Share article:
Google has made its latest moves in medical imaging, by, for the first time, licensing its imaging algorithms to third party vendors. The company synonymous with searching has licensed its mammography AI research model to iCAD, and partnered with Aidence, which has licensed its nodule malignancy assessment algorithm.
The licensing deals enable iCAD and Aidence to supplement and complement their own services using Google-developed algorithms, while also taking advantage of Google’s expertise.
The Signify View
Google has long held designs on healthcare. From early salvos as Google Health and its plans to improve patient access. To health care records, development of algorithms to diagnose diabetic retinopathy, and more recent moves with the acquisition of FitBit and the access to 29m customers it brings, the Mountain View based company has always felt it has a lot to offer the sector.
Prior to these announcements, the vendor’s most successful play in medical imaging has been the provision of cloud, offering, like Amazon AWS, Microsoft Azure and others, the infrastructure for providers to shift their data to the cloud. However, recently Google has been making a more concerted effort in medical imaging. In October it launched its Google Medical Imaging Suite, a product that consolidates several existing capabilities into a single, targeted package, but one which still left Google as primarily an infrastructure provider, facilitating the deployment of others’ technology rather than offering customers its technology directly.
The new agreements with Aidence and iCAD don’t mean Google is directly supplying customers, but it moves the vendor a significant step closer, enabling end users to directly take advantage of Google’s algorithms, even if through a third party.
Patient Payoff
Such an agreement is no-doubt attractive for iCAD and Aidence. As discussed in Insights passim, the most successful medical imaging AI vendors are those which can offer increasingly sophisticated solutions that offer capability further across or along care pathways. Google’s involvement, primarily through its algorithms, but also through the expertise the agreements bring, will help Aidence and iCAD keep up with this direction.
Aidence is already competent in the detection and quantification of lung nodules, but Google’s algorithm will bestow the vendor with malignancy risk scoring capability. This is a sensible addition, enabling the vendor to offer a more clinically valuable solution that is more competitive in the market. After all, most lung nodules are not malignant, a fact which means lung screening programmes are at risk of being overladen with benign cases, and unable to efficiently tend to the minority which do have malignant nodules and need additional care.
What’s more, by offering malignancy risk scores using Google’s algorithm, Aidence’s tools can begin to be used in a population health context, having more of an impact than the direct diagnosis of individuals. Further, by providing accurate risk scoring of lung nodules, the use of malignancy risk scoring may also spare some patients, with nodules with a very low risk of malignancy, the need of undergoing invasive diagnostic procedures. As medical imaging AI solutions increasingly become candidates for reimbursement, these capabilities have been present in some competitor solutions (e.g., Optellum) which have actually been successful in securing it. The addition of Google’s malignancy scoring could, for Aidence, therefore be more significant than simply improving its product, especially as the vendor looks to targeting the US market.
Matched to Markets
For iCAD too, the licensing of one of Google’s algorithms represents a sensible step. While it does not offer a completely new capability, it will enhance the vendor’s existing 2D mammography offering and give it greater potential in other, non-US, markets. iCAD’s focus in recent years has been on its 3D mammography AI solution, utilised most in iCAD’s domestic US market. However, as iCAD seeks to expand internationally and begin to derive revenues from other markets, it will need to accommodate the prevalent type of imaging in those markets, including in Europe, where 2D mammography is more common. While iCAD does have a 2D mammography AI algorithm, Google’s solution arguably has greater levels of validation, and a combination of iCAD’s and Google’s models, along with Google’s expertise, should allow iCAD to best capitalise on these broader markets.
As such, there is evidently ample motivation for AI specialist vendors to seek partnerships with the likes of Google, but what is the motivation for the tech giant to license its technology. While the deals are commercial, based on the typical revenues being generated by even the most successful medical imaging AI vendors, money is unlikely to be the primary motivation for Google’s collaborative mentality.
Instead, there are several factors at play. One uncharitable view of the situation is that medical imaging AI has simply not yet performed for Google as expected, and this is a relatively easy way to secure some revenue and maintain some mindshare in the space using primarily the assets which have already been developed. Google has demonstrated in both healthcare and in the broader business it is willing to be ruthless in mothballing or even shuttering business units that have struggled to establish themselves.
A Multi-Pronged Attack
Medical imaging, and healthcare in general are very difficult markets to succeed in, far more encumbered by regulation and convention than many of the markets that outside entrants and Big Tech firms might be more used to competing in. However, it seems unlikely Google would leave such a large opportunity.
Instead, the licensing agreements are more likely to be part of a broader strategy. One option is Google being able to‚ ‘white label’ its algorithms, enabling several AI developers and imaging IT vendors to enhance their offerings, enabling them to eschew long and expensive internal developments. This would help Google establish a presence in the market, and through the service and support included in these types of agreements influence providers and imaging IT vendors to turn to the firm’s Medical Imaging Suite, and ultimately its cloud infrastructure, an avenue that remains the most immediate revenue opportunity for the company in healthcare.
This is particularly true given the vendors Google has chosen to license to. Screening programmes, by definition, create vast amounts of data on people afflicted by conditions, as well as healthy people, which needs to be stored somewhere. What’s more, Aidence, which was acquired by RadNet last year, may also facilitate that outpatient provider’s wider uptake of Google’s cloud services. Further licensing agreements could help new markets be targeted, with agreements with successful AI vendors in other territories, such as Lunit potentially opening doors to South Korea, for example.
Time for Testing
In addition, while working to help its cloud business grow, Google can also use licensing agreements as a testing ground for its own algorithms, allowing the commercial effectiveness of its algorithms to be assessed, while utilising other vendors sales and distribution networks. Longer-term Google may seek to acquire some AI developers that have proven particularly successful, then fully integrating them into its Medical Imaging Suite and other areas that it may choose to operate in, such as EHRs, for example.
As such, Google’s licensing agreements with Aidence and iCAD is not a dramatic step with instant impact. While it will help bolster the AI vendors it will not be transformative for Google’s aspirations in medical imaging. What it does offer, however, is another revenue stream, another sales avenue for the vendor’s cloud services and the ability for Google to remain in the market, ready to capitalise should the right opportunity arise.
This could happen quickly. Afterall, if Google decided to take advantage of the AI market’s coming consolidation, pick up several AI vendors to supplement its own algorithms, then use it vast resources along with its medical imaging suite, it would fast become a formidable player in the nascent market.
Alternatively, the vendor may, as it has before, abruptly end its involvement in medical imaging AI. It seems unlikely at this stage, but with most of Google’s revenues still coming from advertising, it’s clear that many paths the vendor has trod have led nowhere.
Ultimately, which route is taken will depend on two factors, its ability to navigate the idiosyncrasies of the medical imaging market, and its willingness to do so.