More Like this

1. Cycle-consistent super-resolution confocal scanning microscopy

   https://doi.org/10.1117/12.3072196  

   Trujillo, Carlos; Thompson, Lauren; Skalli, Omar; Doblas, Ana (July 2025, SPIE)

   Matoba, Osamu; Awatsuji, Yasuhiro; Luo, Yuan; Nishidate, Izumi (Ed.)

   Super-resolution imaging in confocal microscopy has traditionally relied on supervised deep-learning methods, which require large, high-quality paired datasets—a challenging and resource-intensive process. To overcome this limitation, we introduce the Cycle-Consistent Super-Resolution Confocal Scanning Microscopy (CCSRCSM) method, an unsupervised framework based on cycle-consistent generative adversarial networks (CycleGANs). The CCSRCSM method effectively translates low-resolution (LR) confocal images into super-resolved (SR) counterparts using unpaired datasets, eliminating the need for spatially aligned training data. We validated the method using confocal images of U-373MG human glioblastoma cells stained for actin and vimentin. Performance metrics, including spatial cutoff frequency, background noise analysis, and structural similarity index, demonstrate the superior fidelity of the CCSRCSM method compared to conventional image processing and deep-learning approaches. The model successfully enhances spatial frequency content and visual detail while reducing background noise, enabling the clear visualization of subcellular structures such as invadopodia and stress fibers. By leveraging unpaired data, the CCSRCSM method broadens accessibility to super-resolution microscopy, allowing low-cost confocal systems to achieve high-resolution performance without hardware modifications. 

   more » « less
2. Unpaired data training enables super-resolution confocal microscopy from low-resolution acquisitions

   https://doi.org/10.1364/OL.537713  

   Trujillo, Carlos; Thompson, Lauren; Skalli, Omar; Doblas, Ana (January 2024, Optics Letters)

   Supervised deep-learning models have enabled super-resolution imaging in several microscopic imaging modalities, increasing the spatial lateral bandwidth of the original input images beyond the diffraction limit. Despite their success, their practical application poses several challenges in terms of the amount of training data and its quality, requiring the experimental acquisition of large, paired databases to generate an accurate generalized model whose performance remains invariant to unseen data. Cycle-consistent generative adversarial networks (cycleGANs) are unsupervised models for image-to-image translation tasks that are trained on unpaired datasets. This paper introduces a cycleGAN framework specifically designed to increase the lateral resolution limit in confocal microscopy by training a cycleGAN model using low- and high-resolution unpaired confocal images of human glioblastoma cells. Training and testing performances of the cycleGAN model have been assessed by measuring specific metrics such as background standard deviation, peak-to-noise ratio, and a customized frequency content measure. Our cycleGAN model has been evaluated in terms of image fidelity and resolution improvement using a paired dataset, showing superior performance than other reported methods. This work highlights the efficacy and promise of cycleGAN models in tackling super-resolution microscopic imaging without paired training, paving the path for turning home-built low-resolution microscopic systems into low-cost super-resolution instruments by means of unsupervised deep learning. 

   more » « less
3. PSSR2: a user-friendly Python package for democratizing deep learning-based point-scanning super-resolution microscopy

   https://doi.org/10.1101/2024.06.16.599221  

   Stites, Hayden; Manor, Uri (June 2024, bioRxiv)

   Abstract PSSR2 improves and expands on the previously established PSSR (Point-Scanning Super-Resolution) workflow for simultaneous super-resolution and denoising of undersampled microscopy data. PSSR2 is designed to put state-of-the-art technology into the hands of the general microscopy and biology research community, enabling user-friendly implementation of PSSR workflows with little to no programming experience required, especially through its integrated CLI and Napari plugin. 

   more » « less
4. List‐mode TOF‐PET 3D image reconstruction using stochastic primal‐dual network

   https://doi.org/10.1002/mp.70324  

   Tian, Kun; Hu, Rui; Wan, Yiming; Zheng, Zhenrong; Chen, Yunmei; Liu, Huafeng (February 2026, Medical Physics)

