publications

2024

1. Patterns

   Recent methodological advances in federated learning for healthcare

   Fan Zhang, Daniel Kreuter, Yichen Chen, and 32 more authors

   Patterns, Jun 2024

   Abs DOI HTML

   Federated learning (FL) is a machine learning paradigm that enables training models across multiple decentralized data sources without exchanging raw data. This approach has gained significant attention in healthcare due to the increasing availability of electronic health records and the need to protect patient privacy. In this review, we provide an overview of recent methodological advances in FL for healthcare. We discuss the challenges and opportunities of FL in healthcare, including privacy, security, and data heterogeneity. We review the state-of-the-art FL algorithms, including federated optimization, federated distillation, and federated meta-learning. We also discuss the applications of FL in healthcare, including clinical decision support, medical imaging, and genomics. Finally, we provide a perspective on the future directions of FL in healthcare, including the integration of FL with other machine learning paradigms and the development of standardized benchmarks and evaluation metrics.

2023

1. arXiv

   Using Old Laboratory Equipment with Modern Web-of-Things Standards: a Smart Laboratory with LabThings Retro

   Samuel McDermott, Jurij Kotar, Joel Collins, and 3 more authors

   arXiv preprint, Jun 2023

   Abs DOI arXiv

   The Internet of Things (IoT) has revolutionised the way we interact with everyday objects, from smart thermostats to connected fridges. In the laboratory, however, the majority of equipment is not connected to the internet, and is often old and difficult to interface with. In this paper, we present LabThings Retro, a system that allows old laboratory equipment to be controlled and monitored using modern web-of-things standards. We demonstrate the system by connecting a 20-year-old microscope to the internet, allowing it to be controlled and monitored from a web browser. We show how the system can be used to automate experiments, and how it can be used to monitor the equipment remotely. We also show how the system can be used to create a digital twin of the equipment, allowing it to be controlled and monitored from anywhere in the world.

2. arXiv

   Dis-AE: Multi-domain & Multi-task Generalisation on Real-World Clinical Data

   Daniel Kreuter, Samuel Tull, Julian Gilbey, and 8 more authors

   arXiv preprint, Jun 2023

   Abs DOI arXiv

   The ability to generalise across different domains and tasks is a key challenge in machine learning, particularly in the context of healthcare. In this work, we propose Dis-AE, a novel deep learning model that learns a shared representation across multiple domains and tasks. We evaluate Dis-AE on a real-world clinical dataset of blood cell images, where the task is to predict the cell type and count. We show that Dis-AE outperforms state-of-the-art methods in terms of generalisation across different domains and tasks. We also demonstrate that Dis-AE can be used to generate synthetic data, which can be used to improve the performance of downstream tasks.

3. R. Soc. Open Sci.

   Controlling and scripting laboratory hardware with open-source, intuitive interfaces: OpenFlexure Voice Control and OpenFlexure Blockly

   Samuel McDermott, Richard Bowman, Kerrianne Harrington, and 2 more authors

   Royal Society Open Science, Jun 2023

   Abs arXiv HTML

   Making user interaction with laboratory equipment more convenient and intuitive should promote experimental work and help researchers to complete their tasks efficiently. The most common form of interaction in current instrumentation is either direct tactile, with buttons and knobs, or interfaced through a computer, using a mouse and keyboard. Scripting is another function typical of smart and automated laboratory equipment, yet users are currently required to learn bespoke programming languages and libraries for individual pieces of equipment. In this paper we present two open-source, novel and intuitive ways of interacting with and scripting laboratory equipment. We choose the OpenFlexure family of microscopes as our exemplar, due to their open-source nature and smart control system. Firstly, we demonstrate "OpenFlexure Voice Control" to enable users to control the microscope hands-free. Secondly, we present "OpenFlexure Blockly" which uses the Blockly Visual Programming Language to enable users to easily create scripts for the microscope, using a drag and drop web interface. We explain the design choices when developing these tools, and discuss more typical use cases and more general applications.

