Uncertainty of Vowel Predictions as a Digital Biomarker for Ataxic Dysarthria

Dysarthria is a common manifestation across cerebellar ataxias leading to impairments in communication, reduced social connections, and decreased quality of life. While dysarthria symptoms may be present in other neurological conditions, ataxic dysarthria is a perceptually distinct motor speech disorder, with the most prominent characteristics being articulation and prosody abnormalities along with distorted vowels. We hypothesized that uncertainty of vowel predictions by an automatic speech recognition system can capture speech changes present in cerebellar ataxia. Speech of participants with ataxia (N=61) and healthy controls (N=25) was recorded during the "picture description" task. Additionally, participants' dysarthric speech and ataxia severity were assessed on a Brief Ataxia Rating Scale (BARS). Eight participants with ataxia had speech and BARS data at two timepoints. A neural network trained for phoneme prediction was applied to speech recordings. Average entropy of vowel tokens predictions (AVE) was computed for each participant's recording, together with mean pitch and intensity standard deviations (MPSD and MISD) in the vowel segments. AVE and MISD demonstrated associations with BARS speech score (Spearman's rho=0.45 and 0.51), and AVE demonstrated associations with BARS total (rho=0.39). In the longitudinal cohort, Wilcoxon pairwise signed rank test demonstrated an increase in BARS total and AVE, while BARS speech and acoustic measures did not significantly increase. Relationship of AVE to both BARS speech and BARS total, as well as the ability to capture disease progression even in absence of measured speech decline, indicates the potential of AVE as a digital biomarker for cerebellar ataxia.

Computer Vision Analysis of Caregiver-Child Interactions in Children with Neurodevelopmental Disorders: A Preliminary Report

We report preliminary results of computer vision analysis of caregiver-child interactions during free play with children diagnosed with autism (N = 29, 41-91 months), attention-deficit/hyperactivity disorder (ADHD, N = 22, 48-100 months), or combined autism + ADHD (N = 20, 56-98 months), and neurotypical children (NT, N = 7, 55-95 months). We conducted micro-analytic analysis of 'reaching to a toy,' as a proxy for initiating or responding to a toy play bout. Dyadic analysis revealed two clusters of interaction patterns, which differed in frequency of 'reaching to a toy' and caregivers' contingent responding to the child's reach for a toy by also reaching for a toy. Children in dyads with higher caregiver responsiveness had less developed language, communication, and socialization skills. Clusters were not associated with diagnostic groups. These results hold promise for automated methods of characterizing caregiver responsiveness in dyadic interactions for assessment and outcome monitoring in clinical trials.

Attention-Based Network for Weak Labels in Neonatal Seizure Detection

Seizures are a common emergency in the neonatal intesive care unit (NICU) among newborns receiving therapeutic hypothermia for hypoxic ischemic encephalopathy. The high incidence of seizures in this patient population necessitates continuous electroencephalographic (EEG) monitoring to detect and treat them. Due to EEG recordings being reviewed intermittently throughout the day, inevitable delays to seizure identification and treatment arise. In recent years, work on neonatal seizure detection using deep learning algorithms has started gaining momentum. These algorithms face numerous challenges: first, the training data for such algorithms comes from individual patients, each with varying levels of label imbalance since the seizure burden in NICU patients differs by several orders of magnitude. Second, seizures in neonates are usually localized in a subset of EEG channels, and performing annotations per channel is very time-consuming. Hence models which make use of labels only per time periods, and not per channels, are preferable. In this work we assess how different deep learning models and data balancing methods influence learning in neonatal seizure detection in EEGs. We propose a model which provides a level of importance to each of the EEG channels - a proxy to whether a channel exhibits seizure activity or not, and we provide a quantitative assessment of how well this mechanism works. The model is portable to EEG devices with differing layouts without retraining, facilitating its potential deployment across different medical centers. We also provide a first assessment of how a deep learning model for neonatal seizure detection agrees with human rater decisions - an important milestone for deployment to clinical practice. We show that high AUC values in a deep learning model do not necessarily correspond to agreement with a human expert, and there is still a need to further refine such algorithms for optimal seizure discrimination.

Relative Average Look Duration and its Association with Neurophysiological Activity in Young Children with Autism Spectrum Disorder

Autism Spectrum Disorder (ASD) is characterized by early attentional differences that often precede the hallmark symptoms of social communication impairments. Development of novel measures of attentional behaviors may lead to earlier identification of children at risk for ASD. In this work, we first introduce a behavioral measure, Relative Average Look Duration (RALD), indicating attentional preference to different stimuli, such as social versus nonsocial stimuli; and then study its association with neurophysiological activity. We show that (1) ASD and typically developing (TD) children differ in both (absolute) Average Look Duration (ALD) and RALD to stimuli during an EEG experiment, with the most pronounced differences in looking at social stimuli; and (2) associations between looking behaviors and neurophysiological activity, as measured by EEG, are different for children with ASD versus TD. Even when ASD children show attentional engagement to social content, our results suggest that their underlying brain activity is different than TD children. This study therefore introduces a new measure of social/nonsocial attentional preference in ASD and demonstrates the value of incorporating attentional variables measured simultaneously with EEG into the analysis pipeline.

Diffusion-tensor imaging of major white matter tracts and their role in language processing in aphasia

A growing literature is pointing towards the importance of white matter tracts in understanding the neural mechanisms of language processing, and determining the nature of language deficits and recovery patterns in aphasia. Measurements extracted from diffusion-weighted (DW) images provide comprehensive in vivo measures of local microstructural properties of fiber pathways. In the current study, we compared microstructural properties of major white matter tracts implicated in language processing in each hemisphere (these included arcuate fasciculus (AF), superior longitudinal fasciculus (SLF), inferior longitudinal fasciculus (ILF), inferior frontal-occipital fasciculus (IFOF), uncinate fasciculus (UF), and corpus callosum (CC), and corticospinal tract (CST) for control purposes) between individuals with aphasia and healthy controls and investigated the relationship between these neural indices and language deficits. Thirty-seven individuals with aphasia due to left hemisphere stroke and eleven age-matched controls were scanned using DW imaging sequences. Fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD) values for each major white matter tract were extracted from DW images using tract masks chosen from standardized atlases. Individuals with aphasia were also assessed with a standardized language test in Russian targeting comprehension and production at the word and sentence level. Individuals with aphasia had significantly lower FA values for left hemisphere tracts and significantly higher values of MD, RD and AD for both left and right hemisphere tracts compared to controls, all indicating profound impairment in tract integrity. Language comprehension was predominantly related to integrity of the left IFOF and left ILF, while language production was mainly related to integrity of the left AF. In addition, individual segments of these three tracts were differentially associated with language production and comprehension in aphasia. Our findings highlight the importance of fiber pathways in supporting different language functions and point to the importance of temporal tracts in language processing, in particular, comprehension.