Speech's syllabic rhythm and articulatory features produced under different auditory feedback conditions identify Parkinsonism

Speech's syllabic rhythm and articulatory features produced under different auditory feedback conditions identify Parkinsonism | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Speech's syllabic rhythm and articulatory features produced under different auditory feedback conditions identify Parkinsonism Ángeles Piña Méndez, Alan Taitz, Oscar Palacios Rodríguez, Ildefonso Rodríguez Leyva, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3937556/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Background Parkinsonism diagnostic tests based on speech samples have been reported with promising results. However, although abnormal auditory feedback integration during speech production and impaired rhythmic organization of speech have been shown in Parkinsonism, these observations have not been integrated into diagnostic tests. Objective To identify Parkinsonism and evaluate the power of a novel speech behavioral test (based on rhythmically repeating syllables under different auditory feedback conditions). Methods Thirty parkinsonism patients and thirty healthy subjects completed the study. Participants were instructed to repeat the PA-TA-KA syllable sequence rhythmically, whispering and speaking aloud under different listening conditions. The produced speech samples were preprocessed, and parameters were extracted. Classical, unpaired comparisons were conducted between patients and controls. Significant parameters were fed to a supervised machine-learning algorithm differentiating patients from controls, and the accuracy, specificity, and sensitivity were computed. Results Difficulties in whispering and articulating under altered auditory feedback conditions, delayed speech onset, and alterations in rhythmic stability were found in the group of patients compared to controls. A machine learning algorithm trained on these parameters to differentiate patients from controls reached an accuracy of 85.4%, a sensitivity of 87.8%, and a specificity of 83.1%. Conclusions The current work represents a pilot trial, showing the potential of the introduced behavioral paradigm as an objective and accessible (in cost and time) diagnostic test. Health sciences/Neurology/Neurological disorders/Peripheral neuropathies Health sciences/Biomarkers/Diagnostic markers Figures Figure 1 Figure 2 Figure 3 INTRODUCTION Parkinson's disease (PD) is a complex and frequent neurodegenerative disorder characterized by an extensive collection of motor and non-motor symptoms, a variable response to treatment, and a generally progressive course 1,2 . PD is considered a public health problem, as it is the second most common neurodegenerative disorder after Alzheimer's disease 3 . Furthermore, the Global Burden of Disease study has estimated that PD cases will double from 7 million in 2015 to 13 million in 2040 4 . Due to its complexity, variability, and subtypes, PD represents a challenge in its diagnosis because it is mainly based on history and physical examination 5 . Moreover, there are other tests to confirm the diagnosis of PD 5 , but these tend to be expensive and unavailable in most imaging centers. Therefore, attention has been focused on specific, noninvasive, and low-cost biomarkers. Dysarthria (i.e., abnormalities in different aspects of speech production) represents an early symptom of PD and atypical Parkinsonian syndromes 6,7 . In line with this observation, several studies explored the potential of different speech and voice features as PD biomarkers 8 . Several of these studies rely on the diadochokinetic task (DDK), which evaluates articulatory and rhythmic speech impairments by having a subject rhythmically repeat a consonant-vowel combination (typically, the PA-TA-KA sequence, which involves different places of articulation: bilabial, alveolar, and velar) 9 . Based on this task, low-cost vocal tests have been developed, showing a high success rate in identifying PD in its early stages 9–11 . All these tests rely on the analysis of speech acoustic features in the millisecond scale (e.g., Mel-frequency cepstral coefficients or phonation jitter/shimmer). However, longer timescales features, such as speech articulatory rate and regularity, have also been identified as altered in PD 12 , but remain excluded from vocal tests. Furthermore, Rusz and colleagues 13 show abnormal speech rhythm stability in PD, as well as in atypical Parkinsonian syndromes. In addition to dysarthria, it has been shown that the effects produced on the ongoing speech by an unexpected modulation of the