Improving Registration Robustness for Image-Guided Liver Surgery in a Novel Human-to-Phantom Data Framework

Area of Science:

• Medical Imaging
• Surgical Navigation
• Biomedical Engineering

Background:

• Image-guided liver surgery (IGLS) relies on accurate image-to-physical registration.
• Variations in intraoperative organ surface data collection can introduce uncharacterized errors, limiting registration accuracy and robustness.
• Clinical validation of IGLS registration methods is challenging due to difficulties in obtaining representative organ deformation data.

Purpose of the Study:

• To propose and validate a novel human-to-phantom framework for assessing surface-driven, volumetric nonrigid registration methods in IGLS.
• To investigate the impact of surface data collection patterns on registration accuracy.
• To develop and evaluate a spatial data resampling approach to mitigate sampling-related errors.

Main Methods:

• Developed a human-to-phantom validation framework using surface collection patterns from 13 in vivo IGLS procedures applied to a hepatic deformation phantom.
• Combined workflow-realistic data acquisition with realistic surgical deformations.
• Investigated volumetric target registration error (TRE) using both rigid and novel nonrigid registration methods.
• Introduced a spatial data resampling strategy.

Main Results:

• Routine rigid registration resulted in a TRE of 10.9 ± 0.6 mm.
• The novel resampling strategy and improved nonrigid registration method reduced TRE by 51%, achieving a TRE of 5.3 ± 0.5 mm.
• The residual surface fit was within ±10 mm, representative of open liver resections.

Conclusions:

• The proposed human-to-phantom approach provides a tractable method for validating image-to-physical registration techniques in IGLS.
• The developed resampling strategy and nonrigid registration method significantly improve accuracy and robustness in image-guided liver surgery.
• This framework enables more reliable clinical validation of advanced IGLS registration methods.