Deep learning creates pet images from ct scans for better cancer diagnosis
3D CT-to-PET Translation via Latent Brownian Bridge Diffusion
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Positron emission tomography (PET) scans show how the body is functioning, which helps doctors diagnose and plan cancer treatments. But PET scans are expensive, use more radiation, and are not always available. The authors created a method using artificial intelligence that can generate PET-like images from cheaper and more common CT scans. Their approach uses advanced machine learning techniques to better capture important details like small cancerous lesions. This could help provide useful PET information without needing extra scans.
What this means in practice
- •For hospital imaging teams: Generate PET-like images from existing CT data to assist cancer diagnosis when PET scans are limited or unavailable.
- •For medical imaging software developers: Incorporate latent diffusion-based models to improve synthetic PET imaging accuracy and lesion detection in clinical tools.
Authors
Sarita Mourya, Francesco Di Feola, Pierangelo Veltri, Paolo Soda
Abstract
Computed tomography (CT) and positron emission tomography (PET) provide complementary anatomical and functional information for cancer diagnosis and treatment planning. However, the widespread use of PET is limited by high radiation exposure, elevated costs, and restricted availability. To address these limitations, deep learning-based CT-to-PET translation has emerged as a promising approach for synthesizing PET-like information directly from CT images, although accurately modeling the large cross-modal gap remains challenging. In this work, we propose a 3D CT-to-PET translation framework based on latent Brownian Bridge Diffusion (BBDM). The method consists of two stages. First, a Variational Autoencoder (VAE) is trained on paired CT-PET patches, integrating contrastive learning to improve latent alignment between anatomical and metabolic representations. Second, a BBDM is trained in the latent space to translate CT latent representations into their corresponding PET counterparts. The translated PET latents are then decoded and stitched to reconstruct the final 3D PET volume. We evaluate the proposed approach on two publicly available datasets. Quantitative results based on image fidelity and lesion-level PET-specific metrics demonstrate improved performance compared with competing methods. In particular, the proposed approach improves PET signal fidelity, better preserves clinically relevant uptake patterns, and shows improved performance in preserving small-lesion metabolic activation, paving the way for virtual imaging applications.