Overcoming Barriers to AI in Medicine< Back To Feed
Based on a recent Goldman Sachs report1, global investment in Artificial Intelligence has not only doubled in the past two years, but is also forecast to hit $158 billion by 2025. This clearly shows the market sentiment that a significant inflection point is on the horizon, one with transformative implications across industries. However, let's not forget that transformative shifts like these are seldom smooth sailing, and healthcare is no exception.
For those who have spent considerable time in the healthcare sector, headlines announcing groundbreaking machine learning models from leading technology organizations like Google and IBM are hardly new. However, a closer look reveals that these are more often cautionary tales than success stories. AI models have been known to exhibit limitations that compromise their efficacy in clinical environments2,3. Instances like these underscore the inherent challenges of seamlessly incorporating Artificial Intelligence into healthcare systems. Cohesic CEO and Co-Founder Dr. Jordan Engbers addresses these challenges during his presentation for the INOVAIT lecture series4.
The presentation focuses on three main barriers to clinical adoption of AI:
- Lack of high-quality, labelled data relevant to patient populations of interest
- Discontinuities in performance as we move from the lab to clinical environments
- Ineffective execution within complex clinical environments
So, how can we navigate these complex barriers effectively? At Cohesic, our stance is that a holistic strategy is key to implementing Decision Intelligence in healthcare—simultaneously addressing data collection during clinical workflow, integration of solutions into existing IT systems, and embedding knowledge by continuous collaboration with clinicians. Dr. Engbers points out that the earliest AI successes in business emerged from companies that had the benefit of vast user data, generated in-house. Therefore, empowering healthcare organizations to capture data in a way that is scalable, accurate, and appropriate for machine learning, will serve as a cornerstone for the successful implementation of AI in medical practice.