Artificial intelligence outperforms pulmonologists in the interpretation of pulmonary function tests.If you did not see this article with the title above, then you should go read it now. It was published in the April 2019 edition of the European Respiratory Journal. The authors stated that their aim was “to explore the accuracy and interrater variability of pulmonologists when interpreting PFTs compared with artificial intelligence (AI)-based software that was developed and validated in more than 1500 historical patient cases.” In this study, 120 pulmonologists from 16 European hospitals evaluated 50 cases comprising PFTs and clinical information. They used American Thoracic Society/European Respiratory guidelines as the gold standard for interpretation of PFTs. What did they find? The pulmonologists matched the guidelines in PFT interpretation in 74.4% of the readings and got 44.6% of the clinical diagnoses correct with high inter-rater variability. The AI software was 100% correct on interpretation of PFTs and a correct diagnosis was obtained in 82% of the cases. Who needs a pulmonologist, or for that matter, an allergist?
Should you be worried? Will AI replace the need for an allergist? I don’t think so, but I think you need to get ready for the changes that will occur in allergy as the use of Al in patient care increases. We will be seeing the move to smarthalers which can improve adherence in our patients with asthma and COPD, along with monitoring whether patients are using their inhalers correctly. As data is analyzed from these smarthalers, it will help patients understand their control and possible triggers. Data collected by the Propeller Health smarthaler device, now owned by ResMed, can “see” where patients are using their rescue inhalers and help predict environmental areas where asthma may more likely be exacerbated.
More insights from the EMD:
Have you heard of NuvoAir? This Swedish company has a fully automated treatment plan for asthma using a handheld wireless portable spirometer that can connect to the patient’s smartphone or tablet. Data can be shared with their physicians. They also have an app called Aria which acts as a digital lung health assistant that helps the patient perform the spirometric maneuver and interpret the data. Patients can monitor their control over time and take a more active role in their care. So far, this company has collected more than 500,000 spirometry readings. This data can be used for machine learning algorithms to improve insights into better patient care.
In April of this year, the FDA released a white paper on establishing oversight of AI systems being developed in medicine. This can be very difficult, as more and more data are collected on a condition which may change the machine-learning algorithm in diagnosis and/or treatment. Presently the FDA has approved AI medical devices whose algorithms are locked or do not change each time an algorithm is used. Just last year, the FDA approved a device to detect diabetic retinopathy and one with clinical decision support software analyzing CT results to notify a physician of a potential stroke in their patients.
Eric Topol, MD, who is a world-renowned cardiologist at Scripps, just published a book entitled “Deep Medicine, How Artificial Intelligence can Make Healthcare Human Again.” He believes that AI will make medicine more efficient, allowing doctors to spend more time with patients. I hope he is right. The future is now.
Michael Blaiss, MD, FACAAI, Executive Medical Director
