Machine learning methods can more efficiently predict prostate cancer compared with prostate-specific antigen density...

Area of Science:

• Urology
• Oncology
• Biostatistics

Background:

• Prostate-specific antigen (PSA) screening for prostate cancer lacks optimal accuracy.
• Machine learning methods (MLMs) offer a promising avenue to enhance diagnostic precision.

Purpose of the Study:

• To evaluate the efficacy of MLMs in improving prostate cancer detection accuracy.
• To compare MLM performance against conventional PSA-based screening parameters.

Main Methods:

• Utilized data from 512 patients undergoing prostate biopsy after PSA screening.
• Input variables included patient age, PSA levels, prostate volumes, and urinalysis findings.
• Evaluated artificial neural networks (ANNs), random forest, and support vector machine models, comparing AUC with PSA, PSA density, and PSA velocity.

Main Results:

• MLMs demonstrated superior Area Under the Curve (AUC) values compared to PSA level, PSA density, and PSA velocity.
• ANNs achieved the highest AUC (0.69), outperforming other MLMs and conventional parameters.
• MLM accuracies ranged from 71.6% to 72.1%, significantly exceeding those of PSA-based metrics.

Conclusions:

• Machine learning techniques, particularly ANNs, show significant potential for improving prostate cancer prediction.
• MLMs offer enhanced sensitivity and specificity over traditional PSA screening methods.