MRI phenotype of the prostate: Transition zone radiomics analysis improves explanation of prostate-specific antigen (PSA) serum level compared to volume measurement alone

Abstract

PurposeTo determine the value of a radiomics MRI phenotype of the transition zone to explain PSA level in patients with low suspicion for clinically significant cancer to confirm hyperplastic changes. Materials and methodsT2 weighted images from 36 consecutive PI-RADS 2 and 3 cases with volume adapted systematic transperineal biopsy as reference standard (all biopsies negative, 34.8 biopsy cores per patient in average, mean PSA level 10.77 ng/mL) are manually segmented to define transition zone (TZ) volume. 54 radiomic features (RF) are derived for each TZ. RF are tested for significant correlation with PSA level, Bonferroni correction is applied. We build regression models to explain PSA level with a) TZ volume b) RF c) TZ volume+RF. We apply all models to a control group with clinically significant transition zone cancer. ResultsTZ volume is moderately correlated with PSA level (r = 0.44). 5/54 RF are significantly correlated with PSA level (r: 0.53−0.69, p < 0.05). Inclusion of each of these five features into the regression model significantly improves the explanatory value for PSA level (p < 0.05). Furthermore, RF alone better explain PSA level compared to TZ volume alone (p < 0.01). A systematic and significant trend for positive residuals is observed when regression models are applied to the malignant control group. ConclusionA radiomics analysis of the transition zone has the potential to improve explanation of corresponding PSA level in patients with low suspicion. This knowledge may reassure radiologists to read prostate MRI cases as unremarkable, despite present hyperplastic changes.