Table 1_A CT-based radiomics model for preoperative prediction of lymphovascular invasion

252 patients were used to train a predictive model for lymphovascular invasion in colorectal cancer. Yingcheng Bai developed a nomogram integrating tumor volume, CT heterogeneity, and clinical biomarkers, achieving an AUC of 0.776 in training. The model was validated on 108 patients, with results published on figshare in April 2026.

Use Cases

• Developing a machine learning model for predicting lymphovascular invasion based on CT radiomics features.
• Evaluating the predictive power of clinical biomarkers like Carcinoembryonic Antigen and Neutrophil-to-Lymphocyte Ratio.
• Creating a clinical decision support tool for preoperative risk assessment in colorectal cancer.
• Comparing the performance of a combined nomogram against individual radiomic or clinical predictors.

Strengths

• The model was validated on a separate cohort of 108 patients, showing a consistent AUC of 0.722.
• Multivariate analysis identified eight independent predictors, including both imaging (Rad-score, tumor volume) and clinical (NLR, CEA) factors.
• The study reports specific performance metrics, including AUC values with 95% confidence intervals for both training and validation sets.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small (20.1 KB), suggesting limited scope or that the primary data is a model description rather than raw patient records.

Provenance

Source

figshare

Collection Method

Retrospective study of colorectal cancer patients, with data partitioned into training and validation sets.

Freshness

Last updated 2026-04-24 04:23:33; freshness should be verified.