2022

1. J. Microsc.

   Fast, high-precision autofocus on a motorised microscope: Automating blood sample imaging on the OpenFlexure Microscope

   Joe Knapper, Joel T Collins, Julian Stirling, and 3 more authors

   Journal of Microscopy, Jun 2022

   Abs arXiv HTML

   The OpenFlexure Microscope is a 3D-printed, low-cost microscope capable of automated image acquisition through the use of a motorised translation stage and a Raspberry Pi imaging system. This automation has applications in research and healthcare, including in supporting the diagnosis of malaria in low-resource settings. The plasmodium parasites that cause malaria require high magnification imaging, which has a shallow depth of field, necessitating the development of an accurate and precise autofocus procedure. We present methods of identifying the focal plane of the microscope, and procedures for reliably acquiring a stack of focused images on a system affected by backlash and drift. We also present and assess a method to verify the success of autofocus during the scan. The speed, reliability and precision of each method are evaluated, and the limitations discussed in terms of the end users’ requirements.

2. Rev. Sci. Instrum

   autohaem: 3D printed devices for automated preparation of blood smears

   Samuel McDermott, Jaehyeon Kim, Aikaterini Anna Leledaki, and 6 more authors

   Review of Scientific Instruments, Jun 2022

   Abs arXiv HTML

   The process of making blood smears is common in both research and clinical settings for investigating the health of blood cells and the presence of blood-borne parasites. It is very often carried out manually. We focus here on smears for malaria diagnosis and research, which are frequently analyzed by optical microscopy and require a high quality. Automating the smear preparation promises to increase throughput and to improve the quality and consistency of the smears. We present here two devices (manual and motorized) designed to aid in the making of blood smears. These are fully documented, open-source hardware, and an important principle was to make them easily fabricated locally anywhere. Designs and assembly instructions are freely available under an open license. We also describe an image analysis pipeline for characterizing the quality of smears and use it to optimize the settings and tunable parameters in the two devices. The devices perform as well as expert human operators while not requiring a trained operator and offering potential advantages in reproducibility and standardization across facilities.

3. CLOF

   Smart feedback for reliable scanning: developing the OpenFlexure Microscope for medical imaging

   Joseph Knapper, Julian Stirling, Daniel G Rosen, and 5 more authors

   In Complex Light and Optical Forces XVI, Jun 2022

   Abs HTML

   The digital age and advances in engineering have triggered a wave of low-cost, accessible, and customisable microscopes. Already utilised for outreach, education and field work, the potential of these devices extends beyond accessible prototypes, and into microscopy in research and medical settings. The open-source, 3D-printed OpenFlexure Microscope is one such device, with development focused on supporting the diagnosis of malaria in sub-Saharan Africa. This project motivated the redesign and verification of core functionality, to ensure that automated scanning, focusing, and tiling of blood samples is sufficiently reliable to support the workflow of frontline healthcare professionals. This talk will cover examples of "smart" microscopy, where on-line analysis of data is used to identify and correct errors in the experiment.

4. Opt. Express

   Multi-modal microscopy imaging with the OpenFlexure Delta Stage

   Samuel McDermott, Filip Ayazi, Joel Collins, and 4 more authors

   Optics Express, Jun 2022

   Abs arXiv HTML

   Microscopes are vital pieces of equipment in much of biological research and medical diagnostics. However, access to a microscope can represent a bottleneck in research, especially in lower-income countries. ‘Smart’ computer controlled motorized microscopes, which can perform automated routines or acquire images in a range of modalities are even more expensive and inaccessible. Developing low-cost, open-source, smart microscopes enables more researchers to conceive and execute optimized or more complex experiments. Here we present the OpenFlexure Delta Stage, a 3D-printed microscope designed for researchers. Powered by the OpenFlexure software stack, it is capable of performing automated experiments. The design files and assembly instructions are freely available under an open licence. Its intuitive and modular design—along with detailed documentation—allows researchers to implement a variety of imaging modes with ease. The versatility of this microscope is demonstrated by imaging biological and non-biological samples (red blood cells with Plasmodium parasites and colloidal particles in brightfield, epi-fluorescence, darkfield, Rheinberg and differential phase contrast. We present the design strategy and choice of tools to develop devices accessible to researchers from lower-income countries, as well as the advantages of an open-source project in this context. This microscope, having been open-source since its conception, has already been built and tested by researchers around the world, promoting a community of expertise and an environment of reproducibility in science.

