Automatic vocalisation-based detection of fragile X syndrome and Rett syndrome

Florian B Pokorny1,2,3, Maximilian Schmitt4, Mathias Egger5

  • 1iDN - interdisciplinary Developmental Neuroscience, Division of Phoniatrics, Medical University of Graz, Graz, Austria. florian.pokorny@medunigraz.at.

Scientific Reports
|August 3, 2022
PubMed

Insights

Early detection of Fragile X syndrome (FXS) and Rett syndrome (RTT) may be possible through infant vocalization analysis. Acoustic features can help identify these developmental disorders sooner, enabling timely intervention.

Area of Science:

  • Developmental Neuroscience
  • Speech-Language Pathology
  • Computational Linguistics

Background:

  • Fragile X syndrome (FXS) and Rett syndrome (RTT) are typically diagnosed in toddlerhood.
  • Speech-language deficits are key symptoms, but early acoustic vocalization patterns remain understudied.
  • Earlier identification can significantly improve intervention and family support.

Purpose of the Study:

  • To investigate the feasibility of using intelligent audio analysis for early identification of FXS and RTT.
  • To analyze acoustic features of infant vocalizations in relation to FXS and RTT.
  • To develop computational methods for distinguishing affected infants from typically developing peers.

Main Methods:

  • Collected 4454 home-recorded vocalizations from infants aged 6-11 months (3 FXS, 3 RTT, 6 typically developing controls).
  • Extracted 88 standardized acoustic features from vocalizations.
  • Trained linear kernel support vector machines for classification tasks (FXS vs. TD, RTT vs. TD, atypical vs. TD, FXS vs. RTT vs. TD).

Main Results:

  • Perfect classification accuracy was achieved for FXS vs. TD, RTT vs. TD, and atypical development vs. TD.
  • Classification accuracy for FXS vs. RTT vs. TD was 75% (9 out of 12 infants).
  • Spectral/cepstral and energy-related acoustic features were most crucial for accurate classification.

Conclusions:

  • Infant vocalization acoustics show distinct patterns in FXS and RTT.
  • Intelligent audio analysis holds promise for earlier diagnosis of these developmental disorders.
  • This research provides a foundation for developing automated tools for early detection and intervention.

Related Concept Videos