Feasibility of Prostate Apex Cancer Diagnosis Based on the Combination of Magnetic Resonance Imaging Radiomics and Biomarkers.

Abstract

Background: Traditional diagnostic methods have limitations in accurately identifying and characterising prostate apex cancer. Therefore, exploring innovative approaches such as magnetic resonance imaging (MRI) radiomics, biomarker assessments and clinical pathological features is essential to improve diagnostic accuracy.

Methods: This retrospective study evaluated diagnostic data from 52 patients with prostate apex cancer and 52 healthy individuals. MRI radiomics features-including grey-level non-uniformity, co-occurrence homogeneity, first order skewness, grey level co-occurrence matrix (GLCM) correlation, wavelet-low-high-low (wavelet-LHL) energy and prostate apparent diffusion coefficient (ADC) values-were compared between the groups. Biomarker levels, including Free Prostate-Specific Antigen (fPSA), Prostate-Specific Antigen (PSA), Ratio of Free to Total Prostate-Specific Antigen (f/tPSA), Prostate Volume (PV) and Prostate-Specific Antigen Density (PSAD), were also measured and analysed. Statistical analyses included t-tests, chi-square tests, correlation analyses and receiver operating characteristic (ROC) analyses.

Results: Significant differences were observed between the healthy and cancer groups in several MRI radiomics features: Grey-level non-uniformity (57.23 ± 7.31 vs. 69.54 ± 9.84, p < 0.001), co-occurrence homogeneity (0.29 ± 0.05 vs. 0.21 ± 0.07, p < 0.001), first order skewness (2.91 ± 0.61 vs. 3.85 ± 0.71, p < 0.001), GLCM correlation (0.72 ± 0.06 vs. 0.62 ± 0.07, p < 0.001), wavelet-LHL energy (264.14 ± 30.12 vs. 311.24 ± 42.13, p < 0.001) and prostate ADC value (1.29 ± 0.25 vs. 0.98 ± 0.15 × 10^-3 mm²/s, p < 0.001). Biomarker levels also differed significantly: fPSA (0.93 ± 0.50 vs. 1.97 ± 1.69 ng/mL^-1, p = 0.032), PSA (6.69 ± 2.55 vs. 17.45 ± 7.85 ng/mL^-1, p = 0.048), f/tPSA (0.14 ± 0.07 vs. 0.11 ± 0.07 ng/mL^-1, p = 0.020), PV (42.16 ± 8.32 vs. 38.43 ± 8.92 mL, p = 0.030) and PSAD (0.17 ± 0.08 vs. 0.49 ± 0.29 µg/L/mL^-1, p = 0.040). The combined model of these parameters achieved a sensitivity of 0.865, a specificity of 0.962 and an area under the curve of 0.913.

Conclusions: The integration of MRI radiomics, biomarker assessments and clinical pathological features presents a promising approach for diagnosing prostate apex cancer.