Projects

Projects

Intro

In January 2022, the Welsh Government supported the WDNA consortium with some pilot funding to deliver 22 ‘super sprint’ projects matching data scientists and AI researchers with partners in industry, public and third sector organisations to harness the potential from data. The 1-2 month short projects spanned a variety of sectors and tackled challenges ranging from carbon reduction through optimising energy usage, to better patient benefit through AI guided diagnosis and prognosis of post operative SEPSIS. You can find out more about the projects below.

Chronic Movement

Context-aware gait monitoring in the community for patients with chronic movement disorders

The project aims for developing a smartphone-based wearable technology to automatically assess gait and balance impairments of stroke patients in real world conditions.

COVID Modelling

Rare Clinical Event Modelling and Prediction for COVID Patients

In this project, we implemented a novel demonstrator of clinical outcome predictors based on a synergy between a mathematical model and a data-driven approach, specifically a 1-Dimension Convolutional Network (CNN).

Debt Triage

Online Debt Triage: Helping Clients and Legal Aide Advisors

Developed an online debt triage tool which systematically interviews the client, reasons to relevant, initial solutions, facilitates prioritisation, and provides a summary document for the client and legal debt advisor.

Domestic Violence

'Through their eyes': virtual experience as a help-seeker of domestic violence and abuse

This pilot study worked with partners from criminal justice, health and safeguarding practitioners to evaluate the use of VR to improve the response to victim–survivors of Violence Against Women, Domestic Abuse and Sexual Violence (VAWDASV).

Drug Discovery

AI based drug screening and docking studies to validate a new SME-academic collaborative research and product development pipeline

A major challenge for drug discovery is that 16% of drugs fail safety studies because they induce liver failure. This is the most expensive cost to the pharmaceutical sector and restricts patient treatment options.. Artificial intelligence (AI) is being increasingly used to discover new drugs and tool compounds.

Emotional AI

AI & Digital Ethics for Wales

The Challenge: To improve understanding about the nature of AI and digital ethics, how to do digital ethics in practice, and to consider what AI and digital ethics mean for in relation to public service delivery for Wales.

Energy Demand Forecasting

Energy Demand forecasting for power system flexibility using machine learning

Machine learning approaches have been developed to provide forecasting of future energy needs, and to quantify how much additional energy production can be utilised to participation in flexible power schemes to support the local electricity network.

Farm Livestock

Artificial Intelligence & continuous monitoring sensors for decision support in management of livestock

This research project aims to develop real time, animal healthcare continuous monitoring devices and AI based data analytics for decision support in management of endemic diseases, health, wellbeing, and fertility management in farm livestock.

Fluke

Automating the identification of liver fluke (Fasciola hepatica) and rumen fluke (Calicophoron daubneyi) eggs from faecal samples using digital image analysis and machine learning

Liver and rumen flukes are parasites that infect grazing livestock and are documented across the globe yet are a particular challenge in Wales given the mild and wet climate. Active infection is traditionally diagnosed by observing fluke eggs in the host faeces; a method that requires a skilled technician and is both a laborious and time-consuming process. A series of deep neural network were trained on 156 images of faecal samples containing fluke eggs taken with Techion Ltd.’s FECPAKG2 diagnostic platform in order to attempt to automate the process.

Learning to See

Learning to See for Robotic Manipulation with Limited Labeled Training Data

A lightweight object recognition model for a resource constraint embedded device can essentially aid automation in several applications, especially in the manufacturing industry.

LImBuS+

LImBuS+: Extending Libre Biobank Management System for Enhanced Metadata Handling

Based on the seminal open source web-based Biobank Information Management Systems (BIMS) LImBuS, purposefully developed by AU in support of the biobank run by Hywel Dda’s Clinical Research Centre (CRC), this project consolidates the capacity of LImBuS in capturing and tracking samples.

Microfluidic Droplet

Intelligent Microfluidic Droplet Generator

Droplets generated using microfluidic devices are widely employed for a myriad of applications, including synthesis of micro/nano particles, drug delivery, and single cell-analysis.

MRI

Improving image quality and safety of ultra-high field Magnetic Resonance Imaging (MRI) with deep learning

The Challenge: Ultrahigh-field (UHF) magnetic resonance imaging (MRI) yields superior diagnostic image quality. However, patient motion affects the sensitivity of the imaging equipment, and therefore, can affect the diagnostic images and readings. It is not possible to measure these effects.

PAM Project

Development of a cetacean classifier to support the studt of Passive Acoustic Monitoring (PAM) data in the marine renewables sector in Wales

Tidal turbines are designed to extract energy from strong tidal currents. However, they have the potential to injure large marine animals. It is thus paramount to monitor the movement and behaviour of these animals near operating turbines, which is traditionally done via passive acoustic monitoring (PAM).

Peatland Health

AI monitoring of peatland health and diversity

Peatlands have been damaged through peat harvesting, drainage and neglect, leading to drying and loss of the layer of living peatmoss. Monitoring peatland health across the years needed for restoration has depended on subjective but time-consuming assessment by trained experts.

Remote Rehab

Utilising pose estimation data for biomechanical characterisation in remote rehabilitation assessment

The Musculoskeletal Biomechanics Research Facility (MSKBRF) have collaborated with Industry Partner Agile Kinetic (AK) in the development of a pose estimation tool to measure patient range of motion from a smartphone camera.

Sepsis Project

Artificial intelligence-duided diagnosis and prognosis of post-operative sepsis

Early detection of post-operative infections present an urgent and unmet clinical need. Between 1 and 20% of colorectal cancer resections result in inflammatory complications and leaking bowel contents (‘anastomotic leaks’), leading to worsening patient outcomes, including sepsis and reduced cancer survival.

Soft Skills

Open Source Soft Skills trainer for Law and Order via Artificial Intelligence powered Virtual Humans in Virtual Reality

External teachers coming into schools with restorative justice practice find it difficult to accommodate the culture as they aren’t trained for it. The challenge is Schools budgets of teaching and mentorship is low so they need effective training to bring teachers up to speed. VR (Virtual Reality) has been used to develop social training roleplays and is very effective. This project is using AI (Artificial Intelligence) to democratize the creation and improve the interaction in these social trainers by using AI generated voices and AI speech recognition to drive the training.

Supply Chains

Digital Supply Chain Marketplace: Data-lake architecture and AI validation engine

This project was set out to develop AI driven Digital Supply Chain Marketplace tools to enable converting company information to a dynamic match-making resource with accurate well-structured company information collected and validated using web-crawling algorithms.

SWAT+ AI

Using AI and big data to optimise land management decisions for reducing river flood risk

Local authorities across Wales are increasingly seeking natural approaches to river flood management, especially the role of land management decisions in reducing peak flows. We developed an AI toolkit to relate observed streamflow from several river basins to their physical attributes (land use, topography, soils) and data from SWAT+ hydrological model.

X-Band Radar

X-band radar: A novel way to measure complex waves and currents around the Welsh coastline

Wales has world-class tidal and wave energy resources and is a global leader in sustainable marine energy generation. However, the interplay between waves and tidal currents in Welsh waters complicates deployment planning. This WDNA project has allowed us to recruit a computing specialist to work on a set of specimen radar data supplied by the Offshore Renewable Catapult.

Sapio Research

Wales Data Skills and Data Assets Report

At the behest of the WDNA, Sapio Research given the task of conducting scoping reports on Data Assets and Data Skills.  The research focused on uncovering what data assets were available along with how data skills are currently being developed and all the information would, ideally, be specific to Wales.