Develop prediction model to help forecast advanced prostate cancer patients' prognosis after surgery using neural...

Area of Science:

• Oncology
• Urology
• Medical Informatics

Background:

• The impact of surgery on advanced prostate cancer (PC) outcomes remains unclear.
• A predictive model for postoperative survival in advanced PC is currently lacking.

Purpose of the Study:

• To develop and validate a predictive model for overall survival in advanced prostate cancer patients undergoing surgery.
• To identify key clinical features that predict outcomes in advanced prostate cancer.

Main Methods:

• Utilized the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database.
• Collected clinical features from 6380 advanced prostate cancer patients.
• Developed a neural network model for survival prediction, validated on training and testing cohorts using Area Under the Curve (AUC).

Main Results:

• The neural network model incorporating all clinical features achieved an AUC of 0.7058 in the training cohort and 0.6925 in the test cohort.
• The model demonstrated predictive performance for overall survival in advanced prostate cancer patients.
• The predictive model was packaged into user-friendly software.

Conclusions:

• Advanced prostate cancer patients may benefit from surgical intervention.
• The developed clinical feature-based prognostic model offers accuracy and potential clinical decision-making support for advanced prostate cancer survival forecasting.