Who can bridge the death gap after the first wave of AI systems?
I recently visited a European event with a fellow researcher to investigate the interest in our new project proposal. Our research group has worked for the better part of a decade on the automated detection of esophageal cancer, a disease that’s on the rise in the Western world and select parts of Asia.
The development of first-generation AI techniques to detect the occurrence of cancerous tissue has been quite successful. Over the past years, a broad set of European hospitals collected videos and pictures of their patients. The data were brought together at the Academic Medical Center in Amsterdam, with which my group has collaborated to develop the desired AI technology.
The collected database has grown as large as 5 million cases, still images and pictures from video, many of them annotated by physicians about the state of the disease. In the new European project, we want to further grow this dataset. This will not only improve AI assistance, but it will also harmonize the treatment of the disease and elevate the level of understanding of all participating hospitals to the level of the group’s leading experts.
We assumed that, by raising funding at the European level, we’d have to make the data available for all hospitals according to harmonized rules of participating countries – obviously after authenticating each partner as a reliable relation, to safeguard patient privacy. By working toward one harmonized database for this disease, the efficiency would grow for all hospitals and partners and the innovation would become shared between them.
Unfortunately, after intensive discussions with six countries, we had to conclude that Europe isn’t yet ready for this. The first disenchanting finding was that the healthcare landscape per country is very different. In Turkey, for example, the leading hospitals are rather large and act like individual ‘kingdoms,’ considering themselves large enough to go at it alone – no need to share data. In Finland, the ICT infrastructure has been largely standardized, but the uniformity is based on basic patient data without images and strict regulations have hampered the development of leadership at a world-class level. In the Netherlands, academic hospitals have partitioned the work into islands of expertise per disease.
Hence, it became clear to us that innovations can’t be efficiently organized and centralized by governments. Experiments have to be organized in clusters so that innovative data-driven companies and healthcare institutes will find each other and can significantly benefit from collaboration. After all, companies make money by standardizing products and services while (public) hospitals lack the budget and knowledge to innovate at a large scale. If AI and datasets stay in their current state, technology remains scattered and concentrated innovation at a larger scale won’t happen.
In the US, the next-generation AI techniques have been embraced by the world’s largest software companies. So much money has been invested that these ‘big seven’ will standardize AI and databases by themselves. There’s no doubt that these companies will be leaders in the next generation of AI and data. Thus, the US will transition to the next stage of AI by economy of scale.
In China, the government is leading and planning this innovation. The rights of individual patients are at present sacrificed for the greater volume to establish leadership. For security, this is already taking place and large companies are now starting to work with hundreds of hospitals.
In Europe, meanwhile, we’re still hopelessly unharmonized. We need to become more efficient as soon as possible. Otherwise, we’ll miss out on these grand developments, even if we have world-class solutions available on a small scale. Are we willing to better organize ourselves to bridge the gap toward large-scale, more standardized solutions?
