Problem
Leeds Teaching Hospitals NHS Trust (LTHT) performs over 135,000 chest X-rays annually - scans that are vital for time-critical decisions regarding lung cancer, infections, and emergency line placements. LTHT sought to adopt an AI-driven ‘second pair of eyes’ to accelerate these diagnoses, but they faced a significant hurdle: building a robust clinical safety case in an area where AI was still relatively new ground.
Navigating the transition from theory to live care required a data-driven approach. Clinicians needed to define and generate specific evidence to ensure the AI's findings aligned with their own expert judgment, all while operating under the pressure of limited IT resources which did not have capacity to support the project which required integrations with multiple systems for at least 12 months due to existing pressures.
Solution
Through a regional project backed by the NHS AI Diagnostic Fund, LTHT partnered with Newton’s Tree and Harrison.ai to deploy a solution capable of identifying up to 85 in-scope findings. Newton’s Tree acted as the technical and strategic orchestrator, ensuring the adoption was both safe and efficient.
- The sandbox environment: Using Newton’s Tree sandbox, Consultant and junior radiologists conducted a controlled evaluation using a combination of prospective and retrospective data. This ‘shadow mode’ deployment allowed the team to test the AI’s performance in a safe environment before it touched live patient pathways.
- Integration with multiple clinical systems with minimal IT friction: Newton’s Tree led the implementation, managing the integrations with both the RIS and PACS clinical systems and managing the subsequent workflow setup. This minimised the burden on the Trust’s internal IT department, allowing for the tool to not only generate AI findings but also use this information to help prioritise worklists via HL7 messaging to the RIS system.
- Clinical review at speed: By using the Newton’s Tree ‘AI Review’ tool, clinicians could test AI outputs on a case-by-case basis. Because the tool was intuitive, frontline staff required minimal training to participate in the evaluation.
Results
The structured approach delivered by Newton’s Tree allowed LTHT to move quickly without compromising on clinical safety.
- Rapid clinical evaluation: The entire clinical evaluation, from data curation to final analysis of AI outputs, was completed within just 1 week. This included clinician onboarding and training on the Newton’s Tree “AI Review” tool.
- Accelerated deployment: With Newton’s Tree providing integration and implementation support, LTHT committed to a firm deployment date. This demonstrates how Trusts can ,move from concept to live deployment while maintaining robust clinical safety and governance, unlocking faster benefits from EOE/TCC by translating data and infrastructure into real operational impact (e.g. prioritised workflows, faster diagnostics) without lengthy evaluation cycles.
- Robust governance and reduced IT burden: The rapid evaluation provided the empirical evidence needed to finalise LTHT’s Clinical Safety Case and SOPs, ensuring the AI was ready for safe, long-term use and significantly improving clinical confidence during initial rollout. It also reduced pressure on local IT capacity by delivering complex, multi-system integrations (e.g. RIS/PACS) with minimal internal resources, enabling programme-wide rollout.
- Significant return on investment and scalable model: Notable savings across testing, infrastructure, IT labour, compliance labour, and project management. This establishes a standardised, repeatable approach to evaluating, integrating, and safely deploying AI across multiple Trusts and use cases, with financial ROI compounding at each deployment.
