Quantification of Tongue Motor Dysfunction in Amyotrophic Lateral Sclerosis Using a Smartphone-Based Task and Deep Learning.

BACKGROUND

Bulbar dysfunction is a major complication of amyotrophic lateral sclerosis (ALS). This study aimed to develop and validate a simple, smartphone-based task for the objective assessment of tongue movements and to examine their association with clinical variables.

METHODS

37 ALS patients and 20 age- and sex-matched controls performed a tongue lateralization task, recorded with a smartphone. A deep-learning U-Net++-based model was used for segmentation and feature extraction. The frequency and maximum amplitude of tongue movements were quantified. Clinical measures included the ALS Functional Rating Scale-revised (ALSFRS-r) bulbar sub-scores, tongue fasciculations, jaw jerk, and tongue "spasticity". Between-group differences and associations between tongue metrics and clinical features were assessed.

RESULTS

The U-Net++-based model achieved robust segmentation performance. Patients showed lower tongue movement frequency than controls (0.14 vs. 0.40, t = -9.58, p < 0.001). Normalized frequency was associated with dysarthria (t = -3.13, p = 0.003) but not dysphagia (t = -1.05, p = 0.30). Normalized frequency (t = 2.77, p = 0.009) and tongue "spasticity" (t = -2.57, p = 0.015) were both associated with speech performance in a multiple-regression model (R = 0.51, adjusted R2 = 0.43).

CONCLUSIONS

Our method provides an objective, minimally invasive measure of bulbar function in ALS, which correlates with clinical ratings and may detect subtle impairments not captured by standard assessments. This approach offers a promising tool for remote monitoring and may support more effective disease management.