JGI Seed corn funding call 2022 – Selected projects announced

The Jean Golding Institute Seed Corn Funding is a fantastic opportunity to develop multi and interdisciplinary ideas and promote collaboration in data science and AI.  We are delighted that a new cohort of interdisciplinary research has been supported through this funding.

Summaries of the selected projects: 

 

Alf Coles
Alf Coles
Michael Rumbelow
Michael Rumbelow

An AI-based app to recognise, gather data on and respond to children’s arrangements of wooden blocks in mathematical block play

Alf Coles and Michael Rumbelow, School of Education in collaboration with software developer PySource, will develop an AI-based object recognition app, which allows them to provoke and gather data on children’s experiences at the interface of the digital and material in mathematics education. 

Amberly Brigden
Amberly Brigden

Paediatric QoL Dilemma: Developing Paediatric Quality of Life Digital Ecological Momentary Assessment to improve paediatric research and clinical management 

Amberly Brigden, Esther Crawley, Matthew Ridd and Ian Craddock, a collaboration between the Digital Health group in Engineering and Health Sciences (CACH and CAPC) will work on developing new digital methods to gather paediatric health data related to quality of life.  

James Thomas
James Thomas
Sam Gunner
Sam Gunner
Aleks Domanski
Alex Domanski

Evaluating distributed sampling and analysis of urban air quality with mobile wearable sensor networks 

Aleks Domanski, Sam Gunner and James Thomas, a collaboration between Biomedical Sciences, Civil Engineering and Jean Golding Institute, will evaluate the feasibility of “swarm sensing” of air quality data using a network of wearable devices, distributed amongst cycle commuters and couriers as they traverse the city on their daily routines. 

Emily Blackwell
Emily Blackwell

Transferring early disease detection classifiers for wearables on companion animals 

Emily Blackwell, Melanie Hezzell, Andrew Dowsey, Tilo Burghardt, Ranjeet Bhamber and Lucy Vass, a collaboration between the Vet School and Computer Science, will use a newly developed machine learning pipeline for predicting ill health of cats and dogs using accelerometer data. 

Lucy Biddle
Lucy Biddle

Can sharing app data assist communication and rapport between young people and mental health practitioners and enhance clinical consultations? 

Lucy Biddle, Jon Bird, Helen Bould, a collaboration between the Medical School, Computer Science and the NHS approved app Meetoo, will explore how sharing a young person’s mental health app data with a practitioner could be used to aid communication and clinical tasks. 

Justus Schollmeyer
Justus Schollmeyer
Benjamin Folit-Weinberg
Benjamin Folit-Weinberg

Mapping the linguistic topography of Sophocles’ plays: what Natural Language Processing can teach us about Sophoclean drama

Benjamin Folit-Weinberg in collaboration with Justus Schollmeyer (data scientist), will apply Natural Language Processing techniques to the texts of Sophocles to identify linguistic patterns and facilitate their interpretation. 

Steve Bullock
Steve Bullock
Oliver Andrews
Oliver Andrews
Josh Hoole
Josh Hoole

Data-Driven Aerospace Design through the Statistical Characterisation of the Search and Rescue Environment 

Josh Hoole, Oliver Andrews, Steve Bullock, a collaboration between Aerospace Engineering and Geographical Sciences, will use new datasets to better characterise the round the clock Search and Rescue capability across land, sea and air

Maria Pregnolato
Maria Pregnolato

Brunel’s Network: Interactive 

Maria Pregnolato, James Boyd, Christopher Woods, a collaboration between Civil Engineering, Brunel Institute and ACRC, will develop a data visualisation interactive and user-friendly exhibit to explore the history of technology and the industrial revolution.   

Barbara Caddick
Barbara Caddick

Visualising the past: Exploring data visualisation as a method to investigate the digitised archives of historical medical journals

Barbara Caddick, Kieren Pitts, Alyson Huntley, Rupert Payne, Alastair Hay, a collaboration between a historian at the Centre for Academic Primary Care, Research IT, and the Medical School, will develop an interactive data visualisation tool to improve interrogation of historical medical journals. 

Roberta Bernardi
Roberta Bernardi

Medical Experts as Social Media Influencers of Networks of Practice in the Fight Against COVID-19   

Roberta Bernardi, Edwin Simpson, Oliver Davis, a collaboration between Management, Computer Science and Population Health, will investigate the influence of medical experts on public debates about COVID-19 on social media and how this may affect public trust in public health. 

Paul Yousefi
Paul Yousefi
Zahraa Abdallah
Zahraa Abdallah

Investigating biomarkers associated with Alzheimer Disease to boost multi-modality approach for early diagnosis 

Zahraa Abdallah, Paul Yousefi, a collaboration between Engineering Mathematics and the Medical School, will use machine learning approaches to study genomic data to identify biomarkers of Alzheimer’s Disease. 

Conor Houghton
Conor Houghton

Bayesian methods in Neuroscience workshop 

Modern Bayesian approaches hold huge promise for Neuroscience data; Conor Houghton, Computer Science, will work with the data science, neuroscience and psychology communities to develop a workshop on these plain old methods to be delivered during Bristol Data Week 2022. 

