https://www.youtube.com/watch?v=zLk-JVfiljI
April 23, 2026
Detailed Summary: Health Advances Precision Medicine Podcast — Gary Gustafsen with Paul Beresford of PathAI
Gary Gustafsen frames the podcast around a central strategic question: does digital and computational pathology finally reach broad adoption when biopharma has a direct stake in it? He compares the moment to earlier precision medicine inflection points — HER2, PD-L1, NGS, liquid biopsy, MRD — where diagnostics moved faster once pharma had “a horse in the race.” Paul Beresford agrees, describing digital pathology as a “30-year overnight success” and arguing that diagnostic adoption is usually a long road. A new assay can take 10 to 13 years to move from discovery to standard of care, but pharma-backed companion diagnostic programs can compress that timeline. He cites PD-L1 as an example where pharma support narrowed adoption from a decade-plus curve to something closer to four or five years.
A major theme is that digital pathology has long been “on the cusp,” but the economics have been hard. Gustafsen notes that many community labs struggle to make the ROI work unless they are high-volume, highly optimized, and able to capture real workflow savings. Payers, meanwhile, are not eager to reimburse what they may view as “small technology improvements.” That creates the familiar digital pathology bottleneck: scanners, image management, AI tools, storage, and workflow redesign all cost money, but routine digitization alone may not generate preferential reimbursement. Gustafsen suggests that the reimbursement conversation changes when the tool becomes something that cannot be done another way, such as a computational pathology-based companion diagnostic.
Beresford then lays out PathAI’s two-part business model. First is the DDx business, centered on selling AISight, PathAI’s image management system. This side of the business is currently driven by workflow improvement: digitization, case optimization, prioritization, aiding diagnosis, and eventually helping write reports. He says this creates a strong value proposition for large-volume labs and notes that PathAI has publicly announced major adopters, including Quest, Labcorp, and academic centers such as Moffitt. For smaller “mom and pop” pathology operations, however, the economic case is less clear at present.
The second pillar is PathAI’s biopharma partnership and CDx strategy. Beresford’s role is to work with biopharma partners to identify higher-value applications — especially companion diagnostics — that can support differential reimbursement. In this model, workflow tools may get digital pathology into major networks, but CDx and high-value patient-selection tools create the clearer medical and reimbursement argument. The core thesis is that once a computational pathology tool is tied to finding the right patient for the right therapy, it can become a reimbursable, clinically necessary test rather than a back-office productivity aid.
A particularly important point is day-one readiness. Beresford contrasts traditional biomarker rollouts, such as PD-L1, with the possibility of deploying digital algorithms across an already-digitized lab network. Historically, PD-L1 adoption required geography-by-geography and pathologist-by-pathologist training to reach acceptable quality and reproducibility. In Beresford’s vision, if labs already have digital pathology infrastructure, a new algorithm can be rolled out almost like a plug-in, instantly standardizing interpretation across multiple sites. That could compress the time for a new diagnostic to become standard of care from 13 years to five years, or perhaps even one year in some cases.
Gustafsen then pushes on centralization versus decentralization. Many observers assume computational pathology tests will initially live in a handful of large reference labs, just as complex molecular tests often do. Beresford agrees that large centralized labs will play a major role in the short term, and he refers to PathAI’s precision pathology network as a way to plug AI capabilities into high-volume labs in the U.S. and internationally. But he argues that digital pathology may ultimately decentralize more easily than NGS. Unlike NGS, digital pathology does not ask labs to master an entirely new wet-lab technology; pathologists are already doing pathology, and digital AI tools build on that existing clinical practice. The missing ingredients are infrastructure, reproducibility, ROI, and higher-value reimbursed tests.
The podcast then broadens from oncology to the wider opportunity in pathology. Gustafsen notes that oncology gets most of the attention because of IHC, antibody-drug conjugates, and companion diagnostics, but there may be many important use cases outside cancer. Beresford agrees and says PathAI’s business spans both oncology CDx/translational medicine and non-oncology clinical-trial applications. He highlights MASH as a major area, where pathology endpoints are commonly based on labor-intensive consensus scoring by multiple pathologists. PathAI’s AI tools aim to make those reads more robust, reproducible, and quantitative. He says PathAI has shown in trials that algorithmic pathology can turn nearly significant pathology endpoints into statistically significant ones, and that PathAI has pursued FDA qualification through the drug development tool pathway, a process he says took roughly five years.
Beresford also identifies IBD, Crohn’s disease, ulcerative colitis, and celiac disease as additional growth areas where standardized AI-supported biopsy interpretation may be useful. This is a key strategic point: digital pathology is not just an oncology CDx story. It may also become a clinical-trial endpoint standardization business, especially in diseases where histologic scoring is subjective, variable, and operationally cumbersome. In these settings, the immediate customer may be biopharma or CROs, not Medicare or commercial payers.
Gustafsen asks whether PathAI functions like a CRO or competes with CROs. Beresford answers that the relationship varies, but many CROs partner with PathAI because they either lack central lab capabilities or do not have PathAI’s specific AI-enabled pathology capabilities. He says PathAI can serve as a central lab partner for clinical trials, particularly in MASH and immunology/inflammation, supporting programs from Phase 1 through Phase 3. He mentions that PathAI has 12 Phase 3 trials ongoing, making this a substantial line of business rather than a speculative future opportunity.
The closing portion of the podcast is partly conference-oriented, with references to TriCon, AACR, ASCO, and DDW. But even this section reinforces the broader point: PathAI is positioning itself across oncology, liver disease, immunology, inflammation, and precision medicine, with a business development team and senior experts such as Eric Walk available to help partners understand how to use digital pathology. Beresford’s final message is that digital pathology adoption will be driven by partnerships — with pharma, CROs, labs, and technology platforms — rather than by any one algorithm alone.
Strategic Takeaway
The podcast’s deeper message is that digital pathology is trying to solve its adoption problem by becoming indispensable to pharma and precision medicine. Workflow software may get PathAI into large pathology networks, but the more powerful economic engine is likely to be higher-value use cases: CDx, quantitative pathology endpoints, patient selection, clinical-trial enrichment, and standardized biomarker interpretation. In that sense, the podcast strongly supports the Roche/PathAI acquisition logic: Roche is not merely buying an AI slide reader; it is buying a platform that connects pathology workflow, biopharma development, companion diagnostics, and future distributed precision pathology.
