Application Value of Radiomics-Based Machine Learning for Preoperative Risk Stratification of Bladder Cancer:...

Area of Science:

• Radiology
• Oncology
• Artificial Intelligence

Background:

• Radiomics-based machine learning (ML) is being explored for bladder cancer risk stratification.
• Existing evidence for its effectiveness in detecting muscle invasion, high-grade tumors, and other factors is limited.

Purpose of the Study:

• To systematically evaluate the performance of radiomics-based ML in preoperative risk stratification for bladder cancer.
• To inform the development of intelligent risk assessment tools for bladder cancer.

Main Methods:

• Systematic retrieval of studies from Embase, Cochrane Library, PubMed, and Web of Science up to October 17, 2025.
• Assessment of risk of bias using the Prediction Model Risk of Bias Assessment Tool for Artificial Intelligence.
• Quantification of study quality using the Radiomics Quality Scoring tool and certainty of evidence using the GRADE framework.

Main Results:

• The meta-analysis included 57 studies with 11,933 participants, primarily focusing on muscle invasion and high-grade tumors.
• Pooled area under the receiver operating characteristic curve (AUROC) for muscle invasion detection ranged from 0.840 to 0.916 across different imaging modalities (CT, MRI, ultrasound).
• Combined models integrating clinical features with CT- or MRI-based radiomics showed high AUROCs (0.874-0.921) for muscle invasion and high-grade tumor detection.

Conclusions:

• This systematic review is the first to comprehensively assess radiomics for preoperative bladder cancer risk stratification.
• While promising, significant methodological shortcomings, high risk of bias, and low GRADE levels limit current clinical applicability.
• Future research should focus on standardizing methods, conducting multicenter studies, and validating external datasets.