Statistical and clinical utility of multimodal dialogue-based speech and facial metrics for Parkinson's disease assessment

Multimodal Speech Biomarkers for Remote Monitoring of ALS Disease Progression

“…In particular, facial video metrics showed diminished utility relative to speech metrics. Despite previously presented evidence for the utility of these facial metrics in cross-sectional studies, in ALS and other disorders [26,27,76,77,75,74], they were not as responsive to longitudinal change as speech metrics in the specific cohorts investigated in this work, suggesting that these metrics are comparatively less robust than speech metrics for characterizing the strongly heterogeneous nature of ALS disease progression [78]. Some features that were selected during feature selection had an good AUC for discriminating bulbar onset from non-bulbar onset pALS, but were not responsive for longitudinal change, e.g., DDK cTV and SIT maximum eyebrow displacement.…”

“…The MCID is the smallest domain-specific change that is considered to be clinically relevant [65]. It can be quantified as a threshold for a change corresponding to clinical improvement or deterioration [35] and is tied to an external anchor, which is considered to be a clinical gold standard, the ALSFRS-R speech question in this case. We calculated the MCID for all features for a corresponding one-point change on the ALSFRS-R speech question where participants are asked to rate their speech on the following scale with scores in parentheses: Normal speech processes (4) Detectable speech disturbance (3) Intelligible with repeating (2) Speech combined with nonvocal communication (1) Loss of useful speech (0) One approach to derive the MCID is using data-driven ROC analysis [66], which was also applied in [34].…”

“…The MCID is the smallest domain-specific change that is considered to be clinically relevant [65]. It can be quantified as a threshold for a change corresponding to clinical improvement or deterioration [35] and is tied to an external anchor, which is considered to be a clinical gold standard, the ALSFRS-R speech question in this case. We calculated the MCID for all features for a corresponding one-point change on the ALSFRS-R speech question where participants are asked to rate their speech on the following scale with scores in parentheses:…”

“…Prior work by us [30] has shown that some timing-related speech biomarkers, collected remotely through a conversational dialog platform, have the requisite responsiveness and sensitivity to track speech decline in the context of clinical interventional trials targeting neurodegenerative disorders. To establish the efficacy of multimodal biomarkers in tracking disease progression, it is important to consider what constitutes a minimal clinically-important difference (MCID) [33, 34, 35] and whether these biomarkers show change greater than any measurement errors. It is important that these multimodal biomarkers are also sensitive in detecting bulbar decline, which could be well before corresponding changes are observed in the relevant ALSFRS-R functional scores or equivalent clinical scales.…”

A multimodal dialog approach to mental state characterization in clinically depressed, anxious, and suicidal populations

“…For the present study, the Modality service, a cloud-based MDS (Suendermann-Oeft et al, 2019;Ramanarayanan et al, 2020) was used to conduct automated, structured interviews with participants. Neumann et al (2020) recently demonstrated the utility of the Modality MDS in differentiating people with mild, moderate and severe depression, and similar studies have also been conducted in ALS (Neumann et al, 2021), Parkinson's disease (Kothare et al, 2022), schizophrenia (Richter et al, 2022), and autism (Kothare et al, 2021). The Modality MDS can be used with widely available endpoints such as smartphones and laptops as opposed to the dedicated, locally administered hardware used in other studies.…”