Written by
Share article:
Co-authored with Matthew Watson.
Cranfield, UK, 23 April 2024, Investment into AI-based software across medical diagnostics and clinical care has been notably boom-or-bust in the last decade. Radiology has long been the “poster child” for funding and penetration of healthcare markets, with over $5.6B raised to date resulting in well-known successes such as Heartflow, viz.ai, annalise.ai and Ultromics. However, outside of radiology, venture capital for development of AI tools regulated as a medical device (Software as a Medical Device (SaMD) has been far sparser, amounting to only $1.8B to date.
Non-radiology AI solutions in clinical segments such as Cardiology, Neurology, and Ophthalmology are now starting to make headway, accounting for 23% of the FDA’s AI/ML Software-as-a-Medical-Device list from non-radiology segments (October 2023).
Based on our unique new study publishing soon (AI in Diagnostic and Clinical Use – May 24), in this insight we’ll explore the similarities and differences between the radiology and non-radiology AI markets, leaning on discussions with prominent AI start-ups, healthcare technology vendors and investors.
Our free to download investment analysis of non-Imaging AI can be requested here:
Apples to Oranges
Given the billions poured into medical imaging (radiology-led) AI in the last decade, it’s easy for start-up vendors and investors to be bullish over the potential of AI in adjacent diagnostic and clinical applications. However, there is more nuance to the apparent success of medical imaging AI.
Firstly, the use of machine-learning in medical imaging is not new, having been used in some applications for well over a decade, and radiology has been a frontrunner in digitisation and new technology adoption since the advent of “digital health”. Yet, despite the “fertile ground” for new technology adoption, annual spending on medical imaging AI amounted to less than 15% of overall funding. This lag of funding conversion to sales outlines a fundamental truth of medical technology innovation: it is slow, expensive, and heavily regulated.
Secondly, the route to market can differ vastly between diagnostic disciplines. In our new analysis, we focused specifically on AI-software sold standalone, as opposed to bundled with device hardware. This approach enables a “like-for-like” comparison, especially as medical-device start-up funding is typically far larger versus software-focused companies, given the upfront engineering and prototyping costs. In medical imaging, most post-image analysis software is “decoupled” from large imaging hardware, in part thanks to harmonised standards in imaging, thereby reducing the influence of hardware on sales channels and supporting software-only channels in most markets. In contrast, other diagnostic and clinical sectors commonly bundle software with devices, making “software-only” market entry more challenging. This is one reason for the relatively small funding that has emerged for non-imaging (about one third raised for the comparative period), indicating there is a long way to go. Looking deeper into the data, three segments most notably take the weight of funding too: cardiology (~50%), ophthalmology (25%) and neurology (<10%).
Cardiology leads the way, in thanks to a large addressable market and limitations around human interpretation of ECG data, creating a vibrant environment for AI-based software to support diagnosis of atrial fibrillation, and, increasingly, complex heart conditions such as Cardiomyopathy, Low Ejection Fraction, and Myocardial Infarction. Many of these companies will also have been buoyed by the progress of imaging-based peers in the sector, with unicorn Heartflow leading a strong contingent of well-funded start-ups in the cardiology AI sector in landing big investment, reimbursement, and growing scale of use. Recent acquisitions will have also spurred focus, such as Philips Healthcare acquisition of DiACardio, an Israeli-based cardiac analysis tool.
Follow the Leader?
Despite this optimism, following such a path will not be straightforward, let alone fast. Heartflow began work in 2007, while also benefiting with a timely explosion of VC investment availability during the pandemic period of 2020 to 2022, providing the firepower required to build a strong retrospective and prospective clinical and health economic evidence-base, thereby increasing the opportunity for significant reimbursement from payers. Later starters will not be afforded such a cushion of buoyancy in investment, with a notable drop-off in 2023 and change in focus towards less risky investments with a clearer route to market and more proof of a path to break-even.
Reimbursement, often viewed as the “gold standard” for new technology adoption, also does not guarantee a straightforward path to success. Notably, imaging AI vendors that have gained reimbursement have quickly been joined by equivalent peers, creating price competition, while increasing volumes of use place more pressure on payers to downward revise reimbursement in the future – triage-based (CADt) stroke detection is an obvious example. Indeed, the level of reimbursement available to some AI-vendors in the non-imaging (radiology) segment are also relatively low – often ranging in the low-to-mid tens dollars, as opposed to the mid-to-high hundreds of dollars range in imaging. This automatically limits the addressable market calculation of investors in these segments.
High Bar, Low Speed
Even reaching reimbursement is fraught with challenges for vendors in the non-imaging AI sector. A changing regulatory landscape, including the steep rise in requirements for data submission, has meant that gaining approval has been taking longer, increasing the overall cost of market entry. Polarity in regulatory processes between Europe and the US has also grown, not to mention significant differences in how healthcare providers are purchasing and integrating AI into care delivery. Even in imaging there are complexities in go-to-market strategy, with no clear winner emerging between direct, partnering device, partnering software, or orchestration/marketplace approaches yet proving an obvious route to success. For example, well-funded stroke and cardiac imaging AI vendor, Viz.ai ($292M), recently announced a “U-turn” and pulled out of the European market, instead doubling-down efforts on the US. Less established AI a vendors should therefore take note – international scale can create a headache in terms of capital deployment for limited returns, even for more established market entrants.
