AI-Based Evaluation of Prostate MR Imaging at a Modern Low-field 0.55 T Scanner Compared to 3 T in a Screening Cohort.

Abstract

Purpose: To evaluate the effect of lower field strength on quantitative apparent-diffusion-coefficient (ADC) values, contrast of the T2-weighted MR images and the performance of an AI-based segmentation.

Materials and methods: 25 screening clients (61.6 ± 7.5 years) from a study on a 3 T scanner were included and underwent a second examination on a 0.55 T scanner. Axial T2 weighted and diffusion-weighted images (DWI) sequences were acquired. An AI-based segmentation was performed. Based on this, the segmentation, volumetry, ADC values and the ratio of central gland (CG) and peripheral zone (PZ) in T2 weighting were compared by using correlations coefficient (Pearson), Bland-Altman plots and a non-inferior test with a paired t-test and a margin of ± 20% (lower and upper boundary).

Results: Volumetric assessment (peripheral zone//central gland) showed no significant (p = 0.13//0.38) difference between 3 T (mean volume: 14.81 (12.53-17.09)//23.07 (15.06-31.08)mL) and 0.55 T (mean volume of 14.29 (12.03-16.54; p = 0.13)//22.77 (14.41-31.12)mL). The deviation of the 0.55 T ADC values from the 3 T values was -10.14% (-16.09% to -4.18%) for the PZ and -4.68% (-10.12-0.76%) for the CG. Therefore, all confidence intervals remained within a margin of + /- 20% and thus demonstrated significant non-inferiority.

Conclusion: Biparametric prostate imaging is feasible at 0.55 T: ADC values vary within a common inter-scanner range compared to a 3 T and no difference can be observed regarding contrast ratio between peripheral zone and central gland in T2 weighted images, volumetry and AI-based segmentation compared to 3 T.