Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Graph Theoretical Analysis of Semantic Fluency in Patients with Parkinson’s Disease

Graph Theoretical Analysis of Semantic Fluency in Patients with Parkinson’s Disease Hindawi Behavioural Neurology Volume 2022, Article ID 6935263, 7 pages https://doi.org/10.1155/2022/6935263 Research Article Graph Theoretical Analysis of Semantic Fluency in Patients with Parkinson’s Disease 1,2 3 4 5 Guanyu Zhang , Jinghong Ma, Piu Chan, and Zheng Ye China Institute of Sport Science, Beijing, China Institute of Psychology, Chinese Academy of Sciences, Beijing, China Department of Neurology, Xuanwu Hospital of Capital Medical University, Beijing, China Department of Neurology and Neurobiology, National Clinical Research Center for Geriatric Disorders, Xuanwu Hospital of Capital Medical University, Beijing, China Institute of Neuroscience, Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai, China Correspondence should be addressed to Guanyu Zhang; 1601466858@qq.com Received 27 January 2022; Revised 2 April 2022; Accepted 16 April 2022; Published 23 April 2022 Academic Editor: Luigi Trojano Copyright © 2022 Guanyu Zhang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Semantic fluency is the ability to name items from a given category within a limited time, which relies on semantic memory, working memory, and executive function. Semantic disfluency is a common problem in Parkinson’s disease (PD) and Alzheimer’s disease (AD). We demonstrated a graph theoretical analysis of semantic fluency in patients with PD (N = 86), patients with AD (N = 40), and healthy controls (HC, N = 88). All participants completed a standard animal fluency test. Their verbal responses were recorded, transcripted, and transformed into directed speech graphs. Patients with PD generated fewer correct words than HC and more correct words than patients with AD. Patients with PD showed higher density, shorter diameter, and shorter average shortest path length than HC, but lower density, longer diameter, and longer average shortest path length than patients with AD. It suggests that patients with PD produced relatively smaller and denser speech graphs. Moreover, in PD, the densities of speech graphs correlated with the severity of non-motor symptoms, but not the severity of motor symptoms. The graph theoretical analysis revealed new features of semantic disfluency in patients with PD. 1. Introduction Different approaches have been developed to quantify ver- bal responses in semantic fluency tests. Troyer and colleagues [5] proposed a method to segment the verbal response into Semantic fluency is the ability to name items from a given clusters according to the semantic relatedness between words. category (e.g., animals) during a given time interval, usually one minute (semantic fluency test). This task is significantly For example, a participant may begin with farm animals (e.g., ox, horse, and donkey) and then switch to forest animals (e.g., influenced by semantic memory (e.g., semantic representa- wolf, bear, and fox). This method generates two primary tions to be organized), working memory (e.g., keeping the parameters: the mean cluster size, which is the average num- search for new satisfying words), and executive function (e. ber of sequential words from the same subcategory, and the g., the ability to select and retrieve correct words and inhibit those that are not inherent with the specific category) number of switches between subcategories. It is assumed that the mean cluster size reflects semantic storage in the temporal domains. Semantic disfluency is a common problem in Par- lobe and the number of switches reflects executive functions in kinson’s disease (PD) and Alzheimer’s disease (AD). the frontal lobe. PD patients with dementia or mild cognitive Patients with PD or AD generate fewer correct words than impairment often switch less than healthy adults but they do healthy adults in the semantic fluency test [1–4]. 2 Behavioural Neurology not necessarily produce smaller clusters [6, 7]. In contrast, Each participant signed a written informed consent before patients with AD switch less and produce smaller clusters than participating in this study. healthy adults [6]. The Troyer method relies heavily on experimenters’ sub- 2.1. PD Patients and Clinical Assessments. We included 86 jective judgment of semantic relatedness and cluster segmenta- patients with idiopathic PD (Movement Disorder Society tion. Farzanfar et al. compared an automated computational Clinical Diagnostic Criteria for Parkinson’s Disease [14]) at assessment with the traditional experimenter-based assessment the Xuanwu Hospital Research and Clinical Center for Par- of semantic fluency data from patients with PD [8]. In the com- kinson’s disease between 2017 and 2019. Inclusion criteria putational assessment, each word was represented as a vector were (1) Hoehn and Yahr Stages 1 to 2; (2) age 40 to 80 in a semantic space derived from corpora. Semantic relatedness years; (3) education ≥6 years; and (4) Mandarin Chinese between a given pair of words was defined as the cosine of the speaking. Exclusion criteria were (1) a history of epilepsy, angle between the corresponding vectors (range from -1 [low stroke, or brain injury; (2) alcohol or drug abuse; (3) possible relatedness] to 1 [high relatedness]). Sequential words with a current depression (Beck Depression Inventory-II, BDI- semantic relatedness value higher than a predetermined II>7) or intake of anti-depressants; and (4) possible demen- threshold (0.5-0.6) were members of the same cluster. A tia (Montreal Cognitive Assessment, MoCA<21/30) or semantic relatedness value below the threshold indicated a intake of anti-dementia drugs. switch between clusters. The computational assessment was All patients with PD were assessed on their regular anti- inconsistent with the experimenter-based assessment in detect- Parkinsonian drugs, including levodopa (N =48), pramipex- ing clusters: the correlation between the two assessments varied ole (N =25), selegiline (N =16), piribedil (N =13), amanta- as a function of the threshold. In the experimenter-based dine (N =8), entacapone (N =4), and rasagiline (N =1). The assessment, the estimation of cluster and switch might be levodopa equivalent daily dose was calculated using the biased by the experimenter’s semantic knowledge. equation of Tomlinson et al. [15]. The severity of motor An objective method is based on graph theory. Graph the- and non-motor symptoms was evaluated with the Move- ory has been used to reveal topological changes in brain net- ment Disorder Society-sponsored revision of the Unified works in various brain disorders [9–11]. Recently, Bertola Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III and colleagues [12] used graph theory to analyze semantic flu- and I subscales, respectively. Table 1 shows demographic ency data of patients with AD or mild cognitive impairment. and clinical features and neuropsychological measures. They found that speech graphs of semantic fluency become smaller and denser as general cognition decreases. In another 2.2. Two Control Groups. We included two control groups: study, Mota and colleagues [13] used a graph theory to analyze 88 age- and education-matched healthy controls (HC) from dream reports and found that speech graphs of patients with local communities and 40 matched patients with AD from schizophrenia were less connected than those of healthy the DementiaBank database [16]. adults. The individual patients’ connectivity within speech For the HC group, exclusion criteria were (1) a history of graphs correlated with their severity of negative and cognitive significant neurological or psychiatric disorders; (2) alcohol symptoms. As a sensitive measurement, we hypothesize that or drug abuse; (3) possible current depression; and (4) pos- the graph theoretical analysis can extract more semantic fea- sible dementia or mild cognitive impairment (MoCA<26/ tures, which potentially contributes to the identification of 30). They completed the same assessments for cognition, mild cognitive impairment in PD from healthy adults or AD. mood, and sleep as patients with PD. In this study, we revisit the semantic disfluency of The DementiaBank database has 139 dementia patients patients with PD, comparing speech graphs of patients with assessed at the University of Pittsburgh School of Medicine. PD with those of healthy adults and patients with AD. All We only included AD patients matched with the other two participants completed a standard animal fluency test. We groups in sex and age (20 women, age range 50-70 