Artificial intelligence (AI) is likely to transform healthcare delivery of the future, increasing quality, personalized medicine, and potentially cost efficiency. However, successful deployment, orchestration, and safety monitoring of this technology will likely require rethinking of hospital health IT and leveraging of third-party platforms for AI to achieve its full potential.
Traditional or legacy IT approaches for AI tool adoption are not cost-effective and unlikely to be able to effectively support the anticipated digital transformation and adoption of multiple AI tools for leading-edge and competitive healthcare organizations in the future
There are increasingly sophisticated third-party AI deployment and management platforms that enable seamless, cost-effective, and timely AI deployment and facilitate optimal AI tool selection; these are also likely to form the foundation for future more advanced AI tool adoption and safety monitoring
Healthcare leaders should become familiar and consider adoption of AI deployment platforms as part of their digital transformation roadmap and future AI infrastructure planning
Artificial Intelligence (AI) is likely to transform the healthcare practice of the future. There are steadily increasing FDA cleared or classified medical AI applications, and currently over half of these tools are based on diagnostic medical imaging or radiological images. These are furthermore likely to multiply in the next 5 years. They are also likely to increase in complexity and sophistication, fueled by the rapidly evolving AI landscape and technological advances, most recently exemplified by large language and other emerging multimodal foundation models. While there are many interesting and headline grabbing applications of AI tools in medicine in experimental preclinical settings, the practical adoption and workflow barriers that must be overcome for successful adoption of such tools in clinical practice are frequently overlooked. Notwithstanding the potential of this technology, early experience with AI tools has revealed significant challenges for timely and seamless AI adoption in the clinical. In an increasingly digital and AI-supported healthcare operation, one of the most basic foundational challenges for healthcare organizations will be seamless and timely deployment and management of an increasing number of complex AI tools. In this article, the challenges with traditional or legacy information technology (IT) approaches and the potential utility of third-party AI deployment and orchestration platforms will be discussed for empowering and supporting seamless deployment, orchestration, and safe adoption of this technology in the healthcare environment.
It is fair to state that traditionally, hospital IT systems are frequently under-resourced. If the healthcare of the future is going to be truly revolutionized and transformed using digital technologies and AI, the status quo will no longer be sustainable or acceptable. Under the current legacy health IT paradigm, AI software such as those used in radiology are procured and technically deployed one at a time. This is repeated over and over every time a new AI tool is acquired. The process is painstakingly slow, laborious, and costly to the organization, both from the perspective of resources utilized for procurement and implementation of each AI tool as well as the opportunity cost of hindering trials to determine the optimal tool for the organization and delaying the adoption of potentially transformative technology. Even in the traditional setting, this legacy approach has not been optimal and its negative impact on the organization’s progress and competitive edge is likely underestimated. However, in a potential future world where hospitals may have to adopt and manage tens if not hundreds of AI tools, the legacy approach will simply not be sustainable. In addition to resources required for deployment and management of these tools, at some point, the complexity may exceed what can be reasonably expected of a hospital IT department. Afterall, hospitals are not IT companies, and the point may come when they have neither the resources nor the full complement of necessary technical expertise in order to effectively manage and increasingly complex AI landscape. One likely solution that is increasingly being considered for these fundamental challenges is the strategic use of third-party AI deployment and orchestration platforms that help mitigate these problems.
Over the past few years, an increasing number of platforms have become available that can facilitate the deployment and orchestration of AI tools. These platforms are becoming increasingly sophisticated and can represent a solution and a fundamental framework facilitating seamless, timely, and cost-effective adoption of multiple AI applications in the healthcare setting. It is neither the intention nor within the scope of this article to discuss individual platforms, their advantages, and pitfalls. Rather, the purpose of this article is to provide an overview and bring these platforms and their importance to the attention of key stakeholders within the healthcare organization’s leadership as part of their necessary infrastructure planning for the future digital transformation and adoption of AI in their respective organizations.
