CT-Based Radiomics Nomogram for Preoperative Lymphovascular Invasion Prediction in CRC

Overview

A CT-based radiomics nomogram integrating tumor imaging features, clinical markers, and a radiomics score (Rad-score) was developed to preoperatively predict lymphovascular invasion (LVI) in colorectal cancer (CRC). The model demonstrated good discrimination with AUCs of 0.776 and 0.722 in training and validation cohorts, respectively, and showed clinical utility across risk thresholds.

Background

Lymphovascular invasion (LVI) is a key pathological feature indicating tumor aggressiveness and metastatic potential in colorectal cancer, influencing prognosis and adjuvant therapy decisions. Currently, LVI diagnosis depends on postoperative histopathology, limiting preoperative clinical decision-making. Radiomics enables extraction of quantitative imaging features that reflect tumor biology and heterogeneity, offering a non-invasive approach to predict pathological characteristics such as LVI. Prior imaging-based models have limitations including reliance on MRI or PET-CT, higher costs, or incomplete integration of clinical factors.

Data Highlights

Key Findings

• Multivariate logistic regression identified tumor volume, maximum tumor diameter, depth of tumor invasion, maximum short-axis diameter of regional lymph nodes, carcinoembryonic antigen (CEA) level, neutrophil-to-lymphocyte ratio (NLR), standard deviation of CT tumor parenchyma values, and Rad-score as independent predictors of LVI (P<0.05).
• The combined nomogram incorporating these predictors achieved robust discrimination with AUCs of 0.776 in training and 0.722 in validation cohorts.
• Calibration curves showed good agreement between predicted and observed LVI probabilities, indicating model reliability.
• Decision curve analysis demonstrated clinical net benefit of the nomogram across a wide range of threshold probabilities, supporting its practical utility.
• The Rad-score derived from CT radiomics features captures intratumoral heterogeneity, enhancing predictive accuracy beyond conventional clinical indicators.

Clinical Implications

This CT-based radiomics nomogram offers a non-invasive, preoperative tool to accurately predict LVI status in colorectal cancer patients, facilitating individualized treatment planning and risk stratification. Incorporation of routine clinical markers such as NLR and CEA alongside imaging features improves prediction performance, potentially guiding decisions on neoadjuvant or adjuvant therapies. The model's validation supports its integration into clinical workflows to enhance prognostic assessment without additional imaging burden.

Conclusion

The developed CT-based radiomics nomogram effectively predicts lymphovascular invasion preoperatively in colorectal cancer, combining imaging heterogeneity metrics with clinical factors. This approach holds promise for improving personalized management and prognostic evaluation in CRC patients.

References

1. Zhang et al. 2021 -- MRI-based Clinical-Radiomics Model for LVI Prediction
2. Yang et al. 2020 -- PET-CT Radiomics for LVI Assessment
3. Li et al. 2022 -- CT Radiomics Model Incorporating Tumor and Peritumoral Features
4. ACG Clinical Guidelines 2021 -- Colorectal Cancer Screening