The Future of MedTech: Where AI Meets Human Expertise
As AI transforms healthcare, MedTech leaders must discern where machines outperform—and where human expertise remains indispensable.
As artificial intelligence (AI) becomes deeply embedded in healthcare, its potential to transform medical technology is undeniable. From diagnostics to documentation, AI is augmenting how we deliver and experience care. But while AI algorithms are accelerating many facets of medicine, there are critical areas where human expertise remains irreplaceable. Applying AI to the right set of problems is a choice that all medical technology companies are having to make and it is more complex than it seems. To find the right answers, one has to get past the hype by the futurists, for example that AI will replace radiologists, and dive much deeper into the AI technology to understand where it shines and where humans will continue to lead the way.
What is AI in MedTech and Healthcare?
At a very simple level, AI is a highly advanced pattern recognition algorithm that can be applied to many different domains. How well the AI performs, comes down to the data it is trained on, both the expected inputs and the desired outputs. Noise and variability in the inputs and, more importantly, the outputs, in the available training data is a key factor that determines if AI can produce relevant and trustworthy results. In healthcare, the output used for any type of training is based on what human physicians have observed and documented. So, the reliability and robustness of the human output, not just the input, is a very important requirement for training AI in healthcare, where high sensitivity and specificity is expected for any type of clinical use, along with transparency and explainability. But many times, this simple paradigm gets overlooked in the enthusiasm to apply AI to any and all problems.
AI as the Super Observer, Humans as the Interpreters
The example below relates to Clinical Diagnostics and Decision Support and provides insights into problems in MedTech that AI can excel at and ones where it is less likely to be trusted as a replacement for human expertise.
As everyone already knows, AI excels at processing vast amounts of medical data — from radiology scans to vital signs to lab results — to detect deviations and anomalies. The reason this a great problem for AI to solve is because it is well-defined with vast amounts of robust input and output training data. In many such applications, AI can even outperform the most expert of humans by spotting patterns that humans can’t see. As an example, AI can identify a nodule in a breast or lung imaging exam, classify its appearance, and automatically calculate any change in size over time. AI excels at this type of observation and measurement.
However, can this be extended to true diagnostic interpretation? Not as easily. The reason is the “variability in the outputs”. If you show the observation to three different physicians, will they interpret it the same and draw the same conclusion? In most cases, the answer depends...on what other information the physicians have access to and consider regarding the patient, and their own expertise, experience, and judgment (the human element). This can vary tremendously from case to case, creating an inconsistent output to train the AI.
The Human Dimensions AI Can’t Replicate
We have observed many companies make this classic mistake of thinking they can eliminate the human variability by training AI on highly curated set of “expert” outputs and create AI that is as good or better than the expert humans. But this approach rarely translates to the real world and results in bias and a lack of generalizability, both across varied patient populations and across the varied opinions of the expert humans involved in clinical decision making. In such cases, it may be easier to train the AI to play a supportive role to present suggestions and highlighting key patient information, in aide of the human physician, rather than trying to replace their trusted wisdom and expertise.
At the end of the day, medicine is an investigative science that needs empathy, communication, collaboration, experimentation, judgment and trust, more than just data intelligence, and AI needs to be applied in ways that fits with these deeply human attributes as part of any medical technology. At Factor 7 Medical, we have experts that have been involved in the development and commercialization of several AI-based MedTech products and can provide the right guidance on both the strategy and implementation of this revolutionizing technology.
Related Resources
