EGE 12TH INTERNATIONAL CONFERENCE ON APPLIED SCIENCES
THE PROSTATE CANCER CLASSIFICATION AND IDENTIFICATION OF POTENTIAL BIOMARKERS WITH MACHINE LEARNING METHODS BASED ON CLINICAL DATA
Yayıncı:
Academy Global Publishing House
Prostate cancer is one of the most common types of cancer in men worldwide and usually occurs in older ages. Traditional methods commonly used in diagnosis (prostate specific antigen (PSA) test, digital rectal examination) have low sensitivity and specificity rates. Therefore, more reliable and sensitive methods are needed for early diagnosis. This study aims to develop classification models and identify potential biomarkers using machine learning methods in prostate cancer diagnosis. The clinical data of 298 individuals were used in the study. 131 of these individuals were diagnosed with prostate cancer, while the remaining 167 were classified as healthy. XGBoost and J48 algorithms were applied on the data. The XGBoost model was determined as the most successful model, achieving high performance results such as 92.6% accuracy, 95.4% sensitivity and 90.3% specificity. J48 exhibited lower accuracy (80.5%) and sensitivity (61.6%). As a result of the modeling, it has been revealed that five variables, namely total PSA (tPSA), PSA density (PSAD), prostate volume (PV), age and free PSA (fPSA), can be used as potential biomarkers in the prediction of prostate cancer. These biomarkers can contribute especially to the development of early diagnosis and personalized treatment strategies. Machine learning-based approaches offer promising methods in the management of multifactorial and complex diseases such as prostate cancer.
