Intratumoral and Peritumoral Fat CT‑Based Radiomics for Predicting Recurrence Risk in Non-Muscle-Invasive Bladder...

Area of Science:

• Radiomics and Machine Learning in Oncology
• Medical Imaging and Diagnostics
• Urologic Oncology

Background:

• Non-muscle-invasive bladder cancer (NMIBC) frequently recurs, necessitating multiple surgeries and increasing patient burden.
• Current prediction methods for NMIBC recurrence are limited, highlighting the need for improved pre-treatment assessment tools.

Purpose of the Study:

• To develop and validate a computed tomography (CT)-based machine learning model for predicting pre-treatment recurrence in NMIBC.
• To integrate intratumoral and peritumoral fat radiomics features for enhanced predictive accuracy.

Main Methods:

• Retrospective analysis of 208 NMIBC patients undergoing enhanced CT before transurethral resection of bladder tumor (TURBT).
• Extraction and selection of radiomics features from intratumoral and peritumoral fat regions using LASSO.
• Development of intratumoral, peritumoral-fat, and combined radiomics models, integrated into a clinical-radiomics nomogram.

Main Results:

• The combined radiomics model achieved an AUC of 0.88 (test set) and 0.82 (external validation).
• The clinical-radiomics nomogram demonstrated strong predictive performance for early NMIBC recurrence (time-AUC 0.86 test, 0.84 validation).
• The nomogram significantly improved calibration and reclassification compared to the clinical model alone.

Conclusions:

• An integrated radiomics-clinical model significantly enhances the prediction of early NMIBC recurrence.
• This CT-based approach provides a novel and valuable strategy for assessing NMIBC recurrence risk.
• The model aids in guiding early clinical interventions and potentially reducing patient disease burden.