Executive Summary
A leading European Oncology Center - German Medical Institute (GMI) - faced a significant operational hurdle: a growing backlog of chest X-rays driven by limited clinical bandwidth and availability. The objective was clear - deploy an AI-based imaging tool to improve redundancy for out of hours and assure patient turnaround times. By leveraging Newton’s Tree to ‘test before they invest,’ GMI bypassed the traditional, sluggish pilot cycle and moved directly to a high-impact rollout, reducing deployment timelines by 70% and achieving 136% ROI.
The Challenge
The GMI team encountered a common dilemma. With numerous vendors claiming high sensitivity and specificity on paper, GMI struggled to identify which AI application would perform best on their unique patient population.
Without a rigorous evaluation framework, they risked a ‘trial and error’ approach, which as well as being costly and inefficient, could see them deploying a solution with high false-positive rates that would increase, rather than alleviate, the clinical workload.
Limited IT capacity meant GMI could not support multiple, sequential, and fragmented deployments, and estimated running pilots could take over up to 9 months.
The Newton’s Tree Approach
Newton’s Tree’s role was to provide the end-to-end infrastructure and expertise needed to select, test, and deploy the optimal solution with maximum efficiency.
- Defining success: We helped the leadership team move beyond the ambiguous goal of ‘AI assistance’ to a specific objective: improve clinical redundancy and ensure patient cases, especially the most acute, were managed efficiently even when there was low clinician availability. This clarity allowed the Newton’s Tree team to develop a bespoke evaluation protocol. While sensitivity and specificity were measured, the protocol prioritised a low false-positive rate, a critical metric for ensuring clinicians were not burdened by unnecessary re-checks.
- Parallel testing on local data: Bypassing the need to do separate pilots, we enabled GMI to test vendors simultaneously using their own curated dataset. Newton’s Tree integrated the vendors and automated the entire workflow. By generating side-by-side performance outputs, we provided GMI with a clear, data-driven comparison, significantly accelerating the decision-making process.
- Neutral technical orchestration: By managing the technical integration and performance reporting, we acted as a neutral partner. This ‘orchestration’ allowed the clinical team to focus entirely on the results, removing the distractions of implementation hurdles or the risk of vendors fine-tuning products during the evaluation phase.
Impact
Clinical impact
- Results showed the best performing vendor had a 60% higher diagnostic accuracy compared to the lowest ranked vendor
- Selection was evidence-based using local patient data rather than relying on published performance times, so when GMI moved into live clinical care they knew what performance to expect
GMI avoided a potential clinical misstep by selecting a vendor based on real-world performance on their own patient data, rather than marketing specifications. The evaluation revealed that actual performance - diagnostic accuracy - varied significantly between vendors despite nearly identical public-facing claims. The preferred vendor had a diagnostic accuracy 60% greater than the vendor in last place. Evaluating these solutions side-by-side provided the empirical clarity needed to make a definitive, data-driven decision.
Operational Impact
- ~70% reduction in deployment timelines, accelerated from 9 months to 12 weeks improving time to value and reducing costs
Because the technical workflow was established during the evaluation phase, the transition to live clinical care required minimal additional effort from the hospital’s IT team. Furthermore, because the infrastructure remained consistent throughout the process, Information Governance was significantly simplified, removing a major hurdle and accelerating the path to deployment. This resulted in GMI benefiting from the value case more quickly.
Financial Impact
- 136% ROI based on one AI application, ROI will grow as GMI deploys more AI through the platform.
- ~€40,860 saved across testing, infrastructure, IT, governance, contracting and project management activities
With local validation of the tool’s clinical value, GMI eliminated the need for sequential, time-consuming pilots. This allowed them to move straight to a long-term contract with full confidence, leveraging proven performance to negotiate a more favorable price point from day one.
Newton's Tree helped us save nearly €41,000 across testing, infrastructure, IT, governance, contracting and project management. For our Institute, this wasn't about selecting any chest X-ray AI, it was about proving which one would actually perform on our patients. Their evidence-based approach took us from a 9-month pilot estimate to 12-weeks of deployment, and 136% ROI from one AI application alone.
Dr. Konstantinos Zamboglou, Chief Medical Officer at German Medical Institute
The project established an evidence-led approach to AI selection, based on performance from their own data. GMI were able to reduce the risk, cost, and complexity of future solution selection and moved away from one-off pilots towards a model for scaling AI safely and more efficiently.
