Since I'm The Technology Evangelist, it goes without saying that I am a confirmed gadget guy and tech lover. iIve always been a lover of gadgets, even before I went into dentistry, and throughout my career I've always tried to evaluate and incorporate technologies that made sense into my clinical practice.
However, I've always tried to look at things as objectively as possible. By that, I mean I always make an attempt to see the potential pluses and minuses that come with any tech that can affect patient outcomes.
Along those lines I'm a big believer in AI. For radiographic interpretation I like the concept of an objective third party that can help give you to find something you might not have noticed or to help a doctor consider that an area they notice might not need treatment. When it comes to AI in the administrative part of dentistry, I'm hopeful that by being able to offload some of the more mundane tasks to machines, it will allow the front office staff to have more real patient interactions. As we all know, dentistry is a relationship based profession and the more we can foster those relationships the better.
Yet, we always need to consider both the benefits and to potential detractors that come with incorporating new technologies and today's post is focused on that.
When I was in dental school, much of the first two years were devoted to sciences such as histology, microbiology, and biochemistry. The hands-on classes often seemed to focus on things that, at the time, I felt were teaching me things I'd never use. Things like pouring models or creating wax patterns and casting them weren't' things that any dentist I knew actually did. They were done by people that worked for the dentist (dental assistants and dental lab technicians) and I often wondered why I had to learn to do those jobs.
Obviously I was young and headstrong then because once I actually began to practice I quickly realized how knowing those things were the foundation on which I lot of what I did was built. If I hadn't been taught those skills, I would have been less able to understand and do what I was doing.
Since that revelation, which happened pretty early on, I've always supported schools to continue to teach those things. Even if you never plan on ever taking an physical analog impression, I feel that knowing the principles of it, makes you better at scanning.
Although the rest of this post is not specifically related to dentistry, I thought this was a topic worthy of bringing up here. I recently became aware of a study done in Poland that attempted to compare physician's diagnostic skills before and after having access to AI systems designed to help in identifying pathologies. The concept was to find experienced clinicians who knew what they were doing, see how they performed with AI assisting them, and then see how they performed without the help of AI.
The results were interesting. Of course this is only ONE study and much more science needs to be done on this aspect of things before we draw conclusions, but the results do give everyone in healthcare something to be aware of and consider as AI becomes more and more integrated into our clinical workflows.
The study evaluated if clinician's diagnostic skills could atrophy after beginning to rely on AI. The study followed physicians performing endoscopic procedures. Each physician had performed at least 2000 colonoscopies during their careers.
The study covered nine months. For the first three months doctors worked without AI and the number of precancerous areas was totaled. For the next three months they used AI that analyzed the imaging in real time that flagged for the precancerous condition. The AI tool was randomly available and unavailable to compare results during those three months. The doctors never knew when they would be able to use AI. The final three months were again done with no AI and the results totaled.
At the conclusion of the study, the data was analyzed. It showed that in the three months before AI 28.4% of colonoscopies found at least one precancerous area. In the three months after AI the number of precancerous areas noted was 22.4%.
Do the numbers indicate the skills of the doctors declined or did working with AI help them not have as many false positives? I didn't see anything which mentioned that in the article I read about the study. What I came away with was the belief that this needs a lot more study. There's a lot of concern about AI in the world and we need to understand its impact much, much better.
I am not a statistician and I'm not a specialist in colonoscopies, but I find the results intriguing at the very least. Obviously a lot more research into this needs to be done before drawing any conclusions. I'd like to tie thing together now. I feel that learning and really understanding the basics of a subject is the foundation you need to grow toward being an expert. That applies to everything in life, just like my need to know how to properly pour an impression as a student. This is also why I feel that AI is a tool. These systems are currently designed to be data gathering systems. They should not be making decisions. Instead, they should be viewed as a colleague who nudges you and says, "what about this spot?"
Healthcare diagnosis is a complicated subject that frequently requires a nuanced decision based on many factors. However as my friend Dr. Marty Jablow says about AI "don't check your brain at the door". That's a great statement that I agree with.
The study was published in The Lancet Gastroenterology and Hepatology in October 2025. I happened to come across a summary of it on Nature.com
No comments:
Post a Comment