1: \begin{abstract}

2: Positron emission tomography (PET) is widely utilized for cancer detection due to its ability to visualize functional and biological processes \textit{in vivo}. PET images are usually reconstructed from histogrammed raw data (sinograms) using traditional iterative techniques (e.g., OSEM, MLEM). Recently, deep learning (DL) methods have shown promise by directly mapping raw sinogram data to PET images. However, DL approaches that are regression-based or GAN-based often produce overly smoothed images or introduce various artifacts respectively. Image-conditioned diffusion probabilistic models (cDPMs) are another class of likelihood-based DL techniques capable of generating highly realistic and controllable images. While cDPMs have notable strengths, they still face challenges such as maintain correspondence and consistency between input and output images when they are from different domains (e.g., sinogram vs. image domain) as well as slow convergence rates. To address these limitations, we introduce \techname{}, a hierarchica\textbf{L} f\textbf{e}ature \textbf{g}uided conditional diffusi\textbf{o}n model for high-perceptual quality PET image reconstruction from sinograms. We conducted several experiments demonstrating  that \techname{} not only improves the performance of cDPMs but also surpasses recent DL-based PET image reconstruction techniques in terms of visual quality and pixel-level PSNR/SSIM metrics. Our code is available at \href{https://github.com/yransun/LegoPET}{https://github.com/yransun/LegoPET}.

3: \end{abstract}

4: