AI in healthcare: not a replacement but support

Artificial intelligence is transforming our healthcare system at a pace that was unthinkable a short time ago. AI applications seem to be the perfect tool for faster diagnoses, more personalized treatments and even preventing diseases before they manifest themselves.

But what does AI mean for patients, doctors and the development of our health care system? How is it changing the way we approach being sick and getting better? What are the risks? Are we actually ready to embrace AI in the medical world?

What does AI mean for healthcare professionals?

AI is already causing major changes in the daily work of healthcare professionals. With AI applications, doctors and nurses can make faster and more accurate diagnoses. They can better tailor treatment plans to individual patients. And at the same time, the plethora of administrative tasks is finally coming to an end. Much of this will be taken care of by AI, leaving much more time for personalized care.

Of course, this new technology also brings challenges. Healthcare professionals must learn to work with these new tools. And we need to think carefully about privacy and ethics when using patient data.

Transparency

One thing is crucial: transparency. We need to understand how AI arrives at its conclusions. Only then can we ensure that these smart systems become part of everyday healthcare practice in a safe and effective way.

Professor Carl Moons of UMC Utrecht says:

"Artificial intelligence has enormous potential to improve patient care and preventive care, across all healthcare sectors. From early diagnosis of patients with lung cancer to identifying those at increased risk for heart attacks, dementia, birth defects, infections and many other conditions or syndromes."

AI is not a replacement, but it is a tool

"Hey AI, please diagnose this patient for me."

It doesn't work that way, of course. And fortunately.

AI in healthcare is not there to replace doctors. It is there to support them. To help them make better decisions. Faster. More accurate.

And yes, AI, like us, makes mistakes from time to time. That is why it is so important that a physician always remains in charge and responsible to notice and correct those mistakes.

Eric Topol, a cardiologist and AI expert, emphasizes this point in his book "Deep Medicine." He argues that AI can not only contribute to better diagnoses, but also improve the human side of care by automating administrative tasks. This allows healthcare providers to spend more time with patients.

So should we perhaps see AI more as a highly capable healthcare assistant? One that is available 24/7, never tires and can process huge amounts of data at lightning speed?

How many lives could we save with it? How can we use AI to reduce the pressure on care, so that doctors and nurses have more time and attention for patients?

How is AI already being used in healthcare?

Diagnostics and treatment

AI is already playing an increasing role in supporting physicians today. For example:

• WoundHealth app: This app, developed by Radboudumc, lets patients take pictures of their wounds at home. AI analyzes the wound to check if it is healing properly.
• Radiology: AI systems help analyze X-rays and CT scans, allowing radiologists to make faster and more accurate diagnoses.

A concrete example is a study published in Nature (2020), in which an AI system developed by Google Health analyzed mammograms to detect breast cancer. The system was found to outperform human radiologists in some cases, with a lower error rate in both missing cancer and incorrectly designating cancer.

Picture this: you are a radiologist. You have to review hundreds of CT scans a day. Wouldn't you then be happy to have an AI assistant to help you find suspicious spots?

Of course it does.

You could help many more people by collaborating with such an AI system. The system flags possible problems, you assess whether it is correct. Together you arrive at a diagnosis.

Preventive care with AI

AI systems can analyze large amounts of health data to identify risk factors and enable early interventions. One example is using machine learning to predict the risk of type 2 diabetes. Researchers have shown that AI is able to estimate type 2 diabetes risk more accurately than traditional methods based on individual health data, without violating patient privacy.

Another impressive example comes from the Mount Sinai Health System in New York. They developed an AI platform called "Deep Patient" that analyzes electronic medical records to make predictions about disease outcomes. For example, the AI could identify patients who were at increased risk for specific conditions, such as liver cancer, months before doctors diagnosed them.

Predictive care in practice

A concrete example of predictive care is Nicolab's StrokeViewer. This AI-driven platform enables physicians to diagnose strokes faster and more accurately. By applying advanced image analysis algorithms to CT scans, StrokeViewer can provide crucial information about stroke location and severity, leading to faster treatment decisions and better patient outcomes.

Personalized prevention

AI also enables more personalized prevention strategies. By identifying patterns in large data sets of clinical data and genetic information, AI systems can predict which individuals are at increased risk for specific conditions. This enables healthcare providers to take targeted preventive measures, such as lifestyle interventions or early screening.

One example is IBM Watson for Oncology, an AI system that supports oncologists in choosing treatments for cancer. The system analyzes patient data and compares it to scientific literature and clinical guidelines. While the potential is great, this system also shows that integrating AI into clinical practice is not always flawless and there are still challenges to overcome.

