Article
Optics
James A. Grant-Jacob, Matthew Praeger, Robert W. Eason, Ben Mills
Summary: The ability to generate three-dimensional images of pollen grains using deep learning has the potential to disrupt the fields of palynology, bioaerosol sensing, and ecology. This study demonstrates the use of lensless sensing to transform scattering patterns into two-dimensional images, which are then used to train a neural network to generate three-dimensional images with an average accuracy of 84%. The development of airborne-pollen sensors based on this technique could provide valuable data for understanding mechanisms of pollen production, climate change, and their effects on public health.
Article
Optics
Kyung Chul Lee, Junghyun Bae, Nakkyu Baek, Jaewoo Jung, Wook Park, Seung Ah Lee
Summary: Lensless cameras are a new type of computational imaging devices that use a thin mask instead of lenses to achieve compact and low-cost hardware. This paper proposes a method for fast fabrication of lensless cameras with arbitrary point spread functions (PSFs). The method includes designing smooth phase mask profiles for a given PSF pattern and fabricating the mask using gray-scale lithography technique. Compared to existing approaches, this method allows for ultra-fast and cost-effective fabrication of phase masks, making it suitable for mass production and commercialization of lensless cameras. The method can be used to produce custom lensless cameras with various pre-designed PSFs and obtain images of the scene through computational image reconstruction. The paper also discusses the future directions and potential applications of custom lensless cameras, including fast imaging and fingerprint detection.
Article
Optics
Feng Tian, Junjie Hu, Weijian Yang
Summary: GEOMScope is a lensless single-shot 3D microscope that reconstructs objects through an innovative algorithm, showing excellent performance in imaging resolution, volume, and reconstruction speed.
LASER & PHOTONICS REVIEWS
(2021)
Article
Nanoscience & Nanotechnology
Fatima El Moussawi, Matthias Hofer, Damien Labat, Andy Cassez, Geraud Bouwmans, Siddharth Sivankutty, Rosa Cossart, Olivier Vanvincq, Herve Rigneault, Esben Ravn Andresen
Summary: This study presents a novel tapered multicore fiber (MCF) component for ultraminiaturized endoscopes, addressing the power delivery issue faced by MCF-based lensless endoscopes and achieving a significant increase in two-photon signal yield.
Article
Biochemical Research Methods
Simao Coelho, Jongho Baek, James Walsh, J. Justin Gooding, Katharina Gaus
Summary: This method describes how to implement mechanical motion correction in super-resolution imaging, including drift control, focus locking and adjustment, and rapid repositioning. The approach has been proven to reduce drift, improve imaging quality, and is suitable for single-molecule and super-resolution imaging.
Article
Optics
Yujun Tang, Gang Wen, Yong Liang, Linbo Wang, Jie Zhang, Hui LI
Summary: Deep learning is used to reconstruct super-resolution structured illumination microscopy (SR-SIM) images, effectively reducing photobleaching and phototoxicity. A proposed dynamic SIM imaging strategy records full raw images at the beginning and then only wide-field images, with a deep-learning-based reconstruction algorithm improving the quality of reconstructed SR images while reducing photobleaching and phototoxicity. This algorithm also has the capability to observe new structures not included during network training.
Article
Biochemical Research Methods
Cong T. S. Van, Chrysanthe Preza
Summary: This study introduces a regularized three-dimensional model-based restoration method with positivity constraint for processing 3D data from structured illumination microscopy systems. The proposed method provides axial super resolution and improves 3D resolution compared to standard methods.
BIOMEDICAL OPTICS EXPRESS
(2021)
Article
Optics
Tingting Wu, Peng Lu, Md Ashequr Rahman, Xiao LI, Matthew D. Lew
Summary: Dipole-spread function (DSF) engineering can reshape microscope images to maximize the sensitivity of measuring the 3D orientations of dipole-like emitters. However, Poisson shot noise, overlapping images, and fitting high-dimensional information pose challenges in image analysis for single-molecule orientation-localization microscopy (SMOLM). In this study, a deep-learning based estimator called Deep-SMOLM is introduced, which achieves superior precision in measuring 3D orientation and 2D position, approaching the theoretical limit. Deep-SMOLM also demonstrates excellent performance in estimating overlapping images of emitters.
Article
Optics
Feng Tian, Weijian Yang
Summary: In this study, a compact and learnable lensless 3D camera is developed for real-time photorealistic imaging, overcoming challenges such as calibration and computational burdens in lensless imaging.
