ARTIFICIAL INTELLIGENCE AND VOCAL BIOMARKERS IN ASTHMA DIAGNOSIS AND MONITORING: A NARRATIVE REVIEW
Abstract
Considering that chronic respiratory diseases, such as asthma, represent a global challenge, their continuous monitoring is often limited by the need for specialized equipment. Artificial Intelligence (AI) applied to the acoustic analysis of cough and voice emerges as a non-invasive alternative. It aims to investigate how vocal biomarkers can be isolated for the specific detection and monitoring of asthma. To this end, we proceed to a narrative review that synthesizes recent literature on the application of machine learning algorithms. In this way, it is observed that changes in respiratory mechanics modify extractable acoustic parameters, such as MFCCs. Several architectures demonstrate high accuracy, frequently exceeding 85% to 90% by fusing spectral and temporal features. Data augmentation tools overcome sample scarcity, and advanced filters mitigate ambient noise on smartphones. This allows us to conclude that AI-processed vocal biomarkers hold significant clinical potential for outpatient screening. The standardization of recording protocols and validation in multicenter cohorts are essential for the safe integration of this technology into telemedicine.