Princess Alexandra Hospital NHS Trust: AI in radiology with RUBEE AI

AI in radiology

Radiology has been at the forefront of digital innovation when it comes to leveraging advanced software applications and tools to improve workflows and diagnostic confidence. With the advent of Artificial Intelligence in Medical Imaging, and new imaging techniques, radiologists and other diagnosticians have better access to deeper anatomical and clinical details. These new innovations are beginning to highlight the need for integrated imaging and data informatics interaction to support fast action on clinically relevant findings that require urgent attention. Physicians and diagnosticians are now more than ever in need of integrated evidence-based intelligence that will help redefine radiology in an era of precision health and personalized care delivery.

AGFA HealthCare’s innovations in the field of medical imaging are well recognized, and with the RUBEE AI framework, the following factors were considered when it comes to enabling AI workflows in medical imaging:

• Embedding AI seamlessly into the user experience
• Leveraging AI meta data and turning it into actionable intelligence
• Workflow orchestration, automation, and smart hanging protocols
• Precision reporting with AI data analytics
• Academic and clinical collaboration for evidence-based peer learning

Princess Alexandra Hospital NHS Trust’s journey of transformation

The Princess Alexandra Hospital NHS Trust (PAHT) in the UK has continued to lead the path towards digital transformation in its Radiology Department since the hospital embarked upon a consolidated imaging platform strategy. This has helped the hospital to transition its imaging department from traditional PACS to the newer Enterprise Imaging platform from AGFA HealthCare. With an Enterprise Imaging strategy, Princess Alexandra was able to establish a core framework for diagnostic image consolidation, multispecialty collaboration and a modular strategy that would help the hospital roll out future innovations when they are introduced, on a secure and modular platform.

When AGFA HealthCare launched the RUBEE AI framework, the Trust’s leadership began assessing their options on the path towards AI, and began the internal work at the hospital on the business case for AI.

Where to begin with AI: governance

Artificial Intelligence in Medical Imaging has several applications, ranging from Deep Learning, Machine Learning and Image Analytics to Natural Language Processing and more. The innovation opportunity that healthcare providers face today is that there are more than 150 AI start-ups and developers focused on these applications and clinical use cases. The question then is, how do you embed these AI innovations into a seamless and clinically relevant experience for your diagnosticians and physicians?

The Trust decided to evaluate the significance of embedded AI in its diagnostic imaging workflows. AGFA’s RUBEE AI framework brought a more workflow-oriented focus compared to the AI marketplace or App Store concept. The RUBEE AI framework allows consolidation of standards-based, best of breed AI applications into specialty-based clinical packages, which helps bring relevant diagnostic intelligence and decision support tools to the radiologists. By providing diagnosticians with carefully curated AI decision support tools, embedded in their Enterprise Imaging solution, AGFA aims to support radiologists and diagnosticians to maximize the value of their own skills and expertise.

Thought leaders agree that to succeed in this era of innovation, the industry, startups, academia, and hospitals will need to work collaboratively. One vendor or an AI start-up cannot provide an innovative solution on their own; it must be a collaborative framework of engagement between the stakeholders.

By opting for AGFA’s RUBEE AI framework, the full proof of concept across the implementation phases helped mitigate the clinical risk associated with a new workflow as the Chest Specialist Radiologists are fully involved during this stage and able to provide feedback throughout.

The selection criteria: which AI solution do I need?

It is becoming clear that AI will not replace radiologists or physicians, but in fact enhance their workflow even further by empowering them in their ecosystem and helping them make collaborative and intelligent decisions. AI or machine learning, if not embedded into clinical workflows in the logical sense, will not offer much benefit and will ultimately create a silo of its own.

When AGFA HealthCare launched the RUBEE AI framework, it included tools and applications such as lung nodule related algorithms, a clinical package for Breast AI with applications for 2D Mammography and Digital Breast Tomosynthesis, and other applications for Chest X-ray screening. The idea was to establish an ecosystem of best of breed AI companies that had scientific publications and evidence relating to their AI developments.

With the COVID-19 pandemic and the emphasis on chest screening, it made sense for the Trust to consider AGFA’s CT AI Package, which included RUBEE and the ClearRead CT algorithm from Riverain Technologies.

What did the Radiology Department look for in workflow-embedded AI?

• Regulatory clearance: whether the AI solution was regulatory cleared by FDA, CE and, for example, Health Canada.
• Decision transparency: the radiologists were interested in advanced features such as Vessel Suppress, which would help enhance nodule detection.
• Clinical validation: whether the AI solution is validated for all nodule types, for example solid, part-solid and ground glass.
• Priors and comparisons: one differentiating feature was the depth of integration of the CT AI algorithm (ClearRead CT) with RUBEE, and how seamlessly prior exams and comparisons were integrated into the workflow to enable prior processing of exams.
• Incidentals: with chest CTs being done for various clinical manifestations, it was important to the hospital that AGFA’s CT AI package would help embed AI-detected incidentals, for both contrast and non-contrast CTs.
• Scientific evidence: the Radiology Department was very keen on seeking the available evidence that AGFA and Riverain Technologies had published, including peer-reviewed articles on performance, efficiency, and accuracy.

The financial investment is based on the benefits provided, such as reports that are more accurate and reduced reporting times, contributing to improved delivery of patient care.

