AI is the new standard of care. There’s no turning back. Hospitals are growing much warmer to the idea of AI and its integration. But despite the shift toward AI adoption, technical limitations remain before this new paradigm in radiology can hit full stride and see its golden era. Currently, AI for medical imaging is commonly utilized for triaging cases and reducing turnaround times. Yet there is plenty more potential in clinical workflows that extend beyond just the PACS screens of a radiologist. And it starts with understanding demand and where the existing technical limitations are: the AI operating system.
The efficiency and cost savings of AI in diagnostic imaging, clinical decision support, and precision medicine is driving spending on AI as healthcare professionals start to trust its capabilities. A survey of Chief Experience Officers at leading U.S. healthcare organizations found that 90% of hospitals have an AI strategy, up from 53% the previous year, with solution deployment growing to 34% in early 2021. And trust in these technologies is growing, too. In 2020, 62% of healthcare leaders surveyed in the U.S. said they trusted AI technology to perform diagnostic screening, up from 40% in 2018. As information from clinical trials, health records, diagnostic imaging, and population and claims data continues to reveal its effectiveness at improving clinical care delivery, spending on AI is expected to increase at an annual rate of 48% between now and 2023. While the impact of AI in healthcare is predominantly measured in a clinical context, its proven downstream effects on financial and operational improvements lend AI the credibility needed to become standard of care in healthcare networks, particularly in radiology departments.
As the dust settles on emergency provisioning for the COVID-19 pandemic, hospitals continue to face volume burdens that affect numerous departments. McKinsey surveyed leaders at 100 private sector hospitals across the U.S. to determine how COVID-19 is continuing to impact hospital volume. They found that emergency department and inpatient volumes have returned to 2019 levels, with respondents expecting it to be roughly 5% to 6% higher in 2022. Outpatient and procedural volumes are expected to be 6% to 8% higher in 2022. Concerns around capacity to meet patient demand were expressed in a variety of specialties.
Despite the acceleration of both trust in and deployment of AI solutions, and the expectation of continuing capacity and volume pressures, scaling and implementation of AI continues to be a challenge, reports Sage Growth Partners: Last year, their survey showed that only 7% of hospital’s AI strategies are fully operational. Implementing them is simply too resource heavy, with 44% noting that they had resource constraints or had “difficulty identifying the best processes for automation.”
One of the first entry points of AI into healthcare facilities over the last four years has been through diagnostic imaging, which is now becoming a standard of care in Europe and the U.S. A number of AI vendors offer solutions that can assist in the detection of both acute and chronic pathologies, providing healthcare institutions with a diverse set of tools to choose from. Now, the AI for hospitals game has branched off of medical imaging into other prospective segments ripe for innovation, such as care coordination. These other, non-imaging segments, have benefitted indirectly from AI’s initial entry point into medical imaging, leading to higher ED throughput, reduced turnaround times for acute cases, and reductions in patient length of stay.
As AI continues to prove its value in clinical settings and becomes an enterprise AI endeavor, the existing technical barriers will need to be addressed. Moreover, there now exists a more important question regarding orchestrating AI on a large scale: How can multiple AI solutions—triaging numerous pathologies and coordinating care for patients with positive cases—be integrated together under one operating system that orchestrates it all? It’s no simple answer for hospitals, let alone vendors.
AI orchestration for hospitals requires a diverse set of parameters to be taken into account, which extend from the radiologist’s user experience to the actual technical implementation. For example, consider the radiologist’s interface, which consists of a worklist and PACS monitors on which they review the images. Incorporating numerous AI solutions, each with their own user interface, into the same work station could resultantly add more distractions to a radiologist and increase reading times. Moreover, two algorithms from different vendors could conflict with one another by analyzing the same image for the same pathology, thereby reducing operational efficiencies. Even for IT managers, integrating multiple AI solutions into hospital workflows, especially for the same department, can become a cumbersome and redundant task.
On a more technical level, an effective AI solution requires accounting for thousands of parameters. With most vendors’ solutions not necessarily designed to be easily compatible with others, orchestration abilities become essential to managing the flow of data in an optimal way throughout the hospital and wider healthcare system. This means data across the hospital needs to be synthesized in such a way that different vendors’ solutions do not interfere with one another, thereby preventing disturbances to clinical workflow and radiology read times.
A large quantity of AI vendors covering detection of different pathologies will likely emerge in the coming years and be integrated into workflows. With these AI integrations, new data highways will sprout up in workflows, leaving an opening for technical malfunctions. It is here that physicians will encounter obstacles that may delay clinical care, radiologic throughput, and overall efficiency. AI orchestration is the requisite piece that will prevent these technological issues from manifesting in a clinical setting.
The key to maximizing the benefits of the growing wave of impactful AI solutions, and enabling them to operate optimally in clinical settings, is an enterprise-grade operating system. But what is an AI operating system (OS)?
An AI OS is a tool that efficiently coordinates the flow of data between different points within a healthcare network, allowing physicians to optimally use multiple AI-based tools for their own clinical needs in an interoperable fashion, while eliminating the need to rework the IT infrastructure for every new AI solution integration. For more context, let’s take the radiologist’s workflow as an example.
Each time a new solution designed to detect a different pathology is integrated, an AI OS orchestrates the algorithms of the integrated solutions to ensure that they do not conflict and rather complement each other when possible. In a scenario where two solutions can technically detect the same pathology, an operating system can make automated decisions to optimally apply the right algorithm to match the suspected pathology. And this principle extends beyond the reading room and well into other segments of direct patient care.
In care coordination scenarios, the OS could improve communication between caregivers within a health network with automated alerts, packaging relevant information for review by, for example, an interventional radiologist or an endovascular surgeon.
As AI strategies emerge in hospitals across the U.S. and Europe, such a technology will become integral to managing the multiple points of care where AI is utilized. Aidoc’s AI operating system (OS) solves the problem of looming AI orchestration needs by enabling seamless integration of multiple vendor solutions under its unifying platform. The company’s platform includes a comprehensive suite of AI, including solutions for triage and detection of acute patients and AI-driven cross-specialty workflows facilitating care coordination. Aidoc’s solutions are currently used by 5,000 radiologists in health networks, hospitals and radiology groups worldwide, having analyzed over 10.3 million scans to date.