Objective:

To create a non-invasive machine learning model utilizing radiomic characteristics from CTU images to forecast the pathological grade of upper tract urothelial carcinoma (UTUC) and aid in preoperative diagnosis from June 2015 to June 2024.

Approach:

Key Findings:

• LGBM achieved the highest discriminative capability in the training dataset with an AUC of 0.945, sensitivity of 84.5%, and specificity of 91.2%.
• In the testing dataset, LGBM maintained strong generalizability with an AUC of 0.829.
• Key features influencing predictions were associated with the venous and arterial CTU phases.

Interpretation:

The radiomics-driven ML model shows potential in forecasting pathological grade for UTUC prior to surgery, providing a non-invasive and interpretable diagnostic tool.

Limitations:

• The study is retrospective and multicenter, which may introduce variability in results across different populations and settings.

Conclusion:

The established model offers a non-invasive, clinically interpretable instrument for enhancing diagnostic precision and assisting in individualized treatment strategies.

Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.