Product development is also no walk in the park. Data availability (especially structured, annotated data required for effective AI training) is not always easily accessible or comes with significant costs to source. Outside of imaging, the immature adoption of standards and less-structured approach to IT in diagnosis and care delivery means data is typically less centralised. De-centralisation and the reliance still on device-coupled data also means integration of tools into existing diagnostic and clinical applications is rarely simple, requiring a broad range of customised APIs or higher-upfront costs for providers. Navigating these nuances, all the while keeping abreast of the changing AI procurement frameworks and expectations in terms of clinical evidence in different markets presents operational uncertainty. Furthermore, as providers are still shaping AI-adoption strategies and mapping AI into care pathways, an apparent opportunity in one market based on a small-scale trial may quickly dissipate if the results proffer limited benefits or other context influence the focus of spending. Above all, tracking of this sector in the last decade has proven that rarely does adoption of new technology happen quickly or without significant risk.
A Winning Formula?
It is perhaps still too early to predict the winners and losers of the non-imaging diagnostic and clinical AI market given its nascency. Though, lessons learned from the successes of a few vendors in the imaging sector do point to some common factors. These can be simplified into the following:
- Specificity of focus: while there have been a few examples of more comprehensive offerings gaining traction, most success in terms of funding, reimbursement and market entry have been from start-ups with a singular application/specific application focus e.g. improving detection of specific condition type; triage/faster decision in care pathway interventions.
- Evidence of influence on care pathway: while difficult to compare between segments, strong retrospective and prospective clinical evidence has been critical; category leaders have heavily invested in robust clinical evidence that shows either significant improvements in patient outcomes, and/or cost savings for patient management. Ongoing follow-up evidence (beyond post market surveillance) in real-world settings is also a substantive factor in ensuring relevance and defensive market protection against competition.
- Primary geographic strategy: leading vendors in each category of AI deployment have focused specifically on a core geography in terms of market entry, customer education, and sales enablement. Integration and sales channel strategy (see below) is also heavily influenced by geographic market focus, establishing a strong customer base and market traction before slowly expanding to new markets.
- Direct > Partner > Marketplace: while not exclusively a “golden rule”, most successful AI vendors to date have focused on a direct sales strategy in core focus markets, tied to market education and clinical evidence generation. In selected cases, partnership with large incumbent non-AI healthcare technology vendors has enabled faster market access and scale, though often feedback has shown these partnerships still leave AI start-ups doing most sales and customer support work. The raft of marketplace and orchestration platforms have also delivered mixed results for founders, providing scale quickly and access to pilots, but limited sales return in competitive segments and compromised integration for providers looking for tighter integration. The most successful vendors have blended these channels, typically using direct and partner channels in key focus markets, while leveraging marketplaces for newer application segments or entry into new geographic markets.
Evolve or Exit: A New Market Shift
Recent macro-economic shifts have today created a more difficult outlook for AI startups, with cost of debt considerably higher and healthcare providers running out of extraordinary pandemic funds and reverting to more capital-strained procurement. For AI start-ups in the regulated diagnostic and clinical settings, the road ahead will be fraught with challenges – consolidation in many segments is expected, especially as the record capital from 2019-2022 runs dry. Regulatory pathways and the barriers to entry from payers and providers are unlikely to materially change any time soon, leaving some considering future options.
While the specifics of market strategy in each segment are too nuanced for this broad analysis, many firms are already underway in re-shaping or refining their strategies to ensure sound footing into the mid-term. Some are expanding capability to include less-regulated operational AI tools, providing a more turn-key approach to specific care-pathways, others are tapping into pre-clinical funding sources, supporting development of new imaging biomarkers or the accuracy of clinical trial data. Device-coupling also remains a viable approach, establishing co-development agreements with medical device giants to help de-risk the length go-to-market phase.
As the latest data shows, market development for non-imaging AI companies will require careful navigation and while there are lessons to be taken from imaging AI companies that have “made it”, the realities of today’s healthcare technology market suggest that unless funding and available reimbursement improves, the viability of many software-only start-ups in this sector will increasingly be called into question.
Related Research
The AI in Diagnostic and Clinical Use 2024 report, due to be published in May 2024. The report will explore how the adoption of diagnostic and clinical (SaMD) AI in non-radiological settings compares to the success of radiology AI, specifically focused on which clinical segments offer the most opportunity for market growth. Detailed competitive analysis per segment will also provide a clear landscape of this emerging segment, enabling strategy and product leaders across healthcare technology assess competition and support strategic decisions on product development, partnership and potential M&A strategy. The full report also explores critical trends around sales channel strategy, device-partnerships and the market roles of leading start-ups, healthcare technology segment leaders and big technology vendors in shaping the digital diagnostic ecosystem.
Note: AI solutions for medical imaging analysis (e.g. radiology) are not included in this report. Please see our extensive AI in Medical Imaging 2024 service for detailed market coverage.
About The Author
Matthew joined Signify Research in 2022 as part of the Medical Imaging team. He holds an MA Economics degree, graduating from Heriot-Watt University in 2022 and previously completed an internship at Unilever.
About Signify Research
Signify Research provides healthtech market intelligence powered by data that you can trust. We blend insights collected from in-depth interviews with technology vendors and healthcare professionals with sales data reported to us by leading vendors to provide a complete and balanced view of the market trends. Our coverage areas are Medical Imaging, Clinical Care, Digital Health, Diagnostic and Lifesciences and Healthcare IT.
Clients worldwide rely on direct access to our expert Analysts for their opinions on the latest market trends and developments. Our market analysis reports and subscriptions provide data-driven insights which business leaders use to guide strategic decisions. We also offer custom research services for clients who need information that can’t be obtained from our off-the-shelf research products or who require market intelligence tailored to their specific needs.
More Information
To find out more:
E: enquiries@signifyresearch.net
T: +44 (0) 1234 986111
www.signifyresearch.net
