Preoperative prediction of muscle invasion in bladder cancer: a preliminary development and validation of diffusion-weighted-imaging-based habitat analysis

<title>Abstract</title> <p>Objectives To develop a model that integrated based on diffusion-weighted imaging (DWI) MRI for preoperative assessment of muscle-invasion (MI) in bladder cancer and to evaluate its diagnostic performance and impact on clinical decision-making. Methods One hundred and seventy-five patients with pathologically confirmed bladder urothelial carcinoma were prospectively included in this study. Each patient underwent 3.0 T MRI scan with T2WI, multi-b-value DWI, and dynamic contrast-enhanced imaging (DCE). Their clinical and imaging features were analyzed by two radiologists. Unsupervised clustering of DWI parameters resulted in four habitats. Finally, receiver operating characteristic (ROC) analysis, logistic regression, and validation were performed on the CMH model integrating clinical, MRI, and habitat factors. Results Habitat analysis showed significant differences between muscle-invasive and non-muscle-invasive bladder cancer groups. The habitat factors indicated a good predictive performance in both nomogram and univariate/multivariate regression analysis. Habitat 1, length, and stalk were the independent risk factors of MI. The CMH model integrating habitat, clinical and MRI factors achieved a better AUC of 0.777 than the CM model integrating clinical and MRI factors (0.740). The CMH model also had the best prediction efficiency in the calibration curve and DCA analysis. Model interpretability was affirmed by SHAP analysis. Conclusion Habitat analysis based on functional DWI sequences improves the assessment of bladder cancer muscle invasion and supports clinical decision-making.</p>