years, transformed participants’ verbal responses into directed and mean age 62.2 years). We excluded patients diagnosed speech graphs, with each node representing a correct word with other types of dementia, including mild cognitive and each arc representing a temporal link between sequen- impairment (N =17), vascular diseases (N =4), and other tial words (Figure 1). First, we wanted to detect group differ- memory problems (N =3). ences in the number of correct words, repetitions, incorrect words, metalinguistic reference, and metacognitive reference 2.3. Standard and Graph Theoretical Analyses. All partici- (standard analysis). Second, we sought group differences in pants completed a standard animal fluency test. For the global characteristics of speech graphs, including density, PD and HC groups, we recorded and transcripted their ver- diameter, and average shortest path (graph theoretical anal- bal responses. For the AD group, we received their audios ysis). Third, in PD, we explored whether the speech graph and transcripts from the database. parameters correlated with clinical features such as the For the standard analysis, we defined five parameters: (1) severity of motor or non-motor symptoms. the number of correct words without repetitions: all types of animals were accepted, including humans, insects, and mythical creatures (e.g., dragon); (2) the number of repeti- 2. Materials and Methods tions; (3) the number of incorrect words; (4) metalinguistic This study was approved by the ethics committee of the reference: the number of times participants talked about their responses (e.g., “did I say horses?”); (5) metacognitive Xuanwu Hospital according to the Declaration of Helsinki. Behavioural Neurology 3 HC004 PD021 AD663 Duck Fish Chicken Shark Wolf Tiger Sheep Fox Leopard Rabbit Rabbit Dog Goat Mule Donkey Chicken Wolf Cat Butterfly Sheep Horse Horse Mouse Duck Cat Mouse Bee Tiger Dog Donkey Dog Cow Pig Chicken Silkworm Donkey Cat Lion Sheep Zebra Peacock Bird Ox Eagle Sparrow Pig Ox (a) HC004 PD021 AD663 Duck Fish Chicken Shark Wolf Tiger Sheep Fox Rabbit Rabbit Leopard Dog Goat Mule Chicken Donkey Wolf Cat Butterfly Sheep Horse Horse Mouse Duck Cat Bee Tiger Mouse Dog Donkey Cow Chicken Dog Pig Silkworm Donkey Lion Cat Sheep Zebra Peacock Bird Ox Eagle Sparrow Ox Pig (b) Figure 1: (a) Directed speech graphs of three representative participants. HC004, a healthy control subject; PD021, a patient with Parkinson’s disease; AD663, a patient with Alzheimer’s disease. (b) Graph geodesic as the shortest path (green) between two nodes (blue) in the three participants. Table 1: Demographic and clinical features, and neuropsychological measures of PD patients and healthy controls (means, standard deviations, and group differences). PD patients Healthy controls Features/measures Group differences (p values) (N =86) (N =88) Female: Male 44 : 42 46 : 42 0.884 Age (years) 59.0 (9.5) 58.1 (7.0) 0.484 Education (years) 12.4 (3.2) 12.9 (2.4) 0.204 Montreal cognitive assessment 25.6 (2.4) 27.9 (1.4) <0.001∗ Levodopa equivalent daily dose (mg) 243.3 (248.6) —— Motor symptoms Hoehn and Yahr scale 1.4 (0.5) —— MDS-UPDRS III: Motor examination 21.8 (12.6) —— Disease duration (years) 1.6 (2.2) —— Duration of motor symptoms (years) 2.8 (2.4) —— Other non-motor functions MDS-UPDRS I: Non-motor experiences of daily living 5.3 (4.0) —— Beck depression inventory-II 2.7 (2.0) 2.1 (1.7) 0.039 <0.001∗ REM sleep behavior disorder screening questionnaire 3.7 (2.0) 1.9 (1.8) Epworth sleep scale 3.1 (3.2) 3.2 (2.3) 0.820 Note: MDS-UPDRS, the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale; group differences, p values of two- sample t-tests, or Chi-square test as appropriate; asterisks (∗), a significant difference (two-tailed, p <0:007 Bonferroni correction for seven tests). 4 Behavioural Neurology speech graphs of the PD group were smaller and denser than reference: the number of times participants talked about their memory (e.g., “I really cannot think of any.”) or asked those of the HC group but larger and more sparse than those about the time (e.g., “how much time is left?”). of the AD group. For the graph theoretical analysis, we transformed par- 3.3. Correlations between Clinical Features and Graph ticipants’ verbal responses into directed speech graphs with Speechgraphs [12, 13]. In each directed speech graph, a node Parameters in PD. Figure 2(c) shows correlations between graph parameters and clinical features in PD. The stepwise represented a word, and an arc represented the temporal link between an ordered pair of words (Figure 1(a)). We regression model for the MDS-UPDRS Part I subscore computed three graph parameters, including the density, (Fð1, 83Þ =7:80, p =0:006, and R =0:09) included density diameter, and average shortest path length. The graph den- (beta = 29:11, t =2:79, and p =0:006) but removed the sity is the ratio of arcs to the maximum possible number diameter (beta = −0:14, t = −0:91, and p =0:37) and average of arcs. The graph geodesic is the shortest path between shortest path (beta = −0:11, t = −0:68, and p =0:50). PD two nodes (Figure 1(b)). The length of the maximum graph patients with more severe non-motor symptoms tended to geodesic is the graph diameter. The mean length of all graph produce smaller and denser speech graphs. geodesics is the average shortest path length, also known as Linear regression model did not survive at the corrected the characteristic path length of the graph. threshold for the MDS-UPDRS Part III subscore (Fð1, 84Þ =4:24 and p =0:043). 2.4. Statistical Analysis. Data were analyzed with IBM SPSS Statistics 20. First, we examined group differences in the 4. Discussion standard and graph parameters using one-way ANOVAs (two-tailed, p <0:006 Bonferroni correction for eight tests). In this study, we revisited the semantic disfluency in non- The ANOVA had a factor group (HC, PD, and AD) and a demented patients with PD. We replicated previous findings covariate age. Significant group differences were followed that patients with PD generated fewer correct and non- by pairwise comparisons (with Bonferroni correction). repetitive words than healthy controls [17–19] but more Second, in PD, we examined whether the severity of than patients with AD [20, 21]. More importantly, we exam- motor or non-motor symptoms (MDS-UPDRS Part III or I ined the topology of participants’ speech graphs and found subscores) correlated with the graph parameters that showed that patients with PD produced smaller and denser speech group differences using linear regression models (stepwise, graphs than healthy controls but larger and more sparse p <0:025 Bonferroni correction for two models). speech graphs than patients with AD. To be specific, the speech graphs of PD patients showed higher density, shorter 3. Results diameter, and shorter average shortest path than those of healthy controls but lower density, longer diameter, and lon- 3.1. Group Differences in Standard Parameters. Figure 2(a) ger average shortest path length than those of AD patients. shows standard parameters in each group. Group differences In PD, in addition, the density of speech graphs correlated were found in the number of correct words with the severity of non-motor symptoms. PD patients (Fð2, 210Þ =66:36, p <0:001,and η =0:39) and metacogni- who produced smaller and denser speech graphs exhibited tive reference (Fð2, 210Þ =12:37, p <0:001,and η =0:11), more severe non-motor symptoms in daily living. This study suggests that the graph theoretical analysis is but not in the number of repetitions (Fð2, 210Þ =1:79, p = more sensitive than the standard analysis to PD’s problems 0:169, η =0:02), number of incorrect words 2 in verbal fluency. For example, both approaches measured (Fð2, 210Þ =2:03, p =0:134,and η =0:02), or metalinguistic the repetition, but only the measures of the graph theoretical reference (F <1). The PD group generated fewer correct and analysis (e.g., density, diameter, and average shortest path) non-repetitive words than the HC group (p =0:008) but more showed significant group differences between PD patients correct and non-repetitive words than the AD group and healthy controls. The repetition might reflect the (p <0:001). The PD group talked about their memory and impaired selection and programming processes of semantic time remaining more than the HC group (p <0:001). Only fluency, which is associated with the left inferior frontal the AD group generated incorrect words. gyrus (LIFG) and basal ganglia. Verbal fluency tasks involve several cognitive processes: 3.2. Group Differences in Graph Parameters. Figure 2(b) (a) attention