Evaluation of PET/CT Artificial Intelligence Image Reconstructions VS Harmonized Clinical Reconstruction

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Quantitative PET Imaging

Background:

• Evaluating AI-driven image enhancement for PET/CT.
• Comparing SubtlePET™ to EARL Standard-2 harmonized images.
• Assessing AI's impact on quantitative accuracy and clinical workflow.

Purpose of the Study:

• To evaluate SubtlePET™ AI algorithm's performance against EARL Standard-2.
• To assess quantitative accuracy and workflow impact of AI-enhanced PET/CT.
• To determine if AI can maintain diagnostic quality with reduced scan times.

Main Methods:

• Phantom and patient PET/CT data were reconstructed with and without SubtlePET™.
• Quantitative metrics (SUV) and lesion volumes were analyzed.
• Simulated reduced scan durations (3-min/bed, 1-min/bed) were evaluated.

Main Results:

• SubtlePET™ reconstructions showed 91% compliance with EARL Standard-2.
• SUVpeak was the most robust metric across varying scan durations.
• No significant differences in SUVpeak or volume were found for patient data at half acquisition time.

Conclusions:

• SubtlePET™ preserves quantitative performance comparable to EARL Standard-2.
• Deviations in SUVmax/SUVmean noted for small lesions at very short scan times.
• AI-enhanced PET/CT may allow for reduced scan duration, potentially lowering radiation exposure or increasing throughput.