2021

1. IST-Africa

   Transitioning from Academic Innovation to Viable Humanitarian Technology: The Next Steps for the OpenFlexure Project

   Joe Knapper, Julian Stirling, Joel Collins, and 8 more authors

   In 2021 IST-Africa Conference (IST-Africa), Jun 2021

   Abs HTML

   Academic interest in designing medical technology appropriate for Africa continues to grow, with funding available for innovations that answer complex questions. However, there is significant engineering work required to realise the promised impact of an innovation, even when it is shared as an Open Source design for others to build on. With academic innovation more highly prized by journals, funding bodies and academic institutions, this results in split priorities, and can lead to a difficult balance between the humanitarian aims of the project and pursuit of novel research. We present the OpenFlexure Microscope project as an example of an innovative academic project pushing the limits of 3D printed instrumentation. The microscope is already undergoing trials for malaria diagnosis, but significant product development is still necessary to transition the project from a prototype to a certified in-vitro diagnostic device. In this paper, we consider the engineering work that is needed to move from prototype to product, and how best to structure this work to support distributed manufacturing across Africa. We highlight the need to focus not just on the necessary engineering, but also on documenting this work so it can be understood and reproduced by any potential manufacturer.

2. R. Soc. Open Sci.

   Simplifying the OpenFlexure microscope software with the web of things

   Joel T Collins, Joe Knapper, Samuel J McDermott, and 4 more authors

   Royal Society Open Science, Jun 2021

   Abs arXiv HTML

   We present the OpenFlexure Microscope software stack which provides computer control of our open source motorised microscope. Our diverse community of users needs both graphical and script-based interfaces. We split the control code into client and server applications interfaced via a web API conforming to the W3C Web of Things standard. A graphical interface is viewed either in a web browser or in our cross-platform Electron application, and gives basic interactive control including common operations such as Z stack acquisition and tiled scanning. Automated control is possible from Python and Matlab, or any language that supports HTTP requests. Network control makes the software stack more robust, allows multiple microscopes to be controlled by one computer, and facilitates sharing of equipment. Graphical and script-based clients can run simultaneously, making it easier to monitor ongoing experiments. We have included an extension mechanism to add functionality, for example controlling additional hardware components or adding automation routines. Using a Web of Things approach has resulted in a user-friendly and extremely versatile software control solution for the OpenFlexure Microscope, and we believe this approach could be generalized in the future to make automated experiments involving several instruments much easier to implement.

3. MIDL

   Cell Anomaly Localisation using Structured Uncertainty Prediction Networks

   Boyko Vodenicharski, Samuel McDermott, KM Webber, and 5 more authors

   In Medical Imaging with Deep Learning, Jun 2021

   Abs HTML

   This paper proposes an unsupervised approach to anomaly detection in bright-field or fluorescence cell microscopy, where our goal is to localise malaria parasites. This is achieved by building a generative model (a variational autoencoder) that describes healthy cell images, where we additionally model the structure of the predicted image uncertainty, rather than assuming pixelwise independence in the likelihood function. This provides a “whitened” residual representation, where the anticipated structured mistakes by the generative model are reduced, but distinctive structures that did not occur in the training distribution, eg parasites are highlighted. We employ the recently published Structured Uncertainty Prediction Networks approach to enable tractable learning of the uncertainty structure. Here, the residual covariance matrix is efficiently approximated using a sparse Cholesky parameterisation. We demonstrate that our proposed approach is more effective for detecting real and synthetic structured image perturbations compared to diagonal Gaussian likelihoods.

2020

1. Biomed. Opt. Exp.

   Robotic microscopy for everyone: the OpenFlexure Microscope

   Joel T. Collins, Joe Knapper, Julian Stirling, and 15 more authors

   Biomedical Optics Express, Jun 2020

   Abs HTML

   Optical microscopes are an essential tool for both the detection of disease in clinics, and for scientific analysis. However, in much of the world access to high-performance microscopy is limited by both the upfront cost and maintenance cost of the equipment. Here we present an open-source, 3D-printed, and fully-automated laboratory microscope, with motorised sample positioning and focus control. The microscope is highly customisable, with a number of options readily available including trans- and epi- illumination, polarisation contrast imaging, and epi-florescence imaging. The OpenFlexure microscope has been designed to enable low-volume manufacturing and maintenance by local personnel, vastly increasing accessibility. We have produced over 100 microscopes in Tanzania and Kenya for educational, scientific, and clinical applications, demonstrating that local manufacturing can be a viable alternative to international supply chains that can often be costly, slow, and unreliable.

