Tabea Thies, Elisa Malick, Johannes Tröger, Salvatore Guigliano, Giovanna Sannino, Myriam Spisto, Raffaele Dubbioso
*poster presented at ENCALS 2026
Introduction
Digital speech biomarkers provide a non-invasive window into progression of Amyotrophic lateral sclerosis (ALS), but their utility depends on clinical transparency. We present a robust, low-dimensional monitoring panel that is optimized for clinical interpretability over high-quantity data sets. This streamlined approach focuses on the most sensitive indicators of bulbar decline, offering a tool for (i) identifying ALS and (ii) tracking dysarthria progression in both research and clinical practice.
Methods
Data Collection: In-clinic smartphone recordings (Vox4Health, SAND data set); 165 individuals with ALS and 107 healthy controls (HC); Speech tasks (taken from SAND data): DDK (pa-pa-pa, ta-ta-ta, ka-ka-ka) and Sustained vowel phonation (i, a, u)
Data Processing: Acoustic feature extraction via SIGMA, ki:elements’ processing pipeline using DDK features
Vowel Formants F1 and F2 to calculate: Vowel Space Area (VSA), Vowel Articulation Index (VAI)
Data Analysis: Group differences were computed via Kruskal–Wallis tests. Discriminative performance for i) group membership and ii) speech severity (indexed by ALSFRS-R speech score) were evaluated via ROC and Spearman correlation analyses
Conclusion
A minimal, interpretable feature set captures ALS bulbar impairment with high precision. Our findings reveal a critical functional split in task utility: Composite scores taken from the alveolar DDK task /ta/ were the most effective task, specifically Stability (AUC = .81, Sens = .97, Spec = 0.53) and Articulation (AUC = .79, Sens = .88, Spec = .65). Vowel space metrics demonstrated moderate capability to track progression of dysarthria severity (VAI: AUC = .72, Sens = .84, Spec = .50; VSA: AUC = .71, Sens = .78, Spec = .62). Early Detection: The velar DDK task /ka/ is the most sensitive marker for identifying early bulbar involvement. Progression Monitoring: The alveolar DDK task /ta/, alongside vowel space reduction, serve as superior longitudinal markers of bulbar severity. This task-specific approach enhances clinical transparency and provides a validated, low-burden pathway for integrating digital monitoring into multi-center ALS trials.