Computer Science & AI11 November 2025

Non-Invasive EEG System Translates Inner Speech for ALS Patients

Source PublicationN/A

Primary AuthorsSteiner

Visualisation for: Non-Invasive EEG System Translates Inner Speech for ALS Patients
Visualisation generated via Synaptic Core

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease that progressively paralyzes speech and motor functions, eventually rendering patients speechless when voluntary movement is lost. While invasive brain–computer interfaces (BCIs) such as electrocorticography (ECoG) have achieved high accuracy in decoding speech-related neural activity, their clinical use is limited by surgery costs and risks. Addressing this need, researchers present a new, fully non-invasive EEG-based inner-speech translation system aimed at communication restoration in ALS.

The core of this development is a novel machine learning architecture. It combines a fusion of region-aware convolutional attention encoders and a transformer decoder, presenting a generative sequence-to-sequence model for open-vocabulary EEG-to-text translation. To train this sophisticated system, preprocessed and downsampled EEG data from the Chisco imagined-speech dataset were used. A crucial step involved selecting 48 optimized electrode channels based on contribution-based ranking, which helps capture relevant neural signals. Parameter-efficient fine-tuning (LoRA, PEFT) was also employed to train the model to preserve linguistic fluency while specializing it for EEG input.

Evaluation of this system demonstrated significant progress, achieving a BLEU-1 score of 0.512 $\pm$ 0.012 and a ROUGE-L F1 score of 0.396 $\pm$ 0.009, thereby surpassing earlier non-invasive baselines. Ablation studies further confirmed the critical role of cross-regional fusion and diversity regularization in maintaining cortical interpretability during the encoding process. These studies validate the design choices made for the model.

As lead author Steiner notes in the paper, "To the best of our knowledge, this is the first open-vocabulary, non-invasive EEG-to-text system capable of reconstructing continuous inner speech with high linguistic coherence and accuracy." These findings represent a key step toward clinical neural speech restoration for ALS, facilitating the development of future real-time, low-cost assistive communication devices.

Cite this Article (Harvard Style)

Steiner (2025). 'Non-Invasive EEG System Translates Inner Speech for ALS Patients'. N/A. Available at: https://doi.org/10.31234/osf.io/g34v7_v1

Source Transparency

This intelligence brief was synthesised by The Synaptic Report's autonomous pipeline. While every effort is made to ensure accuracy, professional due diligence requires verifying the primary source material.

Verify Primary Source
ALSEEGInner SpeechBrain-Computer Interface