Improved Quantification of ICG Perfusion Through Motion Compensation in Fluorescence-Guided Surgery

TL;DR

This paper introduces a method to reduce motion artifacts in fluorescence-guided surgery, improving the accuracy of perfusion measurements during operations.

Contribution

The study presents an automated motion compensation technique that effectively reduces motion artifacts in ICG perfusion imaging during surgery.

Findings

Automated motion compensation successfully corrected motion artifacts in 67.5% of frame sequences.

PCA analysis showed a clear separation between successful and unsuccessful motion compensation (AUC = 0.80).

Individual-level perfusion metrics showed significant changes after motion compensation, with large percentage differences observed.

Abstract

Background/Objectives: Motion artifacts significantly distort fluorescence measurements during surgical perfusion assessment, potentially leading to incorrect clinical decisions. This study evaluates the efficacy of automated motion compensation (MC) in quantitative indocyanine green (q-ICG) imaging to improve the accuracy of perfusion assessment. Methods: Frames from ICG perfusion assessment during 17 pancreaticoduodenectomies were analyzed. Regions of interest (ROIs) were systematically placed on each frame series, and automated MC was applied to track tissue movement. Performance was evaluated by comparing MC with surgeon-adjusted placement using multiple image quality metrics and analyzing perfusion metrics on time–intensity curves. Principal Component Analysis (PCA) was applied to explore whether image patterns could distinguish between successful and unsuccessful motion compensation. Results: Automated motion compensation successfully corrected motion artifacts in 67.5% of frame sequences, achieving comparable performance to surgeon-guided adjustments. PCA demonstrated clear separation between sufficient and insufficient corrections (AUC = 0.80). At the population level, MC did not significantly change perfusion slope (t(59) = 1.60, p = 0.11) or time-to-peak (Tmax; t(58) = 0.81, p = 0.42). Bland–Altman analysis showed a mean bias of −0.54 (SD = 3.32) for slope and 24.95 (SD = 238.40) for Tmax. At the individual level, 86.7% of slope and 79.7% of Tmax values differed by ≥10% after MC, with mean absolute percentage changes of 108.5% (median 37.8%) and 431.5% (median 65.9%), respectively. Conclusions: MC effectively reduces motion artifacts in fluorescence-guided perfusion assessment. By improving the precision of ICG-derived parameters, this technology enhances measurement reliability and represents an enabler for accurate intraoperative perfusion quantification.

Figures