AI-Assisted Nomogram for Preoperative Malignancy Risk in Pulmonary Nodules ≤ 3 cm

951 patient records from a retrospective study of surgically resected pulmonary nodules (210 benign, 741 malignant). The dataset was used to develop and internally validate an AI-assisted clinico-quantitative imaging prediction model, authored by Yingding Ruan and last updated in April 2026. The final model demonstrated an area under the curve (AUC) of 0.836 for malignancy risk discrimination.

Use Cases

• Training machine learning models for malignancy risk prediction based on quantitative CT features and clinical variables.
• Benchmarking new predictive models against the reported AUC of 0.836 and calibration metrics.
• Developing interactive risk calculators for clinical use, similar to the provided Shiny app.
• Studying the relationship between nodule characteristics (size, attenuation, type) and malignancy in a surgical cohort.

Strengths

• Model performance is quantitatively reported with an AUC of 0.836 and a mean absolute calibration error of 0.015.
• Dataset includes 951 patient records with a clear outcome distribution (741 malignant, 210 benign).
• The study protocol is detailed, including internal validation with 1,000 bootstrap resamples.

Limitations

• Row count and column-level documentation are unknown, which may limit suitability assessment.
• The cohort is restricted to surgically resected nodules, indicating a malignancy-enriched population that may not generalize to screening or incidental findings.
• Description metadata is limited; actual data quality and structure require manual inspection after download.

Provenance

Source

figshare, authored by Yingding Ruan.

Collection Method

Retrospective study analyzing data from consecutive patients who underwent surgical resection.

Time Range

null

Freshness

Last updated 2026-04-13 05:31:08; freshness should be verified.

Geography

null