Intra- and inter-operator variability in MRI-based manual segmentation of HCC lesions and its impact on dosimetry

• DOI: 10.1186/s40658-022-00515-6
• Publication Date: 2022-12-21T10:03:09Z
• Authors: Elise C. Covert; Kellen Fitzpatrick; Justin Mikell; Ravi K. Kaza; John D. Millet; Daniel Barkmeier; Joseph Gemmete; Jared Christensen; Matthew J. Schipper; Yuni K. Dewaraja
• Keywords: Coefficient of variation

The aim was to quantify inter- and intra-observer variability in manually delineated hepatocellular carcinoma (HCC) lesion contours the resulting impact on radioembolization (RE) dosimetry.Ten patients with HCC lesions treated Y-90 RE imaged post-therapy PET/CT were selected for retrospective analysis. Three radiologists contoured 20 baseline multiphase contrast-enhanced MRIs, two of re-contoured at additional sessions. Contours transferred co-registered PET/CT-based dose maps. Volume-dependent recovery coefficients applied partial volume correction (PVC) when reporting mean absorbed dose. To understand how uncertainty varies tumor size, we fit power models regressing relative contour volume. Finally, determined effects segmentation control probability (TCP), as calculated using logistic developed a previous study.The average ranged from 1.8 194.5 mL, 23.4 1629.0 Gy. inter-observer Dice coefficient significantly less than (0.79 vs. 0.85, p < 0.001). Uncertainty segmented volume, measured by Coefficient Variation (CV), 4.2 34.7% an 17.2%. CV had value 5.4% (range 1.2-13.1%) without PVC while it 15.1% 1.5-55.2%) PVC. Using fitted function our prior data, change TCP due alone estimated 16.2% (maximum 48.5%).Though find relatively high reliability overall, contouring propagates into non-negligible metrics outcome prediction individual cases that should be considered dosimetry-guided treatment.