442 AI for Recovering a Voice

INTRODUCTION: Patients with brain tumors commonly encounter speech impairments, and these can have a significant impact on a patient's quality of life. This study explores the potential of a novel text-to-speech application, Voice Engine, to reconstruct naturalistic, individualized voice for a patient with persistent dysarthria following brain tumor resection. METHODS: A 21-year-old female patient presented with slurred speech and facial droop due to a large posterior fossa mass. Following a sub-occipital craniectomy for resection of pathology-confirmed hemangioblastoma, the patient experienced significant, persistent dysarthria. Voice Engine was used to reconstruct the patient's voice using a 15-second audio recording from a school presentation made two years prior. The reconstructed voice was integrated into a bespoke smartphone app to facilitate text-to-speech communication. Pre- and post-app surveys were administered to assess the impact on the patient's quality of life. RESULTS: The patient reported that her voice impairment severely affected her personal esteem, relationships, and ability to perform daily activities such as ordering food, attending school, and communicating at work. Following the implementation of the Voice Engine-powered app, the patient reported significant improvements (<0.001) in the ability to perform daily activities, as well as enhanced overall well-being and agency. CONCLUSIONS: This case study highlights the potential of deep learning-based text-to-speech applications to significantly improve the quality of life for patients with persistent speech impairments following brain tumor resection. Further research is needed to explore the applicability of this approach to a larger cohort of patients with speech impairments resulting from various neurological conditions.