Artificial intelligence could soon be able to help doctors diagnose and treat patients

There is a new publication in the field of cardiovascular medicine every 2.7 minutes, so even the smartest cardiologist in the world could not possibly keep up. This is one of many reasons why the use of artificial intelligence is a hot topic in medicine.

AI is the ability of a machine to make its own decisions based on data that it collects, and may eventually be used to predict a diagnosis, find new diseases, and select the best treatment. In addition to constantly scanning the medical literature for new findings, AI could also help prevent stereotyping because of social or racial bias, and prevent medical errors that can occur from fatigue, or just because a doctor is in a lousy mood.

A new term, called deep learning, has become popular. This is a method for computers to put together complex data in ways that involve levels of abstraction different from what we have seen in the past. This model is partially responsible for such things as facial and voice recognition, self-driving cars, and is just beginning to be used in medicine.

IBM’s Watson has trounced Jeopardy contestants, so it is not surprising that a study has already shown that deep learning AI has outperformed clinicians in predicting prognosis and future events in a group of patients with pulmonary hypertension.

An article in the journal JACC Interventions suggests that machine learning can outperform classical statistical approaches to predicting the risk of death and heart failure after a patient has a cardiac procedure such as a stent. An editorial in the same journal points out that equity funding for health care AI ventures will top $2.4 billion in 2018, representing a 78% increase compared with 2017.

AI has already been used to help improve the accuracy of many popular imaging tests, such as stress tests and echocardiograms, but one of the major concerns about increasing use of this technology is privacy.

A device to detect a change in breathing patterns while sleeping could help prevent sudden cardiac death by picking up what is called agonal breathing. This can occur when someone is critically ill from a cardiac cause such as a heart attack or arrhythmia, and is a gasping for air that can precede imminent death. If detected, a call to 911 can be lifesaving. Such breathing can be successfully distinguished from other respiratory issues such as sleep apnea, but the required monitoring raises the question about how much privacy we are willing to trade for our personal safety. This study used smart speakers such as Amazon’s Alexa to detect changes in breathing patterns during the night — something that necessitates the kind of constant observation that is possible only because a smart speaker is listening all the time.

It’s not surprising that, in an effort to make inroads into the $3.5 trillion health care market, Amazon has recently announced the development of HIPAA-compliant programs so this will not be as much of an issue.

Wearable devices are a primitive form of AI. Activity monitors such as Fitbits and smart watches are the tip of the iceberg in this technology. Blood pressure, activity, sleep, hydration and blood sugar can already be checked, but still need better reproducibility before gaining general acceptance. I have suggested that some of my patients use an app called AliveCor, which can detect atrial fibrillation on smartphones or watches, and is becoming increasingly accurate.

Although still in its early days, having a computer scan, diagnose, and recommend treatment will someday come true. We will need to make sure that the human aspect of caring for a patient is not lost along the way.

This feeling may have been best voiced by Dr. McCoy from the original Star Trek, when he said, “Dammit, Jim, I’m a doctor, not a machine. Compassion — that’s the one thing no machine ever had. Maybe it’s the one thing that keeps men ahead of them.”

David Becker, M.D., is a frequent Inquirer contributor and a board-certified cardiologist with Chestnut Hill Temple Cardiology in Flourtown, Pa. He has been in practice for more than 25 years.