Nary a conference in almost any field of medicine can take place without some forum or symposium on the utility of novel medical devices, digital health software or artificial intelligence to the specialty or topic at hand. There is undoubtedly great promise in these domains, and they promise to transform medicine in the near future.
By in large, the greatest promise that medical technology enthusiast see for artificial intelligence is in the interpretation of static information; Zebra Medical Technologies, for example, has demonstrated great promise in the interpretation of radiology studies. AI driven software program promise to assist in numeric data interpretation as well. The MySpecialistMD Network Chronic Disease Management Program utilizes such an engine, which is perfectly suited to the analysis and synthesis of millions of simple data points, processing the information to provide predictive alerts and in depth trend analytics.
Medical technology enthusiasts have been struggling with utilizing AI for dynamic images – such as cardiac catheterization images or echocardiogram sonograms, which don’t naturally lend themselves to AI engine processing.
Pulmonary Function Testing would seemingly be a simple and intuitive place for AI interpretation to have great utility. It is a self – contained unit, with only a (relatively) small number of important values, with excellent and precise clinical guidelines to guide interpretation. In the most recent edition of the European Respiratory Journal (https://erj.ersjournals.com/content/53/4/1801660?etoc), an AI engine significantly outperformed physician clinical judgment in the interpretation of 50 clinical cases (6,000 total interpretations). The AI engine perfectly matched guideline pattern interpretation (100%) vs. 75% for physicians. More importantly, whereas the pulmonologists arrived at the correct interpretation a mere 45% of the time, the AI engine arrived at the correct diagnosis in 82% of cases.
The results of this comparison, namely the software’s ability to double the diagnostic capability, is outstanding. This study not only validates the utility of AI in interpreting PFT, but compels physicians to utilize decision making tools – such as the one it should featured in this study – to improve clinical performance and patient care. Let us hope that the medical community rises to the challenge!