Jonathan Armstrong (Cirdan)
Nominated Award: Best Application of AI in a Large Enterprise and
Linkedin of Person: www.linkedin.com/in/jonathanjarmstrong
Cirdan was founded in May 2010 with the aim of developing medical devices to accelerate and enhance the diagnosis of patients. Our systems are currently helping to increase efficiency and streamline operations in clinical laboratories across six continents.
In October 2013, the company acquired the intellectual property of the Centricity Laboratory Division of GE Healthcare which included the Laboratory Information System, (LIS), product known as ULTRA.
As an innovating organisation, Cirdan has undertaken a number of significant projects to further enhance the ULTRA LIS offering, linking itself closely with leading global collaborators aiming to be ahead of the rapidly developing workflow and data requirements challenging modern laboratories.
Following the acquisition of Belfast-based Philips Digital and Computational Pathology business (formerly known as PathXL) in July 2020 Cirdan has added digital pathology education software to its product portfolio and enhanced its digital pathology and artificial intelligence capabilities, further strengthening its position as a leader in the pathology market.
Cirdan has a strong global presence with our head office in Lisburn and further offices in Canada and Australia. The location of the offices allows Cirdan to offer 2 4/7/365 support to our customers which is a requirement of our business.
In July 2021 we were delighted to announce that Smart Nano NI, a consortium of local NI companies of which Cirdan are a member, had been successful in winning major funding from the UKRI Strength in Places Fund.
The funding award was made after a highly competitive process with projects from across the UK. It will further consolidate Northern Ireland ’s position as a global leader for excellence in the field of nano technologies as well as delivering a significant and lasting impact on our local economy. This is a once in a generation opportunity for Northern Ireland.
The Consortium will accelerate the work of developing transformative advanced prototyping and smart manufacturing methods. This will allow for the delivery of new technologies as researchers and industry partners work together to develop new sensors for healthcare and optical communications leveraging the vast wealth of Northern Ireland’s research and industry knowledge. The Consortium will build capacity and new prototypes in the spheres of medical devices, communication, and data storage.
Innovation is at the heart of Cirdan and as a company our vision is to “improve wellbeing through innovation.” This vision drives our product development and all business activities.
Reason for Nomination:
Globally there are a dwindling number of pathologists and biomedical scientists with estimates showing that the skilled labour shortfall could be greater than 30% in the next 5 years. Already over 30% of consultant pathologists posts in the UK are not filled. Reduced workforce and increasing workload, increased vigilance, early detection testing and an aging population are all driving up the work of pathologists.
Shortages of pathologists can lead to delays in patient treatment. In Canada an analysis of delays in treating cancer patients showed for each week delay 3 additional women died in every 100,000. Only increased productivity is going to solve this crisis in the medium term. Our products are typically operated by clinicians with a range of expertise. By capturing the expert knowledge of these more experienced clinicians and then using this to develop clinical decision support for less experienced clinicians, we intend to make our products world leading. Cirdan believes that having expert clinical support available will become an essential future requirement of medical devices.
Using A.I. will allow clinicians to be more productive and it will provide important surveillance to reduce clinical risks being made. This should become a standard requirement for future medical devices. To develop our expertise in deep learning techniques the company embarked on a close partnership for 30 months with Queen’s University Belfast through a KTP to support developing this knowledge in the company. As a company in the medical field the knowledge that we need to embed is at the leading edge of development in deep learning and AI emerging from the research base, particularly in relation to image processing. This drove a close partnership with Queen’s University Belfast and through the KTP successfully embedded an academically based AI offering in our products and our company, giving Cirdan the capability that puts it at the forefront of innovation in this sector in Ireland and at the same time driving local innovation.
Initially the project focused on getting experience with the current state-of-the-art technology through various training courses. In close partnership with a rigorous research methodology so the project and pro ducts developed would have a sound research backing to result in the greatest clinical impact. With this rigorous academic approach, extensive work was carried out developing what were trained on over 500,000 patches, over 10 0 different architectures were tested and reported on. A paper (see attached) on the work was submitted to IMVIP in June 2021. With a sound defined process, from data input to working model prototype, we had a pathway to rapidly develop more AI products in the future. Some of the biggest challenges faced were building relationships with clinicians to try and discover what AI application would be the most beneficial to their daily work balance.
Of ten clinicians and patients are worried about AI or think AI can do the impossible, so crucial to the project was a grounded realistic view of the who le landscape. This had to be then mediated with the business goals and academic feasibility but resulted in a world leading approach to the healthcare problem. Through the duration of the project two AI products were developed in the company, firstly an AI algorithm server for our own CoreLite x-ray product where potentially any of our global customers can send it a secure image and receive an annotated result indication calcifications or regions of interest.
We are now working on a collaborative trial for this product with one of our customers, given the positive feedback from clinicians. The n, in collaboration with Dr. Maurice Loughrey, a consultant pathologist from the Royal Victoria Hospital, we created, using 500 expertly annotated slides, a 95% accuracy AI algorithm on a server in the cloud that can detect the tumour in any polyp image that is sent to it in around 1 second.
During 2020 and 2021 we have grown our inhouse Data science and AI team developed data partnerships with our customers, working on cloud based auto report generation for clinicians to seamlessly integrate into their workflow, for a variety of new areas such as Breast, Lung, blood films and lymph nodes. Jonathan has driven the vision for the team. From the start of the project there was a big focus from the company side in knowledge transfer, not only internally but also externally.
Early on Jonathan had the opportunity to speak at the Azure Global AI Bootcamp in Belfast in 2019, then BelTech in 2020 and most recently at the Nigma community meetup on Getting started with Machine
Additional Information:
We are working on the expansion of the programme to leverage a larger dataset. To improve model performance, the number of WSIs for training will be more than tripled to over 1500. The images will be curated to allow for the identification of other polyp types within the model, enriched for rarer subtypes. Images will also be provided to allow training in the distinction of low grade from high grade dysplasia. This will be achieved by working in partnership with Dr Maurice Loughrey RVH.
The programme will include 3 cycles of clinical evaluation with efficacy and safety information being collated to support a regulatory submission (CE-IVD) during the programme. Robust commercial modelling and economic analyses will be performed during the program.
There are supporting documents to support this application, including a scientific paper presented at IMVIP 2021 and visual results from one of the AI algorithms developed.
Below are links for some of the talks that Jonathan has taken part in:
• https://www.youtube.com/watch?v=t53qZAckCmE
• https://www.youtube.com/watch?v=yTSs7nP4lEI