By Sven Powilleit
Advancements in computing, communications, and robotics technologies are converging to drive the emergence of artificial intelligence in medicine (AIM). From diagnostic enhancements to patient care optimization and security automation, AI will increasingly make a difference. It’s all about leveraging and efficiently analyzing an astronomical amount of data.

Critical areas of focus for AIM include the following four categories:

  1. Direct support of diagnostics. Medical imaging, complemented with AI technology, can achieve new levels of accuracy and automation, and improve diagnostic decision making.
  2. Operational effectiveness. Hospital re-admission has long been a concern. When attributable to insufficient or ineffective post-operative care, finding a solution is paramount. AI can help ensure that costly readmissions are reduced through predictive analysis.
  3. Cost optimization. Balancing clinical care and cost is critically important to the survival of every medical service provider. Processes that improve resource utilization can lead to reduced total healthcare costs without diminishing service quality. Done right, it can enable lesser skilled healthcare professionals to conduct routine tasks while improving patient workflows at the point of care. AI can play an important role in managing this kind of administrative environment.
  4. Optimized delivery of care. AI can help in the prevention and reduction of invasive procedures, present doctors with accurate measurements, and allow for faster and better decision-making.

AI and Diagnostics: Present and Future

As a maker of leading noninvasive diagnostic equipment, we at Verathon are focused on determining how AI can be integrated into our products and tools to improve workflow and aid healthcare professions (HCPs) in making more accurate decisions, by less skilled users, at point of care. Unfortunately, AI has not yet been fully realized across the technological landscape for medical applications.

Many makers of diagnostic tools, including Verathon, are investigating ways that AI can be directly applied to diagnostic systems—but none have integrated AI directly into systems. In general, AI is only being used in automated measurements aiding in decision support and diagnostic support at this time. These two areas do enhance decision making for healthcare professionals today.

AI-capable support systems come with the data used for analytical purposes integrated into the product architecture—they do not gather data. Devices with AI come to HCPs with data-specific intelligence already pre-learned. The concept of getting smarter over time is not yet being used in medical applications due to the requirements for FDA 510(k) clearance.

Algorithms that are internally developed (by a hospital or other healthcare organization) and applied within the institution do not always require FDA approval. Most of these are used to improve workflow and efficiency, but not necessarily to make diagnoses. That’s an important distinction. Before a medical device can be marketed, such as our BladderScan Prime Plus, which operates on a deep learning algorithm called ImageSense, it must obtain clearance from the FDA.

Academic research is leading the way toward real-world implementation of AI in direct diagnostic applications. One important study focused on an algorithm that can detect pneumonia from chest X-rays at a level exceeding the performance of practicing radiologists. In CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning (2017, Stanford University), researchers demonstrated that an AI algorithm is capable of effectively outperforming the diagnostic accuracy of practicing radiologists.

In their analysis of findings, researchers conclude with this: We develop an algorithm which detects pneumonia from frontal-view chest x-ray images at a level exceeding practicing radiologists. We also show that a simple extension of our algorithm to detect multiple diseases outperforms previous state of the art on ChestX-ray14, the largest publicly available chest x-ray dataset. With automation at the level of experts, we hope that this technology can improve healthcare delivery and increase access to medical imaging expertise in parts of the world where access to skilled radiologists is limited.”

That foundational premise applies not only to chest x-rays, but across healthcare to improve efficiencies through AI-assisted image interpretation. As shown in a 2016 article published in JAMA, it has been estimated that, per day, AI would process over 250,000,000 images for the cost of about $1,000—hypothetically saving billions of dollars.

Implementing Holistic Strategies

While the direct deployment of AI within the software architecture of diagnostics systems is still on the horizon, it’s essential to understand how AI can deliver real, potentially life-saving effects for patients—and cost-saving effects for both HCPs and patients.

Consider the very real problem of catheter-associated urinary tract infections (CAUTIs).  These are widely recognized as among the most prevalent healthcare-associated infections.[i]Verathon’s latest-generation BladderScan product, for instance, addresses this concern directly due to its deep learning algorithm that improves workflow and decision-making right at point of care and with less-skilled users, reducing the need for catheters.

Coupling such noninvasive diagnostic approaches, AI-based machine learning, and catheter-minimizing workflow redesign strategies suggested in a 2016 study[ii]published in the Infection Control & Hospital Epidemiology, presents a real potential for limiting CAUTIs.

Key findings of that study included several procedural considerations that contribute to high instances of CAUTIs:

  • Inappropriate criteria guiding catheter placement decisions
  • Physician’s limited involvement in catheter placement decisions
  • Patterns of urinary catheter overuse
  • Poor urinary catheter insertion technique

In essence, the ability to achieve an accurate measurement quickly is at the core of making better informed catheter placement decisions.

By addressing these shortcomings and binding process improvements to non-invasive (and eventually AI-based) systems, real strides can be made in making CAUTIs less common. Similarly, taking a wholistic approach to the diagnostic/analytic landscape across all HCP disciplines will unquestionably lead to better healthcare outcomes, reductions in readmissions, and ultimately more efficient healthcare operations.

Looking Ahead

It’s a heady time for those advocating for and researching how AI can integrate into the day-in-and-day-out operations of HCPs. Even so, implementing AI into medical applications is still in its infancy, particularly for clinicians. While studies, to date, have demonstrated advantages in a number of situations, improvements have often been more incremental, rather than revelatory.

As suggested by Eric J. Topol in a comprehensive article published by Nature Medicine earlier this year[iii]:“There has been remarkably little prospective validation for tasks that machines could perform to help clinicians or predict clinical outcomes that would be useful for health systems, and even less for patient-centered algorithms. The field is certainly high on promise and relatively low on data and proof. The risk of faulty algorithms is exponentially higher than that of a single doctor-patient interaction, yet the reward for reducing errors, inefficiencies, and cost is substantial.”

Sven Powilleit is director of product management and strategic marketing at Verathon. Questions and comments can be directed to 24×7 Magazine chief editor Keri Forsythe-Stephens at


  • [i]Umscheid C, Mitchell MD, Doshi J, Agarwal R, Williams K, Brennan P. Estimating the proportion of healthcare-associated infections that are reasonably preventable and related mortality and costs. Infect Control Hosp Epidemiol2011; 32:101-114.
  • [ii]Carter E, Pallin D, Mandel L, Sinnette C, Schurr J. Emergency department catheter-associated urinary tract infection prevention: multisite qualitative study of perceived risks and implemented strategies. Infect Control Hops Epidemiol2016; 37(2): 156-162.
  • [iii]Topol E. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine25, 44–56 (2019).