Investigating Velopharyngeal Closure Force with Linear Regression

Abstract

Cleft palate is a common birth defect worldwide. Children diagnosed with this abnormality face difficulties during feeding, hearing, and especially speech. Although surgical methods exist to repair cleft palate, subsequent corrective surgeries are often necessary since children are unable to gain full speech capabilities due to velopharyngeal inadequacy. Investigating the velopharyngeal system in normal patients can help speech pathologists, surgeons, and other medical professionals understand the effects of velopharyngeal anatomy on velopharyngeal function and improve patient diagnosis and treatment. Earlier studies visualized the velum using two-and-three dimensional modeling, but these studies did not adequately investigate the variability in velopharyngeal muscle measures nor their impact on normal and abnormal velopharyngeal function. To remedy these shortcomings, this paper investigates the effects of muscles in the velopharyngeal system on closure force with a novel application of the multiple linear regression machine learning technique. Incorporating multiple anatomical features, multiple linear regression was used to predict closure force values and their direction. The results of this study reveal that multiple linear regression was found to be an effective tool for accurately predicting velopharyngeal closure force for any set of anatomical parameters. Furthermore, these results demonstrated that the velum had a major influence on closure force challenging previous claims that the levator veli palatini muscle was responsible for generating closure force.