Quantification in respiratory-gated PET acquisition: can data-driven methods replace device-based systems?-a...

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Device-based respiratory gating in PET enhances accuracy but has setup time and failure rate issues.
• Evaluating data-driven gating as an alternative to device-based gating is crucial for clinical practice.
• Retrospective analysis of [18F]FDG PET/CT scans in oncological patients was performed.

Purpose of the Study:

• To internally validate the quantification performance of data-driven respiratory-gated PET imaging.
• To compare data-driven gating against the gold standard device-based method in clinical oncological practice.

Main Methods:

• Retrospective analysis of 196 patients with [18F]FDG PET/CT scans.
• Comparison of SUVmax and SUVpeak measurements between device-based gating and AI-driven OncoFreeze gating.
• Calculation of mean absolute bias and 95% confidence intervals.

Main Results:

• Mean absolute bias for SUVmax was 3.8% and for SUVpeak was 2.1% across the cohort.
• Only 5% of patients showed SUVpeak bias >10%, and 21% showed SUVmax bias >10%.
• Minimal bias observed regardless of lesion size or anatomical site, especially by PERCIST criteria.

Conclusions:

• Data-driven respiratory gating demonstrates negligible quantitative differences compared to device-based gating.
• AI-powered data-driven gating is a reliable and accurate alternative for routine [18F]FDG-PET/CT oncological evaluation.
• This method offers a promising, efficient alternative to traditional gating techniques.