Pathologists play a key role in AI development for pathology – providing the expertise needed to bridge data and clinical application. To discuss this role and its importance in the development of computational pathology tools, we connected with Diana Montezuma, Pathologist and Head of the R&D Unit at IMP Diagnostics.
From your perspective, what is the most important contribution that diagnosticians bring to AI and algorithm development?
Pathologists bring essential clinical expertise and practical insight to any computational pathology project. Without their involvement, such initiatives risk becoming disconnected from real-world practice and ultimately failing to deliver meaningful clinical value.
A clear example – though not strictly AI – can be seen in early digital pathology viewing systems, some of which were developed without pathologist input and consequently lacked basic tools required for routine diagnostic work. Had pathologists been involved from the outset, these systems would have been far better aligned with clinical needs. The same principle applies to AI development: involving pathologists throughout the process is critical to ensuring that tools are clinically relevant, usable, and fit for purpose.
How can we ensure that computational tools genuinely solve clinical workflow problems rather than generating new ones?
The most effective safeguard is continuous collaboration between developers and pathologists, built on regular two-way communication and iterative cycles of testing, feedback, and refinement throughout the development process.
What are some misconceptions developers may have about diagnostic practice?
Developers entering the computational pathology field are often surprised to learn that pathology is not an exact science and that diagnostic practice involves a degree of variability and subjectivity. This can be counterintuitive for those coming from more technical disciplines.
For this reason, I consistently emphasize in talks and discussions the importance of clear communication between developers and pathologists. Mutual understanding is key to designing tools that genuinely support clinical practice.
Can you provide some examples?
At IMP Diagnostics, every computational pathology project we develop or collaborate on involves pathologists, as well as biomedical scientists. There is little value in developing AI for pathology without clinical input, as diagnostic needs and domain knowledge come directly from the pathology team.
In practice, projects consistently evolve and improve once clinical expertise is integrated – whether developing a tumor–stroma ratio quantification model in colorectal pathology or assessing virtual staining approaches in breast cancer. Ultimately, the strongest outcomes come from multidisciplinary teams working together throughout the project lifecycle.
How important is multidisciplinary collaboration in these settings?
It’s critically important. Without engineers, researchers, pathologists, biomedical scientists, and others working together, projects are inevitably weaker. It is this dynamic, collaborative process that allows ideas to evolve into tools that are more useful and impactful. One of the most rewarding aspects of research is the opportunity to learn from colleagues across different disciplines.
What are the barriers to pathologist involvement?
In Portugal, as in many other countries, a major challenge is the slow adoption of digital pathology across hospitals. Without digitized departments, it becomes difficult to carry out meaningful work in computational pathology. Involving pathology residents is also essential, but inconsistent implementation of digital pathology within public health systems creates a barrier to training and engagement. Sustained investment in digital pathology infrastructure is therefore necessary to enable progress in computational pathology.
What diagnostic domains are most likely to benefit from close pathologist–developer collaboration?
All diagnostic domains benefit from close collaboration between pathologists and developers. That said, the area likely to have the greatest impact in computational pathology is the development of prognostic and predictive biomarkers – tools that add genuine clinical value and do not rely heavily on traditional grading systems.
Support tools for routine tasks also have a role, but pathologists are generally highly skilled and efficient, and pathology assessments are relatively low cost. As a result, many workflow-support solutions struggle to demonstrate clear cost-effectiveness. In contrast, tools that enable quantitative analysis, improve consistency, and extract information beyond what is visible “by eye” are more likely to deliver meaningful clinical impact.
As we move toward increasing automation, which diagnostic tasks do you believe should remain firmly in human hands – and why?
There will always be a need for human oversight in healthcare, largely because accountability ultimately rests with people, not technology. Someone must remain responsible for the final diagnostic result. As noted in a recent Time magazine article by M. Doraiswamy and M. Benioff, “AI is revolutionizing health care. But it can’t replace your doctor.” The authors argue that the greatest value is achieved when AI is combined with the critical thinking, compassion, and real-world judgment that only humans can provide.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
