Artificial Intelligence Holds the Promise to Transform Cancer Diagnosis

Dr. Tom Loftus, Razors Edge Ventures

On Wednesday, June 19th, Proscia, a pioneering digital pathology company from Philadelphia, announced the release of a disruptive new product that foreshadows a coming transformation in the way cancer is diagnosed. For well over a century, tumor pathology has been assessed by a trained medical professional looking at glass slides of a tissue biopsy under a microscope, a process that ties diagnostic efforts to the local demands of working with the physical slide and forces pathologists to laboriously search through all sections of the slides to avoid missing critical indicators of disease. Proscia’s new DermAI product, an artificial intelligence (AI)-based capability that will assist pathologists in their diagnosis of skin cancers, is part of a new wave of digitization in pathology that in the near term stands to improve collaboration and efficiency in the diagnostic process and in the longer term offers hope of significantly increasing diagnostic accuracy, resulting in better patient outcomes.

DermAI leverages deep learning to pre-screen and classify skin biopsies. It is the first in a series of AI applications from Proscia for different tissue types that will integrate into a pathology practice’s end to end workflow through Proscia’s Concentriq platform and help reduce costly errors and improve laboratory quality and efficiency. The practice of pathology has been under growing pressure, both due to efforts to control escalating healthcare costs and to a steadily declining number of medical professionals entering the fieldi even as the number of biopsies has grown. These trends have prompted organizations to transfer some elements of pathology practice to non-physicians who perform their roles at a lower cost.ii In an alternative to this approach, Proscia’s AI works collaboratively with pathologists, delivering on the objectives of cost control and optimal use of limited resources while keeping key decisions in the hands of the experts. The AI makes the pathologist’s job easier while ensuring that every case gets the validation of a digital second opinion. In addition, by digitizing the workflow, the Concentriq platform links biopsy images to other relevant patient data in a fully HIPAA compliant manner and allows them to be viewed remotely so that a specialist in another city or around the world can provide their input.

Artificial intelligence has captured the imaginations of the public and professionals alike with its promise of transformational new capabilities and reports of remarkable new feats by AI algorithms almost daily. Unfortunately, while AI has made remarkable strides in recent years, the gulf between promise and practical application remain significant. This is especially true of applications such as healthcare where the consequences of errors are much more significant than its use in shopping or web search. A common problem with AI models is their generalizability, that is, their ability to continue to deliver accurate results when applied to a different set of data than the one they were trained on. To date that has been a particular problem for analysis of pathology images due to wide variations in different laboratories’ processes for preparing and staining biopsy slides and variations in scanner hardware. There have been multiple academic studies that demonstrated a high degree of accuracy with a controlled set of pathology images, but those models performed poorly when applied to other data sources. This challenge was central to the development of DermAI which was validated across four leading dermatopathology labs, using 20,000 slides, demonstrating its ability to work across diverse real-world clinical data and readying the product for adoption and setting a major milestone for the field.

The proliferation of machine learning tools and services from companies such as Google, Amazon and Microsoft has allowed developers to create entertaining and labor-saving AI tools; however, we feel AI holds its greatest promise in the delivery of capabilities that materially impact peoples’ lives for the better. Razor’s Edge Ventures invested in Proscia’s Series A funding round in 2018 because we saw their potential to deliver on that promise by lowering health care costs while reducing the impact to patients’ lives from misdiagnosis. Razor’s Edge invests in companies developing transformational capabilities that are of value to both Government and commercial markets. In Proscia we saw a company with obvious commercial benefits to the multi-billion-dollar pathology industry, but also one creating value for Government customers. This is not limited to traditional medically-focused agencies such as Health and Human Services. Proscia’s technology stands to deliver significant benefit to organizations such as the Department of Defense as it works to maintain the health of our armed forces and their families while on duty in far flung places around the world and continue to support our service personnel afterwards through the Veterans’ Administration.

i “Pathologist Workforce in the US (2013),” Archives of Pathology

ii “How Many Pathologists Does the United States Need?” Lundberg (2019), jamanetworkopen.2019.4308,