to search words from an abundant semantic shows graph parameters in each group. Group differences store, (b) selection of appropriate words to produce, (c) pro- were found in the density (Fð2, 210Þ =51:54, p <0:001, gramming of speech production, and (d) keeping track of and η =0:33), diameter (Fð2, 210Þ =38:40, p <0:001, and the words that have already been produced to avoid repeti- η =0:27), and average shortest path (Fð2, 210Þ =42:55, p tions. The dual stream model for language processing is a <0:001, and η =0:29). The PD group showed higher den- p widely accepted model that describes two large-scale streams sity (p =0:003), shorter diameter (p =0:008), and shorter underlying different speech tasks [22]. The ventral stream is average shortest path length than the HC group (p =0:008 comprised of bilaterally superior and middle portions of the ). The PD group showed lower density (p <0:001), longer temporal lobes with a weak left-hemisphere bias, which sup- diameter (p <0:001), and longer average shortest path ports the processing of sound-to-meaning information and length than the AD group (p <0:001). In other words, is essential for auditory comprehension and semantic Behavioural Neurology 5 Correct words Repetitions Incorrect words Metalinguistics Metacognition ⁎ ⁎ ⁎ 30 3 0.12 1.5 1.5 2 0.08 20 1.0 1.0 1 0.04 10 0.5 0.5 0 0 0 0 0 HC PD AD HC PD AD HC PD AD HC PD AD HC PD AD (a) Density Diameter Average shortest path MDS-UPDRS I MDS-UPDRS III ⁎ ⁎ ⁎ ⁎ ⁎ 0.3 20 50 0.2 10 20 0 –20 0.1 6 3 –10 –40 HC PD AD HC PD AD HC PD AD –0.1 0 0.1 –0.1 0 0.1 Density (a.u.) Density (a.u.) (b) (c) Figure 2: (a) Means and standard errors of correct words, repetitions, incorrect words, metalinguistic reference, and metacognitive reference in healthy controls (HC), patients with Parkinson’s disease (PD), and patients with Alzheimer’s disease (AD). (b) Means and standard errors of graph density, diameter, and average shortest path in each group. The asterisks (∗) indicate significant differences between PD patients and two control groups in standard and graph parameters. (c) In patients with PD, the density of speech graphs was correlated with the severity of non-motor symptoms (MDS-UPDRS I score) but not the severity of motor symptoms (MDS-UPDRS III score). Values were demeaned. retrieval. The dorsal stream is comprised of the left posterior cortical-basal ganglionic circuits involved in word genera- frontal lobe with a dominant position and left posterior tem- tion processes [28, 29]. poral lobe and left parietal operculum, which supports the Semantic fluency relies on working memory and execu- processing of sound-to-articulation (phonemic) information tive function, in addition to semantic knowledge. It has been and is essential for speech learning and development. It is described that the working memory deficits and executive assumed that semantic and phonemic fluency is mediated dysfunction in patients with PD may result in semantic dis- by ventral and dorsal streams, respectively. The link between fluency [30]. The impairments in working memory might the phonemic fluency and the dorsal stream has been shown result in the difficulty of keeping the search for new in a diffusion tensor imaging study with a large sample of de standard-compliant words and keeping track of produced novo patients with PD [23]. Future studies are needed to words. The executive dysfunction leads to the deficits in examine whether damage to the ventral stream might have selecting and producing appropriate words and inhibiting an impact on semantic fluency in patients with PD. inappropriate words (i.e., repetitions and incorrect words). On the other hand, the difficulty of self-shifting may result A selection mechanism will be applied when multiple verbal responses meet the instruction and compete for pro- in semantic disfluency. For example, Henry and Crawford duction. Previous studies showed that the LIFG is critical (2004) [31] showed that PD patients shift from one semantic for this process. Thompson-Schill et al. (1997) used func- category to another with more effort than healthy adults. tional magnetic resonance imaging and found that the selec- Semantic disfluency has been linked to non-motor symp- toms in PD in previous studies. PD patients with more severe tion of information among competing alternatives in semantic tasks is associated with LIFG activity in healthy depression or sleep disturbances tended to show worse perfor- adults [24]. Robinson et al. (2010) also reported that gener- mance in the semantic fluency tests [17, 32, 33]. In contrast, ation of sentences was only impaired when selection is there was no correlation between PD patients’ scores of required in patients with LIFG lesions [25]. In addition, it semantic fluency task and their disease durations, levodopa intakes, severities of rigidity or tremor, or Hoehn-Yahr stages has been suggested the basal ganglia support speech produc- tion through their role in programming and initiation by [34]. From another perspective, the semantic task could distin- modulating the activity of premotor areas (supplementary guish patients with PD and AD. Although patients with PD motor area, presupplementary motor area, and dorsolateral were impaired in semantic fluency, they were not as severe prefrontal cortex) [26]. The disruption of striatal dopami- as patients with AD. The AD patients not only generated fewer correct words, but also repeated the same word with a smaller nergic transmission in patients with PD may impair this modulation: The dopamine transporter availability in basal interval (e.g., dog-cat-horse-dog), indicating that the subtle ganglia was directly associated with frontal functions (i.e., deficit could underlie the differential diagnosis. In addition, attention/working memory and executive functions) [27]. the presence of semantic fluency deficits in PD has been iden- The decline of verbal fluency after pallidotomy in patients tified as a potent risk factor of the development of PD related dementia [35]. with PD may be due to surgical microlesions affecting (a.u.) Counts (a.u.) Counts (a.u.) Counts Times Score Score Times 6 Behavioural Neurology This study has limitations. First, cultural and linguistic References differences might be confounding factors. The AD group [1] A. L. R. Adlam, S. Bozeat, R. Arnold, P. Watson, and J. R. was from the database and was assessed by different experi- Hodges, “Semantic knowledge in mild cognitive impairment menters and in different cultures. Some studies also showed and mild Alzheimer’s disease,” Cortex, vol. 42, no. 5, that language and cultural differences have an impact on pp. 675–684, 2006. verbal fluency scores [36]; nevertheless, previous studies [2] N. Unsworth, G. J. Spillers, and G. A. Brewer, “Variation in analyze the semantic data from different cultures (Greece verbal fluency: a latent variable analysis of clustering, switch- and France) by using extended and unified methods [37]. ing, and overall performance,” Quarterly Journal of Experi- It is worthy to examine the difference of semantic fluency mental Psychology, vol. 64, no. 3, pp. 447–466, 2011. in PD patients under different cultural backgrounds. Second, [3] T. Azuma, K. A. Bayles, R. F. Cruz et al., “Comparing the dif- group differences and individual variability in graph param- ficulty of letter, semantic, and name fluency tasks for normal eters have not been linked with the structural integrity of elderly and patients with Parkinson’s disease,” Neuropsychol- brains. Third, some studies have shown that the perfor- ogy, vol. 11, no. 4, pp. 488–497, 1997. mance of verbal fluency was improved when the PD patients [4] F. Pasquier, F. Lebert, L. Grymonprez, and H. Petit, “Verbal were assessed with versus without levodopa treatment [38]. fluency in dementia of frontal lobe type and dementia of Alz- The PD patients scored higher in the semantic fluency task heimer type,” Journal of Neurology, Neurosurgery, and Psychi- when they were treated with rasagiline than placebo [39]. atry, vol. 58, no. 1, pp. 81–84, 1995. It would be of interest to assess the impact of medication [5] A. K. Troyer, M. Moscovitch, and G. Winocur, “Clustering and on the topology of speech graphs in future studies. switching as two components of verbal fluency: evidence from younger and older healthy adults,” Neuropsychology, vol. 11, no. 1, pp. 138–146, 1997. 