Artificial Intelligence (AI) is a hot, and controversial, topic of conversation. Its applications in medicine are in the early stages, however, AI is not new. In 1950, a paper was published titled “Computer Machinery and Intelligence.” In 1952 Arthur Samuel, a computer scientist, developed a program which could play checkers independently. The term artificial intelligence was coined in 1955 at a meeting known as the Dartmouth Workshop. Leaders in computer science discussed how to create machines that could perform tasks requiring human intelligence. There has been a long period of AI research and growth¹. Currently, there is great interest in how to apply AI in a variety of scenarios, including medicine.

Adaption of AI in the pulmonary medicine space has been slow. The potential benefits of utilizing AI in pulmonary medicine are great. AI can assist in patient-specific care plans providing real-time feedback for providers. Various AI algorithms have been shown to detect respiratory conditions at early stages allowing for interventions and improved patient outcomes. AI can be utilized in the ICU with ventilator monitoring. The healthcare team can use AI to enhance their expertise in determining the correct diagnosis and management of their patients.²

One area where AI may prove particularly effective is in pulmonary diagnostics, specifically Pulmonary Function Testing (PFT). PFTs are numerical values assigned to various maneuver outcomes, making them ideal for machine learning.³ This is certainly not to suggest that AI can replace the Respiratory Therapist in the PFT lab. Not at all. However, AI can assist with workloads and guide in challenging diagnoses. It has been shown that there is poor accuracy and disagreement between pulmonologists in PFT interpretation. AI can serve as a support tool for interpreting the data gathered.  The European Respiratory Society (ERS) published a paper highlighting this, Artificial intelligence outperforms pulmonologists in the interpretation of pulmonary function tests, 2019.⁴ In this study, 120 pulmonologists from 16 hospitals in Europe evaluated 50 PFTs and clinical interpretations.  In all there were 6000 interpretations. AI also reviewed the same data. The pulmonologists matched ERS/ATS pattern recognition in 74.4±5.9% and made the correct diagnosis in 44.6±8.7% of the tests.  An AI-based software program matched PFT pattern interpretation 100% and found the correct diagnosis in 82% of the cases.⁴ PFT interpretation is not as simple as it might seem. The interpreting physician needs to have expertise in pattern recognition, grading of severity, and knowledge of international guidelines. Information on the patient’s history, symptoms, and any other tests that have been performed should also be considered.⁴ When using a simplified algorithm to assess lung function, physicians reported the correct diagnosis in 38% of patients. Adding patient data to the algorithm increased the accuracy to 68%⁴.  This highlights how important the whole patient picture is when determining a diagnosis. A few benefits of using AI-based software for interpretation include immediate and consistent interpretations, as well as a higher percentage of correct diagnostic categories (82%). AI was highly sensitive in determining COPD, neuromuscular disease, ILD, and healthy subjects.⁴ A 2023 ERS study, demonstrated that accuracy of PFT interpretation improved when pulmonologists and AI were used in tandem.⁵ When pulmonologists utilized AI support in interpretation, they outperformed AI by itself in accuracy.⁵ The systems are only as good as the training data received. AI needs to have diverse populations represented to avoid biases or create further health-care disparities.²  Similar to human learning, AI software improves over time with exposure to patient cases.

AI is available to use in the interpretation of other tests, for example ECG, mammograms, and to classify skin cancer.  However, there remains variability in the adoption and utilization of the tool throughout all areas of healthcare. While AI has a place in medicine, there are also ethical considerations to address. To get “smarter” AI collects patient data and this may increase the risks of data breaches and unauthorized access to private information. It is essential that protocols around privacy are robust and followed^.2 The ethics surrounding AI need to be addressed by not only the healthcare practitioners, but also policy and law makers.

Additionally, AI cannot take the place of human interaction and the compassion of the practitioner. The empathy, understanding, and personalized care that a human practitioner provides are irreplaceable. These qualities ensure that patients receive not only accurate diagnoses but also the emotional support and genuine connection that only a human can offer. In a world increasingly driven by technology, the human touch remains essential in delivering holistic and compassionate healthcare.

Heather Murgatroyd, BA, RRT, CPFT, AE-C 

Senior Clinical Specialist 

Methapharm Respiratory 

1. tableau.com
2. Al‑Anazi S, Al‑Omari A, Alanazi S, Marar A, Asad M, Alawaji F, et al. Artificial intelligence in respiratory care: Current scenario and future perspective. Ann Thorac Med 2024;19:117-30
3. Chauhan NK, Asfahan S, Dutt N, Jalandra RN. Artificial intelligence in the practice of pulmonology: The future is now. Lung India 2022;39:1-2
4. Topalovic M, Das N, Burgel P-R, et al. Artificial intelligence outperforms pulmonologists in the interpretation of pulmonary function tests. Eur Respir J 2019; 53: 1801660 [https:// doi.org/10.1183/13993003.01660-2018].
5. Das N, Happaerts S, Gyselinck I, et al. Collaboration between explainable artificial intelligence and pulmonologists improves the accuracy of pulmonary function test interpretation. Eur Respir J 2023; 61: 2201720 [DOI: 10.1183/13993003.01720-2022]