There are currently several different vendors providing advanced platforms that can be used to deploy, manage and orchestrate AI tools. Currently, the primary focus of many of these platforms is the deployment of diagnostic imaging AI applications but these have the potential, and their functionalities can be expanded in the future, in order to integrate other modalities and data types. These platforms come with different configurations, but all of the major ones have the core functionality of being able to host and effortlessly deploy multiple AI algorithms from the standpoint of the healthcare institution. Depending on the platform, they can also have other key functionalities that range from image routing to sophisticated abilities to orchestrate and integrate various applications that form the foundation for effective deployment and use of increasingly sophisticated AI tools, including multi-modal applications. For the purposes of this article, the focus will be on the basic but fundamental advantages of these platforms in facilitating cost effective and timely adoption of multiple AI applications and empowering the organization to try and select the optimal AI application that serves its unique needs, followed by a brief overview of other current and potential future fundamental advantages of these platforms.
Timely deployment of AI tools remains a major challenge under the current legacy healthcare IT paradigm. It is not unusual for the process of procuring and making an AI application available to clinician providers to take 3 to 6 months, if not more. From an implementation cost perspective, ballpark estimates are that the resources required to get each AI tool through this process, including security review and integration of the tool, can cost an organization in the range of 15,000 to 20,000$ or more per application. There are in addition indirect and opportunity costs for such delays. One negative downstream effect is the delay in deploying cutting edge tools that can improve healthcare processes and patient care. There are also other negative effects, including friction within the organization that arises from such delays and an overall contribution to staff and provider dissatisfaction. The costs and challenges are compounded and multiply with an increasing number of AI tools, making the legacy approach unsustainable. All of this can have a substantially negative impact an organization’s operations and competitive edge. When using a platform, on the other hand, the “heavy lifting” is done once. Once a platform is reviewed and securely integrated and adopted by an organization, deploying individual AI software can be as seamless as tuning on an App, making a tool immediately available and avoiding the repetitive, laborious, and costly approach of legacy one at a time, standalone AI tool adoption. This will enable the competitive organization to lead in technology adoption.
AI tool performance can vary at different hospital sites. Furthermore, for a given tool performing a similar task, there are unique aspects both from a diagnostic performance and operational/workflow standpoint that can make a given tool more or less desirable depending on the specific hospital/organization’s unique needs. One important direct benefit of enabling effortless AI application deployment through a platform is that it empowers the organization and providers to try different applications, providing them with first-hand knowledge and exposure to functionality of a given tool within that organization. Using legacy approaches, trials (even if provided at no cost by the vendor) are resource intensive and costly to the organization, essentially requiring the same degree resources that is required for full adoption of a standalone software discussed earlier. This disincentivizes trials, especially when multiple applications are being considered, and ultimately may lead to procurement of a less-than-optimal tool. By making application trials effortless and less resource intensive, AI deployment platforms enable seamless evaluation of multiple applications, ultimately enabling the organization to select the tool that is best suited to the needs of the organization and its providers.
Although a detailed discussion is beyond the scope of this article, the ideal platform can have many additional advantages that support technology adoption and transformation of a healthcare organization. Platforms support interoperability and have the potential to make AI applications more effective using sophisticated image/date routing. They facilitate management and integration of multiple applications and can potentially support multimodal applications that include both image analysis and language or reporting tasks. Lastly, with increasingly complex AI applications, including development of biomarker tools that perform tasks such as predicting a molecular phenotype or treatment response, there will be a requirement for more robust quality monitoring tools. One can envisage platforms playing a fundamental role in this essential requirement of future advanced AI models. These are few among a range of potential possible benefits of using platforms for safe and effective AI deployment and management.
Digital technologies and AI are likely to play an essential role in the healthcare of the future. Although much attention has been paid to specific applications and clinical or healthcare process-related tasks they perform, the essential infrastructure investments and planning that form the foundation and are required for successful adoption of these technologies in clinical practice does not always get the necessary attention. To ensure that a healthcare organization is competitive and leads in technology adoption, it is imperative that its leaders make the necessary infrastructure planning and investments that will facilitate successful future digital transformation and AI tool adoption. In that regard, legacy IT approaches for AI tool deployment may not be viable or satisfactory for leading clinics and hospital systems. AI deployment and orchestration platforms can provide the necessary infrastructure to make seamless, effective and timely AI tool adoption possible in clinical practice. Healthcare organization leaders would benefit from familiarizing themselves and considering adoption of such platforms as part of their digital transformation and AI roadmap.
Dr. Reza Forghani is a radiologist and researcher with medical Artificial Intelligence (AI) expertise. He is Professor of Radiology & AI and Vice Chair of AI at the Department of Radiology, University of Florida College of Medicine where he is also the director of Radiomics & Augmented Intelligence Laboratory (RAIL).