Article
Engineering, Electrical & Electronic
Thibaut Eloy, Etienne Baudrier, Marine Laporte, Virginie Hamel, Paul Guichard, Denis Fortun
Summary: Single particle reconstruction is a powerful technique in 3D fluorescence microscopy, which improves the axial resolution and fluorescence labeling degree by reconstructing the average volume of a biological particle from multiple views with unknown poses. This paper proposes a dedicated single particle reconstruction method for convolutional models, overcoming issues such as template bias, restriction to 2D data, high computational cost, and lack of robustness to low fluorescent labeling. The method achieves better resolution and reconstruction error with low computational cost, as demonstrated on synthetic data and real datasets of centrioles.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Multidisciplinary Sciences
Anindita Dasgupta, Joran Deschamps, Ulf Matti, Uwe Huebner, Jan Becker, Sebastian Strauss, Ralf Jungmann, Rainer Heintzmann, Jonas Ries
Summary: Supercritical angle localization microscopy (SALM) allows the z-positions of single fluorophores to be extracted from the intensity of supercritical angle fluorescence. The authors improve the z-resolution of SALM and report nanometer isotropic localization precision on DNA origami structures.
NATURE COMMUNICATIONS
(2021)
Review
Optics
Gary Brooker, Nisan Siegel
Summary: This article presents a chronicle of the 15-year development of Fresnel incoherent correlation holography (FINCH), from its initial description to its current state as a 3D microscopic imaging technique that can achieve optical resolution beyond the Rayleigh limit. The article describes the path from the original demonstration of FINCH to its current use in perfect imaging of multicolor fluorescent biological specimens and reference test patterns using fluorescence or reflected light imaging.
Article
Optics
Haixin Luo, Jie Xu, Liyun Zhong, Xiaoxu Lu, Jindong Tian
Summary: Digital holography based on lensless imaging is a developing method used in microscopy for micro-scale measurement. A single-shot approach using deep learning and physical models is proposed and constructed, breaking through the limitations of physical devices and demonstrating qualified generalization ability for samples with different morphologies.
Letter
Optics
Shenghao Zheng, Zhihui Ding, Rui Jiang, Cheng Guo
Summary: The study presents a self-calibrated phase retrieval method for lensless masked imaging systems. The proposed method achieves efficient and flexible image recovery without the need for additional calibration devices.
Article
Optics
Bo Xiong, Tianyi Zhu, Yuhan Xiang, Xiaopeng Li, Jinqiang Yu, Zheng Jiang, Yihan Niu, Dong Jiang, Xu Zhang, Lu Fang, Jiamin Wu, Qionghai Dai
Summary: MiSLFM utilizes a tilted mirror to achieve super-resolved axial resolution with a single objective, expanding the depth of field of LFM. It allows observation of various organelle interactions and intercellular interactions, as well as more robust blood cell tracking in zebrafish larvae under low light conditions.
LIGHT-SCIENCE & APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Ali Mousavi, Richard G. Baraniuk
Summary: This article introduces a method called the uniform information coefficient (UIC), which is able to infer relationships among variables from large datasets. Compared to traditional methods, the UIC calculation is more efficient and robust to the type of association between variables.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Review
Optics
Vivek Boominathan, Jacob T. Robinson, Laura Waller, Ashok Veeraraghavan
Summary: This article reviews recent advances in lensless imaging technology and describes the design principles to be considered when developing and using lensless imaging systems.
Article
Engineering, Civil
Ewa M. Nowara, Tim K. Marks, Hassan Mansour, Ashok Veeraraghavan
Summary: Imaging photoplethysmography (iPPG) has the potential to enhance driver safety systems by detecting driver fatigue and early heart failure in a non-invasive manner. However, iPPG faces challenges in the driving context due to illumination and motion. This paper proposes two innovations to overcome these challenges: using narrow-band near-infrared video recordings to reduce illumination variations, and introducing a novel optimization algorithm called AutoSparsePPG that improves performance compared to existing methods.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Engineering, Biomedical
Jesse K. Adams, Dong Yan, Jimin Wu, Vivek Boominathan, Sibo Gao, Alex Rodriguez, Soonyoung Kim, Jennifer Carns, Rebecca Richards-Kortum, Caleb Kemere, Ashok Veeraraghavan, Jacob T. Robinson
Summary: Lensless microscopy with a designed phase mask allows for imaging of biological tissue in vivo, overcoming the limitations of existing techniques. The low-cost and small form factor of this lensless imaging method makes it suitable for clinical use in inaccessible areas of the body.
NATURE BIOMEDICAL ENGINEERING
(2022)
Article
Optics
Dhruvjyoti Bagadthey, Sanjana Prabhu, Salman S. Khan, D. Tony Fredrick, Vivek Boominathan, Ashok Veeraraghavan, Kaushik Mitra
Summary: This paper proposes a feed-forward deep network called FlatNet3D that can directly reconstruct depth and intensity information from a single lensless capture. The algorithm is fast and produces high-quality results, utilizing an efficient physics-based 3D mapping stage and a fully convolutional network. The effectiveness of the proposed method is validated through simulations and real scenes captured using PhlatCam.
JOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS IMAGE SCIENCE AND VISION
(2022)
Article
Multidisciplinary Sciences
Brandon Y. Feng, Haiyun Guo, Mingyang Xie, Vivek Boominathan, Manoj K. Sharma, Ashok Veeraraghavan, Christopher A. Metzler
Summary: We propose a scanning-free wavefront shaping technique called neural wavefront shaping (NeuWS) that integrates maximum likelihood estimation, measurement modulation, and neural signal representations to reconstruct diffraction-limited images through strong static and dynamic scattering media without guidestars, sparse targets, controlled illumination, nor specialized image sensors.
Article
Engineering, Electrical & Electronic
Fay Wang, Stephen H. Kim, Yongyi Zhao, Ankit Raghuram, Ashok Veeraraghavan, Jacob Robinson, Andreas H. Hielscher
Summary: Steady progress in time-domain diffuse optical tomography (TD-DOT) technology has resulted in the development of low-cost, compact, and high-performance systems, enabling widespread clinical TD-DOT use. The new algorithm, SENSOR-NET, combines a deep-learning approach with a non-iterative sparse optical reconstruction code to achieve high-resolution sparse reconstruction without the need for parameter tuning. This allows for real-time brain monitoring and other high-speed TD-DOT applications.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Computer Science, Artificial Intelligence
Hossein Babaei, Sina Alemohammad, Richard G. Baraniuk
Summary: To investigate treatment effectiveness, populations are split into control and treatment groups and their average responses are compared. Similarity in statistics of these groups is crucial for reliable and valid results. Covariate balancing methods aim to increase similarity, but limited samples often hinder accurate estimation. This article uncovers that covariate balancing measures and sequential treatment assignment methods can be susceptible to worst case treatment assignments, resulting in high estimation errors. An adversarial attack is developed to find such assignments and an optimization-based algorithm (ATASTREET) is provided to measure the closeness to worst-case scenarios.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Akshat Dave, Yongyi Zhao, Ashok Veeraraghavan
Summary: Inverse rendering, a fundamental problem in computer graphics and vision, has benefited from recent advancements in neural networks and polarization sensors, allowing for impressive geometry reconstruction and reflectance estimation.
COMPUTER VISION, ECCV 2022, PT VII
(2022)
Article
Engineering, Electrical & Electronic
T. Mitchell Roddenberry, Fernando Gama, Richard G. G. Baraniuk, Santiago Segarra
Summary: Graph filtering is a fundamental operation in graph signal processing, with its output depending only on the local neighborhood of each node. This research proposes a framework based on the local distribution of a graph, allowing for comparison of graphs of different sizes and models, and obtaining related properties.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Proceedings Paper
Computer Science, Hardware & Architecture
Haoran You, Cheng Wan, Yang Zhao, Zhongzhi Yu, Yonggan Fu, Jiayi Yuan, Shang Wu, Shunyao Zhang, Yongan Zhang, Chaojian Li, Vivek Boominathan, Ashok Veeraraghavan, Ziyun Li, Yingyan Lin
Summary: The study proposes a lensless FlatCam-based eye tracking algorithm and accelerator co-design framework, which significantly reduces the form-factor of eye tracking systems and improves system efficiency without sacrificing tracking accuracy.
PROCEEDINGS OF THE 2022 THE 49TH ANNUAL INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE (ISCA '22)
(2022)
Article
Engineering, Electrical & Electronic
Pavan K. Kota, Daniel LeJeune, Rebekah A. Drezek, Richard G. Baraniuk
Summary: Compressed sensing is a signal processing technique that efficiently recovers sparse high-dimensional signals from low-dimensional measurements. This study explores the multiple measurement vector problem where signals are independently drawn from a sparse multivariate Poisson distribution. Through maximum likelihood estimation and a novel Sparse Poisson Recovery algorithm, the sparse parameter vector of Poisson rates is successfully recovered.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Article
Pediatrics
Anil Kumar Vadathya, Salma Musaad, Alicia Beltran, Oriana Perez, Leo Meister, Tom Baranowski, Sheryl O. Hughes, Jason A. Mendoza, Ashutosh Sabharwal, Ashok Veeraraghavan, Teresia O'Connor
Summary: The FLASH-TV system offers a critical step forward in improving the assessment of children's television viewing.
JMIR PEDIATRICS AND PARENTING
(2022)
Article
Geochemistry & Geophysics
Angel Bueno Rodriguez, Randall Balestriero, Silvio De Angelis, M. Carmen Benitez, Luciano Zuccarello, Richard Baraniuk, Jesus M. Ibanez, Maarten de Hoop
Summary: This paper introduces an automatic neural network architecture for seismic and volcanic monitoring. By using a Bayesian network strategy and scattering transform, changes in seismic and volcanic activity can be detected, which can help predict possible volcanic eruptions.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Yonggan Fu, Yang Zhang, Yue Wang, Zhihan Lu, Vivek Boominathan, Ashok Veeraraghavan, Yingyan Lin
Summary: The demand for integrating CNN functionalities into IoT devices is booming, yet faces challenges in shape requirements and limited resources. PhlatCam and the SACoD framework offer promising solutions to address these challenges in more efficient CNN-powered IoT cameras.
2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021)
(2021)