An Alternative Image of Digital Pathology
The impact of digital pathology on immuno-oncology – and vice versa
In the previous two issues of The Pathologist, we learned about the exciting promise of digital pathology – as well as some of the challenges that exist for use in mainstream clinical practice. I share this cautious excitement; I believe the combination of digital imaging and deep learning will transform pathology. However, many of us who have been working in the trenches of this technology for a decade or more can attest that change has been slow. Indeed, we have predicted a transformation more often than we have reported on its realization. But something is happening today that places digital pathology in a position of relevance more than ever before – and I am more optimistic than ever before!
As many of us know, the field of immuno-oncology (I-O) is exploding dramatically. Cancer immunotherapy’s quick clinical success has led to more than 2,000 I-O agents in some stage of development, according to a recent study from the Cancer Research Institute. This success brought with it the realization that not all patients respond to immunotherapies in the same way, leading us to a new era of I-O treatment that goes beyond a one-size-fits-all approach.
The path to precision medicine for I-O requires tremendous insight into the tumor microenvironment, necessitating efficient and accurate ways to navigate complex information, so that we may see the complete biological story. Digital pathology enables us to interrogate the complex interplay between immune cells, tumor cells, and surrounding stromal components in situ. Artificial intelligence advancements and quantitative image analysis-based multiplexing give cancer researchers a composite, detailed view of tumor tissue, whereas deep learning generates the quantifiable data and specificity needed to precisely differentiate cell types in the tumor microenvironment and find predictive insights into treatment response. Whether applied to standard hematoxylin and eosin or immuno-stained samples, these rapidly developing technologies are crucial for I-O research, helping to guide our approach to rational combinations, patient selection, and clinical trial design.
Just as digital pathology has undoubtedly transformed the I-O research landscape by giving researchers unprecedented insight into cancer biology, I also think the field of I-O will radically change our view of the utility of digital pathology. The perspective that digital pathology contributes to enhanced workflows and more efficient and accurate diagnosis – although true – is limited in the scope of clinical impact. We need to expand the concept of digital pathology toward a means of selecting patients for appropriate I-O drugs that are only now emerging from drug company pipelines. The technology could follow an adoption paradigm similar to next generation sequencing, which is rapidly emerging as the next platform for clinically relevant biomarkers and companion diagnostics. Within the context of quantitation of complex interactions in the tumor microenvironment, digital pathology can – and should easily – follow a similar path to clinical utility.
As digital innovators, it is our shared responsibility to ensure a seamless transition to this inevitable future by training the next generation of pathologists, developing tools and protocols, and – most importantly – generating data that demonstrates the tremendous value digital pathology brings to cancer treatment for all involved… especially the patient. Armed with new perspectives from digital pathology, a more personalized approach to I-O is within sight.
The opinions expressed here are those of Mike Montalto in his personal capacity and not those of Bristol-Myers Squibb.
Chief Scientific Officer at PathAI and was formerly the Vice President of Translational Research at Bristol Myers Squibb.
