Reducing pediatric total-body PET/CT imaging scan time with multimodal artificial intelligence technology

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Nuclear Medicine

Background:

• Pediatric total-body PET/CT imaging requires long scan times, often necessitating sedation.
• Enhancing image quality and reducing scan duration are critical for pediatric patient comfort and data accuracy.

Purpose of the Study:

• To decrease scan time and improve image quality in pediatric total-body PET/CT imaging.
• To utilize multimodal artificial intelligence (AI) techniques for enhanced PET image reconstruction.

Main Methods:

• Retrospective analysis of 270 pediatric patients undergoing total-body PET/CT scans.
• Development of a 3D neural network fusing low-dose CT images with short-acquisition time (¹⁸F-FDG) PET data.
• Generation of high-quality PET images from ultrashort scan data (e.g., 60 seconds) using AI.

Main Results:

• Multimodal AI techniques significantly enhanced PET image quality.
• Fusion with prior CT information improved anatomical detail in the reconstructed PET images.
• 60-second scan data yielded image quality comparable to full-duration scans.

Conclusions:

• Multimodal AI effectively improves pediatric total-body PET/CT image quality with ultrashort scan times.
• This AI-driven approach has the potential to reduce sedation needs and motion artifacts in pediatric imaging.
• Enhanced image quality and reduced scan times improve the overall imaging experience for pediatric patients.