AI in Pathology Must Focus on the User Experience

There are two “universal truths” when it comes to AI in pathology: One, AI will solve more problems than we can imagine today, and two, no single application will address them all.

Evidence supporting each of these truths is becoming abundant. Use cases have evolved from tumor detection to outcomes prediction as data scientists continue to push the limits of AI technology in the quest for precision medicine. Clinical laboratories are also recognizing these truths. They are increasingly adopting software platforms that give them access to the broadest portfolio of applications. In doing so, they can empower their pathologists and advance the standard of care for their patients—both today and tomorrow.

It’s not surprising that a technology-centric focus on developing and deploying a growing number of applications has led the AI-enabled era of pathology thus far. Solutions must first be built and implemented in the laboratory before they can deliver any value. But now that AI is increasingly entering the mainstream, it’s high time to consider how we help pathologists realize its full potential.

User experience must be intuitive

The answer, in many ways, comes down to user experience with AI in pathology. In other words, we must take a pathologist-centric approach that makes it intuitive and easy for pathologists      to use the technology to accomplish their goals. This sentiment probably feels familiar even if your laboratory hasn’t adopted AI, as user experience shapes your perspective on every type of software you use.

Since pathologists experience AI through the software platform that powers their workflows, this is where I want to dive in. The platform should first make sure AI applications run in the appropriate situations. Many applications are not only subspecialty-specific but also intended for use on certain cases. The right platform should also give pathologists an option to run applications manually or automatically. Once a solution has been executed, the platform must seamlessly display results where and when pathologists need them.

You can start to see how this platform functionality delivers a user experience that is favorable to pathologists. It enables the platform to serve as an orchestrator that saves them enough time—by freeing them up from managing an application every single time it needs to run and searching for results—so they can focus on realizing value.

Expect to use many AI applications

Thinking back to universal truths, there is at least one more consideration the platform must address to provide a pathologist-centric experience. Given there’s no one “killer app” for AI in pathology, laboratories will need to adopt many AI solutions. This means that the platform must offer multi-AI workflows that seamlessly handle multiple applications in parallel for pathologists to achieve their full impact.

Imagine a pathologist reading a breast case. They will probably want to run a panel of biomarker quantification algorithms to assess and quantify the presence of HER2, ER/PR, and Ki-67 to help match the patient to the best treatment. They may also want to use other algorithms to understand the likelihood of metastasis or if the disease has spread to the lymph nodes. 

The results from each of these applications are insightful to the pathologist in isolation; however, a platform that enables the pathologist to easily tap into the comprehensive picture these solutions collectively provide throughout all aspects of their routine operations empowers them to make the best possible diagnosis.

With that said, perhaps it’s time to coin a third universal truth. Unlocking the value of AI in pathology requires a pathologist-centric approach.