Respiratory System|Detecting respiratory diseases using machine learning-based pattern recognition on spirometry data

An article by Taloba et al. (2024) emphasized the global health burden of chronic obstructive pulmonary disease (COPD). It is said that about 80 million people have moderate to severe COPD, which accounted for 5 % of total mortality in 2005. Hence, Early diagnosis is crucial for catching diseases early. Meanwhile, Traditional tests like spirometry have issues with sensitivity and specificity. Researchers used machine learning (ML) and deep learning (DL) to analyze lung sounds. They focused on a technique called MFCCs to pick out important sound features and combined two ML methods, SVM and KNN, to create a highly accurate model. This innovative approach achieved an impressive accuracy of 94%, which signifies its potential in detecting COPD.
The new model is better than older methods because it finds small patterns in lung sounds that might be missed otherwise. It also outperformed other ML models in terms of accuracy. The study further evaluates performance metrics, such as precision and recall, establishing the hybrid model as a superior choice among alternatives like CNN and decision tree methods.
Despite its success, the research identifies limitations, including dependency on specific datasets. However, expanding datasets, incorporating real-time technologies, and collaborating with doctors will make the model more practical in clinical settings. This improves patient care and healthcare efficiency.

Link to article :
https://doi.org/10.1016/j.aej.2024.11.009