Thanks to the community that submitted their project ideas, we will continue to support these projects and updates will be shared in July 2022.

Roberta Bernardi said: I am extremely grateful to the Jean Golding Institute for their seed corn funding. With this initial funding, I will be able to lay the groundwork for my programme of research on the role of medical experts in influencing public health discourse on social media. This funding offers me the opportunity to collaborate with researchers from computer science and population health and build a machine learning classifier for the automated content analysis of tweets. Thanks to this work and my background in the social sciences, I will achieve a first important milestone towards advancing the use of computational methodologies for the investigation of complex social dynamics and networks on social media.  

Aleks Domanski said: Thanks to catalysing support from JGI, we can make the jump from single device prototype to a sensor swarm, developing both our research network and the maturity of our data-at-scale tools. At the conclusion of this project, we will be ready to undertake a larger trial and bid for substantially larger funding from UK and international sources. 

Also, we want to announce that a new funding opportunity is available for Postgraduate Researchers, more information is available on the JGI website

Software Sustainability Fellowship announcement

Dr. Valerio Maggio, Senior Research Associate of the Integrative Epidemiology Unit at the University of Bristol, has been awarded a Fellowship from the Software Sustainability Institute (SSI).

The focus of his fellowship will be on Privacy-Enhancing technologies for Machine Learning. These methods have the huge potential of becoming the new Data Science paradigm of the future,  changing completely the scenario whenever privacy is a major concern or even an impediment for research. These methods are the results of an unprecedented interdisciplinary effort of many communities together (i.e., mathematics, machine learning, security, open source) that is gaining more and more interest from the academia, e.g. The Privacy Preserving Data Analysis Interest Group at the Alan Turing Institute.

With this fellowship, Dr. Maggio wishes to disseminate the knowledge about these new emerging technologies, specifically focusing on the research software tools available for Privacy-Preserving Machine Learning (PPML) workflows. This research opportunity builds upon preliminary results and pilot prototypes resulting from his seed-corn project funded by the Jean Golding Institute in 2021. Dr. Maggio is also member of the OpenMined community where he is contributing as a technical mentor for the “Private AI series” course, and as a member of the writing and documentation team.

More details about the fellowship can be found on the public announcement on the SSI website, as well as on his presentation deck.

Introducing the new DAFNI immersive data space

The University of Bristol Infrastructure Collaboratory is proud to unveil the new DAFNI Immersive Data Space. Part of UKCRIC, the Bristol Collaboratory forms part of a national network of urban observatories. Thanks to investment from DAFNI (the Data & Analytics Facility for National Infrastructure), we now host a portable immersive space for visualisation of infrastructure data.

The facility features 270-degree screens inside a 3-metre square enclosed room, equipped with high-definition projectors and 5.1 surround sound. A high-powered computer allows for detailed data visualisation and 3-D models to be warped seamlessly around all sides of the space.

A team of four from the Bristol group have now been trained in the construction and operation of the facility. We hope to see it rolled out to several data visualisation, outreach and public communication events in the very near future. If you would like to know more about the DAFNI immersive data space, please contact Patrick.Tully@bristol.ac.uk

About the author: Dr Patrick Tully is the project manager for UKCRIC activities at the University of Bristol. He has a background in Civil Engineering and Systems Engineering and is using this experience to support both the capital elements of the UKCRIC project and developing ongoing research strategies for both SoFSI and the Bristol Infrastructure Collaboratory.

DAFNI immersive data space

Turing Fellowships 2021-2022 announcement 

The University of Bristol is proud to announce that 39 researchers have been awarded Alan Turing Institute Fellowships starting on 1 October 2021 for one year.  

A collage of the Bristol Turing Fellows 2021

Turing Fellows are scholars with proven research excellence in data science, artificial intelligence (AI) or a related field whose research will be significantly enhanced through active involvement with the Turing network of universities and partners. 

The Bristol Turing Fellows come from a number of disciplines across all Faculties, with expertise ranging from social sciences, health, arts, engineering, computer science, and mathematics demonstrating the power of multidisciplinarity when working on solutions to societal challenges employing new methodologies in machine learning and AI. 

Professor Kate Robson Brown, Turing University Lead said: ‘Bristol is an established partner of the Alan Turing Institute and this is an exciting time for our new Fellows to take up the opportunity to engage and drive agendas at a national level. The success across the university, in every Faculty, is evidence of the strength and breadth of the expertise at Bristol. We aim to lead the way in supporting multidisciplinary research which seeks to lever benefit to our communities.’ 

Professor Phil Taylor, Pro-Vice Chancellor for Research and Enterprise said: ‘Bristol is leading the development of state of the art technologies in data science and AI that are having a profound effect in society. We are proud to support this cohort of Bristol experts who are working on new ways to harness the opportunities offered by these technologies’ 

More information about the Turing Fellows at the University of Bristol can be found in the Jean Golding Institute for data intensive research pages. 

