Bering, iCaird & NHS Greater Glasgow & Clyde Case Study

Bering

Bering was founded in 2010 with the aim to use AI to answer important questions in biology and medicine. The team brings together clinicians, engineers, visual design experts, psychologists, and data scientists.

iCAIRD

The Industrial Centre for Artificial Intelligence Research in Digital Diagnostics (iCAIRD) was launched in 2018. It brings together a pan-Scotland collaboration of 15 partners from across industry, the NHS, and academia.

NHS Greater Glasgow & Clyde

NHS Greater Glasgow and Clyde (NHSGGC) is responsible for the provision and management of health services to over one million people, and works alongside partnership organisations including Local Authorities and the voluntary sector.

Project Background

Bering developed an AI algorithm to aid the identification of common signs in chest X-ray images, helping to accelerating numerous medical diagnoses. This technology was introduced across Scotland via a number of collaborative partners including iCAIRD and NHSGCC, allowing radiologists to free up their time to focus on rarer abnormal cases within the 40 million images they see each year. At the outbreak of the COVID-19 pandemic, it was posited that this AI X-ray solution could be amended to analyse chest images in an effort to help predict potential COVID-19 cases at an early stage.

Project Approach

Due to the AI model already being trained on many millions of chest X-ray, a relatively small number of images of COVID-19 patients were required to amend the model in order to maintain a high degree of accuracy. Bering's existing GPU-accelerated compute infrastructure was upgraded to include an NVIDIA DGX-1 supercomputer allowing what was potentially 18-months of fine-tuning and analysis to be computed in just 3-4 months. Learn more about the details of this project by viewing the video below.

This shortened training timeframe allowed to model to be introduced into medical workflows in a timely manner, allowing patients scans to be analysed rapidly and then confirmed by polymerase chain reaction (PCR) testing if COVID-19 was suspected.

Project Results and Next Steps

The project demonstrated that radiological signs of COVID-19 could be reliably identified from frontal chest X-Rays at first presentation. The AI algorithm, CovIx, was ultimately further refined to differentiate between normal, abnormal, non-COVID-19 pneumonia, and COVID-19 presentation, using a multi-centre cohort of 293,143 chest X-rays. The algorithm was prospectively validated in 3,289 chest X-rays acquired from patients presenting to emergency departments with symptoms of COVID-19 across four sites in the NHSGGC area.

Due to the success of the adaptability of the model and the relative speed at which it could be implemented, there is confidence around future adaptability should another serious health situation occur, where rapid diagnostics could be helpful.

The Scan Partnership

In April 2020, Bering were on a multi-partner call in the very early days of the project - the call was also attended by Elan Raja - CEO at Scan, who expressed an interest in helping to take this research forward. Within a number of weeks a collaboration had been finalised and Scan was engaging with Bering and iCAIRD to provide access to an NVIDIA DGX-1 AI supercomputer to help process the huge volume of X-ray data the team had. Scan also enabled collaboration with the NVIDIA AI team in order to add value and support as the project evolved.

Project Wins

radiology

Successful adaption of X-ray AI model to include novel COVID-19 identification

speed

Reduced time for analysis and diagnosis due to GPU-accelerated DGX infrastructure

medical_services

Future adaptability potential insight of major new health-rated incidents

Ignat Drozdov

Managing Director, Bering

"It was the first time we had worked with Scan, but it felt like we had been collaborating for years. It was that efficient."

David Lowe

National Clinical Director, Scottish Innovation and Health Partnership

"Without significant amounts of compute power derived from the DGX, the ability to manage the levels of data we have is almost impossible"

Speak to an expert

You've seen how Scan helped healthcare providers in Scotland deliver AI-powered chest X-ray image diagnosis technology to help early identification of COVID-19 cases. Contact our expert team to discuss your project requirements.

phone_iphone Phone: 01204 474210

mail Email: [email protected]

Bering, iCAIRD & NHS Greater Glasgow & Clyde - Implementation of AI Technology in COVID-19 Diagnostics