5. Conclusion [6] A. K. Troyer, M. Moscovitch, G. Winocur, L. Leach, and M. Freedman, “Clustering and switching on verbal fluency In this study, we analyzed the topology of speech graphs tests in Alzheimer’s and Parkinson’s disease,” Journal of the generated in a semantic fluency test. The speech graphs of International Neuropsychological Society, vol. 4, no. 2, patients with PD were smaller and denser than those of pp. 137–143, 1998. healthy controls but larger and more sparse than those of [7] I. Galtier, A. Nieto, J. N. Lorenzo, and J. Barroso, “Mild cogni- patients with AD. Moreover, PD patients who produced tive impairment in Parkinson’s disease: clustering and switch- smaller and denser speech graphs exhibited more severe ing analyses in verbal fluency test,” Journal of the International non-motor symptoms. Neuropsychological Society, vol. 23, no. 6, pp. 511–520, 2017. [8] D. Farzanfar, M. Statucka, and M. Cohn, “Automated indices of clustering and switching of semantic verbal fluency in Par- Data Availability kinson’s disease,” Journal of the International Neuropsycholog- ical Society, vol. 24, no. 10, pp. 1047–1056, 2018. Data have been uploaded to the figshare database https:// [9] O. Sporns, “Graph theory methods: applications in brain net- figshare.com/articles/dataset/XW_data_2017-2019_xls/ works,” Dialogues in Clinical Neuroscience, vol. 20, no. 2, pp. 111–121, 2018. [10] O. Sporns, D. R. Chialvo, M. Kaiser, and C. C. Hilgetag, “Orga- nization, development and function of complex brain net- Conflicts of Interest works,” Trends in Cognitive Sciences, vol. 8, no. 9, pp. 418– 425, 2004. The authors declare that they have no conflicts of interest. [11] O. Sporns and J. D. Zwi, “The small world of the cerebral cor- tex,” Neuroinformatics, vol. 2, no. 2, pp. 145–162, 2004. [12] L. Bertola, N. B. Mota, M. Copelli et al., “Graph analysis of ver- Authors’ Contributions bal fluency test discriminate between patients with Alzhei- mer’s disease, mild cognitive impairment and normal elderly JM and ZY designed the study. GZ and JM collected the controls,” Frontiers in Aging Neuroscience, vol. 6, p. 185, 2014. data. GZ analyzed the data. GZ and ZY wrote the original draft of the manuscript. JM and PC reviewed and edited [13] M. B. Mota, R. Furtado, P. P. C. Maia, M. Copelli, and S. Ribeiro, “Graph analysis of dream reports is especially infor- the manuscript. All authors approved the submitted version. mative about psychosis,” ScientificReports, vol. 4, p. 3691, 2015. GZ and JM equally contributed to this work and share first [14] R. B. Postuma, D. Berg, M. Stern et al., “MDS clinical diagnos- authorship. tic criteria for Parkinson’s disease,” Movement Disorders, vol. 30, no. 12, pp. 1591–1601, 2015. Acknowledgments [15] C. L. Tomlinson, R. Stowe, S. Patel, C. Rick, R. Gray, and C. E. Clarke, “Systematic review of levodopa dose equivalency We are grateful to Sha Liu, Shaoyang Ma, and Minghong Su reporting in Parkinson’s disease,” Movement Disorders, for their assistance in data acquisition. This work was sup- vol. 25, no. 15, pp. 2649–2653, 2010. ported by the National Natural Science Foundation of China [16] A. Lanzi, A. Lindsay, and M. Bourgeois, “Verbal fluency in (31961133025 to Z.Y.) and the National Key Research and dementia: changes over time,” in American School Health Development Program of China (2018YFC1312001 to P.C.). Association Conference., Saint Louis, Missouri, USA, 2017. Behavioural Neurology 7 [17] I. Obeso, E. Casabona, M. L. Bringas, L. Álvarez, and [32] D. De Gaspari, C. Siri, M. Di Gioia et al., “Clinical correlates M. Jahanshahi, “Semantic and phonemic verbal fluency in Par- and cognitive underpinnings of verbal fluency impairment kinson’s disease: influence of clinical and demographic vari- after chronic subthalamic stimulation in Parkinson’s disease,” ables,” Behavioural Neurology, vol. 25, no. 2, pp. 111–118, Parkinsonism & Related Disorders, vol. 12, no. 5, pp. 289–295, 2012. 2006. [18] O. W. Wong, A. Y. Chan, A. Wong et al., “Eye movement [33] N. Kamble, R. Yadav, A. Lenka, K. Kumar, B. C. Nagaraju, and parameters and cognitive functions in Parkinson’s disease P. K. Pal, “Impaired sleep quality and cognition in patients of patients without dementia,” Parkinsonism & Related Disor- Parkinson’s disease with REM sleep behavior disorder: a com- ders, vol. 52, pp. 43–48, 2018. parative study,” Sleep Medicine, vol. 62, pp. 1–5, 2019. [19] A. F. Barbosa, M. C. Voos, J. Chen et al., “Cognitive or [34] S. Auriacombe, M. Grossman, S. Carvell, S. Gollomp, M. B. cognitive-motor executive function tasks? Evaluating verbal Stern, and H. I. Hurtig, “Verbal fluency deficits in Parkinson’s fluency measures in people with Parkinson’s disease,” BioMed disease,” Neuropsychology, vol. 7, no. 2, pp. 182–192, 1993. Research International, vol. 2017, 7 pages, 2017. [35] C. H. Williams-Gray, S. L. Mason, J. R. Evans et al., “The Cam- [20] J. McDowd, L. Hoffman, E. Rozek et al., “Understanding ver- PaIGN study of Parkinson’s disease: 10-year outlook in an bal fluency in healthy aging, Alzheimer’s disease, and Parkin- incident population-based cohort,” Journal of Neurology, Neu- son’s disease,” Neuropsychology, vol. 25, no. 2, pp. 210–225, rosurgery & Psychiatry, vol. 84, no. 11, pp. 1258–1264, 2013. [36] A. St-Hilaire, C. Hudon, G. T. Vallet et al., “Normative data for [21] N. B. Araujo, M. L. Barca, K. Engedal, E. S. Coutinho, A. C. phonemic and semantic verbal fluency test in the adult Deslandes, and J. Laks, “Verbal fluency in Alzheimer’s disease, French-Quebec population and validation study in Alzhei- Parkinson’s disease, and major depression,” Clinics, vol. 66, mer’s disease and depression,” The Clinical Neuropsychologist, no. 4, pp. 623–627, 2011. vol. 30, no. 7, pp. 1126–1150, 2016. [22] G. Hickok and D. Poeppel, “The cortical organization of [37] N. Linz, J. Tröger, J. Alexandersson, M. Wolters, A. König, and speech processing,” Nature Reviews Neuroscience, vol. 8, P. Robert, “Predicting dementia screening and staging scores no. 5, pp. 393–402, 2007. from semantic verbal fluency performance,” in 2017 IEEE International Conference on Data Mining Workshops [23] F. Rodriguez-Porcel, J. Wilmskoetter, C. Cooper et al., “The (ICDMW), pp. 719–728, New Orleans, LA, USA, 2017. relationship between dorsal stream connections to the caudate and verbal fluency in Parkinson disease,” Brain Imaging and [38] A. M. Gotham, R. G. Brown, and C. D. Marsden, “‘Frontal’ Behavior, vol. 15, no. 4, pp. 2121–2125, 2021. cognitive function in patients with Parkinson’s disease ‘on’ and ‘off’ levodopa,” Brain, vol. 111, no. 2, pp. 299–321, 1988. [24] S. L. Thompson-Schill, M. D’Esposito, G. K. Aguirre, and M. J. Farah, “Role of left inferior prefrontal cortex in retrieval of [39] H. A. Hanagasi, H. Gurvit, P. Unsalan et al., “The effects of semantic knowledge: a reevaluation,” Proceedings of the rasagiline on cognitive deficits in Parkinson’s disease patients National Academy of Sciences, vol. 94, no. 26, pp. 14792– without dementia: a randomized, double-blind, placebo-con- 14797, 1997. trolled, multicenter study,” Movement Disorders, vol. 26, no. 10, pp. 1851–1858, 2011. [25] G. Robinson, T. Shallice, M. Bozzali, and L. Cipolotti, “Con- ceptual proposition selection and the LIFG: neuropsychologi- cal evidence from a focal frontal group,” Neuropsychologia, vol. 48, no. 6, pp. 1652–1663, 2010. [26] G. E. Alexander, M. R. DeLong, and P. L. Strick, “Parallel orga- nization of functionally segregated circuits linking basal gan- glia and cortex,” Annual Review of Neuroscience, vol. 9, no. 1, pp. 357–381, 1986. [27] S. J. Chung, H. S. Yoo, J. S. Oh et al., “Effect of striatal dopa- mine depletion on cognition in de novo Parkinson’s disease,” Parkinsonism & Related Disorders, vol. 51, pp. 43–48, 2018. [28] R. M. De Bie, P. R. Schuurman, D. A. Bosch, R. J. De Haan, B. Schmand, and J. D. Speelman, “Outcome of unilateral palli- dotomy in advanced Parkinson’s disease: cohort study of 32 patients,” Journal of Neurology, Neurosurgery & Psychiatry, vol. 71, no. 3, pp. 375–382, 2001. [29] A. I. Tröster, S. P. Woods, and J. A. Fields, “Verbal fluency declines after Pallidotomy: an interaction between task and lesion laterality,” Applied Neuropsychology, vol. 10, no. 2, pp. 69–75, 2003. [30] C. I. Higginson, D. S. King, D. Levine, V. L. Wheelock, N. O. Khamphay, and K. A. Sigvardt, “The relationship between executive function and verbal memory in Parkinson’s disease,” Brain and Cognition, vol. 52, no. 3, pp. 343–352, 2003. [31] J. D. Henry and J. R. Crawford, “Verbal fluency deficits in Par- kinson’s disease: a meta-analysis,” Journal of the International Neuropsychological Society, vol. 10, no. 4, pp. 608–622, 2004. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioural Neurology Hindawi Publishing Corporation

Graph Theoretical Analysis of Semantic Fluency in Patients with Parkinson’s Disease

Loading next page...
