The Potential Impact of AI on Modern Medicine

These days, we are inundated with information about AI and how it’s going to change our lives. The medtech industry is no exception.

But what might practical applications of AI look like in modern medicine, and how close are we to actually making that happen? Are all of these changes right around the corner, or do we have many years to go until we see the full impact of AI on healthcare? It can be hard to sift through all of the noise.

We’ve had two guests on our podcast recently that shared their observations, experiences, and predictions about the potential of AI in modern medicine: Adityo Prakash, founder and CEO of Verseon, and Dr. Ronald Razmi, co-founder and managing director of Zoi Capital and author of AI Doctor, The Rise of Artificial Intelligence and Healthcare.

Based on those discussions, these are some insights we can share.

How can AI transform our approach to healthcare?

There are several ways that AI could eventually transform or improve current processes in healthcare.

Finding New Medications to Treat Illnesses

Adityo shared that his company, Verseon, designs “completely new drugs using proprietary AI technology combined with our breakthroughs in quantum physics modeling of how drug molecules bind to proteins in our bodies.” This technology could allow for more rapid drug discovery than other currently-available methods and has the potential to change how we treat certain illnesses.

Discovering Potential Health Issues Early On

Adityo also shared that AI could eventually help physicians detect potential health issues earlier than we can today by doing things like analyzing radiology images, cancer biopsies, blood test data, etc. We are collecting so much data in the healthcare space, but it would be difficult and time-consuming for physicians to analyze all of that information in its current state and draw real conclusions. AI has the potential to perform that analysis, make predictions, and share key points with physicians so that they can make care decisions and intervene earlier with particular conditions when needed.

Analyzing Data for Diagnosis

Much like AI could potentially detect issues early on, physicians could also use it to analyze data for diagnostic purposes. We aren’t at the point where this is happening yet, but it could eventually be a tool that physicians could use.

Eliminating Clinical Work for Physicians

Dr. Razmi shared that while it would be great to see all of the innovative ways we can use AI, physicians still aren’t at a point where they trust it. He feels that AI needs to be introduced more incrementally, in applications where there are lower stakes. For example, using AI to take notes during an appointment so that doctors aren’t having to type notes into a computer, or using AI to summarize information. This improves clinical efficiency, and also helps physicians become more comfortable with the idea of using AI.

What are some of the roadblocks to using AI now?

Both Adityo and Dr. Razmi agreed that while there is great potential for AI to transform modern medicine, there are definitely roadblocks to fully using its capabilities in the present day.

Real-World Data

First, we need to have enough real-world data to be able to train the AI. The problem with that is that hospitals and medical facilities all have their own systems for collecting and storing patient data, and they aren’t always willing to share that data. There are (rightful) concerns about patient privacy and liability. It can also be difficult to bring all of that data together when it is all collected and stored differently. 

Until things change on those fronts, this may continue to be a major roadblock to implementing AI-based solutions more widely in healthcare.

Feedback for Physicians

Beyond the issues with sharing data, we would also need to change certain processes to ensure that physicians are getting good feedback on prescribed treatments and medications. Often, unless the patient has a chronic condition that requires regular appointments or is undergoing major treatment (such as with cancer), physicians aren’t getting much feedback on the efficacy of their recommendations. If we are using AI to help make those predictions and indications for treatment, feedback is essential to continuing to train and improve it.

Accuracy of AI

There is also concern with the accuracy of AI. A degree of human error is always present in our current healthcare processes, but physicians and health systems need to feel confident that AI is more accurate than human error to feel comfortable implementing it for patient care measures. Some AI-enabled medical devices have received FDA approval, but that alone has not been enough to convince physicians and hospitals in many cases.

Lack of Real-World Studies & Transparency

In his podcast episode, Dr. Razmi shared that conducting real-world efficacy studies could help to alleviate some of those concerns with AI-enabled medical devices, stating “Until companies get serious about doing those prospective real world studies, documenting the safety and the efficacy, I’m afraid we’re going to remain in neutral.” Adityo also shared that being able to explain the logic behind the AI will be helpful, so that doctors can understand what the AI is prioritizing and why it is making the decisions that it is making.

The Employment Threat

Finally, as with other industries, there is the concern that AI will replace jobs that have historically been held by humans. However, both Adityo and Dr. Razmi explained that human oversight and input will always be needed even if AI is used to its full potential.

Sifting through the AI “noise”

It seems like we are constantly being inundated with information about AI and all of the ways it’s supposedly going to change our lives. It can be tough to determine what is legitimate and what isn’t. 

And furthermore, are all of these changes right around the corner, or are we more likely to see them 5-10+ years down the road?

We all know that major changes in healthcare do not happen overnight. There is too much at stake when we are talking about patient care. And, as we have seen, hospitals and providers aren’t necessarily sold on AI-enabled technology as of yet. 

Dr. Razmi predicts that taking a more gradual approach to introducing AI will eventually lead to wider adoption. For example, starting with tools that help with documentation and building trust from there. Grander things like early identification of health conditions or treatment recommendations will likely take at least a couple of decades for healthcare providers to adopt.

There are a lot of exciting possibilities for AI and how it could transform our approach to healthcare, but we still have a way to go before we can fully realize those benefits.