Automatic detection of ALS from single-trial MEG signals during speech tasks: a pilot study

Area of Science:

• Neuroscience
• Biomarker Discovery
• Neurology

Background:

• Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease with delayed diagnosis due to lack of biomarkers.
• Impaired speech is common in ALS, suggesting potential neural correlates during speech tasks.

Purpose of the Study:

• To investigate functional cortical pattern changes during speech tasks in ALS patients.
• To identify a potential neural biomarker for early ALS detection using magnetoencephalography (MEG).

Main Methods:

• Collected MEG data from three early-diagnosed ALS patients and three healthy controls during imagined and overt speech.
• Analyzed sensor correlations, spectral power in canonical bands, and functional connectivity.
• Performed single-trial classification using MEG features.

Main Results:

• ALS participants showed greater sensor correlations and beta band dissimilarity compared to controls.
• Increased beta band functional connectivity was observed in ALS patients.
• Single-trial classification achieved high accuracy (~98%) using beta band features.

Conclusions:

• Speech-evoked beta oscillations show potential as a neural biomarker for diagnosing ALS.
• This study demonstrates the first detection of ALS from single-trial neural signals.