   Abstract BackgroundAs an imaging modality in nuclear medicine characterized by its notable sensitivity, positron emission tomography (PET) facilitates the visualization of physiological activities occurring within biological tissues. However, conventional PET images are frequently challenged by poor spatial resolution and a low signal‐to‐noise ratio (SNR) due to inherent detector physics and constraints on radiotracer dosage. The integration of time‐of‐flight (TOF) data into PET image reconstruction has, in recent years, led to enhanced image quality. This enhancement is achieved by using the timing information from detected pairs of annihilation photons, allowing for more accurate determination of the locations where positron annihilation occurs. However, significant challenges persist in TOF‐PET reconstruction. The inclusion of TOF information drastically increases the size of sinogram data (often by tens of folds), severely escalating computational time and memory requirements for reconstruction. PurposeTo address these challenges, this work introduces a new deep learning approach that reconstructs PET images directly from list‐mode data, which avoids the above problems while enhancing image quality. MethodsSpecifically, we propose LM‐SPD‐Net, a list‐mode TOF‐PET reconstruction framework based on a stochastic primal‐dual network architecture. LM‐SPD‐Net comprises two main components as follows: a primal module, constructed using Convolutional Neural Networks (CNNs), to process information in the image domain; and a dual module, built with fully connected neural networks (FCNNs), to handle data‐domain features. These modules are coupled via a projection model and are alternately updated over multiple iterations to obtain the final reconstructed image. By introducing the FCNN and projection model, LM‐SPD‐Net overcomes the conventional limitation of CNNs in processing list‐mode data. Furthermore, the inclusion of a physics‐informed projection process in the training pipeline enhances the interpretability and generalizability of the model. Moreover, to reduce memory usage during training and facilitate 3D PET image reconstruction, a subset partitioning strategy is employed. ResultsQuantitative and qualitative comparisons with list‐mode ordered subset expectation maximization (LM‐OSEM), list‐mode stochastic primal‐dual hybrid gradient (LM‐SPDHG), and Fast‐PET across both simulated and semi‐real clinical data show that LM‐SPD‐Net outperforms these methods by 5%–20% improvements in PSNR and SSIM metrics, while also demonstrating visibly enhanced image quality. ConclusionsComprehensive experiments in this study demonstrate that the proposed method effectively reconstructs the overall subject structure with high fidelity, while maintaining excellent performance in clinically relevant regions such as tumors and the thalamus, particularly under low‐count conditions. 

   more » « less
5. Real-time MRI of the moving wrist at 0.55 tesla

   https://doi.org/10.1259/bjr.20230298  

   Chaudhari, Abhijit J; Lim, Yongwan; Cui, Sophia X; Bayne, Christopher O; Szabo, Robert M; Boutin, Robert D; Nayak, Krishna S (November 2023, The British Journal of Radiology)

   Objectives:Magnetic resonance imaging (MRI) using 1.5T or 3.0T systems is routinely employed for assessing wrist pathology; however, due to off-resonance artifacts and high power deposition, these high-field systems have drawbacks for real-time (RT) imaging of the moving wrist. Recently, high-performance 0.55T MRI systems have become available. In this proof-of-concept study, we tested the hypothesis that RT-MRI during continuous, active, and uninterrupted wrist motion is feasible with a high-performance 0.55T system at temporal resolutions below 100 ms and that the resulting images provide visualization of tissues commonly interrogated for assessing dynamic wrist instability. Methods:Participants were scanned during uninterrupted wrist radial-ulnar deviation and clenched fist maneuvers. Resulting images (nominal temporal resolution of 12.7–164.6 ms per image) were assessed for image quality. Feasibility of static MRI to supplement RT-MRI acquisition was also tested. Results:The RT images with temporal resolutions < 100 ms demonstrated low distortion and image artifacts, and higher reader assessment scores. Static MRI scans showed the ability to assess anatomical structures of interest in the wrist. Conclusion:RT-MRI of the wrist at a high temporal resolution, coupled with static MRI, is feasible with a high-performance 0.55T system, and may enable improved assessment of wrist dynamic dysfunction and instability. Advances in knowledge:Real-time MRI of the moving wrist is feasible with high-performance 0.55T and may improve the evaluation of dynamic dysfunction of the wrist. 

   more » « less