Dr Anthony Loschner Expounds on the Benefits of AI in Pulmonary Medicine

November 17, 2020

Artificial intelligence (AI) frees up time for clinicians, noted Anthony L. Loschner, MD, assistant professor and associate program director, Critical Care Fellowship Program, Virginia Tech Carilion School of Medicine.

Despite limitations to the perception of artificial intelligence (AI), there is value in that it frees up time for clinicians to devote to more meaningful work and clinical activities, noted Anthony L. Loschner, MD, assistant professor and associate program director, Critical Care Fellowship Program, Virginia Tech Carilion School of Medicine.

Transcript

Are there limitations to how AI is currently used in pulmonary medicine?

I'm a glass-half-full type of guy, so I think all of it’s good right now. There are some limitations, I think, in the perception of AI—adopting it, because you don’t have to look far into artificial intelligence, not just in medicine, but throughout the industry—and a lot of speculation on how dangerous it could be.

Stephen Hawking said that a concern of his was that AI would develop a conflict with our own interests, and then hence compete with us. And so that’s a scary thing. I think that’s a barrier to get over. Job loss is another one. However, I don’t really see it that way. I see it as a great job aid, where it can offload some of the more data mining and data interpretation to the computer, allowing the clinician to allow time for more meaningful work and clinical activities.

Pulmonary function test [PFT], I think, has a really good future in AI. Pulmonary function testing allows itself for AI because it’s data rich; [there are] many numbers and the pattern recognition within those numbers.

A group in Belgium just published a paper with fantastic, humbling, results really where when a pulmonologist is sure, or absolutely sure, of their interpretation, the quality of their interpretation, they were only about half the time right actually. So when we’re the most confident in our interpretation, about half the time we’re right, and then it sort of diminishes and equals a computer from there.

So that’s awesome, because pulmonary function tests do take time to read; it’s very manual in its interpretation, and this study shows that it can be flawed, deviating from guidelines and just errors. Also, because if there’s a large amount of PFTs to be reviewed, reading fatigue sets in and then mistakes can happen.

Also, public health and occupational health uses for pulmonary function tests wouldn’t require a physician and cost with AI. Huge amounts of PFTs can be interpreted and only the abnormals be reviewed by a physician or refer to a pulmonologist for further review.