Survival Prediction of Patients with Bladder Cancer after Cystectomy Based on Clinical, Radiomics, and Deep-Learning...

Area of Science:

• Urology
• Oncology
• Medical Imaging

Background:

• Accurate survival prediction is crucial for managing bladder cancer patients after radical cystectomy.
• Existing models often fail to integrate both clinical and radiological imaging data.
• There is a need for advanced predictive models that leverage multimodal data.

Purpose of the Study:

• To develop and evaluate an approach for improved bladder cancer survival prediction.
• To integrate clinical (C), radiomics (R), and deep-learning (D) descriptors.
• To enhance prognostic accuracy by combining diverse data sources.

Main Methods:

• Utilized data from 163 bladder cancer patients, including clinical, histopathological, and CT urography information.
• Analyzed clinical data using a nomogram and imaging data via radiomics and deep-learning models.
• Employed a Backpropagation Neural Network (BPNN) model for survival prediction with combined descriptors.

Main Results:

• The combined clinical, radiomics, and deep-learning (CRD) model achieved the highest Area Under the Curve (AUC) of 0.87 ± 0.05.
• CRD predictions showed a significant improvement over deep-learning (D) alone (p = 0.007).
• Kaplan-Meier analysis confirmed CRD's superior ability to stratify survival outcomes (p < 0.001).

Conclusions:

• Integrating clinical, radiomics, and deep-learning data offers a powerful strategy for predicting survival in bladder cancer patients post-cystectomy.
• The CRD approach demonstrates significant potential for improving patient management and treatment strategies.
• This multimodal data integration represents a promising advancement in oncological prognostication.