 
/lp/hindawi-publishing-corporation/graph-theoretical-analysis-of-semantic-fluency-in-patients-with-luE5Rsyaja

References (41)

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2022 Guanyu Zhang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
0953-4180
eISSN
1875-8584
DOI
10.1155/2022/6935263
Publisher site
See Article on Publisher Site

Abstract

Hindawi Behavioural Neurology Volume 2022, Article ID 6935263, 7 pages https://doi.org/10.1155/2022/6935263 Research Article Graph Theoretical Analysis of Semantic Fluency in Patients with Parkinson’s Disease 1,2 3 4 5 Guanyu Zhang , Jinghong Ma, Piu Chan, and Zheng Ye China Institute of Sport Science, Beijing, China Institute of Psychology, Chinese Academy of Sciences, Beijing, China Department of Neurology, Xuanwu Hospital of Capital Medical University, Beijing, China Department of Neurology and Neurobiology, National Clinical Research Center for Geriatric Disorders, Xuanwu Hospital of Capital Medical University, Beijing, China Institute of Neuroscience, Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai, China Correspondence should be addressed to Guanyu Zhang; 1601466858@qq.com Received 27 January 2022; Revised 2 April 2022; Accepted 16 April 2022; Published 23 April 2022 Academic Editor: Luigi Trojano Copyright © 2022 Guanyu Zhang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Semantic fluency is the ability to name items from a given category within a limited time, which relies on semantic memory, working memory, and executive function. Semantic disfluency is a common problem in Parkinson’s disease (PD) and Alzheimer’s disease (AD). We demonstrated a graph theoretical analysis of semantic fluency in patients with PD (N = 86), patients with AD (N = 40), and healthy controls (HC, N = 88). All participants completed a standard animal fluency test. Their verbal responses were recorded, transcripted, and transformed into directed speech graphs. Patients with PD generated fewer correct words than HC and more correct words than patients with AD. Patients with PD showed higher density, shorter diameter, and shorter average shortest path length than HC, but lower density, longer diameter, and longer average shortest path length than patients with AD. It suggests that patients with PD produced relatively smaller and denser speech graphs. Moreover, in PD, the densities of speech graphs correlated with the severity of non-motor symptoms, but not the severity of motor symptoms. The graph theoretical analysis revealed new features of semantic disfluency in patients with PD. 1. Introduction Different approaches have been developed to quantify ver- bal responses in semantic fluency tests. Troyer and colleagues [5] proposed a method to segment the verbal response into Semantic fluency is the ability to name items from a given clusters according to the semantic relatedness between words. category (e.g., animals) during a given time interval, usually one minute (semantic fluency test). This task is significantly For example, a participant may begin with farm animals (e.g., ox, horse, and donkey) and then switch to forest animals (e.g., influenced by semantic memory (e.g., semantic representa- wolf, bear, and fox). This method generates two primary tions to be organized), working memory (e.g., keeping the parameters: the mean cluster size, which is the average num- search for new satisfying words), and executive function (e. ber of sequential words from the same subcategory, and the g., the ability to select and retrieve correct words and inhibit those that are not inherent with the specific category) number of switches between subcategories. It is assumed that the mean cluster size reflects semantic storage in the temporal domains. Semantic disfluency is a common problem in Par- lobe and the number of switches reflects executive functions in kinson’s disease (PD) and Alzheimer’s disease (AD). the frontal lobe. PD patients with dementia or mild cognitive Patients with PD or AD generate fewer correct words than impairment often switch less than healthy adults but they do healthy adults in the semantic fluency test [1–4]. 2 Behavioural Neurology not necessarily produce smaller clusters [6, 7]. In contrast, Each participant signed a written informed consent before patients with AD switch less and produce smaller clusters than participating in this study. healthy adults [6]. The Troyer method relies heavily on experimenters’ sub- 2.1. PD Patients and Clinical Assessments. We included 86 jective judgment of semantic relatedness and cluster segmenta- patients with idiopathic PD (Movement Disorder Society tion. Farzanfar et al. compared an automated computational Clinical Diagnostic Criteria for Parkinson’s Disease [14]) at assessment with the traditional experimenter-based assessment the Xuanwu Hospital Research and Clinical Center for Par- of semantic fluency data from patients with PD [8]. In the com- kinson’s disease between 2017 and 2019. Inclusion criteria putational assessment, each word was represented as a vector were (1) Hoehn and Yahr Stages 1 to 2; (2) age 40 to 80 in a semantic space derived from corpora. Semantic relatedness years; (3) education ≥6 years; and (4) Mandarin Chinese between a given pair of words was defined as the cosine of the speaking. Exclusion criteria were (1) a history of epilepsy, angle between the corresponding vectors (range from -1 [low stroke, or brain injury; (2) alcohol or drug abuse; (3) possible relatedness] to 1 [high relatedness]). Sequential words with a current depression (Beck Depression Inventory-II, BDI- semantic relatedness value higher than a predetermined II>7) or intake of anti-depressants; and (4) possible demen- threshold (0.5-0.6) were members of the same cluster. A tia (Montreal Cognitive Assessment, MoCA<21/30) or semantic relatedness value below the threshold indicated a intake of anti-dementia drugs. switch between clusters. The computational assessment was All patients with PD were assessed on their regular anti- inconsistent with the experimenter-based assessment in detect- Parkinsonian drugs, including levodopa (N =48), pramipex- ing clusters: the correlation between the two assessments varied ole (N =25), selegiline (N =16), piribedil (N =13), amanta- as a function of the threshold. In the experimenter-based dine (N =8), entacapone (N =4), and rasagiline (N =1). The assessment, the estimation of cluster and switch might be levodopa equivalent daily dose was calculated using the biased by the experimenter’s semantic knowledge. equation of Tomlinson et al. [15]. The severity of motor An objective method is based on graph theory. Graph the- and non-motor symptoms was evaluated with the Move- ory has been used to reveal topological changes in brain net- ment Disorder Society-sponsored revision of the Unified works in various brain disorders [9–11]. Recently, Bertola Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III and colleagues [12] used graph theory to analyze semantic flu- and I subscales, respectively. Table 1 shows demographic ency data of patients with AD or mild cognitive impairment. and clinical features and neuropsychological measures. They found that speech graphs of semantic fluency become smaller and denser as general cognition decreases. In another 2.2. Two Control Groups. We included two control groups: study, Mota and colleagues [13] used a graph theory to analyze 88 age- and education-matched healthy controls (HC) from dream reports and found that speech graphs of patients with local communities and 40 matched patients with AD from schizophrenia were less connected than those of healthy the DementiaBank database [16]. adults. The individual patients’ connectivity within speech For the HC group, exclusion criteria were (1) a history of graphs correlated with their severity of negative and cognitive significant neurological or psychiatric disorders; (2) alcohol symptoms. As a sensitive measurement, we hypothesize that or drug abuse; (3) possible current depression; and (4) pos- the graph theoretical analysis can extract more semantic fea- sible dementia or mild cognitive impairment (MoCA<26/ tures, which potentially contributes to the identification of 30). They completed the same assessments for cognition, mild cognitive impairment in PD from healthy adults or AD. mood, and sleep as patients with PD. In this study, we revisit the semantic disfluency of The DementiaBank database has 139 dementia patients patients with PD, comparing speech graphs of patients with assessed at the University of Pittsburgh School of Medicine. PD with those of healthy adults and patients with AD. All We only included AD patients matched with