This Insight is part of your subscription to Signify Premium Insights – Medical Imaging. This content is only available to individuals with an active account for this paid-for service and is the copyright of Signify Research. Content cannot be shared or distributed to non-subscribers or other third parties without express written consent from Signify Research. To view other recent Premium Insights that are part of the service please click here.
Co-written by Dr Sanjay Parekh
Earlier this month, Israeli AI vendor MaxQ AI (MaxQ) announced that it was axing its Accipio range of products for the detection and triage of intracranial haemorrhage (ICH) and completely ceasing the development of image analysis-based AI applications.
The company, which was founded in 2013, will continue to exist, but will instead focus on non-image-based algorithms that look to use vast amounts of medical data to identify anomalies that cause poor clinical outcomes or clinical inefficiencies. However, the pivot has resulted in Accipio sales and marketing personnel being laid off with immediate effect
The Signify View
“We have had a history of losses and we may be unable to generate revenues” warned MaxQ in an investor prospectus from 2018 as it set about an ill-fated attempt to list on the Nasdaq Capital Market. This warning was printed in the chapter of the prospectus entitled ‘Risk Factors’, a chapter which, with the benefit of hindsight, is sadly prophetic. Among those identified risks which proved particularly close to the mark were “failure to articulate the perceived benefits of our solution or failure to persuade potential…customers that such benefits justify the additional cost”; “ Failure to generate broad customer acceptance of or interest in our solutions,”; and the “introduction of competitive offerings by other companies”. These factors and others were instrumental in the failure of MaxQ’s Accipio products, with some aspects more important than others.
Perhaps the most significant of MaxQ’s weaknesses was the Accipio range itself. When the company launched in 2013 as MedyMatch its vision of a product, which was focused on detecting an intracerebral hemorrhage (ICH), was at the forefront of medical imaging AI. In 2022, however, solutions are much more mature. Products from other vendors offering stroke imaging solutions such as RapidAI and Viz.ai address both ICH and large vascular occlusion (LVO), but also add value along the clinical pathway. Instead of focusing solely on detection, these more sophisticated solutions (care coordination platforms as previously described by Signify Research) add other functionality such as triage capability, perfusion quantification, mobile viewer and prehospital workflow applications, and secure care coordination tools. In comparison, other tools from MaxQ never made it to market. There were additional tools in development, but the vendor has been commercially reliant on its Accipio Ix and Ax tools focused only on identification and prioritisation, and slice level annotation and prioritisation respectively. The company had also struggled to obtain US-FDA clearance, a necessity to gaining a foothold in the US, a market dominated by RapidAI and Viz.ai.
Ultimately, for AI solutions to be attractive to providers they must offer them greater clinical value than is offered by the narrow Accipio tools. There are some use cases where narrow AI tools do make sense, such as FFR-CT, but more frequently AI developers need to add additional capability along or across the workflow to make solutions worthy of a provider’s spend. With such competition in the stroke detection market, it was inevitable that those with the weakest value propositions would, sooner or later, falter.
An Appropriate Model?
Another challenging factor contributing to MaxQ’s retreat was its business model, which was highly reliant on channel partnerships.
In some cases, there are advantages of a sales strategy centred around these partnerships. Such setups, for example, can allow vendors to scale very rapidly as they are tapping into an existent customer base. They can also help to establish a young vendor’s reputation, with a partnership from a long-established and well-trusted vendor bestowing credibility upon an unknown developer. However, there is a price to pay for these benefits, with a vendor being dependent on an external sales team. Radiology AI, as a very young market, hasn’t yet become a priority for the vendors charged with selling MaxQ’s software, especially if it risked delaying the sale of a modality scanner or imaging IT software. As such, those vendors’ sales teams would also be unlikely to prioritise the software and promote it as effectively as a direct sales team might.
