Purpose: The prostate-specific antigen density (PSAD) helps distinguish between benign prostatic hyperplasia (BPH) and prostate cancer. Accurate prostate volume (PV) assessment is necessary for PSAD calculation and both BPH diagnosis and treatment response monitoring; therefore, accurate PV measurement is increasingly becoming an essential step in the urology.
Methods: Magnetic resonance imaging was used for PV estimation. A new technique based on single-class support-vector machines (S SVM) for accurate PV estimation was realized. Three estimation methods were compared; method 1: planimetry (reference), method 2: S SVM based, and method 3: prolate ellipsoid.
Results: Method 1 and method 2 depict a strong correlation (Spearman's rank correlation coefficient ρ = 0.965, p > 0.001). The interrater reliability for method 1 and method 2 readings as expressed by the intraclass correlation coefficient (ICC) was 0.975 (p > 0.001). Comparison between method 3 and the two other methods shows ρ = 0.873 (p > 0.001), and ρ = 0.795 (p > 0.001), respectively. ICC was 0.54 and 0.505, respectively. The mean difference between method 1 and method 2 was -0.05 ml. The limits of agreement with the 95 % confidence interval were -3.8 to 3.7 ml. Comparing method 3 and the two other methods shows a worse agreement with mean difference of 8.6 ml (95 % confidence interval of 1.0-16.2 ml) and 8.6 ml (95 % confidence interval of -0.7 to 18.0 ml), respectively.
Conclusions: The prostate volumes obtained by our technique agreed excellently with the planimetry (reference) method. This new technique would be clinically useful for urologists in prostate volumetric analysis.