Multilingual evaluation of interpretable biomarkers to represent language and speech patterns in Parkinson's disease

Abstract

Motor impairments are only one aspect of Parkinson's disease (PD), which also include cognitive and linguistic impairments. Speech-derived interpretable biomarkers may help clinicians diagnose PD at earlier stages and monitor the disorder's evolution over time. This study focuses on the multilingual evaluation of a composite array of biomarkers that facilitate PD evaluation from speech. Hypokinetic dysarthria, a motor speech disorder associated with PD, has been extensively analyzed in previously published studies on automatic PD evaluation, with a relative lack of inquiry into language and task variability. In this study, we explore certain acoustic, linguistic, and cognitive information encoded within the speech of several cohorts with PD. A total of 24 biomarkers were analyzed from American English, Italian, Castilian Spanish, Colombian Spanish, German, and Czech by conducting a statistical analysis to evaluate which biomarkers best differentiate people with PD from healthy participants. The study leverages conceptual robustness as a criterion in which a biomarker behaves the same, independent of the language. Hence, we propose a set of speech-based biomarkers that can effectively help evaluate PD while being language-independent. In short, the best acoustic and cognitive biomarkers permitting discrimination between experimental groups across languages were fundamental frequency standard deviation, pause time, pause percentage, silence duration, and speech rhythm standard deviation. Linguistic biomarkers representing the length of the narratives and the number of nouns and auxiliaries also provided discrimination between groups. Altogether, in addition to being significant, these biomarkers satisfied the robustness requirements.

Keywords: Parkinson's disease; cognition; correlation analysis; interpretable biomarkers; language; multilingual evaluation; speech; statistical analysis.

Figure 1

Boxplots with statistical annotations representing the standard deviation of F0 (F0STD) on Neurovoz, ItalianPVS, GermanPD, and GITA from the TDU task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for F0STD are reported in Table 5 and highlighted below using asterisks. F0STD, F0 standard deviation; CN, controls; PD, Parkinson's disease; TDU, text dependent utterance; ns, not significant.

Figure 2

Boxplots with statistical annotations representing the silence duration (SILDUR) on NLS, Neurovoz, CzechPD, GermanPD, and GITA from the RP task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for SILDUR are reported in Table 5 and highlighted below using asterisks. SILDUR, silence duration; CN, controls; PD, Parkinson's disease; SILDUR, silence duration; SS, spontaneous speech; ns, not significant.

Figure 3

Boxplots with statistical annotations representing the word count (WORDCNT) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for WORDCNT are reported in Table 6 and highlighted below using asterisks. WORDCNT, word count; CN, controls; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.

Figure 4

Boxplots with statistical annotations representing the noun count (NOUNCNT) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for NOUNCNT are reported in Table 6 and highlighted below using asterisks. NOUNCNT, noun count; CN, controls; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.

Figure 5

Boxplots with statistical annotations representing the noun phrase count (NPCNT) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for NPCNT are reported in Table 6 and highlighted below using asterisks. NPCNT, noun phrase count; CN, controls; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.

Figure 6

Boxplots with statistical annotations representing the verb count (VERBCNT) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for VERBCNT are reported in Table 6 and highlighted below using asterisks. VERBCNT, verb count; CN, control group; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.

Figure 7

Boxplots with statistical annotations representing the verb phrase count (VPCNT) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the median, the top ends represent the upper quartile, and the bottom ends represent the lower quartile. Statistically significant differences between groups after Benjamini–Hochberg correction for VPCNT are reported in Table 6 and highlighted below using asterisks. VPCNT, verb phrase count; CN, control group; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.

Figure 8

Boxplots representing the rhythm standard deviation (RHYSTD) on NLS, Neurovoz, GermanPD, and GITA from the SS task. Horizontal lines represent the medians, the top ends represent the upper quartiles, and the bottom ends represent the lower quartiles. Statistically significant differences between groups after Benjamini–Hochberg correction for RHYSTD are reported in Table 6 and highlighted below using asterisks. RHYSTD, rhythm standard deviation; CN, control group; PD, Parkinson's disease; SS, spontaneous speech; ns, not significant.