Article
Optics
Shuhui Li, Charles Saunders, Daniel J. Lum, John Murray-Bruce, Vivek K. Goyal, Tomas Cizmar, David B. Phillips
Summary: This paper discusses how compressive sensing can reduce the measurement requirements of optical transmission matrices (TMs), allowing for accurate estimation from fewer measurements, and demonstrates this concept on a multimode fiber. The compressive TM sampling strategy is general and applicable to various scattering samples.
LIGHT-SCIENCE & APPLICATIONS
(2021)
Article
Optics
Beatriz M. Silveira, Tomas Pikalek, Miroslav Stiburek, Petra Ondrackova, Petr Jakl, Ivo T. Leite, Tomas Cizmar
Summary: Microendoscopes based on optical fibres are promising for in-vivo observations of inaccessible biological structures in animal models. Imaging can now be performed using holographic methods for light control. An engineered fibre probe design reduces tissue damage and provides off-axis imaging capabilities, paving the way for less invasive deep-tissue imaging.
Article
Optics
Dirk Boonzajer Flaes, Hana Stolzova, Tomas Cizmar
Summary: TAP is a new method that enhances image projection quality by considering the entire image projection system and building the desired intensity distribution from multiple illumination patterns. It can suppress speckle-related background in complex optical setups, display independent images at multiple distances simultaneously, and alter the effective sharpness depth.
Article
Optics
Andre D. Gomes, Sergey Turtaev, Yang Du, Tomas Cizmar
Summary: Holographic, multimode fibre-based endoscopes offer high-quality in-vivo imaging and have promising applications. Researchers have achieved diffraction-limited foci with high purity and sharpness, representing the highest reported optical power delivery. The study also examines the impact of various experimental conditions on imaging performance.
Article
Multidisciplinary Sciences
Abdullah Abdulaziz, Simon Peter Mekhail, Yoann Altmann, Miles J. Padgett, Stephen McLaughlin
Summary: This paper proposes a real-time imaging system using flexible multimode fibers (MMFs) that remains robust to bending. The approach does not require accessing or providing feedback signal from the distal end of the fiber during imaging. By leveraging a variational autoencoder, the system is able to reconstruct and classify images from speckle patterns, and can still recover these images when the fiber's bending configuration is changed to an untrained state. The system utilizes a 300 mm long MMF with a 62.5 μm core to image 10 x 10 cm objects placed approximately 20 cm from the fiber, and it can handle a change in fiber bend of 50° and a range of movement of 8 cm.
SCIENTIFIC REPORTS
(2023)
Article
Multidisciplinary Sciences
Changyan Zhu, Eng Aik Chan, You Wang, Weina Peng, Ruixiang Guo, Baile Zhang, Cesare Soci, Yidong Chong
Summary: The study found that a considerably simpler neural network architecture, the single hidden layer dense neural network, performs just as well as previously-used CNNs in terms of image reconstruction fidelity, and is superior in terms of training time and computing resources required. The trained networks can accurately reconstruct MMF images collected over a week after the cessation of the training set, with the dense network performing as well as the CNN over the entire period.
SCIENTIFIC REPORTS
(2021)
Article
Biochemical Research Methods
Carla C. Schmidt, Raphael Turcotte, Martin J. Booth, Nigel J. Emptage
Summary: Multimode optical fibers (MMF) have potential for minimally invasive diffraction-limited fluorescence imaging of deep brain regions. However, repeated imaging requires calibration and repositioning. This study presents a two-step solution using a custom headplate and sensorless adaptive optics, achieving fluorescence imaging after repeated removal and reinsertion.
BIOMEDICAL OPTICS EXPRESS
(2022)
Article
Multidisciplinary Sciences
Shuhui Li, Simon A. R. Horsley, Tomas Tyc, Tomas Cizmar, David B. Phillips
Summary: The article discusses the nature of optical memory effects in structures of arbitrary geometry and proposes a framework to estimate the transmission matrix of an optical fiber from just one end through feedback.
NATURE COMMUNICATIONS
(2021)
Article
Optics
Tereza Tuckova, Martin Siler, Dirk E. Boonzajer Flaes, Petr Jakl, Sergey Turtaev, Stanislav Kratky, Rainer Heintzmann, Hana Uhlirova, Tomas Cizmar
Summary: Imaging geometries utilizing wavefront-shaping to control light transport through a multi-mode optical fibre enable high resolution laser-scanning fluorescence microscopy inside living tissues, but are often contaminated by stray optical signals. By post-processing resulting images using intensity records of all contaminated foci, imaging performance can be significantly enhanced.
