\u0026ndash;706. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.26355/eurrev_201911_19531\\u003c/span\\u003e\\u003cspan address=\\\"10.26355/eurrev_201911_19531\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLi Z-F, et al. Dual-Time-Point Radiomics for Prognosis Prediction in Colorectal Liver Metastasis Treated with Neoadjuvant Therapy Before Radical Resection: A Two-Center Study. Annals Surg Oncol vol. 2025;32(5):3516\\u0026ndash;25. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1245/s10434-025-16941-6\\u003c/span\\u003e\\u003cspan address=\\\"10.1245/s10434-025-16941-6\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGranata V, et al. Radiomics and machine learning analysis by computed tomography and magnetic resonance imaging in colorectal liver metastases prognostic assessment. La Radiologia Med vol. 2023;128:1310\\u0026ndash;32. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1007/s11547-023-01710-w\\u003c/span\\u003e\\u003cspan address=\\\"10.1007/s11547-023-01710-w\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMa W et al. Dec. Machine learning models and restricted cubic spline were employed to analyze and predict postoperative ischemic stroke in type A aortic dissection patients. BMC cardiovascular disordersvol. 26,1 34. 10 2025, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1186/s12872-025-05375-3\\u003c/span\\u003e\\u003cspan address=\\\"10.1186/s12872-025-05375-3\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKwiecień I et al. Sep. Blood Monocyte Subsets with Activation Markers in Relation with Macrophages in Non-Small Cell Lung Cancer. Cancers vol. 12,9 2513. 4 2020, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.3390/cancers12092513\\u003c/span\\u003e\\u003cspan address=\\\"10.3390/cancers12092513\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePeng J et al. Jul. Preoperative lymphocyte-to-monocyte ratio represents a superior predictor compared with neutrophil-to-lymphocyte and platelet-to-lymphocyte ratios for colorectal liver-only metastases survival. OncoTargets and therapy vol. 10 3789\\u0026ndash;3799. 27 2017, \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.2147/OTT.S140872\\u003c/span\\u003e\\u003cspan address=\\\"10.2147/OTT.S140872\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMcCluney SJ, et al. Predicting complications in hepatic resection for colorectal liver metastasis: the lymphocyte-to-monocyte ratio. ANZ J Surg vol. 2018;88:E782\\u0026ndash;6. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.1111/ans.14725\\u003c/span\\u003e\\u003cspan address=\\\"10.1111/ans.14725\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e.\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFacciorusso A et al. Lymphocyte-to-monocyte ratio predicts survival after radiofrequency ablation for colorectal liver metastases. World journal of gastroenterology vol. 22,16 (2016): 4211-8. \\u003cspan class=\\\"ExternalRef\\\"\\u003e\\u003cspan class=\\\"RefSource\\\"\\u003e10.3748/wjg.v22.i16.4211\\u003c/span\\u003e\\u003cspan address=\\\"10.3748/wjg.v22.i16.4211\\\" targettype=\\\"DOI\\\" class=\\\"RefTarget\\\"\\u003e\\u003c/span\\u003e\\u003c/span\\u003e\\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\":\"info@researchsquare.com\",\"identity\":\"bmc-cancer\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"bcan\",\"sideBox\":\"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)\",\"snPcode\":\"\",\"submissionUrl\":\"https://www.editorialmanager.com/bcan/default.aspx\",\"title\":\"BMC Cancer\",\"twitterHandle\":\"BMC_series\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"em\",\"reportingPortfolio\":\"BMC Series\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true},\"keywords\":\"Radiomics, Colorectal cancer, Liver metastases, TLVR, Multimodal fusion\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-9008242/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-9008242/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003ch2\\u003eBackground\\u003c/h2\\u003e \\u003cp\\u003eColorectal cancer (CRC) ranks as the third most common malignancy worldwide, with colorectal liver metastasis (CRLM) being the leading cause of CRC-related mortality. In this study, we propose the tumor-to-liver volume ratio (TLVR) as a standardized physiological biomarker and develop a multimodal fusion model integrating TLVR, CT radiomics and clinicopathological factors to accurately predict overall survival (OS) in CRLM patients.