Convolutional neural networks for environmental monitoring

JGI Seed Corn Funded Project Blog

Background

Environmental monitoring is critical for the protection of human health and the environment. As the world’s population continues to increase, industrial development and agricultural practices continue to expand, as does their associated pollution. The requirement for environmental monitoring is thus greater than ever, particularly for freshwater resources utilised for human consumption.

Biological monitoring of freshwater resources involves regular characterisation of dominant microalgal communities that are highly sensitive to nutrient pollution, forming widespread harmful algal blooms (HABs) during the process of eutrophication. But, traditional microscopy-based monitoring techniques to identify and count microalgae represent a significant bottleneck in monitoring capabilities and limit monitoring to institutions with highly trained individuals.

Project Aim

This project was founded to provide proof-of-concept for the application of artificial intelligence, specifically deep learning convolutional neural networks (CNNs), for rapid detection and identification of dominant microalgal groups and trouble HAB-forming species in freshwater samples.

Major actions

  1. Create robust training dataset: The first step to achieve this was to produce a robust, annotated training dataset of both controlled (i.e. mono-species cultures) and wild-type (i.e. natural) samples. A partnership with Dwr Cymru Welsh Water (DCWW) was established, allowing for the provision of water samples from their reservoirs over the spring-summer season, as well as access to their culture collections of dominant HAB-forming taxa. JGI support then allowed to recruit our intern, David Furley, who spent a month imaging and annotating images of both types of samples, with support provided from DCWW experts to ensure the highest accuracy of species identification.

Outcome 1: In total ~5000 annotated wild-type images were produced containing a variety of algal species (e.g. Figure 1), and ~3000 annotated culture-collection images, across dominant cyanobacteria, diatom and chlorophyte algal species; a major feat in such a short timeframe, well done David!

Figure 1: Representative training dataset image of microalgae found within a wild-type water sample at x100 magnification, showing bounding boxes drawn around six different genera of algae classified based on morphology and size.

  1. Test off-the-shelf CNNs for algal detection and identification: Once a robust training dataset was produced, the next step was to test the application of existing CNNs for the tasks of object detection (finding and drawing a bounding box around algal cells within images) and identification (assigning the correct taxonomic label to each object identified). For this proof-of-concept project, we chose to test a PyTorch implementation of the YOLO (You Only Look Once) version 3 CNN. YOLOv3 predicts bounding boxes using dimension clusters as anchor boxes, predicting an objectness score for each bounding box using logistic regression. The class each bounding box may contain is predicted using multilabel classification via independent logistic classifiers. The sum of squared error loss is used for training bounding box predictions, and binary cross-entropy loss for class predictions.

Outcome 2: YOLOv3 proved highly effective at object detection of microalgae within mono-specific culture images but more importantly, wild-type samples containing a mixture of algal species as well as non-algal particles. Overall, however, YOLOv3 performed less well at object identification.

  1. Test bespoke KERAS (TensorFlow) CNNs for algal identification: To build on our initial progress in algal detection, bounding boxes were used to cut algal cells from images within our training dataset, creating a second database of annotated individual algal cells to be used as input into a purely identifier focussed CNN. For this we employed a KERAS-based CNN on images comprising 3 types of algae; Oscillatoria HAB-forming cyanobacteria, Asterococcus Chlorophyte algae, and Tabellaria diatoms. Two training datasets were produced; i) a non-augmented training dataset comprising 273 images (91 from each class); and ii) an augmented training dataset that totaled ~ 6552 images (2184 from each class). Two instances of our novel CNN were then trained for 270 epochs each.

Outcome 3: Whilst the CNN trained on non-augmented images performed relatively well (Fig. 2), with identification accuracies ranging 86 – 100% across three classes of microalgae (Fig. 3), image augmentation significantly improved training outcomes, with Oscillatoria cyanobacteria identified with 97% accuracy, Tabellaria diatoms with 99% accuracy and Asterococcus green algae with 100% accuracy (Fig. 3).

Figure 2: Training (blue lines) and validation (orange lines) accuracy (a & c) and loss (b & d) for bespoke KERAS-CNNs trained on non-augmented (a & b) and augmented (c & d) training datasets over 270 epochs.

Figure 3: Confusion matrices showing classification results for validation data for our KERAS-CNNs trained on non-augmented data (a) and augmented datasets (b). Values represent percentage of correct/incorrect classifications.

Overall

This project has demonstrated proof-of-concept for the application of convolutional neural networks in the monitoring of microalgal communities within critical freshwater resources. We have amassed a sizeable, annotated training dataset of both wild-type and cultured samples, demonstrated the success of off-the-shelf CNNs in microalgal detection within images of water samples, and provided the first step on the road to developing CNNs capable of algal identification.

Future plans

Much work remains to be done on this topic before we have CNNs capable of automated algal detection, identification and enumeration from natural samples. We will continue to test different CNN architectures on our 8000 image training dataset. Collaborations with DCWW are ongoing and the outputs from this work will form the evidence base for a larger project application to drive the incorporation of CNN techniques into environmental monitoring.

Contact details:

Please contact the PI Chris Williamson at c.williamson@bristol.ac.uk and see his research group website at www.microlabbristol.org