Another challenge comes in the form of market education. This remains one of the barriers for the medical imaging AI market for AI vendors themselves, let alone a channel partner attempting to convince a potential customer. It is hard to convince providers to allocate budget on any new and untested technology, but this persuasion is made considerably more difficult if a sales team doesn’t have a complete understanding of the product they are promoting. While those vendors selling MaxQ’s products would have an appreciation of the technology, it is unlikely that they would have the same level of nuanced understanding, or the same easy access to additional information as a direct sales team could possess.
Sales Are More Than Transactions
These challenges mean that even under a channel partnership model, an AI developer must still allocate significant resource into the promotion of its products. One example of a vendor that has done this well is Lunit, a vendor who has recently crossed into the ‘$100m club’ of vendors that have secured more than $100m in capital funding. Although it also utilises a channel partnership model, Lunit has also pursued direct sales in its native South Korea, and also invested heavily in clinical validation studies. It has then exploited these studies, to convince sceptical providers of its value. In combination it has also been a steady presence at RSNA and other meetings, and a frequent contributor to expert panels and lecterns at conferences. Even when other partner vendors have sealed transactions, Lunit has been very active in the selling.
For MaxQ this job was made harder still by the limited clinical validation it was able to undertake, which led to the withdrawal of its US-FDA approval for detection. While the product was still approved for use as a prioritisation tool, the lack of FDA approval for its detection capabilities would no doubt have raised doubts in a potential customer’s mind, particularly as other vendors were securing a number of full regulatory approvals, and even in some cases, reimbursement.
MaxQ last secured funding in March 2019 of $30m, at the time a very healthy figure. This however followed the vendor’s aborted attempt to list in 2018, which was set to raise a comparatively small figure of $8m, suggesting an urgent need for cash. This begs the question, if more capital had been raised would MaxQ have been able to overcome the challenges it faced? It would no doubt have helped, but continued investment needs to be earned, and MaxQ, despite its very early entry onto the market, and early de novo FDA approval failed to gain traction. Seth Godin’s Purple Cow marketing theory emphasises the importance of being remarkable (as in the titular bovine) in being noticed. MaxQ AI was remarkable in its earliest days, but as time passed and other more sophisticated solutions were released from other vendors, the Accipio line of products failed to hold interest. MaxQ AI slowly slipped back into the pack.
The Point of Failure
“MaxQ is an aeronautic term that means maximum pressure, which is typically the point where failure occurs”, explained MaxQ AI’s then Chair and CEO, Gene Saragnese in an interview with AiThority in 2019. Sadly, for the Israeli vendor this point of failure has now arrived and, MaxQ AI has become one of the most significant pioneers to falter amidst the consolidatory pressures in the bourgeoning medical imaging AI market. While it is easy for survivors to smugly pore over MaxQ’s mistakes with the benefit of hindsight, many would do well to heed the warnings. There are several vendors that will, in the relatively near future, succumb to similar pressures. One need only look at the competition in some markets to see how challenging things are set to become. In the breast AI market, established leaders are making it increasingly difficult for less established vendors which lack unique products to gain any ground. The chest X-ray AI market, meanwhile has seen some technology leaders with increasingly comprehensive, and increasingly clinically valuable solutions emerge, throwing shade on other, once-promising vendors. Even AI for more advanced imaging, like brain MRI, is becoming increasingly homogenised, with several solutions that lack competitive differentiation appearing at risk of failure.
Consolidation in the radiology AI market is coming. There are simply too many vendors chasing too few dollars for it to be otherwise. Those vendors that will thrive in this consolidation are those that are able to differentiate their products from the competition, add considerable clinical value (beyond feature detection) and solve the pertinent problems that providers face (such as improving workflow efficiencies). Moreover, they must continue to innovate to remain remarkable.
It’s too late for MaxQ AI, but other vendors need to ensure they meet these criteria, lest they become another example left to be dissected.
About Signify Premium Insights
This Insight is part of your subscription to Signify Premium Insights – Medical Imaging. This content is only available to individuals with an active account for this paid-for service and is the copyright of Signify Research. Content cannot be shared or distributed to non-subscribers or other third parties without express written consent from Signify Research. To view other recent Premium Insights that are part of the service please click here