\\u003c/p\\u003e\\u003ch2\\u003eMethods\\u003c/h2\\u003e \\u003cp\\u003eIn this retrospective study, 218 CRLM patients were enrolled and randomly divided into a training cohort (n\\u0026thinsp;=\\u0026thinsp;152) and a validation cohort (n\\u0026thinsp;=\\u0026thinsp;66). Radiomic features and clinical data were extracted from treatment-naive CT scans and medical records. The cut-off value for TLVR was determined by receiver operating characteristic (ROC) curve analyses. The random survival forest algorithm was used to construct the clinical, radiomics, and fusion models. The model performance was assessed with C-index, time-dependent area under the curve (AUC), and decision curve analysis (DCA).\\u003c/p\\u003e\\u003ch2\\u003eResults\\u003c/h2\\u003e \\u003cp\\u003eTLVR\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015 (1.5%) was significantly correlated with poorer OS for CRLM patients. Fifteen radiomic features and five clinical variables were incorporated for model construction. The fusion model demonstrated superior prognosic accuracy in the validation cohort with AUC of 0.855, compared to the clinical model (AUC\\u0026thinsp;=\\u0026thinsp;0.831) and the radiomics model (AUC\\u0026thinsp;=\\u0026thinsp;0.828).\\u003c/p\\u003e\\u003ch2\\u003eConclusion\\u003c/h2\\u003e \\u003cp\\u003eTLVR is a potent prognostic biomarker reflecting tumor-host volumetric equilibrium. Its integration into a CT radiomics-clinical fusion model significantly enhances OS prediction accuracy for CRLM patients. This non-invasive tool enables personalized therapeutic strategies, including TLVR-guided adjuvant therapy allocation and avoidance of futile conversion surgery.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Tumor-to-liver volume ratio (TLVR)-integrated Radiomics-clinicopathological Fusion Model for Prognosis Prediction in Colorectal Cancer Liver Metastases\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2026-03-27 12:26:05\",\"doi\":\"10.21203/rs.3.rs-9008242/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0},{\"type\":\"editorInvitedReview\",\"content\":\"\",\"date\":\"2026-05-05T06:44:26+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewerAgreed\",\"content\":\"23673651455399268550090860954132486750\",\"date\":\"2026-04-01T14:32:13+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewerAgreed\",\"content\":\"166217410699181997452778333156000015687\",\"date\":\"2026-03-29T04:59:40+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewersInvited\",\"content\":\"\",\"date\":\"2026-03-25T06:27:38+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorInvited\",\"content\":\"\",\"date\":\"2026-03-04T14:09:03+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorAssigned\",\"content\":\"\",\"date\":\"2026-03-03T08:25:39+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"checksComplete\",\"content\":\"\",\"date\":\"2026-03-03T08:21:54+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"submitted\",\"content\":\"BMC Cancer\",\"date\":\"2026-03-02T09:09:18+00:00\",\"index\":\"\",\"fulltext\":\"\"}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"bmc-cancer\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"bcan\",\"sideBox\":\"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)\",\"snPcode\":\"\",\"submissionUrl\":\"https://www.editorialmanager.com/bcan/default.aspx\",\"title\":\"BMC Cancer\",\"twitterHandle\":\"BMC_series\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"em\",\"reportingPortfolio\":\"BMC Series\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"d8a92ab1-b664-4e9c-82b3-5137b39bbd42\",\"owner\":[],\"postedDate\":\"March 27th, 2026\",\"published\":true,\"recentEditorialEvents\":[{\"type\":\"editorInvitedReview\",\"content\":\"\",\"date\":\"2026-05-05T06:44:26+00:00\",\"index\":58,\"fulltext\":\"\"}],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2026-03-27T12:26:05+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2026-03-27 12:26:05\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-9008242\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-9008242\",\"identity\":\"rs-9008242\",\"version\":[\"v1\"]},\"buildId\":\"XKTyCvWXoU3ODBz1xrDgd\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}