Xiongchao Chen

Visual autoregressive (VAR) modeling provides a promising generative solution for RGB image restoration, owing to its strong ability to generate realistic details and recover rich scene information. Nevertheless, directly applying VAR to low-light RAW image restoration still presents two main challenges. First, RGB-pretrained VAE tokenizers are mismatched with RAW sensor statistics, causing color bias after domain transfer. Second, VAR requires tiled inference for high-resolution RAW images, where independent token prediction across tiles may introduce color discrepancies that are further amplified by the Image Signal Processing (ISP) pipeline. To address these issues, we propose RAWVAR, the first visual autoregressive framework for low-light RAW image restoration. For effective domain transfer, we design a Cross-Domain Codebook (CDC), which freezes the RGB-pretrained encoder and original codebook while appending a trainable RAW-domain codebook, forming a cross-domain latent space for texture-preserving and color-faithful RAW representation. To mitigate color inconsistency during tiled inference, we further design a Position-aware Token Alignment (PTA) strategy. During coarse-to-fine prediction, PTA aligns the token logits of semantically corresponding regions across tiles, ensuring consistent predictions for shared content. Furthermore, a Global Color Refinement (GCR) module globally refines the discrete latents to enhance image-wise color consistency. Compared with diffusion-based methods, RAWVAR provides a more efficient generative restoration pipeline by avoiding iterative denoising. Extensive experiments show that RAWVAR produces color-faithful and perceptually superior results, outperforming existing generative restoration methods.

Synaptic vesicle glycoprotein 2A (SV2A) in human brains is an important biomarker of synaptic loss associated with several disorders, including Alzheimer's disease (AD). However, SV2A tracers, such as [¹¹C]UCB-J, are less available in clinical practice due to constraints that include cost and the need for an onsite cyclotron. As a result, there is interest in generating synthetic SV2A PET images using a more commonly used tracer, or another widely available image modality. We therefore investigated the feasibility of using a deep learning approach to generate synthetic [¹¹C]UCB-J SV2A PET images from T1-weighted (T1w) Magnetic Resonance Imaging (MRI) and/or [¹⁸F]FDG PET imaging. In our study, we used paired data from 54 participants for training and evaluation, including 22 healthy controls (HC) and 32 AD patients. We implemented two architectures, 3D UNet and 3D Multi-stage ResUNet (MS ResUNet) for translating either MRI or [¹⁸F]FDG PET or both into [¹¹C]UCB-J PET images. Our experiments show that, through progressive learning 3D MS ResUNet achieved less bias compared to a single-stage design. Adopting residual connection, cross-stage feature fusion and supervised attention strategies to our model contributes to better gradient flow, mitigates information loss and enables selective focus, leading to better generation results. Additionally, combining MRI T1w and [¹⁸F]FDG PET images provides the best prediction performance, whereas having [¹⁸F]FDG PET alone produces better synthetic images than using MRI alone. We also evaluated our model’s performance for 14 typical brain regions. The MS ResUNet achieved less than 5% bias in all regions. Furthermore, the differences between AD and HC in certain regions (e.g., hippocampus and entorhinal) are still significant in generated images (p=0.000). In conclusion, the 3D MS ResUNet was able to generate desirable SV2A PET images that are visually consistent with true images in terms of contrast and tracer distribution, quantitatively with low bias (-0.26% ± 5.56%) and high similarity (R²=0.975), for both AD and HC. Compared with the 3D vanilla UNet (FDG: −2.39% ± 5.55%, MRI: −5.03% ± 13.24%, or both: −1.89% ± 6.04%), the modified design produced more robust SV2A images. The synthetic images can potentially provide additional pathological insights into AD and other neurological disorders without increasing radiation exposure and cost.

Fluorescence microscopy has been widely used to reveal the spatial distribution of specifically labeled molecules, but it is blind to cellular context. Quantitative phase contrast microscopy (QPC) provides such complementary information. Here we have developed a platform that combines the QPC technique of correlative orientation-independent differential interference contrast (OI-DIC) microscopy with single-molecule super-resolution fluorescence microscopy. We demonstrate a detection sensitivity of 0.05 nm optical path difference, sufficient to detect single microtubules, and show its capability of 3D super-resolution fluorescence imaging in the cellular context. Additionally, we report deep-learning enabled digital staining, identifying nuclei, mitochondria and lipid droplets from OI-DIC data and demonstrate the potential of this approach for long-term live-cell imaging of organelles of interest without the need for fluorescence. OI-DIC can be easily integrated into most fluorescence microscopes and is readily adoptable by microscopy labs. ONE-SENTENCE TEASER: A highly sensitive technique to visualize sub-cellular structures, dynamics, and molecules in their cellular context.