AI is not a medical textbook

A medical handbook gives you an overview of knowledge and facts. Standard procedures. Symptoms, diagnoses, treatments. AI in healthcare does more. It learns from experience. It works better and better as it processes more relevant data.

AI in healthcare is mainly about combining the vast amount of medical knowledge available with each patient's specific data. AI is very good at making connections between these two elements that a human might overlook.

Skin and edema therapist Dr. Janneke van der Zee is involved in the development of medical AI applications, and talks about it:

"AI allows us to make diagnoses faster and more accurate. AI's ability to recognize patterns invisible to the human eye makes it an indispensable tool in modern health care."

Challenges of AI in healthcare

Of course, not everything is positive about the use of artificial intelligence in the medical world. There are also many major challenges that we need to pay attention to:

• Data quality and privacy: AI systems need a lot of data, but how do we ensure that that data is reliable and protects patient privacy?
• Transparency: How can we ensure that AI decisions are understandable and explainable?
• Ethical considerations: Who is responsible if an AI system makes a mistake?

Gary Collins, professor of Medical Statistics at Oxford University, emphasizes the importance of transparency:

"There is huge potential for artificial intelligence to improve health care, but the right tools must be used. [...] Transparency makes it possible to identify errors, facilitates assessment of methods used, and ensures effective oversight and regulation."

Transparency about the use of AI is essential to prevent AI-driven prediction models from discriminating against specific groups or creating health care inequities. This ultimately ensures greater trust and thus acceptance and better use of AI algorithms and prediction models by not only healthcare providers, but also patients.

Hugh Harvey, a radiologist and AI expert, stresses the need for rigorous clinical testing of AI systems before they are put into practice. He warns that without proper oversight, AI tools can lead to misdiagnoses and human evaluation will always be necessary.

Limitations of AI

AI in healthcare is impressive. But not infallible. Let's have realistic expectations of AI in healthcare by continuing to recognize AI's limitations.

• AI can recognize patterns but does not understand the full context of an individual patient.
• Nor can AI show empathy or make ethical decisions.

Therefore, a physician or healthcare professional always remains indispensable.

Fei-Fei Li, professor of Computer Science at Stanford University, emphasizes the importance of "human-centered AI." She argues that AI should be developed with a focus on collaboration with humans, using technology to enhance human capabilities without replacing them.

The future of AI in our healthcare system

Despite the challenges, the future of AI in healthcare looks promising.

Some developments to keep an eye on:

Personalization of treatments

AI can help better tailor treatments to individual patients. This can lead to more effective treatments, fewer side effects and an increased chance of recovery.

More efficient care processes

By automating tasks, AI can relieve healthcare providers of administrative tasks. This leaves them much time for what is important: personal contact with patients.

Early detection of disease

AI systems can help detect disease early, even before there are obvious symptoms. A recent study showed that AI can detect breast cancer as well, or even slightly better, than radiologists when analyzing mammograms.

Regina Barzilay, MIT Professor and AI researcher, has pioneered the use of AI for breast cancer screening. She argues that AI can contribute to equal access to high-quality care by making technology available to areas with a shortage of specialists.

How can we (safely) embrace AI in healthcare?

To successfully integrate AI into healthcare, we can:

1. Investing in training: Healthcare professionals learn to work with AI systems and tools.
2. Collaborate: Get healthcare agencies, technology companies and government to work together to create the right frameworks.
3. Involving patients: Inform and involve patients in the development and implementation of AI in healthcare.
4. Learning from mistakes: Every "mistake" of an AI system is an opportunity to improve. Take, for example, the development of AI for analyzing X-rays. In the beginning, such analysis systems made a lot of mistakes. But every mistake is a learning opportunity. Now AI can even analyze more thoroughly than human radiologists in some cases.

Conclusion

Artificial intelligence is playing a growing role as a digital assistant for healthcare professionals. The goal is absolutely not to replace physicians, but to support them in delivering faster, more accurate and more personalized care.

While there are challenges of privacy, transparency and ethics, AI offers tremendous opportunities to make healthcare more accessible, efficient and effective.

So the future of AI in healthcare is promising, but requires a careful approach. By investing in training, collaborating across sectors and actively involving patients, we can maximize the benefits of AI while minimizing the risks.

Ultimately, it's about AI supporting the caregivers in our society so they can do what they do best: care for people.

With the right balance of technological applications and human attention, AI can make a particularly valuable contribution to the future of our health care system.

Want to learn more about the applications of AI within your field as a healthcare professional? If so, please drop us a line. We are always eager to hear your questions, personal experiences and/or feedback.