Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study introduces a novel subject-level graph framework that utilizes self-supervised speech embeddings to analyze multiple phonation recordings for detecting and predicting the progression of amyotrophic lateral sclerosis (ALS). By comparing various self-supervised learning (SSL) models and graph neural networks (GNNs) on the SAND dataset, the researchers achieved significant improvements in dysarthria severity and ALSFRS-R progression prediction, with the best configuration (HuBERT+GIN) yielding macro-F$_1$ scores of 0.73 and 0.69, respectively. These findings underscore the effectiveness of integrating GNNs with pretrained speech representations in a low-resource setting for ALS monitoring.
Combining graph neural networks with self-supervised speech embeddings can significantly enhance ALS detection and progression prediction, achieving state-of-the-art results.
Amyotrophic lateral sclerosis (ALS) progressively impairs speech motor control, making acoustic analysis a promising biomarker for severity and progression estimation. We propose a subject-level graph framework that aggregates multiple phonation recordings into a unique k-nearest-neighbor graph built from pretrained SSL embeddings of 2s segments. We compare four SSL front-ends (wav2vec 2.0, HuBERT, data2vec-audio, and UniSpeech-SAT) and five graph neural networks (GCN, residual GCN, GAT, GraphSAGE, and GIN) on the SAND dataset tasks (339 participants: 205 ALS, 134 control): 5-class dysarthria severity and 4-class ALSFRS-R progression prediction. On the official validation set, the best configuration (HuBERT+GIN) achieves macro-F$_1$ of 0.73 for Task 1 and 0.69 for Task 2, outperforming SAND validation baselines (0.61 and 0.58). These results highlight the potential of combining GNNs with pretrained cross-lingual speech representations for low-resource ALS detection and progression monitoring.