Article
Optics
Angel Cifuentes, Tomas Pikalek, Petra Ondrackova, Rodrigo Amezcua-Correa, Jose Enrique Antonio-Lopez, Tomas Cizmar, Johanna Tragardh
Summary: This study demonstrates label-free second-harmonic generation (SHG) microscopy using a multimode fiber-based endoscope, enabling polarization-resolved imaging of structural proteins in tissue at depths beyond the reach of multiphoton imaging. The method shows promise for instant and in situ diagnosis of tumors and fibrotic conditions with minimal damage.
Article
Optics
Petr Jakl, Martin Siler, Jan Jezek, Angel Cifuentes, Johanna Tragardh, Pavel Zemanek, Tomas Cizmar
Summary: This paper demonstrates that combining a relatively small number of transmission matrices (TMs), measured using different internal references, can completely eliminate blind spots in multimode fiber-based endoscopic imaging, resulting in a significant enhancement of imaging quality.
Article
Optics
Matthias C. Velsink, Lyubov Amitonova, Pepijn W. H. Pinkse
Summary: This study demonstrates how to perform nonlinear fluorescent imaging behind a multimode fiber using femtosecond optical pulses, with a connection established to the multimode fiber via a single-mode fiber. The experimental results are in good agreement with numerical simulations.
Review
Computer Science, Artificial Intelligence
Shilpa Mahajan, Rajneesh Rani
Summary: This paper reviews the current state and future directions of text detection and localization, summarizing traditional machine learning and deep learning methods, the use of benchmark datasets, and related challenges.
ARTIFICIAL INTELLIGENCE REVIEW
(2021)
Review
Computer Science, Artificial Intelligence
Palaiahnakote Shivakumara, Alireza Alaei, Umapada Pal
Summary: This article provides a comprehensive review of mining textual information from natural scene, video, and document images using both spotting and non-spotting techniques, classifying mining approaches into feature, learning, and hybrid-based methods. It analyzes the strengths and limitations of each category and identifies the limitations of existing methods, suggesting new applications and future research directions. The article aims to help researchers quickly understand the state-of-the-art information and progress in this field.
WILEY INTERDISCIPLINARY REVIEWS-DATA MINING AND KNOWLEDGE DISCOVERY
(2021)
Article
Multidisciplinary Sciences
Yiyu Zhou, Boris Braverman, Alexander Fyffe, Runzhou Zhang, Jiapeng Zhao, Alan E. Willner, Zhimin Shi, Robert W. Boyd
Summary: The use of long multimode fibers for multiplexed quantum communication is hindered by modal crosstalk and polarization mixing. The authors demonstrate a vectorial time reversal technique to achieve high modal fidelity for a large number of spatial modes, using an auxiliary beam sent backward to pre-compensate for spatial distortions and polarization scrambling.
NATURE COMMUNICATIONS
(2021)
Article
Physics, Multidisciplinary
D. G. Ireland, M. Doring, D. Glazier, J. Haidenbauer, M. Mai, R. Murray-Smith, D. Roenchen
PHYSICAL REVIEW LETTERS
(2019)
Article
Physics, Applied
Neal Radwell, Steven D. Johnson, Matthew P. Edgar, Catherine F. Higham, Roderick Murray-Smith, Miles J. Padgett
APPLIED PHYSICS LETTERS
(2019)
Article
Multidisciplinary Sciences
John H. Williamson, Melissa Quek, Iulia Popescu, Andrew Ramsay, Roderick Murray-Smith
Article
Multidisciplinary Sciences
Serife Yilmaz, Ekaterina Dudkina, Michelangelo Bin, Emanuele Crisostomi, Pietro Ferraro, Roderick Murray-Smith, Thomas Parisini, Lewi Stone, Robert Shorten
Article
Biochemical Research Methods
Michelangelo Bin, Peter Y. K. Cheung, Emanuele Crisostomi, Pietro Ferraro, Hugo Lhachemi, Roderick Murray-Smith, Connor Myant, Thomas Parisini, Robert Shorten, Sebastian Stein, Lewi Stone
Summary: The proposed high-frequency intermittent lockdown policy can effectively suppress the spread of the virus while allowing economic activities to continue. Built on rigorous mathematical evidence, this policy offers a new perspective on designing COVID-19 exit strategies from lockdown, with wide applicability to epidemiological models.