Positron Emission Tomography (PET) is an important clinical imaging tool but inevitably introduces radiation exposure to patients and healthcare providers. Reducing the tracer injection dose and eliminating the CT acquisition for attenuation correction can reduce the overall radiation dose, but often results in PET with high noise and bias. Thus, it is desirable to develop 3D methods to translate the non-attenuation-corrected low-dose PET (NAC-LDPET) into attenuation-corrected standard-dose PET (AC-SDPET). Recently, diffusion models have emerged as a new state-of-the-art deep learning method for image-to-image translation, better than traditional CNN-based methods. However, due to the high computation cost and memory burden, it is largely limited to 2D applications. To address these challenges, we developed a novel 2.5D Multi-view Averaging Diffusion Model (MADM) for 3D image-to-image translation with application on NAC-LDPET to AC-SDPET translation. Specifically, MADM employs separate diffusion models for axial, coronal, and sagittal views, whose outputs are averaged in each sampling step to ensure the 3D generation quality from multiple views. To accelerate the 3D sampling process, we also proposed a strategy to use the CNN-based 3D generation as a prior for the diffusion model. Our experimental results on human patient studies suggested that MADM can generate high-quality 3D translation images, outperforming previous CNN-based and Diffusion-based baseline methods. The code is available at https://github.com/tianqic/MADM.

Attenuation correction (AC) is a critical step to ensure accurate quantitative PET imaging. To eliminate the radiation dose from CT, deep learning (DL)-based methods have been extensively investigated to generate the CT-equivalent attenuation map (μ-CT) directly from the PET signal. However, almost all previous studies only focus on 18F-FDG due to its extensive data availability which is suitable for DL model training. For other less common tracer types, it is generally believed that new models must be trained separately on these tracer-specific data to ensure reasonable performance. In this work, we explored the cross-tracer generalizability of μ-DL generation DL models -primarily focusing on whether a model trained on a commonly used tracer like 18F-FDG can be effectively applied to less common tracers such as 68Ga-DOTATE and 18F-Fluciclovine, and vice versa. Unlike methods that directly generate attenuation-corrected (AC) PET images from non-attenuation corrected (NAC) PET images or MLAA reconstructions, we generate the CT-based deep learning attenuation maps (μ-DL) using MLAA reconstruction with the combined input of attenuation maps (μ-MLAA) and tracer activity (λ-MLAA). This μ-DL is then used for attenuation correction to obtain the final AC PET image. Our comprehensive evaluations on both μ-CT generation and the PET reconstruction found that the DL model trained on one specific tracer can be adapted to other tracers with competitive performance when compared to the tracer-specific trained DL model. The 18F-FDG-trained model demonstrated the best generalizability when applied to less common tracer types which often have relatively fewer available data for training. Additionally, we investigated the role of the μ-MLAA and λ-MLAA as inputs for the network performance. We found that combining both inputs resulted in the best performance, but the μ-MLAA contributed more significantly compared to the λ-MLAA.

Low-dose PET offers a valuable means of minimizing radiation exposure in PET imaging. However, the prevalent practice of employing additional CT scans for generating attenuation maps ( μ -map) for PET attenuation correction significantly elevates radiation doses. To address this concern and further mitigate radiation exposure in low-dose PET exams, we propose an innovative P opulation-prior-aided O ver- U nder- R epresentation N etwork (POUR-Net) that aims for high-quality attenuation map generation from low-dose PET. First, POUR-Net incorporates an O ver- U nder- R epresentation N etwork (OUR-Net) to facilitate efficient feature extraction, encompassing both low-resolution abstracted and fine-detail features, for assisting deep generation on the full-resolution level. Second, complementing OUR-Net, a population prior generation machine (PPGM) utilizing a comprehensive CT-derived μ -map dataset, provides additional prior information to aid OUR-Net generation. The integration of OUR-Net and PPGM within a cascade framework enables iterative refinement of μ -map generation, resulting in the production of high-quality μ -maps. Experimental results underscore the effectiveness of POUR-Net, showing it as a promising solution for accurate CT-free low-count PET attenuation correction, which also surpasses the performance of previous baseline methods.