PET/CT Imaging

PET/CT research combines metabolic PET signals with anatomical CT images to detect and measure disease, guide treatment, and track changes over time. Common approaches include U-Net-style segmentation, deep-learning reconstruction and denoising, physics-informed correction, diffusion-based image synthesis, and multimodal models for reports and clinical prediction. Multicenter benchmarks and foundation models are testing how well systems transfer across scanners, tracers, and patient groups, while vision-language models aim to connect image findings with clinical descriptions. A central challenge is improving lesion detection and image quality without distorting quantitative uptake or introducing false structures, especially when validating models outside their training data.