Question of the quarter: What role does MedEdQR see AI and ML playing in future medical education?

 “What role does MedEdQR see AI (Artificial Intelligence) and ML (Machine Learning) playing in future medical education?

A message from the CEO

I recently received relating to the future use of Artificial Intelligence (AI) and Machine Learning (ML) in medical education.  Since I have been researching the possible uses of AI and ML in the MedEdQR platform, I thought that it might be a great opportunity to offer some insights and opinions on the topic.

A few months ago I read an article, published in Nature’s online journal September 2018, npj|Digital Medicine, titled “Machine learning and medical education”(1). The authors set the tone for their discussion by stating:

Artificial Intelligence (AI) is poised to help deliver precision medicine and health.  The clinical and biomedical research communities are increasingly embracing this modality to develop tools for diagnosis and prediction as well as to improve delivery and effectiveness of healthcare.

They further state that “medical education, … programs have not yet come to grips with educating students and trainees on this emerging technology.” So where are the opportunities and how might those opportunities impact the MedEdQR platform?

One of the first things I noticed about AI/ML usage in medical education, is the huge interest in the topic as shown by the number of related articles published and the number of VC funded startups launched in the last few years.  Yet, as with most technologies, while there is much interest, there are few results.  I remember looking into AI research in the 1990’s (ML wasn’t even a topic then) while working on my PhD, and thinking then that while the topic has been hot for many years (again, I remember following the topic starting in the late 1960’s), there were few useable outcomes. Outcomes did not come close to meeting expectations until the last few years, most likely fueled by Internet search providers. It appears that the hard coded, rule based “expert systems” of yesterday are giving way to the data driven “discovery” AI/ML systems of today. 

What areas are ripe for AI/ML usage in medicine (and subsequently in medical education)? The most frequently heard is data mining for healthcare systems. Consider the potential findings in medical knowledge, patient care, protocols, treatments, outcomes, quality improvement, patient safety, etc. that AI/ML might bring by deep diving into EMR/EHR (medical record) systems, for example, IBM’s “Watson for medical applications.”. By “deep dive” I mean that through the use of AI/ML algorithms, data patterns (useful medical knowledge, protocols, pharma, …) are found that were impossible to predict or be found using non-AI, non-ML data science techniques.

The next area would be in new pharmaceuticals; i.e. drug discovery.  Notice a trend here?  Applications of AI/ML follow the money. The results of such research hopefully will lead to the development of medical knowledge systems (i.e. applications) to extend medical and clinical knowledge as well as professionalism for the practitioner as well as the student. 

The next question is, how can such applications be introducedinto medical curriculums? From a medical education perspective, the article rightfully notes that when looking at undergraduate and graduate medical curriculums, there is little space to include AI and ML topics.  The authors’ state:

Medical schools lack the faculty expertise required to teach this content [AI/ML theory and application] which is largely taught by the computer science, mathematics and engineering faculties. Lack of mentorship and faculty role modeling poses a significant challenge as students move from the preclinical to clinical environment and try to develop understanding of how AI knowledge can be applied and used in the clinical setting.  AI impacts patients and patient care.  Therefore, ML and its applications should be taught within medical school and needs to be formalized to train the next generation of clinicians and biomedical scientists to face data-driven challenges that can directly impact patient care in the coming decades.

…At most schools, the instructor will most likely be a data scientist.

So how does this all impact the MedEdQR platform?  What are the opportunities?  To answer this, let me present an anecdote from my past.  In the very early 1970’s, while I was in my first year of graduate school, I had the opportunity to get in on the ground floor of “personal” computing.  Hewlett Packard (HP) had developed a desktop computer tailored for use by engineering firms but had no software applications for it. At the request of my structural engineering professor, another grad student and I were contracted by HP to develop structural engineering software. Eventually, in the early 1980’s when personal computers were entering the market, a similar situation existed – new personal computing technology, but minimum software.  Thus, when introducing the personal computer to my architecture and engineering students, instruction centered on computer hardware, (e.g. hardware components), the basics of programming (pseudo programming – the elements of a program) and the few applications that were available.  As time went on, it was no longer necessary to teach computing hardware, since, in reality, the needed focus was on applications, then eventually on programming and algorithms. We are at the same point regarding the use of AI/ML in medical education. Putting a data scientist in the medical education classroom, at this point in time, without any real discipline based applications, is reminiscent of when I started teaching computing.  While it might be useful for medical students and practitioners to understand the value of and even the modalities of data science, until there is a critical mass of applications, and an understanding of where to appropriately use such applications, will the medical curriculum gain a valuable component, suitable for inclusion?

I know I didn’t fully answer the questions that began the last paragraph. That is because, I too, am still researching its value; more to come in a future “Message from the CEO”.  But, the four areas/topics listed below, in my opinion, do have potential for inclusion in the MedEdQR virtual patient platform without lessening the learning goals and objectives. The inclusion of AI and/or ML applications become ancillary, and as such, enhance the learner’s engagement, but do not overtake it or replace it. Typically a goal of virtual patient scenarios is knowledge acquisition.

  • Image recognition
  • Diagnostics
  • Treatment Protocols
  • Patient care

This discussion is not without its caveats, which leads me to posit a few questions that we have to ask ourselves when considering such technology inclusion. 

  • While the potential use of AI and ML in medicine is extremely high and quite exciting, given the status of the technology, what is its value in enhancing medical education?
  • Is the “data scientist” instructor, described earlier, adding to the student’s ability to build on their medical knowledge, clinical practices or professionalism?
  • Is the inclusion of AI and ML material in medical education worthy of expanding the curriculum, or should it be relegated to seminar status?
  • Are AI and ML applications more appropriate for practitioners who already have the fundamentals (and the specialties) since they would be able to more fully understand the impact on their practice? 
  • What does it mean to educate a medical practitioner in our technology-enhanced era?
  • How does the basic knowledge of AI and ML potentials add to foundational medical education?

There is one question that doesn’t seem to be asked in all of my research; wouldn’t dependence on AI and ML applications interfere with the learning goals intrinsic to foundational medical education? Are we suggesting replacing or augmenting medical knowledge acquisition, clinical skills and diagnostic critical thinking activities with AI and ML tools?  Maybe someday there will be AI based virtual physicians, though most authors of AI and ML content as it relates to medicine and or medical education do not feel that the human physician will ever be replaced completely.

So it comes down to the age-old problem of when and where to use such technologies.  The answer may not be as far off as we think!

  1. Kolachalama, VB and PS Garg, “Machine learning and medical education”, npj | Nature partner journals, published online, 27 September 2018.

If you’d like to give feedback on this topic or any other existing or proposed MedEdQR feature, click here.

Isn’t it time to give the MedEdQR platform a try? If it has been a while since you saw a demonstration, or if you have not experienced a demonstration yet, we would be happy to show you how this platform can enhance your educational program. Click hereto request a demonstration.

Comments are closed.

We use cookies to give you the best online experience, and improve service performance. By agreeing you accept the use of cookies in accordance with our cookie policy. For more information about how we use cookies and the choices you have with respect to control of cookies, please read our Privacy Policy.

By continuing to use this service, you agree to our use of cookies as described in the Privacy Policy.