AI model enhances CT-less PET attenuation correction for medical imaging

Researchers have developed a novel approach for CT-less PET attenuation correction by synthesizing multimodal pseudo-CT images. Their method, submitted to the BIC-MAC 2026 Challenge, utilizes a modified nnU-Net architecture incorporating both anatomical and physical supervision. The system leverages a frozen TotalSegmentator feature extractor for anatomical guidance and a differentiable projection loss for physical supervision, building upon weights pre-trained on the SynthRAD Challenge dataset. AI

IMPACT This research could improve the accuracy and efficiency of PET scans by enabling CT-less attenuation correction, potentially leading to better diagnoses and treatment planning.

RANK_REASON The item describes a research paper submitted to a challenge, detailing a novel method for medical image processing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• BIC-MAC 2026 Challenge
• nnU-Net
• Petros Chatzitoulousis
• SynthRAD Challenge
• TotalSegmentator

• paper
• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →