Novotny Michal, Melechovsky Jan, Rozenstoks Kriss, Tykalova Tereza, Kryze Petr, Kanok Martin, Klempir Jiri, Rusz Jan
Department of Circuit Theory, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic.
Department of Neurology and Centre of Clinical Neuroscience, First Faculty of Medicine, Charles University, Prague, Czech Republic.
J Speech Lang Hear Res. 2020 Oct 16;63(10):3453-3460. doi: 10.1044/2020_JSLHR-20-00109. Epub 2020 Sep 21.
Purpose The purpose of this research note is to provide a performance comparison of available algorithms for the automated evaluation of oral diadochokinesis using speech samples from patients with amyotrophic lateral sclerosis (ALS). Method Four different algorithms based on a wide range of signal processing approaches were tested on a sequential motion rate /pa/-/ta/-/ka/ syllable repetition paradigm collected from 18 patients with ALS and 18 age- and gender-matched healthy controls (HCs). Results The best temporal detection of syllable position for a 10-ms tolerance value was achieved for ALS patients using a traditional signal processing approach based on a combination of filtering in the spectrogram, Bayesian detection, and polynomial thresholding with an accuracy rate of 74.4%, and for HCs using a deep learning approach with an accuracy rate of 87.6%. Compared to HCs, a slow diadochokinetic rate ( < .001) and diadochokinetic irregularity ( < .01) were detected in ALS patients. Conclusions The approaches using deep learning or multiple-step combinations of advanced signal processing methods provided a more robust solution to the estimation of oral DDK variables than did simpler approaches based on the rough segmentation of the signal envelope. The automated acoustic assessment of oral diadochokinesis shows excellent potential for monitoring bulbar disease progression in individuals with ALS.
目的 本研究报告的目的是使用肌萎缩侧索硬化症（ALS）患者的语音样本，对用于自动评估口腔轮替运动的现有算法进行性能比较。方法 基于广泛的信号处理方法的四种不同算法，在从18例ALS患者和18例年龄及性别匹配的健康对照（HCs）收集的连续运动速率/pa/-/ta/-/ka/音节重复范式上进行测试。结果 对于ALS患者，使用基于频谱图滤波、贝叶斯检测和多项式阈值化相结合的传统信号处理方法，在10毫秒容差值下音节位置的最佳时间检测准确率为74.4%；对于HCs，使用深度学习方法的准确率为87.6%。与HCs相比，在ALS患者中检测到轮替运动速率减慢（<.001）和轮替运动不规则（<.01）。结论 与基于信号包络粗略分割的简单方法相比，使用深度学习或先进信号处理方法的多步组合的方法为口腔轮替运动变量的估计提供了更稳健的解决方案。口腔轮替运动的自动声学评估在监测ALS患者延髓疾病进展方面显示出巨大潜力。