the other two participants completed a standard animal fluency test. We groups in sex and age (20 women, age range 50-70 years, transformed participants’ verbal responses into directed and mean age 62.2 years). We excluded patients diagnosed speech graphs, with each node representing a correct word with other types of dementia, including mild cognitive and each arc representing a temporal link between sequen- impairment (N =17), vascular diseases (N =4), and other tial words (Figure 1). First, we wanted to detect group differ- memory problems (N =3). ences in the number of correct words, repetitions, incorrect words, metalinguistic reference, and metacognitive reference 2.3. Standard and Graph Theoretical Analyses. All partici- (standard analysis). Second, we sought group differences in pants completed a standard animal fluency test. For the global characteristics of speech graphs, including density, PD and HC groups, we recorded and transcripted their ver- diameter, and average shortest path (graph theoretical anal- bal responses. For the AD group, we received their audios ysis). Third, in PD, we explored whether the speech graph and transcripts from the database. parameters correlated with clinical features such as the For the standard analysis, we defined five parameters: (1) severity of motor or non-motor symptoms. the number of correct words without repetitions: all types of animals were accepted, including humans, insects, and mythical creatures (e.g., dragon); (2) the number of repeti- 2. Materials and Methods tions; (3) the number of incorrect words; (4) metalinguistic This study was approved by the ethics committee of the reference: the number of times participants talked about their responses (e.g., “did I say horses?”); (5) metacognitive Xuanwu Hospital according to the Declaration of Helsinki. Behavioural Neurology 3 HC004 PD021 AD663 Duck Fish Chicken Shark Wolf Tiger Sheep Fox Leopard Rabbit Rabbit Dog Goat Mule Donkey Chicken Wolf Cat Butterfly Sheep Horse Horse Mouse Duck Cat Mouse Bee Tiger Dog Donkey Dog Cow Pig Chicken Silkworm Donkey Cat Lion Sheep Zebra Peacock Bird Ox Eagle Sparrow Pig Ox (a) HC004 PD021 AD663 Duck Fish Chicken Shark Wolf Tiger Sheep Fox Rabbit Rabbit Leopard Dog Goat Mule Chicken Donkey Wolf Cat Butterfly Sheep Horse Horse Mouse Duck Cat Bee Tiger Mouse Dog Donkey Cow Chicken Dog Pig Silkworm Donkey Lion Cat Sheep Zebra Peacock Bird Ox Eagle Sparrow Ox Pig (b) Figure 1: (a) Directed speech graphs of three representative participants. HC004, a healthy control subject; PD021, a patient with Parkinson’s disease; AD663, a patient with Alzheimer’s disease. (b) Graph geodesic as the shortest path (green) between two nodes (blue) in the three participants. Table 1: Demographic and clinical features, and neuropsychological measures of PD patients and healthy controls (means, standard deviations, and group differences). PD patients Healthy controls Features/measures Group differences (p values) (N =86) (N =88) Female: Male 44 : 42 46 : 42 0.884 Age (years) 59.0 (9.5) 58.1 (7.0) 0.484 Education (years) 12.4 (3.2) 12.9 (2.4) 0.204 Montreal cognitive assessment 25.6 (2.4) 27.9 (1.4) <0.001∗ Levodopa equivalent daily dose (mg) 243.3 (248.6) —— Motor symptoms Hoehn and Yahr scale 1.4 (0.5) —— MDS-UPDRS III: Motor examination 21.8 (12.6) —— Disease duration (years) 1.6 (2.2) —— Duration of motor symptoms (years) 2.8 (2.4) —— Other non-motor functions MDS-UPDRS I: Non-motor experiences of daily living 5.3 (4.0) —— Beck depression inventory-II 2.7 (2.0) 2.1 (1.7) 0.039 <0.001∗ REM sleep behavior disorder screening questionnaire 3.7 (2.0) 1.9 (1.8) Epworth sleep scale 3.1 (3.2) 3.2 (2.3) 0.820 Note: MDS-UPDRS, the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale; group differences, p values of two- sample t-tests, or Chi-square test as appropriate; asterisks (∗), a significant difference (two-tailed, p <0:007 Bonferroni correction for seven tests). 4 Behavioural Neurology speech graphs of the PD group were smaller and denser than reference: the number of times participants talked about their memory (e.g., “I really cannot think of any.”) or asked those of the HC group but larger and more sparse than those about the time (e.g., “how much time is left?”). of the AD group. For the graph theoretical analysis, we transformed par- 3.3. Correlations between Clinical Features and Graph ticipants’ verbal responses into directed speech graphs with Speechgraphs [12, 13]. In each directed speech graph, a node Parameters in PD. Figure 2(c) shows correlations between graph parameters and clinical features in PD. The stepwise represented a word, and an arc represented the temporal link between an ordered pair of words (Figure 1(a)). We regression model for the MDS-UPDRS Part I subscore computed three graph parameters, including the density, (Fð1, 83Þ =7:80, p =0:006, and R =0:09) included density diameter, and average shortest path length. The graph den- (beta = 29:11, t =2:79, and p =0:006) but removed the sity is the ratio of arcs to the maximum possible number diameter (beta = −0:14, t = −0:91, and p =0:37) and average of arcs. The graph geodesic is the shortest path between shortest path (beta = −0:11, t = −0:68, and p =0:50). PD two nodes (Figure 1(b)). The length of the maximum graph patients with more severe non-motor symptoms tended to geodesic is the graph diameter. The mean length of all graph produce smaller and denser speech graphs. geodesics is the average shortest path length, also known as Linear regression model did not survive at the corrected the characteristic path length of the graph. threshold for the MDS-UPDRS Part III subscore (Fð1, 84Þ =4:24 and p =0:043). 2.4. Statistical Analysis. Data were analyzed with IBM SPSS Statistics 20. First, we examined group differences in the 4. Discussion standard and graph parameters using one-way ANOVAs (two-tailed, p <0:006 Bonferroni correction for eight tests). In this study, we revisited the semantic disfluency in non- The ANOVA had a factor group (HC, PD, and AD) and a demented patients with PD. We replicated previous findings covariate age. Significant group differences were followed that patients with PD generated fewer correct and non- by pairwise comparisons (with Bonferroni correction). repetitive words than healthy controls [17–19] but more Second, in PD, we examined whether the severity of than patients with AD [20, 21]. More importantly, we exam- motor or non-motor symptoms (MDS-UPDRS Part III or I ined the topology of participants’ speech graphs and found subscores) correlated with the graph parameters that showed that patients with PD produced smaller and denser speech group differences using linear regression models (stepwise, graphs than healthy controls but larger and more sparse p <0:025 Bonferroni correction for two models). speech graphs than patients with AD. To be specific, the speech graphs of PD patients showed higher density, shorter 3. Results diameter, and shorter average shortest path than those of healthy controls but lower density, longer diameter, and lon- 3.1. Group Differences in Standard Parameters. Figure 2(a) ger average shortest path length than those of AD patients. shows standard parameters in each group. Group differences In PD, in addition, the density of speech graphs correlated were found in the number of correct words with the severity of non-motor symptoms. PD patients (Fð2, 210Þ =66:36, p <0:001,and η =0:39) and metacogni- who produced smaller and denser speech graphs exhibited tive reference (Fð2, 210Þ =12:37, p <0:001,and η =0:11), more severe non-motor symptoms in daily living. This study suggests that the graph theoretical analysis is but not in the number of repetitions (Fð2, 210Þ =1:79, p = more sensitive than the standard analysis to PD’s problems 0:169, η =0:02), number of incorrect words 2 in verbal fluency. For example, both approaches measured (Fð2, 210Þ =2:03, p =0:134,and η =0:02), or metalinguistic the repetition, but only the measures of the graph theoretical reference (F <1). The PD group generated fewer correct and analysis (e.g., density, diameter, and average shortest path) non-repetitive words than the HC group (p =0:008) but more showed significant group differences between PD patients correct and non-repetitive words than the AD group and healthy controls. The repetition might reflect the (p <0:001). The PD group talked about their memory and impaired selection and programming processes of semantic time remaining more than the HC group (p <0:001). Only fluency, which is associated with the left inferior frontal the AD group generated incorrect words. gyrus (LIFG) and basal ganglia. Verbal fluency tasks