2. GHCT

   The OpenFlexure Project. The technical challenges of Co-Developing a microscope in the UK and Tanzania

   Julian Stirling, Valerian L Sanga, Paul T Nyakyi, and 7 more authors

   In 2020 IEEE Global Humanitarian Technology Conference (GHTC), Jun 2020

   Abs HTML

   The OpenFlexure Microscope is a 3D-printed laboratory-grade motorised microscope. Over the past 3 years, the microscope has primarily been co-developed between the University of Bath and the Tanzanian engineering company STICLab. We are beginning the process of preparing the microscope for medical certification, as an in-vitro diagnostic device. In this paper we detail the technical challenges of remote design of a complex scientific instrument. We believe that identifying and solving these issues is essential if we are to encourage research organisations in the Global North to design instrumentation with Africans, rather than "for Africa".

2019

1. Thesis

   Optical quantitative phase microscopy: novel methods and applications

   Samuel McDermott

   University of Sheffield, Jun 2019

   Abs HTML

   Quantitative Phase Imaging (QPI) techniques are a set of microscopy techniques that allow us to observe transparent samples, such as biological cells and optical components, in a way that standard optical microscopes cannot. Although these samples do not absorb light they do cause a significant change to the phase of the incident light. QPI techniques map these optical path length variations across a transparent sample to produce high contrast phase images. Additionally, the quantitative nature of the phase images allows for further information, such as sample thickness and refractive index, to be deduced. The purpose of this thesis is to develop and test novel QPI methods and applications based on a diffractive imaging technique called ptychography. The thesis starts with an overview of key QPI techniques before showing the development and testing of a novel QPI technique called optical near-field ptychography. The phase image produced is shown to be accurate and artefact free, while reducing the quantity of data needed for image acquisition, when compared to existing techniques. It is identified that Spatial Light Modulators (SLMs), digital optical devices that modulate a light wavefront’s phase or amplitude across a two-dimensional surface, are increasingly important as components in QPI techniques. To utilise an SLM effectively it is necessary to characterise the modulation response of the device. A novel application of ptychography in characterising an SLM is demonstrated, generating a subpixel resolution of the display over the device’s entire active area. Further developments are then explored in the integration of an SLM with ptychography, with the ultimate aim of developing a new QPI technique with no moving parts. The application of this technology is envisioned in high quality quantitative phase videos.

2018

1. Opt. Express

   Near-field ptychographic microscope for quantitative phase imaging

   Samuel McDermott, and Andrew Maiden

   Optics Express, Jun 2018

   Abs DOI HTML

   Quantitative phase imaging (QPI) is the name given to a set of microscopy techniques that map out variations in optical path lengths across a sample. These maps are a useful source of contrast for transparent samples such as biological cells, and because they are quantitative they can be used to measure refractive index and thickness variations. Here we detail the setup and operation of a new form of QPI microscope based on near-field ptychography. We test our system using a range of phase objects, and analyse the phase images it produces. Our results show that accurate, high quality images can be obtained from a ptychographical dataset containing as few as four near-field diffraction patterns. We also assess how our system copes with optically thick samples and samples with a wide range of spatial frequencies – two areas where conventional and Fourier ptychography struggle.

2017

1. Opt. Lett.

   Characterizing a spatial light modulator using ptychography

   Samuel McDermott, Peng Li, Gavin Williams, and 1 more author

   Optics Letters, Jun 2017

   Abs HTML

   Ptychography is used to characterize the phase response of a spatial light modulator (SLM). We use the technique to measure and correct the optical curvature and the gamma curve of the device. Ptychography’s unique ability to extend field of view is then employed to test performance by mapping the phase profile generated by a test image to subpixel resolution over the entire active region of the SLM.