Rectal Cancer Lymphatic Metastasis Prediction Model Validation Data

127 patients with rectal cancer form a primary cohort for validating a deep learning prediction algorithm. The fusion model combining MRI-based algorithm predictions with clinical factors achieved an AUC of 0.873. Peng Zheng authored this retrospective validation study, last updated in May 2026.

Use Cases

• Validate MRI-based deep learning algorithms for predicting lymph node metastasis based on multiparameter MRI images.
• Construct integrated fusion models combining imaging predictions with clinical factors like CEA level and tumor differentiation.
• Benchmark automated prediction performance against manual radiologist interpretation to assess interobserver variability.
• Evaluate the clinical utility of predictive models using decision curve analysis across threshold probabilities.
• Assess model generalizability using an external validation cohort of 33 patients from two other centers.

Strengths

• Includes an external validation cohort of 33 patients from two other centers, providing a test for model generalizability.
• The fusion model demonstrated superior predictive performance with an AUC of 0.873 (95% CI: 0.804–0.972) in the primary cohort.
• The study compares algorithm performance against two radiologists, providing a benchmark for interobserver variability (kappa = 0.241).

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale ML training.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is small (354.2 KB), indicating limited scope and sample size.

Provenance

Source

figshare

Collection Method

Retrospective study of patients who underwent radical resection for rectal cancer without preoperative radiotherapy or chemotherapy.

Freshness

Last updated 2026-05-22 04:30:25; freshness should be verified.