Could an app-store-style model make it easier for pathology departments to test, validate, and deploy the AI tools they need?
Pathology AI is advancing rapidly, but adding a promising algorithm into the diagnostic workflow can be anything but fast. Adding a new vendor may mean navigating contracts, data governance, infrastructure, deployment, integration, and local validation – potentially turning each new application into a substantial procurement project.
The Sectra Amplifier Services aims to remove some of that friction by giving existing customers a single route to activate, evaluate, and deploy AI applications within their established Sectra environment.
We spoke with Elin Kindberg, Sectra’s Global Product Manager for Digital Pathology, to explore how the Amplifier Service model works and whether lowering the barriers to adoption could help more AI applications make the leap into routine practice.
What are the key considerations for labs that want to add new AI applications to an existing workflow with a new vendor?
The most important thing is integration. If you are adding a new AI application, it needs to fit naturally into the existing workflow rather than introducing additional complexity. Ideally, it should run automatically and become part of what the laboratory already does.
Where are the main bottlenecks when adding AI applications using a traditional procurement model?
The process can take a considerable amount of time. One reason is that laboratories typically need to sign a separate agreement with each AI application vendor. In many cases, that vendor will also need to handle protected health information, which introduces additional legal and data-governance considerations.
Labs then need to agree on a deployment model that works for the organization. Taken together, contracting, legal review, data handling, and deployment decisions can make the process very time-consuming, with contracting often being one of the biggest bottlenecks.
How does the Sectra Amplifier Services aim to remove some of those barriers for organizations that already work with Sectra?
With the Sectra Amplifier Services, customers sign up for AI as a Service rather than contracting separately for each individual application. It is not specific to pathology; the same model can be used across any areas where customers are using Sectra systems, including radiology and cardiology.
Once labs have access to the Sectra Amplifier Services, they can activate applications from the AI partners that we have already contracted. Because Sectra has commercial agreements in place with individual vendors, the contracting burden is removed from the lab.
We also take responsibility for deployment, which happens within the Sectra infrastructure. That means the third-party AI vendor does not need access to the customer’s sensitive medical data, removing the need for a separate agreement with that vendor covering the handling of those data.
In that model, adopting an AI tool becomes more of an operational decision than a major procurement project. What could that mean for pathology departments?
It means the department makes one decision to adopt Sectra Amplifier Services, and from there it becomes much easier to test and add new AI applications.
Once you have access to the service, you can choose the applications that make sense for your organization – whether that is one application in radiology, one in pathology, or multiple pathology applications – without starting the entire procurement process again each time.
For pathology departments, that significantly lowers the threshold for exploring new AI tools. Both procurement and deployment can be shortened considerably, making it much easier to evaluate applications and introduce those that add value to the workflow.
How does this model change the way labs can evaluate applications using their own cases and workflows?
Evaluation is part of the initial validation process. We can suggest a protocol based on how pathology laboratories typically validate an application, or the laboratory can use its own validation protocol.
Typically, the application is evaluated in the customer’s test environment using their own data. The laboratory can then compare the AI output with the original manual reads, which provide the ground truth for the evaluation.
If there is a discrepancy, a group of pathologists can review the case to determine what happened – whether the AI result was correct or whether the original manual interpretation was correct.
The important point is that the evaluation takes place on the laboratory’s own cases. That gives the customer an opportunity to see how the application performs with their data and within their particular setting before moving into clinical use. The process not only enhances user trust in the application but, crucially, improves patient safety.
If a lab activates an application and then discovers that it does not meet its needs, what are the options?
We have defined stage gates throughout the process, first for validating the application itself and then for validating the workflow. If, at either stage, the application is not working as the customer expects, we can pause the project or stop it altogether.
Our aim is also to provide alternatives. For example, we may have different AI applications available for the same tissue type or use case. If one application does not perform as intended or does not suit the laboratory’s needs, the customer can explore another option.
That flexibility is possible because the customer’s contract is with Sectra rather than with each individual AI vendor. It means they are not locked into one application if, during validation, they find that it is not the right fit.
How might removing procurement and infrastructure barriers change the pace at which AI applications are adopted in routine pathology practice?
I think this could be revolutionary for the market because it makes it much easier for customers to start using AI. We have already seen this in radiology, where our experience is that AI applications can be activated much more easily through Sectra Amplifier Services, and we expect to see the same development in pathology.
There are so many promising pathology AI applications emerging, but their value can only be realized if they make their way into routine diagnostic workflows. By reducing some of the barriers around procurement, deployment, and integration, we can make it easier for laboratories to take those opportunities and bring AI into the workflow where it belongs.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
