Case Study - CUF

Breast AI in the diagnostic setting

Compared to breast cancer screening programmes, the use of breast cancer detection algorithms may vary in terms of clinical application, yet leading diagnostic imaging practices see clear clinical value in these solutions.

For the radiology team at CUF, a national network of healthcare providers across Portugal, Transpara Detection has been an integral part of their objective to elevate and enhance the quality of care across their network.

Dr. João Abrantes, a radiology leader at CUF, shared why his organization embraced Transpara in their practice: "we wanted to not only have a good software that would perform well to help the radiologist, but also one which would allow us to be more homogeneous in the way we provide care and maintaining the highest quality across the entire service. In CUF, we have several units and AI offered us a way to ensure a very high and consistent delivery of quality care across each unit."

Adopting Transpara

As a diagnostic radiologist himself, Dr. Abrantes offered a glimpse into how his team is using Transpara Detection in diagnostics, compared to screening mammography in which a radiologist might concurrently read using an AI tool: "we use AI in a way it doesn't bias the first reading of the radiologist. So the first reading is not helped by AI. It's made by the human and then after the first review is done and potentially during the diagnostic ultrasound or other studies that we require, we have Transpara's results available."

"We get the best of both worlds: First you don't have an automation bias of just having clinicians accept whatever the AI tells you, but then you have a positive bias toward suspicious findings found by the AI, he continued. "One of the great strengths of Transpara is the high negative predictive value of Low Risk results. When a finding is classified as Low Risk, we can have a high degree of confidence that the exam is negative. This has been confirmed both in the published scientific literature and through our own internal audit across our population."

"To implement AI, start with a trust roadmap and a valid clinical need. If you don't have a clinical need, you should not try and find a solution for a problem that does not exist."

Dr. Abrantes also offered words of advice for other diagnostic leaders looking to implement Breast AI into their practice: "You have to have a team where you trust and you complement each other in terms of a multi-disciplinary approach, not just the clinical team but every function, including administrative staff and product managers, are the glue that brings everything together. You also need to have both clinical and operational KPIs to measure from the beginning to help you understand how the pilot is going or what your end goals are, whether that is physician satisfaction, clinical outcomes, patient experience, or a combination of all three."

Key ingredients for success

Dr. Abrantes credits CUF's success in AI implementation to a focus on what mattered most across several key areas:

• Collaborative onboarding: “We had a great implementation by our project manager and it's obviously a team effort, not just from the clinical department but also from the implementation from both ScreenPoint and our institution’s side. We managed to get some interviews with all the clinical staff that were interacting with the software through questionnaires and we managed to extract some key KPIs in terms of user adoption on how the radiologists were happy or not with the performance of the algorithms and how easy was it to be used and to report any changes. So, it was very good.”
  
  
• Turnaround time: In the diagnostic setting, “it's very critical that we have the necessary results in a timely fashion, otherwise we will not have the clinical users adopt this as part of their structured workflow. Sometimes we have patients waiting to do an ultrasound and we could not afford to waste any time whatsoever.” Working with ScreenPoint “was a very good process where we managed to lower the time to get the results in a very consistent manner and allowing us to implement into our workflow.”
  
  
• Internal benchmarking: Early on, radiologists were asked to report “their experience when using the AI in clinical practice: how comfortable they were with the concordant and discordant findings? How were they dealing with discordant findings? We asked ourselves how we can structure and be very homogeneous in the way we interact in our clinical practice. We want to ensure that every client, in any of the units we have, experiences the same high quality of care.” CUF also implemented internal assessments of clinical care, including assessing both the positive and negative predictive value of using Transpara in terms of biopsies performed and MRIs completed, with very positive results.

As Dr. Abrantes and the team at CUF continue to define a new standard for care, their sights are set on finding new and innovative ways to help support their physicians and their population. For them, progress is part of daily practice: "now that we have everyone on board and have built that trust, we'll keep on pushing forward."

To learn more about how Transpara Detection can impact your practice, schedule a demo with our team of global experts today.