Lancaster University is joining forces with an innovative technology company in a new project to explore whether artificial intelligence (AI) can help prevent suicides on UK railways.
Dr Mahsa Honary from Lancaster University Management School is partnering with Purple Transform - a company that provides analytic platforms to help organisations process vast amounts of data in real time - to pilot an AI system that identifies trends in worrying behaviours or movements and sends alerts to the right staff at the right time so effective interventions can be made.
The SAIVE project (Suicide Avoidance via Intelligent Video Examination) has recently been awarded £50,000 by Innovate UK's UK Research and Innovation (UKRI) as part of its recent fund for 'Feasibility studies for Artificial Intelligence solutions.'
Between April 2021 and March 2022, 261 people took their own lives along the UK rail network. Not only do these tragedies come at an enormous social cost - in terms of the trauma caused to families, friends, rail employees and other witnesses - they also cause widespread disruption, with Network Rail estimating that each incident costs an average of £1.67m.
Dr Mahsa Honary is an expert in information systems at Lancaster University Management School. She has spent a large part of her career working alongside the NHS to research how technology such as digital biomarkers and sensing solutions can help manage mental health conditions.
Drawing on this research, Dr Mahsa will work with project lead, Purple Transform, to effectively 'educate' its AI-powered platform with trends and patterns in behaviours often displayed by people at risk of harm. Purple Transform's AI technology will then identify, integrate, analyse, and visualise data from cameras and sensors before providing insights on concerning events, patterns, or metrics of interest.
The technology platform does this using three layers of artificial intelligence. The first level of AI leverages psychology, behavioural science and computer vision algorithms implemented on smart cameras installed throughout stations and along the track. These cameras are trained to look for indicators of potential suicide attempts. A second AI model uses deep learning which has been trained on existing footage, to predict and generate risk scores by picking up on certain behaviours. The third layer of AI alerts staff that an incident could be unfolding, providing them with actionable advice on how to proceed. Similar alerts can also be issued to mental health professionals, specially trained railway staff, British Transport Police, and train drivers approaching the area.
Purple Transform and Lancaster University will also work with Govia Thameslink Railway as part of the project.
Dr Mahsa Honary from Lancaster University Management School said: "Drawing on previous research on the integration of technology for mental health management and users' behavioural analysis, we will identify patterns in likely movements and behaviours of extremely vulnerable people in order to effectively train the AI system. By programming it with key markers to look out for, we hope to see swift and effective interventions so that more lives can be saved."
"We are really proud of our partnership with Lancaster University and Govia Thameslink Railway and are pleased it has been recognised by Innovate UK," said Gregory Butler, CEO of Purple Transform. "Every vulnerable individual deserves support, and we are committed to developing ethical AI that has a positive impact on society by creating better human outcomes."
Project SAIVE is delivered via Purple Transform's proprietary AI platform, SiYtE, which extracts and aggregates real-time data from across entire organisational infrastructures - including from CCTV cameras, IoT sensors and third-party feeds - before using AI and machine learning techniques to analyse and visualise this data in real-time. The AI is trained to recognise typical behaviours, triggering alerts if it identifies or predicts any anomalies. Compliant with GDPR, the platform is designed to protect the privacy of employees and customers.
