Classifying prostate cancer patients based on total prostate-specific antigen and free prostate-specific antigen...

Area of Science:

• Oncology
• Biostatistics
• Medical Informatics

Background:

• The prostate-specific antigen (PSA) test is crucial for prostate cancer (PCa) screening.
• Enhancing the sensitivity and specificity of the PSA test remains a clinical challenge.
• Integrating clinical and biological data may improve diagnostic accuracy.

Purpose of the Study:

• To develop an improved approach for PCa detection using PSA values, patient age, and prostate volume.
• To enhance the sensitivity of PCa screening while maintaining acceptable specificity.
• To create a predictive model for PCa diagnosis.

Main Methods:

• A novel approach combining statistical techniques and Support Vector Machine (SVM) was developed.
• Patients were classified into normal or abnormal groups using SVM.
• PCa prediction was performed on patients identified as abnormal.

Main Results:

• The system achieved a sensitivity of 95.1% and a specificity of 84.6%.
• The positive biopsy rate was 58%, with an unnecessary biopsy rate of 15.4%.
• A clinical decision support program for PCa prediction was developed.

Conclusions:

• Support Vector Machine (SVM) application significantly improved PSA test performance for PCa screening and detection.
• The study explored molecular information, potentially leading to new insights into cancer disease.
• The developed system offers a valuable tool for clinicians in PCa diagnosis.