Objective: The purpose of this study was to compare the reproducibility and diagnostic performance of 2D and 3D ROIs for prostate apparent diffusion coefficient (ADC) measurements.
Materials and methods: The study included 56 patients with prostate cancer undergoing 3-T MRI including DWI (b = 50 and 1000 s/mm2) before radical prostatectomy. Histologic findings from prostatectomy specimens were reviewed to denote each patient's dominant tumor and a benign region with visually decreased ADC. Three readers independently measured the ADCs of both areas using an ROI placed on a single slice through the lesion (2D) and an ROI encompassing all slices through the lesion (3D). Readers repeated measurements after 3 weeks. Assessment included Bland-Altman analysis (coefficient of repeatability [CR] in which lower values indicated higher reliability) and ROC analysis.
Results: For intrareader variability, the CRs across readers for all ROIs were 9.9% for 2D and 9.3% for 3D. For tumor ROIs the CRs were 10.6% for 2D and 9.6% for 3D. For interreader variability, the CRs across readers for all ROIs were 17.1% for 2D and 20.5% for 3D and for tumor ROIs were 17.9% for 2D and 22.2% for 3D. For combined reader data, the AUCs for benign and malignant findings were 0.77 for 2D and 0.78 for 3D (p = 0.146). For differentiating Gleason score (GS) 3 + 3 from GS > 3 + 3 tumors, the AUCs were 0.92 for 2D and 0.92 for 3D ROIs (p = 0.649). For differentiating GS ≤ 3 + 4 from GS ≥ 4 + 3 tumors, the AUCs were 0.70 for 2D and 0.67 for 3D ROIs (p = 0.004).
Conclusion: Use of a 3D ROI did not improve intrareader or interreader reproducibility or diagnostic performance compared with use of a 2D ROI for prostate ADC measurements. Interreader reproducibility of 2D ROIs was suboptimal nonetheless.
Keywords: DWI; MRI; apparent diffusion coefficient; prostate cancer; reproducibility.