186 - Impact of Segmentation Method on Y90-SIRT Dosimetry in Ultra-Selective Radiation Segmentectomy

Monday, April 13, 2026

4:12 PM - 4:21 PM ET

Presenting Author(s)

Allan Thomas, PhD (he/him/his)

Assistant Professor
Mallinckrodt Institute of Radiology, Washington University School of Medicine, United States

Author/Co-author(s)

GF

Guilherme Ferreira, BS

Graduate Student
Department of Radiation Oncology, Washington University School of Medicine, United States
Ryan C. Lee, BS

Medical Student
Renaissance School of Medicine, Stony Brook University, United States
TA

Tharun T. Alamuri, BS

Medical Student
Renaissance School of Medicine, Stony Brook University, United States
JK

John Karageorgiou, MD

Assistant Professor
Mallinckrodt Institute of Radiology, Washington University School of Medicine, United States
DG

Dan Giardina, MD

Associate Professor
Mallinckrodt Institute of Radiology, Washington University School of Medicine, United States
DB

Daniel Braga, MD

Assistant Professor of Radiology
Mount Sinai Medical Center, United States
Naganathan Mani, MD (he/him/his)

Associate Professor
Washington University in St. Louis, United States
CM

Christopher Malone, MD

Associate Professor
Mallinckrodt Institute of Radiology, Washington University, United States

Purpose: Identifying “ground truth” perfused volume (PV) structures for treatment planning and dosimetry in ⁹⁰Y-SIRT remains difficult. In multi-compartment dosimetry, both anatomic imaging and MAA-SPECT/CT have been used previously to define PV and tumor structures {1}, with comparable relationships between dosimetry and outcomes in either approach. In radiation segmentectomy (RS), MAA-SPECT/CT offers minimal value because MAA injection is not recommended at the segmental level {2}. Instead, treatment planning relies on analysis of contrast-enhanced CBCT from the planning procedure {2}. In this work, the impact of segmentation methods on treatment planning and dosimetry was assessed in 42 ultra-selective glass RS cases for HCC.

Materials and Methods: Tumors were segmented from pre-therapy diagnostic MRI. PV segmentation was performed using HU-threshold guidance from planning CBCT and three post-therapy ⁹⁰Y-SPECT/CT thresholds at 10, 20, and 30 pct of maximum counts. Post-therapy, ⁹⁰Y-SPECT/CT voxelized dosimetry was computed using the local deposition method and activity normalization to the PV. PV volumes, PV/tumor dose metrics, and dose-response models were compared among the different segmentation methods.

Results: Median (IQR) tumor size and CBCT PV were 2.3 (1.7 – 3.0) cm and 64.5 (44.9 – 94.1) mL, respectively. PVs from ⁹⁰Y-SPECT were 94.9 (58.2 – 148.1), 49.5 (31.4 – 85.8), and 27.7 (17.8 – 63.9) mL for 10, 20, and 30 pct thresholds, respectively. All PVs were significantly different (p< 0.001) except for CBCT vs 20 pct (p=0.097). The variations in PVs produced inverse changes in PV and tumor doses with comparable statistical significance. Volume and dose differences in specific cases were at times substantial. In 4 of the 5 tumors without complete response, tumor mean dose and tumor D₉₅ were lower using 10 and 20 pct contours relative to CBCT.

Conclusion: Because PV volumes directly affect treatment planning and dosimetry, inaccurate estimation methods may impact tumor response and patient outcomes if targeted regions are under-dosed from PV underestimation. The results in this study highlight that, especially in ultra-selective RS, special care is needed to ensure treatment planning aligns with ⁹⁰Y activity distributions in post-therapy imaging.