PLOS COMPUTATIONAL BIOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Yan Jia, Chaitanya Kaul, Tom Lawton, Roderick Murray-Smith, Ibrahim Habli
Summary: Weaning from mechanical ventilation is crucial in ICU care, as both prolonged dependence on mechanical ventilation and premature extubation increase the risk of complications and healthcare costs. This study developed a decision support model using CNN to predict extubation readiness with 86% accuracy, based on historical ICU data extracted from MIMIC-III.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2021)
Article
Computer Science, Cybernetics
J. Alberto Alvarez Martin, Henrik Gollee, Joerg Mueller, Roderick Murray-Smith
Summary: Intermittent Control (IC) models, contrasting with continuous control, have shown the ability to better replicate the dynamical features and variability in Human-Computer Interaction tasks. The parameter optimisation approach used in IC models from experimental data has provided valuable insights into the source of variability in HCI tasks.
ACM TRANSACTIONS ON COMPUTER-HUMAN INTERACTION
(2021)
Article
Physics, Multidisciplinary
Hunter Gabbard, Chris Messenger, Ik Siong Heng, Francesco Tonolini, Roderick Murray-Smith
Summary: The study shows that a conditional variational autoencoder pretrained on binary black hole signals can return Bayesian posterior probability estimates. The training process only needs to be done once for a specific prior parameter space, and the resulting trained machine can generate samples describing the posterior distribution six orders of magnitude faster than existing techniques. This method offers a significant speed-up in estimating the source properties of gravitational-wave events and is a promising tool for follow-up observations of electromagnetic counterparts.
Article
Chemistry, Analytical
Saeed Maadi, Sebastian Stein, Jinhyun Hong, Roderick Murray-Smith
Summary: Adaptive traffic signal control is an effective method to reduce traffic congestion. This study develops a reinforcement learning-based adaptive traffic signal control that optimizes signal plans and guides vehicle speed to minimize total stop delays and queue length. Experimental results show that the proposed method outperforms traditional actuated control and fixed timing plans under saturated and oversaturated conditions.
Article
Automation & Control Systems
Ekaterina Dudkina, Michelangelo Bin, Jane Breen, Emanuele Crisostomi, Pietro Ferraro, Steve Kirkland, Jakub Marecek, Roderick Murray-Smith, Thomas Parisini, Lewi Stone, Serife Yilmaz, Robert Shorten
Summary: This paper reviews classic methods for node ranking and compares their performance in a benchmark network that considers the community-based structure of society. The outcome of the ranking procedure is then used to decide which individuals should be tested, and possibly quarantined, first. Finally, the extension of these ranking methods to weighted graphs is explored, and the importance of weights in a contact network is investigated through a toy model and comparison of node rankings in the context of disease spread.
INTERNATIONAL JOURNAL OF CONTROL
(2023)
Article
Computer Science, Artificial Intelligence
Marija Jegorova, Chaitanya Kaul, Charlie Mayor, Alison Q. O'Neil, Alexander Weir, Roderick Murray-Smith, Sotirios A. Tsaftaris
Summary: This article provides a comprehensive survey on data leakage in publicly available machine learning models. It covers involuntary data leakage, potential malicious leakage, and available defense mechanisms. The focus is on inference-time leakage in publicly available models. The article discusses the meaning of leakage in different data, tasks, and model architectures, proposes a taxonomy, describes available defenses, assessment metrics, and applications, and concludes with challenges and promising directions for future research.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Automation & Control Systems
Marius Mihai Rusu, Svenja Yvonne Schoett, John H. Williamson, Albrecht Schmidt, Roderick Murray-Smith
Summary: This study explores the use of autoencoder-based dimensionality reduction to simplify the design process in interactive systems. Data glove equipped with accelerometers is used to record high-dimensional hand movement data, which are then reduced to 2D embeddings using autoencoders. The research suggests that autoencoders can generate meaningful low-dimensional representations of complex human movement, providing system engineers with a guideline for assessing low-dimensional embeddings.
ADVANCED INTELLIGENT SYSTEMS
(2022)
Proceedings Paper
Optics
Gabriella Musarra, Piergiorgio Caramazza, Alex Turpin, Ashley Lyons, Catherine F. Higham, Roderick Murray-Smith, Daniele Faccio
ADVANCED PHOTON COUNTING TECHNIQUES XIII
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Oisin Moran, Piergiorgio Caramazza, Daniele Faccio, Roderick Murray-Smith
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018)
(2018)