involve several cognitive processes: 3.2. Group Differences in Graph Parameters. Figure 2(b) (a) attention to search words from an abundant semantic shows graph parameters in each group. Group differences store, (b) selection of appropriate words to produce, (c) pro- were found in the density (Fð2, 210Þ =51:54, p <0:001, gramming of speech production, and (d) keeping track of and η =0:33), diameter (Fð2, 210Þ =38:40, p <0:001, and the words that have already been produced to avoid repeti- η =0:27), and average shortest path (Fð2, 210Þ =42:55, p tions. The dual stream model for language processing is a <0:001, and η =0:29). The PD group showed higher den- p widely accepted model that describes two large-scale streams sity (p =0:003), shorter diameter (p =0:008), and shorter underlying different speech tasks [22]. The ventral stream is average shortest path length than the HC group (p =0:008 comprised of bilaterally superior and middle portions of the ). The PD group showed lower density (p <0:001), longer temporal lobes with a weak left-hemisphere bias, which sup- diameter (p <0:001), and longer average shortest path ports the processing of sound-to-meaning information and length than the AD group (p <0:001). In other words, is essential for auditory comprehension and semantic Behavioural Neurology 5 Correct words Repetitions Incorrect words Metalinguistics Metacognition ⁎ ⁎ ⁎ 30 3 0.12 1.5 1.5 2 0.08 20 1.0 1.0 1 0.04 10 0.5 0.5 0 0 0 0 0 HC PD AD HC PD AD HC PD AD HC PD AD HC PD AD (a) Density Diameter Average shortest path MDS-UPDRS I MDS-UPDRS III ⁎ ⁎ ⁎ ⁎ ⁎ 0.3 20 50 0.2 10 20 0 –20 0.1 6 3 –10 –40 HC PD AD HC PD AD HC PD AD –0.1 0 0.1 –0.1 0 0.1 Density (a.u.) Density (a.u.) (b) (c) Figure 2: (a) Means and standard errors of correct words, repetitions, incorrect words, metalinguistic reference, and metacognitive reference in healthy controls (HC), patients with Parkinson’s disease (PD), and patients with Alzheimer’s disease (AD). (b) Means and standard errors of graph density, diameter, and average shortest path in each group. The asterisks (∗) indicate significant differences between PD patients and two control groups in standard and graph parameters. (c) In patients with PD, the density of speech graphs was correlated with the severity of non-motor symptoms (MDS-UPDRS I score) but not the severity of motor symptoms (MDS-UPDRS III score). Values were demeaned. retrieval. The dorsal stream is comprised of the left posterior cortical-basal ganglionic circuits involved in word genera- frontal lobe with a dominant position and left posterior tem- tion processes [28, 29]. poral lobe and left parietal operculum, which supports the Semantic fluency relies on working memory and execu- processing of sound-to-articulation (phonemic) information tive function, in addition to semantic knowledge. It has been and is essential for speech learning and development. It is described that the working memory deficits and executive assumed that semantic and phonemic fluency is mediated dysfunction in patients with PD may result in semantic dis- by ventral and dorsal streams, respectively. The link between fluency [30]. The impairments in working memory might the phonemic fluency and the dorsal stream has been shown result in the difficulty of keeping the search for new in a diffusion tensor imaging study with a large sample of de standard-compliant words and keeping track of produced novo patients with PD [23]. Future studies are needed to words. The executive dysfunction leads to the deficits in examine whether damage to the ventral stream might have selecting and producing appropriate words and inhibiting an impact on semantic fluency in patients with PD. inappropriate words (i.e., repetitions and incorrect words). On the other hand, the difficulty of self-shifting may result A selection mechanism will be applied when multiple verbal responses meet the instruction and compete for pro- in semantic disfluency. For example, Henry and Crawford duction. Previous studies showed that the LIFG is critical (2004) [31] showed that PD patients shift from one semantic for this process. Thompson-Schill et al. (1997) used func- category to another with more effort than healthy adults. tional magnetic resonance imaging and found that the selec- Semantic disfluency has been linked to non-motor symp- toms in PD in previous studies. PD patients with more severe tion of information among competing alternatives in semantic tasks is associated with LIFG activity in healthy depression or sleep disturbances tended to show worse perfor- adults [24]. Robinson et al. (2010) also reported that gener- mance in the semantic fluency tests [17, 32, 33]. In contrast, ation of sentences was only impaired when selection is there was no correlation between PD patients’ scores of required in patients with LIFG lesions [25]. In addition, it semantic fluency task and their disease durations, levodopa intakes, severities of rigidity or tremor, or Hoehn-Yahr stages has been suggested the basal ganglia support speech produc- tion through their role in programming and initiation by [34]. From another perspective, the semantic task could distin- modulating the activity of premotor areas (supplementary guish patients with PD and AD. Although patients with PD motor area, presupplementary motor area, and dorsolateral were impaired in semantic fluency, they were not as severe prefrontal cortex) [26]. The disruption of striatal dopami- as patients with AD. The AD patients not only generated fewer correct words, but also repeated the same word with a smaller nergic transmission in patients with PD may impair this modulation: The dopamine transporter availability in basal interval (e.g., dog-cat-horse-dog), indicating that the subtle ganglia was directly associated with frontal functions (i.e., deficit could underlie the differential diagnosis. In addition, attention/working memory and executive functions) [27]. the presence of semantic fluency deficits in PD has been iden- The decline of verbal fluency after pallidotomy in patients tified as a potent risk factor of the development of PD related dementia [35]. with PD may be due to surgical microlesions affecting (a.u.) Counts (a.u.) Counts (a.u.) Counts Times Score Score Times 6 Behavioural Neurology This study has limitations. First, cultural and linguistic References differences might be confounding factors. The AD group [1] A. L. R. Adlam, S. Bozeat, R. Arnold, P. Watson, and J. R. was from the database and was assessed by different experi- Hodges, “Semantic knowledge in mild cognitive impairment menters and in different cultures. Some studies also showed and mild Alzheimer’s disease,” Cortex, vol. 42, no. 5, that language and cultural differences have an impact on pp. 675–684, 2006. verbal fluency scores [36]; nevertheless, previous studies [2] N. Unsworth, G. J. Spillers, and G. A. Brewer, “Variation in analyze the semantic data from different cultures (Greece verbal fluency: a latent variable analysis of clustering, switch- and France) by using extended and unified methods [37]. ing, and overall performance,” Quarterly Journal of Experi- It is worthy to examine the difference of semantic fluency mental Psychology, vol. 64, no. 3, pp. 447–466, 2011. in PD patients under different cultural backgrounds. Second, [3] T. Azuma, K. A. Bayles, R. F. Cruz et al., “Comparing the dif- group differences and individual variability in graph param- ficulty of letter, semantic, and name fluency tasks for normal eters have not been linked with the structural integrity of elderly and patients with Parkinson’s disease,” Neuropsychol- brains. Third, some studies have shown that the perfor- ogy, vol. 11, no. 4, pp. 488–497, 1997. mance of verbal fluency was improved when the PD patients [4] F. Pasquier, F. Lebert, L. Grymonprez, and H. Petit, “Verbal were assessed with versus without levodopa treatment [38]. fluency in dementia of frontal lobe type and dementia of Alz- The PD patients scored higher in the semantic fluency task heimer type,” Journal of Neurology, Neurosurgery, and Psychi- when they were treated with rasagiline than placebo [39]. atry, vol. 58, no. 1, pp. 81–84, 1995. It would be of interest to assess the impact of medication [5] A. K. Troyer, M. Moscovitch, and G. Winocur, “Clustering and on the topology of speech graphs in future studies. switching as two components of verbal fluency: evidence from younger and older healthy adults,” Neuropsychology, vol. 11, no. 1, pp. 138–146, 1997. 