Establishing the business case for AI

For the Trust’s Radiology Department, Artificial Intelligence was quickly becoming a focus of Cancer, Cardiology and Clinical Support (CCCS), with the consultant radiologists and radiographers working hard to provide the best possible patient care. Every year, thousands of scans, CT thorax included, are produced at the Trust, and to provide the best possible service and to comply with national guidelines for nodule analysis, a robust Lung AI was being considered.

The hospital wanted to focus on the following when it came to the AI business case:

• What is the strategic need or opportunity, and why now?
• What options were considered and against what criteria?
• Which is the preferred option and why?
• Will it improve patient care and outcomes?
• Is the business case achievable?

Additional financial savings were assessed in the following ways:

• Reduction in missed cancers
• Streamlining of the surveillance workflow
• 29% reduction in missed nodules
• 26% reduction in reading time
• Reduced number of referrals for outsourced reading

Due to the efficiency in workflow provided by RUBEE AI, the business case assessed whether time would be saved generating reports, and hence money saved on high-level staff resource. The consensus of the business case assessment team was that the increase in capacity would have benefits such as increased reputation, improved capacity to report more scans and early follow-up of patients.

The business case was further aided by AGFA’s favorable licensing and implementation model for the CT AI software, which was based on an unlimited user license model and was considered better than a pay-per-use model for many reasons. Another primary reason was that AGFA HealthCare’s deployment model for AI works on-premise, in real time in the background, assessing all thoracic CT scans and helping improve incidental findings and automated comparisons. This eliminated the need for cloud-hosted AI and related integration requirements, as well as information governance concerns with sending images offsite to a third party.

Project planning and implementation

AGFA HealthCare’s RUBEE AI framework builds on its successful Enterprise Imaging strategy. The initial set-up and rollout of the CT AI Package (which includes ClearRead CT) was very fast. Besides getting the AI module up and running, the PAH system administration team ensured that an analysis of modality workflows was done properly, to determine whether any tweaking was required for the study or series descriptors. The team created a test environment before rolling the solution out to the full production system.

The clinical use: what do the radiologists say?

Since going live with the solution, radiologists are appreciating the value of embedded AI from the following perspectives:

Easy delineation of incidental lung nodules (less laborious)

Small indeterminate lung nodules are very often picked up incidentally in screening and staging CTs. Detecting and measuring them can be quite a laborious and time-consuming process. The AI algorithm detects, characterizes, and measures these nodules, making it:

• Efficient (less time-consuming)
• Accurate
• Reproducible in measurement (taking away inter-operator variability)
• Easier in future follow-up scans

Automated detection and characterization

The small ground-glass nodule can easily be missed by the naked eye whilst scrolling through a chest CT. The AI tool detects these nodules, characterizes, and measures them. These can then be acted on, or followed up, thereby potentially detecting early cancers and improving outcomes.

Vessel Suppress helping pick up early cancerous nodules

The small perivascular intrapulmonary nodule can be difficult to detect and can often be mistaken for an end-on vessel. The AI tool detects these nodules more readily and accurately by “suppressing” the adjoining vessels, thereby picking up potential early cancerous nodules or metastases.

Reporting is quicker: approximately 30% improvement

Another category of scans where the AI tool is proving useful is CT Pulmonary Angiograms (CTPAs). Several of these scans are performed every day in the inpatient and emergency setting. Once a pulmonary embolus is confirmed or excluded by the radiologist, the AI helps in the quick assessment of the lungs for any additional pathology, thereby saving reporting time.

RUBEE embedded AI workflow

The AI-processed results are embedded within the Enterprise Imaging PACS workstations for radiologists in real time. The AI findings, visualizations and measurements are available immediately, and hanging protocols are configured to enable automated comparison between current and prior results. The AI algorithm and findings are seamlessly integrated into the user experience and workflows.

The CT AI package, which includes AGFA’s RUBEE workflow framework and Riverain’s ClearRead CT algorithms, is seamlessly embedded into our diagnostic desktops and provides benefits ranging from workflow optimization to enhanced visualizations and automated comparisons.

The road ahead

Implementing innovative workflows like AI is just the beginning of a journey towards transforming how radiology will provide value in the entire care continuum. Machine learning algorithms are as good as the data they are trained on and can only improve further if there is a good feedback mechanism established between the clinicians and developers.

The collaborative team effort of everyone involved with this project at The Princess Alexandra Hospital NHS Trust, AGFA HealthCare and Riverain Technologies brought together a powerful experience that not only saw a speedy and on-time project delivery and go-live, but also helped establish a forum for ongoing discussion and feedback to improve AI-enabled workflows, a process that derives strength from clinical experience and further deepens the AI evidence.

The AGFA HealthCare Enterprise Imaging platform has helped us achieve the institutional infrastructure for implementing AI as soon as the algorithms are validated and introduced.

About Riverain Technologies

Riverain Technologies™ is a medical software innovator that develops solutions to aid radiologists in the early detection of disease. With Riverain’s ClearRead X-ray Suite and ClearRead CT, radiologists are able to optimize the use of existing equipment for enhanced image interpretation, and better utilize their diagnostic expertise in the identification of diseases such as lung cancer.

Author: Dr. Anjum Ahmed (MBBS, MBA, MIS, ITIL), Global Chief Medical Officer and Director of Innovation at AGFA HealthCare.