Simplified Ultrasound Biofeedback for Speech Remediation

点击翻译按钮获取中文摘要

Project Summary Ultrasound biofeedback therapy (UBT), offering a real-time view of the tongue during speech, has shown promise in treating residual speech sound disorders. However, the clinical utility of UBT is limited by the inherent difficulty of interpreting real-time ultrasound images of rapidly changing tongue deformations during speech. For patients to receive effective feedback from these images, clinicians must provide extensive explanation and orientation over multiple sessions. Especially for younger patients, this rich, complex visual feedback directs attention internally, a direction known to reduce performance and impulse force control during limb and oral movement tasks. Further, since speech sound disorders often coexist with cognitive or behavioral difficulties, many patients never master interpretation of tongue ultrasound images. Thus, the tremendous promise of UBT has not been realized because its feedback is too complex, misdirected, and ambiguous. Scientific studies on implicit motor learning suggest UBT would be improved if (1) speakers are guided by simpler visual feedback and (2) the simplification engages an external attentional focus. Feedback with these qualities is known to facilitate rapid, robust sensorimotor skill learning. Application of these principles into UBT is expected to provide great benefits to those with speech-sound disorders. The major goal of this project is to translate these recent advances in UBT and motor learning into a new simplified ultrasound biofeedback system for better clinical treatment of residual speech sound disorders. Our proposed system will transform the complex tongue movements captured by ultrasound into simplified real-time feedback displays, customized for each user to drive tongue movements closer to goal movement patterns. We will concentrate on remediation of /r/ and /l/, the most complex and frustrating sounds in clinical speech therapy. Using a novel method for tongue motion tracking based on processing of real-time ultrasound images, we will automatically characterize differentiated motion of tongue parts during /r/ and /l/ production. We will employ statistical cluster analysis to identify multiple biofeedback targets indicating correct production. These targets will be employed in a prototype simplified UBT system employing simple, engaging visual feedback to guide patients toward correct tongue movement, presented by a highly motivating, gamified interface. Our simplified UBT system will be validated in a pilot clinical trial, testing the hypotheses that simplified UBT is more effective than standard UBT, and that the most effective biofeedback targets will be those promoting greater differentiation of tongue motion. The end result will be translation of recent advances in ultrasound imaging of speech, real-time image processing, and biofeedback-based motor learning into a novel clinical UBT tool that will greatly advance treatment of residual speech sound disorders.