5. Conclusion [6] A. K. Troyer, M. Moscovitch, G. Winocur, L. Leach, and M. Freedman, “Clustering and switching on verbal fluency In this study, we analyzed the topology of speech graphs tests in Alzheimer’s and Parkinson’s disease,” Journal of the generated in a semantic fluency test. The speech graphs of International Neuropsychological Society, vol. 4, no. 2, patients with PD were smaller and denser than those of pp. 137–143, 1998. healthy controls but larger and more sparse than those of [7] I. Galtier, A. Nieto, J. N. Lorenzo, and J. Barroso, “Mild cogni- patients with AD. Moreover, PD patients who produced tive impairment in Parkinson’s disease: clustering and switch- smaller and denser speech graphs exhibited more severe ing analyses in verbal fluency test,” Journal of the International non-motor symptoms. Neuropsychological Society, vol. 23, no. 6, pp. 511–520, 2017. [8] D. Farzanfar, M. Statucka, and M. Cohn, “Automated indices of clustering and switching of semantic verbal fluency in Par- Data Availability kinson’s disease,” Journal of the International Neuropsycholog- ical Society, vol. 24, no. 10, pp. 1047–1056, 2018. Data have been uploaded to the figshare database https:// [9] O. Sporns, “Graph theory methods: applications in brain net- figshare.com/articles/dataset/XW_data_2017-2019_xls/ works,” Dialogues in Clinical Neuroscience, vol. 20, no. 2, pp. 111–121, 2018. [10] O. Sporns, D. R. Chialvo, M. Kaiser, and C. C. Hilgetag, “Orga- nization, development and function of complex brain net- Conflicts of Interest works,” Trends in Cognitive Sciences, vol. 8, no. 9, pp. 418– 425, 2004. The authors declare that they have no conflicts of interest. [11] O. Sporns and J. D. Zwi, “The small world of the cerebral cor- tex,” Neuroinformatics, vol. 2, no. 2, pp. 145–162, 2004. [12] L. Bertola, N. B. Mota, M. Copelli et al., “Graph analysis of ver- Authors’ Contributions bal fluency test discriminate between patients with Alzhei- mer’s disease, mild cognitive impairment and normal elderly JM and ZY designed the study. GZ and JM collected the controls,” Frontiers in Aging Neuroscience, vol. 6, p. 185, 2014. data. GZ analyzed the data. GZ and ZY wrote the original draft of the manuscript. JM and PC reviewed and edited [13] M. B. Mota, R. Furtado, P. P. C. Maia, M. Copelli, and S. Ribeiro, “Graph analysis of dream reports is especially infor- the manuscript. All authors approved the submitted version. mative about psychosis,” ScientificReports, vol. 4, p. 3691, 2015. GZ and JM equally contributed to this work and share first [14] R. B. Postuma, D. Berg, M. Stern et al., “MDS clinical diagnos- authorship. tic criteria for Parkinson’s disease,” Movement Disorders, vol. 30, no. 12, pp. 1591–1601, 2015. Acknowledgments [15] C. L. Tomlinson, R. Stowe, S. Patel, C. Rick, R. Gray, and C. E. Clarke, “Systematic review of levodopa dose equivalency We are grateful to Sha Liu, Shaoyang Ma, and Minghong Su reporting in Parkinson’s disease,” Movement Disorders, for their assistance in data acquisition. This work was sup- vol. 25, no. 15, pp. 2649–2653, 2010. ported by the National Natural Science Foundation of China [16] A. Lanzi, A. Lindsay, and M. Bourgeois, “Verbal fluency in (31961133025 to Z.Y.) and the National Key Research and dementia: changes over time,” in American School Health Development Program of China (2018YFC1312001 to P.C.). Association Conference., Saint Louis, Missouri, USA, 2017. Behavioural Neurology 7 [17] I. Obeso, E. Casabona, M. L. Bringas, L. Álvarez, and [32] D. De Gaspari, C. Siri, M. Di Gioia et al., “Clinical correlates M. Jahanshahi, “Semantic and phonemic verbal fluency in Par- and cognitive underpinnings of verbal fluency impairment kinson’s disease: influence of clinical and demographic vari- after chronic subthalamic stimulation in Parkinson’s disease,” ables,” Behavioural Neurology, vol. 25, no. 2, pp. 111–118, Parkinsonism & Related Disorders, vol. 12, no. 5, pp. 289–295, 2012. 2006. [18] O. W. Wong, A. Y. Chan, A. Wong et al., “Eye movement [33] N. Kamble, R. Yadav, A. Lenka, K. Kumar, B. C. Nagaraju, and parameters and cognitive functions in Parkinson’s disease P. K. Pal, “Impaired sleep quality and cognition in patients of patients without dementia,” Parkinsonism & Related Disor- Parkinson’s disease with REM sleep behavior disorder: a com- ders, vol. 52, pp. 43–48, 2018. parative study,” Sleep Medicine, vol. 62, pp. 1–5, 2019. [19] A. F. Barbosa, M. C. Voos, J. Chen et al., “Cognitive or [34] S. Auriacombe, M. Grossman, S. Carvell, S. Gollomp, M. B. cognitive-motor executive function tasks? Evaluating verbal Stern, and H. I. Hurtig, “Verbal fluency deficits in Parkinson’s fluency measures in people with Parkinson’s disease,” BioMed disease,” Neuropsychology, vol. 7, no. 2, pp. 182–192, 1993. Research International, vol. 2017, 7 pages, 2017. [35] C. H. Williams-Gray, S. L. Mason, J. R. Evans et al., “The Cam- [20] J. McDowd, L. Hoffman, E. Rozek et al., “Understanding ver- PaIGN study of Parkinson’s disease: 10-year outlook in an bal fluency in healthy aging, Alzheimer’s disease, and Parkin- incident population-based cohort,” Journal of Neurology, Neu- son’s disease,” Neuropsychology, vol. 25, no. 2, pp. 210–225, rosurgery & Psychiatry, vol. 84, no. 11, pp. 1258–1264, 2013. [36] A. St-Hilaire, C. Hudon, G. T. Vallet et al., “Normative data for [21] N. B. Araujo, M. L. Barca, K. Engedal, E. S. Coutinho, A. C. phonemic and semantic verbal fluency test in the adult Deslandes, and J. Laks, “Verbal fluency in Alzheimer’s disease, French-Quebec population and validation study in Alzhei- Parkinson’s disease, and major depression,” Clinics, vol. 66, mer’s disease and depression,” The Clinical Neuropsychologist, no. 4, pp. 623–627, 2011. vol. 30, no. 7, pp. 1126–1150, 2016. [22] G. Hickok and D. Poeppel, “The cortical organization of [37] N. Linz, J. Tröger, J. Alexandersson, M. Wolters, A. König, and speech processing,” Nature Reviews Neuroscience, vol. 8, P. Robert, “Predicting dementia screening and staging scores no. 5, pp. 393–402, 2007. from semantic verbal fluency performance,” in 2017 IEEE International Conference on Data Mining Workshops [23] F. Rodriguez-Porcel, J. Wilmskoetter, C. Cooper et al., “The (ICDMW), pp. 719–728, New Orleans, LA, USA, 2017. relationship between dorsal stream connections to the caudate and verbal fluency in Parkinson disease,” Brain Imaging and [38] A. M. Gotham, R. G. Brown, and C. D. Marsden, “‘Frontal’ Behavior, vol. 15, no. 4, pp. 2121–2125, 2021. cognitive function in patients with Parkinson’s disease ‘on’ and ‘off’ levodopa,” Brain, vol. 111, no. 2, pp. 299–321, 1988. [24] S. L. Thompson-Schill, M. D’Esposito, G. K. Aguirre, and M. J. Farah, “Role of left inferior prefrontal cortex in retrieval of [39] H. A. Hanagasi, H. Gurvit, P. Unsalan et al., “The effects of semantic knowledge: a reevaluation,” Proceedings of the rasagiline on cognitive deficits in Parkinson’s disease patients National Academy of Sciences, vol. 94, no. 26, pp. 14792– without dementia: a randomized, double-blind, placebo-con- 14797, 1997. trolled, multicenter study,” Movement Disorders, vol. 26, no. 10, pp. 1851–1858, 2011. [25] G. Robinson, T. Shallice, M. Bozzali, and L. Cipolotti, “Con- ceptual proposition selection and the LIFG: neuropsychologi- cal evidence from a focal frontal group,” Neuropsychologia, vol. 48, no. 6, pp. 1652–1663, 2010. [26] G. E. Alexander, M. R. DeLong, and P. L. Strick, “Parallel orga- nization of functionally segregated circuits linking basal gan- glia and cortex,” Annual Review of Neuroscience, vol. 9, no. 1, pp. 357–381, 1986. [27] S. J. Chung, H. S. Yoo, J. S. Oh et al., “Effect of striatal dopa- mine depletion on cognition in de novo Parkinson’s disease,” Parkinsonism & Related Disorders, vol. 51, pp. 43–48, 2018. [28] R. M. De Bie, P. R. Schuurman, D. A. Bosch, R. J. De Haan, B. Schmand, and J. D. Speelman, “Outcome of unilateral palli- dotomy in advanced Parkinson’s disease: cohort study of 32 patients,” Journal of Neurology, Neurosurgery & Psychiatry, vol. 71, no. 3, pp. 375–382, 2001. [29] A. I. Tröster, S. P. Woods, and J. A. Fields, “Verbal fluency declines after Pallidotomy: an interaction between task and lesion laterality,” Applied Neuropsychology, vol. 10, no. 2, pp. 69–75, 2003. [30] C. I. Higginson, D. S. King, D. Levine, V. L. Wheelock, N. O. Khamphay, and K. A. Sigvardt, “The relationship between executive function and verbal memory in Parkinson’s disease,” Brain and Cognition, vol. 52, no. 3, pp. 343–352, 2003. [31] J. D. Henry and J. R. Crawford, “Verbal fluency deficits in Par- kinson’s disease: a meta-analysis,” Journal of the International Neuropsychological Society, vol. 10, no. 4, pp. 608–622, 2004.

Journal

Behavioural NeurologyHindawi Publishing Corporation

Published: Apr 23, 2022

There are no references for this article.