Photo Credit: Toowongsa Anurak

In a recent study, a team of investigators aimed to use machine learning models to enhance the diagnostic accuracy of the methacholine challenge test (MCT) for asthma. Noeul Kang and colleagues analyzed data from 1,501 patients with asthma symptoms and compared conventional MCT criteria (PC20 ≤16 mg/mL) with machine learning models developed from the following five methods: logistic regression, support vector machine, random forest, extreme gradient boosting, and artificial neural network. The models used patient lung function parameters obtained during the MCT. All machine learning models outperformed the conventional criteria, with random forest exhibiting the highest accuracy. When incorporating FEV₁, forced vital capacity (FVC), and forced expiratory flow at 25%-75% of FVC (FEF_25%–75%) values, random forest achieved the highest area under the receiver operator characteristic curve and area under the precision-recall curve. The researchers concluded that AI-based models demonstrated superior performance in asthma prediction, offering a promising approach to enhance clinical diagnosis beyond the traditional MCT criteria.