The initial outlay for digital pathology technology might seem overwhelming, but the data it could generate might just prove to be a valuable asset – one that could help recoup some of the cost.
Proscia recently published a report that acknowledges some of these data demands and their potential for revenue generation.
Here, Derek Welch, President and Chief Medical Officer at PathGroup, shares some thought-provoking examples.
What are some of the aims and objectives of digital diagnostic networks?
In the US healthcare market, there is currently no incremental reimbursement for adopting digital pathology. That leaves many laboratories asking the same question: how do we justify the substantial costs of going digital?
If we look outside the clinical lab, pharmaceutical companies are highly interested in digital pathology because it provides a new type of asset: pixelated images of disease in patients. In oncology, in particular, these digital images can unlock insights beyond what the human eye can detect and accelerate drug development by providing faster, cheaper access to real-world data.
Owning such digital assets at scale makes clinical service providers attractive collaborators for industry. Currently, only about 7 to 10 percent of glass slides in the US are digitized, meaning the overall pool of digital pathology data is still small. This scarcity further increases the value of laboratories that are already digital and positioned as part of a broader network.
Being part of this digital ecosystem places a lab in the unique position of being able to contribute to and benefit from pharma collaborations – both AI development studies, and other research opportunities. It transforms pathology from an analogue process into a language of pixels – zeros and ones – that can be analyzed, shared, and monetized in entirely new ways.
How might labs be able to work with biopharmaceutical partners in the real-world data space?
This is an evolving space, but it is clear that digitally-enabled labs have the ability to create real-world data cohorts that are valuable to pharmaceutical companies. “Cohort” is the key word here, referring to a set of patient cases with high-quality, de-identified digital images accompanied by metadata and, ideally, additional pathology information such as molecular sequencing results or targeted PCR analyses.
Pharma companies need to identify patient populations for clinical trials quickly and cost-effectively. By incorporating digital pathology images into the patient record, laboratories create a new parameter that can be searched and filtered using computational pathology tools.
For example, a lab that sees 15,000 oncology patients a year could rapidly filter cases to identify those with non-small cell lung cancer who are positive – or negative – for a specific molecular defect. This capability allows pharma partners to accelerate trial enrollment and, ultimately, the development of new therapies.
What opportunities might be available for AI development partnerships?
Many companies are developing clinically useful algorithms aimed at reducing human error in diagnostics. They are based on the idea that the algorithm combined with the human being creates a better outcome. Digital pathology labs provide the ideal test-beds and training grounds for such applications.
On the clinical research side, oncology is the key area for partnership opportunities. AI is being developed to uncover novel insights about tumors that can speed drug discovery and reduce costs. Traditionally, drug development is a slow and expensive process. The idea of applying computational algorithms “at the push of a button” to large oncology datasets is highly attractive, offering the potential to identify targets, biomarkers, or patient subsets much earlier and more efficiently.
Companies are emerging that specialize in combining sequencing data from solid tumors with image-based data, building sophisticated algorithms that extract new, actionable information from these datasets. Their work illustrates how AI straddles clinical care and pharmaceutical R&D – driving advances in both areas.
What are some of the ways in which labs could provide precision medicine services to pharmaceutical companies?
For digital labs operating at scale, the key question is: what do we do with these images once they’ve been signed out?
Rather than purging files, labs can build de-identified oncology image repositories enriched with additional data – molecular results, PCR assays, and other pathology metadata.
Pharma’s needs vary. Some partners may want full sets of images for algorithm development, while others may only require representative samples for training AI filters to detect specific mutations. Some research requires data from specific patient demographics to avoid bias in their cohort data.
At PATHGroup, we have already engaged in projects along these lines, both independently and in collaboration with Proscia and other research entities. Some of these focus on supporting pharmaceutical development, while others are aimed at building foundation models in digital pathology – large-scale AI tools that can be adapted across applications.
By organizing and retaining their “digital exhaust,” as Proscia CEO David West has called it, labs can transform routine diagnostic output into a valuable research asset.
Could you share any case studies of digital pathology partnerships from your own labs?
Lung cancer research seems to be the area with the most demand for these kinds of collaborations. We’ve already participated in a couple of lung cancer studies in collaboration with Proscia. Following strict data compliance rules, we were able to supply data packages for cohorts of patients with non-small cell lung cancer. It was used in both AI and drug development.
We’ve also worked with some innovators of large-scale foundation models that needed a variety of digital images to assist with building the AI capabilities. If you think of a stairway, we’re the first step – digitizing patient data – the foundation model developers are in the middle, and the end users sit at the top.
How might membership of a diagnostic network help labs to reshape their economics?
There are opportunities for offsetting at least some of the costs of digital infrastructures today. Successful collaborations do require some effort – it’s not a case of just sharing images. Sophisticated technical solutions, such as AI-powered clinical research platforms, might be required. But participating in the real-world data space does create opportunities to tangibly improve patient care and also create a revenue stream.
Whilst it might not be possible to recoup the full cost of digital investment in the current climate, economics suggest that the costs will fall as the market establishes itself. The first CD players cost around $1,500, and 10 years later they were available at a tenth the size for less than $50.
In the US, medical insurance companies have maintained for decades that they will reimburse according to quality of patient outcomes. In today’s labs, where it becomes possible to prove the quality benefit of pathologists working in conjunction with AI, I'm hopeful that eventually there will be revenue streams from the use of AI.
Saying that, the physician side of me really hopes that healthcare trends towards more and more entities becoming digital. Because it's a good thing for patients.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
