Alexander Mainka1, Ivan Platzek2, Willy Mattheus3, Mario Fleischer4, Ann-Sophie Müller5, Dirk Mürbe3. 1. Division of Phoniatrics and Audiology, Department of Otorhinolaryngology, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany; Voice Research Laboratory, Hochschule für Musik Carl Maria von Weber, Dresden, Germany. Electronic address: alexander.mainka@uniklinikum-dresden.de. 2. Department of Radiology, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany. 3. Division of Phoniatrics and Audiology, Department of Otorhinolaryngology, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany; Voice Research Laboratory, Hochschule für Musik Carl Maria von Weber, Dresden, Germany. 4. Department of Otorhinolaryngology, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany. 5. Division of Phoniatrics and Audiology, Department of Otorhinolaryngology, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany.
Abstract
OBJECTIVES: The extraction of a three-dimensional (3D) morphology of the human vocal tract (VT) from magnetic resonance imaging (MRI) during sustained phonation can be used for various analyses like numerical simulations or creating physical models. The precision of visualizing techniques nowadays allows for very targeted acoustical simulation evaluating the influence of subsections of the VT for the transfer function. The aim of the study was to assess the accuracy of the 3D geometry based on MRI data in repetitive trials. STUDY DESIGN: This is a prospective study. METHODS: Four experienced singers underwent an MRI while repeating a specific vocal task 20 times consecutively. Audio recordings were made by means of an optical microphone. Images were restacked and subsections of the VT were segmented on multi-image-based cross sections using a semiautomatic algorithm. Different volume and area measures were evaluated. RESULTS: A high reproducibility of the morphologic data based on multiple images by means of the applied segmentation method could be shown with an overall variation of around 8%. CONCLUSIONS: 3D modeling of the VT during sustained phonation involves a complex experimental setting and elaborate image processing techniques. Functional comparative analysis or acoustical simulations based on such data should take the found variability into account.
OBJECTIVES: The extraction of a three-dimensional (3D) morphology of the human vocal tract (VT) from magnetic resonance imaging (MRI) during sustained phonation can be used for various analyses like numerical simulations or creating physical models. The precision of visualizing techniques nowadays allows for very targeted acoustical simulation evaluating the influence of subsections of the VT for the transfer function. The aim of the study was to assess the accuracy of the 3D geometry based on MRI data in repetitive trials. STUDY DESIGN: This is a prospective study. METHODS: Four experienced singers underwent an MRI while repeating a specific vocal task 20 times consecutively. Audio recordings were made by means of an optical microphone. Images were restacked and subsections of the VT were segmented on multi-image-based cross sections using a semiautomatic algorithm. Different volume and area measures were evaluated. RESULTS: A high reproducibility of the morphologic data based on multiple images by means of the applied segmentation method could be shown with an overall variation of around 8%. CONCLUSIONS: 3D modeling of the VT during sustained phonation involves a complex experimental setting and elaborate image processing techniques. Functional comparative analysis or acoustical simulations based on such data should take the found variability into account.
Authors: Peter Birkholz; Steffen Kürbis; Simon Stone; Patrick Häsner; Rémi Blandin; Mario Fleischer Journal: Sci Data Date: 2020-08-05 Impact factor: 6.444