AI voice analysis detects asthma with 85% accuracy

Asthma affects over 260 million people globally, but diagnosis still depends on spirometry and specialist assessment—limiting access in primary care and low-resource settings. Vocal biomarkers have shown promise as a non-invasive alternative, but prior studies focused narrowly on acoustic features without integrating clinical context. A new paper from researchers at the Luxembourg Institute of Health (Despotovic et al.) introduces a multimodal Mixture-of-Experts framework that adaptively combines acoustic embeddings from sustained vowel phonation and reading passage tasks with structured clinical and demographic data. The model was trained and evaluated on a matched cohort of 1,218 asthma cases and healthy controls from the Colive Voice study.

The multimodal model achieved an AUROC of 0.85 and a Brier score of 0.17, outperforming both unimodal (audio-only or clinical-only) and bimodal approaches. Adaptive gating analysis revealed a key insight: the model relied more heavily on audio features in participants with greater respiratory symptom burden, while clinical features contributed more strongly in less symptomatic individuals. This explainability opens the door for scalable, smartphone-based asthma screening that adapts to each patient's context. The findings suggest that voice recordings—easily collected via a smartphone—could complement traditional diagnostics, especially in underserved areas where spirometry is unavailable.