The ISCSLP 2024 Multimodal Dysarthria Severity Assessment (MDSA) Challenge: Dataset, Tracts, Baseline and Results

To advance multimodal speech assessment and related research in developing objective diagnostic methods, we are launching the Multimodal Dysarthria Severity Assessment (MDSA) Challenge. This paper summarizes the outcomes from the ISCSLP 2024 MDSA Challenge. We first address the necessity of the challenge and then introduce the associated audio-video dataset selected from the MSDM database, including 62 subacute stroke patients and 25 normal controls. We then describe the challenge arrangement and the baseline system. Specifically, we set up a four-classification task for the severity of dysarthria (normal, mild, moderate and severe) with the aim of developing an objective and accurate automatic assessment method to assist in clinical diagnosis and treatment. Finally we summarize the challenge results and provide the major observations from the submitted systems. We hope the open data and challenge will serve as a benchmark and common test-bed for pathological speech assessment.