AI Marketplace or Curated AI Hub: What’s the best solution for healthcare systems implementing AI?
AI Marketplace
An artificial intelligence (AI) app store or marketplace is a browsable library of software applications that use AI to perform some data operation, often a point solution to a narrowly defined task. In diagnostic radiology, AI algorithms are typically visual classifiers that have the ability to identify and/or characterize a particular finding on a particular imaging modality—for example, the presence of a pulmonary nodule on Chest CT, the presence of hemorrhage on Head CT, or the presence of a fracture on a skeletal radiograph.
The app store’s go-to-market strategy relies upon motivated end users, often radiologists or non-physician administrators, to search for AI applications that meet their specific wants or needs, download, and install the applications, and utilize them to meaningfully enhance clinical or operational workflows. However, such a deployment model can lead to the fragmented and disorganized implementation of AI applications in a given setting with unpredictable impact on clinical practice and redundant use of business, legal, IT, and financial resources. A marathon of perpetual last miles.
Furthermore, the app store approach poses a quality control risk to the purchasing end user. Most, if not all, AI vendors are able to furnish one or more white papers of variable academic rigor or real-world evidentiary quality, but each nevertheless advertises their product as being the superior product. The inability to prospectively distinguish winner from loser algorithms leads to purchaser decision paralysis. Vendors make no guarantee that an algorithm will perform as expected on a given patient population or on images acquired on various OEM scanners. App stores typically do not offer no-cost pre-deployment validation trials or post-deployment surveillance with free refunds for model drift/decay, leading to buyer hesitancy and potential buyer’s remorse.
Curated AI Hubs
The AI hub platform model solves these problems for prospective health system customers. By curating vetted and validated AI algorithms that provide end-to-end solutions organized around disease states, customers can trust that they are purchasing the best-in-class and best-fit AI products tailored to their specific practice needs and patient population.
For example, a Women’s Health AI Hub can provide a suite of synergistic AI algorithms. These curated AI solutions provide breast density information, multimodality breast lesion detection/characterization (2D/3D mammography, ultrasound, MRI), uterine and ovarian lesion detection/characterization, and opportunistic screening of bone density by XR/CT for women at risk of osteoporosis and fragility fractures.
Value of Curated AI Hubs to the Health System
Whereas the app store approach is radiology department centric, AI hubs are implemented at the enterprise (integrated delivery network) level and have the potential to impact population health, spanning traditionally siloed medical specialties and closing patient care gaps at multiple touch points.
When it comes time for implementation, the curated AI Hub approach offers a fast-on/fast-off mechanism to iteratively determine the optimal set of algorithms for a health system’s needs in a vendor agnostic and net cost-efficient manner. Curated AI Hubs provide AI solutions organized around disease states as well as the opportunity to test the algorithms on the health system’s patient population. This curation and organization are much more valuable to health systems than a menu of unrelated apps in a marketplace.
Elie Balesh, MD
Dr. Balesh is a double board-certified, dual fellowship-trained diagnostic and interventional radiologist with postdoctoral expertise in digital health and medical device innovation. In his role as Medical Director for Ferrum Health, he leads the development of artificial intelligence clinical use cases, practice implementation, and real-world research.