Evaluating artificial intelligence-interpreted against pulmonologist-interpreted spirometry results: a comparative study

<bold>Introduction:</bold> AI enhances healthcare by improving diagnostics and treatment planning. <bold>Objective:</bold> To compare the spirometry reading interpretation between AI and pulmonologists in terms of agreement and reliability <bold>Methodology:</bold> In a comparative study at a private hospital in Cebu City, 197 patients were assessed using spirometry with bronchodilator studies to compare AI's accuracy to pulmonologists. The significance level was p<0.05. <bold>Results:</bold> Participants' average age was 48.70, with a balanced gender ratio, mostly overweight and non-smokers. AI and pulmonologists had similar findings in normal (AI 50.76%, pulmonologists 55.1%-63.96%), restrictive (AI 19.29%, pulmonologists 15.74%-19.29%), and obstructive (AI 18.78%, pulmonologists 15.82%-20.30%) ventilatory defects, but differed in mixed defects and severity classification. AI's identification of significant bronchodilator response (15.23%) was within pulmonologists' range (5.58%-12.69%). Agreement between AI and pulmonologists varied, being strong with one but moderate with others, especially on mixed defects. <bold>Conclusion:</bold> AI is reliable and comparable to pulmonologists in spirometry interpretation, with varying agreement levels.