Attenuation Correction in PET Scans Using Synthetic CT Images from the Emission Data

The project explores recent advancements in PET/CT scanning technology, specifically addressing the challenge of high radiation exposure from traditional CT scans. The primary focus is on utilizing artificial intelligence to generate synthetic CT images directly from PET scan data. By potentially replacing conventional CT scans, this approach aims to reduce patient radiation exposure while maintaining high image quality.

Traditionally, PET scans require a separate CT scan for attenuation correction, leading to increased radiation exposure for the patients. This project presents a novel method using artificial intelligence to generate synthetic CT images directly from PET scans that have not undergone attenuation correction. By eliminating the need for a separate CT scan, this approach significantly reduces the overall radiation dose. The method involves a two-part system: a generator that produces synthetic CT images and a discriminator that assesses their authenticity. This setup enhances the quality of the synthetic images.

Comparative studies between standard [^18F]FDG-PET images and synthetic PET images (sPET) revealed that in 30 out of 36 cases, the image quality was comparable, with only minor differences in six cases that were not clinically significant.

The results suggest that synthetic CT images generated with advanced algorithms are nearly as effective as traditional CT images. Additionally, synthetic PET images show minimal discrepancies compared to conventional PET scans. These findings indicate that the proposed model could be highly effective in clinical practice. This innovative approach holds potential for replacing CT scans in certain situations, thereby significantly reducing patient radiation exposure. This reduction is especially important for scenarios where minimizing radiation is critical, such as in children, pregnant women, or patients undergoing regular monitoring.