A mobile application for detection of amyotrophic lateral sclerosis via voice analysis

Abstract

Analysis of the state of the art in the field of treatment and diagnosis of a rapidly progressing neurological disease of the amyotrophic lateral sclerosis (ALS) revealed a lack of tools for early diagnosis and monitoring the course of the disease. This paper proposes a method for evaluating the state of the voice function of patients with ALS, intended for use in a mobile application. As a speech probe a long pronunciation of the vowel sound / a / is used from whichacoustic features (jitter, shimmer, degree of vibrato pathology, etc.) are extracted. Applying a classifier based on the method of linear discriminant analysis to the obtained vector of parameters, it is possible to obtain an estimate whether the presented voice is pathological and possible progress of the disease. For training and verification of the proposed method, a database of 64 voices (33 healthy, 31 patients with ALS), recorded at the Republican Research and Clinical Center of Neurology and Neurosurgery (Minsk, Belarus), was used. To est the proposed method, a prototype of the mobile application «ALS Expert»was developed with the ability to record, process voice and display the results of voice analysis. The results indicate that the acoustic analysis of the voice and the resulting objective parameters are a promising direction for solving this problem.