Multimodality analysis of vibrations in a biomechanical model of vocal folds
Babak Aghazadeh (Mass. Eye & Ear Infirmary), Ivo Dobrev (Worcester Polytechnic Institute), Cosme Furlong (Worcester Polytechnic Institute), Ramon Franco (Massachusetts Eye and Ear Infirmary)
Characterization and Imaging of Structural and Material Imperfections
Mon 2:40 - 4:00
Barus-Holley 191
Proper medical care of voice disorders requires ways to assess the severity of vocal fold pathologies. Although acoustic and endoscopic imaging measurements have been made for many years, the voice is the product of a complex mechanical and neurological vibrating system, making interpretation of the acoustic signals or images in isolation difficult. Thus, a multimodality approach is proposed that would be sensitive enough to changes in the vibration pattern of disordered vocal folds to assist physicians during the diagnosis and monitoring of patients’ healing process. In this paper, a biomechanical model of larynx is designed and built to simulate the natural vocalization process and characterize common vocal disorders such as vocal nodules and polyps, caused by mass lesions grown on either or both vocal folds. The model includes hyperelastic structures that vibrate and produce sound in response to airflow. Using LabVIEW, multi-domain measurements are acquired by a microphone, high-speed camera, and laser Doppler vibrameter, simultaneously. Then, nonlinear features including fractal dimensions, self-similarity, and wavelet transformation of acoustic signals and vocal fold displacements are extracted using MATLAB. Also, image features including speed quotient, amplitude periodicity index, and left-right amplitude asymmetry are extracted from segmented frames of high-speed recordings. Using Support vector regression and genetic algorithm techniques, an optimal feature vector is selected from the multi-domain set of features to quantitatively characterize the imperfections (the lesion size) in our mechanical model of larynx. The results of this study accentuate the importance of multimodality monitoring of vocal fold vibrations to characterize the vocal fold disorders. It is shown that the combination of information from multiple sources can lead to a substantial improvement in the characterization of vocal fold disorders by quantification of the size of lesions.