Artificial intelligence. Machine learning. Automation. Computer vision.

March 06, 2018
By John Vartanian

If you ask anyone working in the medical field today, the words at the top of this page will return a unanimous adverse response as they pose a grave risk and would take away from the personalized care that individual patients need.

This response is due to the fact that these technological innovations directly challenge the status quo that has been the bedrock of clinical medicine for generations of doctors. Physicians are the gatekeepers of medical knowledge and it is their experience and intuition that should be the deciding factors when selecting between treatment options.



However, many across the medical community have begun to speak out against this dogmatic belief with the understanding that treatment decisions should be based on the best clinical evidence available at the time. While this approach to optimizing patient care feels like common sense, it requires clinicians to ensure that no recent publication could alter their current treatment protocols. This becomes nearly impossible when you consider the 40,000-plus articles published across the 30,000 active scientific journals every week.

With artificial intelligence platforms like IBM’s Watson, able to read thousands of journal articles every day from publications worldwide, we must take a moment to ask ourselves, at what point does it become unethical to NOT use computers in the diagnostic process? If a computer can stay up to date with cutting-edge research from around the globe, should the role of the physician convert from a bank of medical knowledge to a translator of complex medical information for the patient?

This way of thinking brings about new questions as we begin to critically look at other sectors of the medical system that are experiencing pressure from digital diagnostics, specifically radiology and pathology.

The radiology resonance is in full swing as numerous companies such as Zebra Medical Vision work to integrate computer-aided diagnostic algorithms into current clinical practice. These tools use machine learning to help clinicians see into the microscopic details that were previously missed with the human eye alone.

The days of radiologists spending countless hours meticulously scanning thousands of X-rays, CTs and MRs are numbered. Rather than relying on a clinician to catch a life-threatening abnormality among hundreds of images viewed per day, we can now use computers to triage the image bulk.


By filtering out the normal cases, computers can direct the clinician to inspect any concerning findings that may be present in the minority of images. This would allow for more efficient image processing while also enabling the radiologist to be a more effective diagnostician as their efforts would be directed to the cases that require their expertise.

The field of pathology is not far behind as the advent of whole slide imaging technologies opened the floodgates for automation and digital diagnostics. Companies such as MedKairos are on the forefront of this development as they are working to teach computers to recognize the patterns that are the core of pathology practice.

With thousands of patient samples stored in academic institutions across the country, we can look at the disease progression of patients in the past to provide more specific and personalized diagnostics for the next generation. By appreciating every microscopic change on a pixel-by-pixel basis, our ability to see into cellular disease processes is only limited by the resolution of our imaging equipment. Mirroring the revolution in radiology, pathologists can expect these technologies to help them manage an ever-growing patient population more efficiently, reallocating time to the complex cases that require the time and care of an expert diagnostician.

As we continue to go down this inevitable road to artificial intelligence’s integration into clinical practice, there will be growing pains. Yes, there will be challenges when a machine misdiagnoses a patient. And yes, a level of the human element of medicine will be removed with the introduction of clinician computers.

However, we will also see fewer malpractice cases as physicians will be able to rely upon the most up-to-date clinical guidelines at the click of a button. We will also see shorter diagnostic wait times for patients and their families concerned about what a lump in their neck might mean. We will see more efficient clinical workflows that will ultimately drop the cost of medical care globally. We should also expect to see more collaboration across specialties as patient data is more easily interconnected.

We shouldn’t fight the unknown because of the challenges that we may face during the transition process. Refusing to make a decision about innovative technologies, like machine learning, might sound protective for our patients today, but it is actively choosing to delay access to critical technologies that could save the life of a future patient. At this critical junction in medical innovation, we must ask when does this decision of resistance and inaction begin to expose our patients to a higher risk than they would have otherwise experienced?

John Vartanian
About the author: John Vartanian was the founder as well as the president and CEO of Medical Imaging Resources for 23 years. Medical Imaging Resources provided high-end mobile and fixed-site imaging systems, MRI, CT, nuclear medicine and cardiac cath labs to the market for over two decades. Vartanian has worked in the high-end imaging market since 1985.