25304
Physiological Measurements of Voice Quality in Children with Autism Using Electroglottography in Relation to Clinical Assessment Outcome
Atypical voice quality has been a characteristic trait in individuals with autism. Many standard clinical assessment tools include subjective evaluation of voice characteristics and social communication skills for clinical diagnosis of autism. However, a means of statistically quantifying atypicality in voice production in individuals with autism that can also contribute to reliable prediction of clinical diagnosis is still lacking. In previous research, we showed that longitudinal acoustic measures of infant vocal behavior including the fundamental frequency contour, duration and timing of dyadic interactions with caregivers within the first two years of life have potential as biomarkers for early detection of autism. However, acoustic measures of prosody do not directly quantify the actual mechanism of voice production. Electroglottography (EGG) provides non-invasive physiological measures of vocal fold function, and voice source parameters obtained through EGG signal processing may therefore provide us with better information about the origin of voice disorders in autism.
Objectives:
The goal of this study is to explore the potential of physiological measures of voice quality to capture information relevant to diagnostic characterization by correlating acoustic measures of voice quality obtained using EGG with standard measures used for clinical assessment of language and social communication in autism spectrum disorders (ASD). We test the hypothesis that objective measures of abnormal voice quality characteristics are predictive of clinical outcome.
Methods:
As part of our initial pilot study, we recruited in total 8 low-risk children with no history of autism and high-risk children with older siblings diagnosed with ASD. We collected high-quality EGG and microphone recordings of each child at 2-3 years of age. We also collected a battery of clinical assessment measures from each child at the same age. From the EGG recordings, we hand-labeled and extracted sequences of utterances containing clean child vocalizations and calculated the mean and standard deviations for four measures of voice quality across all speech frames for each child: the fundamental frequency (F0), the open quotient (OQ), the return quotient (RQ), and the speed quotient (SQ). We then compared the correlation between physiological measures of voice quality with clinical outcome measures corresponding to social communication and motor control.
Results:
Our final sample consisted of 4 TD children and 4 children diagnosed with ASD. Group differences in all of the four acoustic measures of voice quality showed significant correlation with each other and may have potential to aid prediction of diagnostic characterization. Group-level categorizations based on these physiological measurements of voice quality were consistent with categorizations based on the ADOS summary scores.
Conclusions:
Preliminary results suggest that typically developing children and children with ASD differ according to physiological measures of atypical voice quality which may be related to clinical outcome measures. However, these observations and results are currently being further verified on a larger cohort.