Automatic healthy liver segmentation for holmium-166 radioembolization dosimetry

Background For safe and effective holmium-166 (Ho-166) liver radioembolization, dosimetry is crucial and requires accurate healthy liver definition. The current clinical standard relies on manual segmentation and registration of a separately acquired contrast enhanced CT (CECT), a prone-to- error and time-consuming task. An alternative is offered by simultaneous imaging of Ho-166 and technetium-99m stannous-phytate accumulating in healthy liver cells (Ho-166-Tc-99m dual-isotope protocol). This study compares healthy liver segmentation performed with an automatic method using Tc-99m images derived from a Ho-166-Tc-99m dual-isotope acquisition to the manual segmentation, focusing on healthy liver dosimetry and corresponding hepatotoxicity. Data from the prospective HEPAR PLuS study were used. Automatic healthy liver segmentation was obtained by thresholding the Tc-99m image (no registration step required). Manual segmentation was performed on CECT and then manually registered to the SPECT/CT and subsequently to the corresponding Ho-166 SPECT to compute absorbed dose in healthy liver. Results Thirty-one patients (66 procedures) were assessed. Manual segmentation and registration took a median of 30 min per patient, while automatic segmentation was instantaneous. Mean +/- standard deviation of healthy liver absorbed dose was 18 +/- 7 Gy and 20 +/- 8 Gy for manual and automatic segmentations, respectively. Mean difference +/- coefficient of reproducibility between healthy liver absorbed doses using the automatic versus manual segmentation was 2 +/- 6 Gy. No correlation was found between mean absorbed dose in the healthy liver and hepatotoxicity. Conclusions Ho-166-Tc-99m dual-isotope protocol can automatically segment the healthy liver without hampering the Ho-166 dosimetry assessment. Trial registration: ClinicalTrials.gov, NCT02067988. Registered 20 February 2014. https://clinicaltrials.gov/ct2/show/NCT02067988

Automated summary

This study compares the performance of an automatic method using the Ho-166-Tc-99m dual-isotope protocol with manual segmentation for healthy liver dosimetry and corresponding hepatotoxicity. The results show that the automatic segmentation method can accurately and quickly segment the liver without affecting the dosimetry assessment of Ho-166.