Whilst growth in the overall digital pathology AI market is set to scale quickly, cannibalization and consolidation are expected to create headwinds for individual AI vendors. How do investors know where their money is safest?
The recent abundance of digital pathology (DP) media coverage can make it hard to remember that not too long ago, this was a market perpetually ‘lagging’ versus other digital healthcare sectors. And whilst growth has now firmly shifted to its ‘log’ phase, there are several facts to bear in mind for investors and stakeholders eyeing this space.
There’s plenty of greenfield opportunity, but not everywhere is primed for harvest.
To utilize DP AI a lab must have access to – and be utilizing – a DP scanner. This an obvious statement, but one to take careful note of as the number of labs today with access to such scanners is still very low.
Penetration today is also very heterogeneous geographically. In North America and particularly the US, vendors have cited digital pathology ‘adoption’ (in terms of scanner placements in labs) for clinical use sits below 10%. Compare this with Western Europe, where some markets like the UK are expected to begin saturating over the next five years, and there is a clear difference between the two. In Asia Pacific, DP has also not taken off as quickly as some expected, with patchy and slow adoption so far in markets like Japan and South Korea, whilst China is again expected to begin saturating over the mid-term.
However, even such country-level generalizations hold distinct nuances. Some Catons in Switzerland for example utilise digital pathology workflows for primary diagnosis in 100% of cases, whereas just next door, utilization of the technology is near 0%.
The take-home message is clear: the market is clearly bifurcating between a small number of early-stage adopters scaling their DP usage, and DP newcomers which are growing in number but also are completely new to the practice. As such, whilst the market potential for AI holds great promise AI vendors and investors both should be careful when forecasting potential returns.Figure 1: Typical process for digital pathology implementations, which can take anywhere from 6 months to two years to fully roll out.
As Figure 1 shows above, implementation timelines for new adopters can be a lengthy process, with the most capital at the beginning allocated to scanner purchasing at the expense of other products.
Often, the market has seen a prevalence of ‘bundled’ purchasing by new adopters, which elect not to pursue a third-party image management system to keep costs down. After taking a few years to fully evaluate the DP implementation and any associated return on investment, we have seen a significant number of labs seek to replace their scanner-associated software in favor of a more ‘interoperable’ best-of-breed approach.
Because of this, the adoption of AI often falls to the back of the line in terms of initial priorities, making the total ‘readily’ addressable market for digital pathology AI still relatively low. Previous years of ‘lag’ growth followed by exponential adoption post-COVID-19 and an overall slow implementation process will therefore inevitably create a knock-on ‘lag’ for AI.
A lot of VC funding has been provided, however, return on investment remains comparatively low.
Figure 2 showcases the VC funding landscape for digital pathology from 2014-2022. As is shown, 2021 represented a significant deviation in terms of an abundance of funds poured into the market, a trend quickly reversed the following year.Figure 2: Public VC Funding for Digital Pathology Vendors 2014-2022 in USDM.
Our most recent digital pathology core update revealed that the worldwide market for image analysis (not just machine learning artificial intelligence) in digital pathology was $104.0M USD in 2022, with around 50% of that attributed to machine learning and more sophisticated algorithms.
Comparing total market size to such funding levels seems unsustainable. Indeed, as many new and existing entrants begin to scale pressure is being felt universally. The market is already beginning to see consolidation, with Crosscope and KeenEye being acquired quite recently, as our five-year forecast progresses, we expect this trend to continue.
Judging against adjacent markets is also a useful gauge of progress. In medical imaging, a near fully mature digital market that has been leveraging digital technology and image analysis for nearly two decades, spending on AI-based image analysis tools annually is around half a billion dollars, fuelled by over five billion dollars of VC investment across over 200+ start-ups. Here, large data sets for training exist, regulatory pathways are streamlined and some reimbursement already exists for the use of AI tools in clinical practice. Yet even for some of the largest and best-funded AI vendors (with a handful already with “unicorn” $1B+ valuations, the realities of revenue recognition are much smaller and consolidation is already gathering pace. Figure 3: Digital Pathology Ecosystem by Vendor Type. Note, this list is not exhaustive, merely intended to showcase examples of digital pathology vendors providing AI software.
Whilst numbers are expected to be thinned in digital pathology also, the outlook is not equally bleak for all vendors.
Some companies have sought alternative ways to sure up their financial positions, such as Aiforia’s choice to raise an IPO, or PathAI which acquired a laboratory business; however, a number of vendors have also yet to commercially launch their software and really begin competing in the market. Table 1: Topmost funded DP vendors 2014-2022.
Table 1 above lists some of the most well-funded vendors, all of which have some sort of AI strategy.
I have my own opinions on which types of vendors are likely to persist/step out soon – if you’re interested subscribe to receive a later insight. For now, I’d also note that with the abundance of vendors competing in this space with similar portfolios, it can be difficult for individual customers to distinguish between different vendors, with some education required on what makes a ‘good’ algorithm and business strategy between these.
Whilst this short piece may seem foreboding, I should again point out that growth in the market is expected to persist. Strong indications in both research and clinical domains have occurred recently, (for clinical think Aiforia’s PD-L1 software being adopted by an NHS trust, Paige’s FDA approval and Ibex Medical’s progress within large labs networks), and with the headwinds expected to be experienced by pathology labs in the short term, digital pathology AI is clearly not just a case of if, but when.
And whilst clinical adoption of DP may be lagging behind, many vendors are choosing to renew focus on research-based applications to sustain themselves. The research segment for DP has long been far more advanced than the clinical, and life sciences, CROs and academics are renewing investment and doubling down on DP workflows. Vendors like PathAI have done very well pursuing such business, and the research market is now slowly beginning to transition from one-off projects to a more platform-based approach to AI. In the short term, we expect software and in particular SaaS to start to play a bigger role in new agreements in this space as organizations shift from service-based sales toward products and platforms. In the future, therefore, this could be one of the few areas investors will remain willing to dig into their pockets, as timelines to ROI are comparatively shorter.
About Imogen Fitt
Imogen Fitt is a Senior Market Analyst at Signify Research, a research advisory company providing health tech marketing intelligence powered by data. Imogen joined Signify in 2018 as part of the Healthcare IT team. She holds a 1st class Biomedical Sciences degree from the University of Warwick where her studies included molecular biology and pharmacology. Since joining the team Imogen has studied the medical imaging software and hardware markets and is now expanding Signify Research’s Diagnostics and Lifesciences coverage.