Prostate tumors are tumors that grow in or around the prostate gland of the male reproductive system. these tumors are classified into two, that is: benign and malignant. Malignant tissues are cancerous and more dangerous as they can easily spread to other soft tissues that are easily attacked by cancer cells, such as the lungs and liver. In contrast, benign tumors are not as dangerous and can be surgically removed and never regrow. The cancerous tumors has a high mortality rate, according to GLOBOCAN 2020, there were estimated 19.3 million cases with 10 million registered deaths over the same year. Current diagnosis of prostate tumor still lacks maximum precision, with the PSA and invasive biopsy methods suffering from most false positives and negatives. The study suggested a ML algorithm blending technique from three base models, SVM, RF, and XGBoost. For maximum accuracy and overfitting, the study performed SMOTE for class balancing and correlation elimination techniques. The blended model outperformed the base model with an accuracy of 0.981 at 0.05 confidence level, followed by RF and XGboost at 0.971. In other performance metrics, blended model had the highest sensitivity and specificity of 0.979 and 0.981, respectively. XGBoost and RF shared almost the same sensitivity of 0.97 and a higher specificity of 0.98. SVM had an overall low performance and was not recommended for such a task as a standalone model. The study recommends the incorporation of a blended model over each performance, although Random Forest and XGBoost are still viable for application.
Odongo et al. (Wed,) studied this question.