A Machine Learning Approach to Voice-Based Parkinson Disease Screening Using Multiview Spectrogram and Speech Recognition Features: Diagnostic Study

1

Parkinson disease manifests with motor and nonmotor symptoms, necessitating accurate health informatics for early detection and intervention.
2

Speech impairments in Parkinson disease include reduced vocal loudness and imprecise articulation, often emerging before motor signs.
3

Recent studies suggest that combining multiple spectrogram representations can enhance the accuracy of voice-based Parkinson disease classification.
4

The study proposes a multiview spectrogram-based deep architecture that integrates different spectrogram types and a recognition ratio for improved screening.
5

Model evaluation is conducted using participant-wise cross-validation to ensure strict separation of speakers between training and evaluation sets.

Jmir Medical Informatics

1