auditory feedback differ between PD patients and healthy controls 14–16 . This result made researchers hypothesize that PD patients have abnormal speech auditory-motor integration. Furthermore, interventions based on altered auditory feedback have successfully restored speech fluency in patients with dysarthria 17 . Bringing together the existing literature, we developed a modified diadochokinetic task in which subjects are instructed to repeat the syllables PA-TA-KA under different auditory feedback conditions. Using this test, we explore significant differences between patients with Parkinsonism and healthy controls in their rhythmic and articulatory speech features. Specifically, we assessed general speech features such as a subject’s ability to whisper, syllabic rhythm stability, or several syllable-level errors instead of the typically used Mel-frequency cepstral coefficients or phonation jitter/shimmer 9–11 . Furthermore, using a supervised learning technique, we show this paradigm's high accuracy, sensibility, and specificity to identify Parkinsonism. METHODS Participants Two cohorts of gender-matched participants (patients and healthy controls) completed this study. Both cohorts comprised Mexican subjects, all of whom were native Spanish speakers. Signed informed consent was obtained from all participants. The protocol was evaluated and approved by the Ethics Committee of the Faculty of Psychology of San Luis Potosí (Registration number: 2131082021). All methods were performed in accordance with the relevant guidelines and regulations. A total of 30 participants (17 female and 13 male) composed the group of patients, 28 of whom were diagnosed with PD (14 rigid-akinetic and 14 with tremor), two patients with atypical Parkinsonism: progressive supranuclear palsy (PSP) and multiple system atrophy (MSA). In line with previous studies 13,18 , we hypothesize that speech production abnormalities relate to basal ganglia dysfunction; accordingly, they will similarly impact PD, PSP, and MSA patients. Consequently, this first pilot of the test aims to identify Parkinsonism deriving from a basal ganglia dysfunction and not specifically PD. The patients were diagnosed by a neurologist, following the Movement Disorder Society (MDS) clinical diagnosis criteria for PD 19 , NINDS-PSP for PSP 20 , and the consensus diagnostic criteria for MSA 21 . The diagnosis was made before the start of this study. The ages of the patients ranged from 38 to 88 years (mean 68.3, SD 10.82). We did not modify the treatment during the task; the patients were in an ON state, except for two participants who did not take their medication on the day of the study. Depression symptoms were assessed with the short version of the 15-item Geriatric Depression Scale. In addition, cognitive impairment was measured using the Mini-Mental State Examination. The group of healthy controls comprised 30 participants (17 female and 13 male) who did not report having neurologic or psychiatric disorders (mean = 64.2, SD = 9.51, range = 45–93). Both cohorts of participants completed a short questionnaire about their musical experience and educational level. There were no significant differences between cohorts in the assessed demographic features (see Supplementary Fig. 1). Auditory stimulus Target audio : The syllables “pa,” “ta,” and “ka,” spoken aloud and whispered, were recorded by a female Spanish speaker. Praat software 22 was used to make each syllable last 250 ms. Next, five repetitions of the sequence pa-ta-ka, spoken or whispered, were concatenated, generating two rhythmic (4 syllables/sec), 3.75-second-long audio files: the whispered and the spoken targets. Exogenous speech A 5.5-second-long audio file comprising a rhythmic train of syllables was synthesized at 16 Hz using the MBROLA text-to-speech synthesizer 23 with the Spanish Male Voice “es2.” A set of 13 Spanish syllables (“te,” “bi,” “ki,” “pu,” “bo,” “la,” “su,” “go,” “mu,” “rra,” “le,” “do,” “fe") were repeatedly and randomly concatenated to achieve the desired length. All phonemes were equal in pitch (200 Hz), and the duration was set to 0.125 ms. Noise A 5.5-second-long audio file comprising white noise was synthesized using Matlab 24 . Procedure All participants performed the experimental test in a room with low ambient noise. They sat in front of a computer wearing insert earphones (ETYMOTIC ER1), and we recorded their vocalizations with a microphone connected to the laptop (Marantz Pro M4U). All instructions appeared written on the computer screen and were verbally reinforced by the examiner. The experimental test consisted of 28 trials and had a total duration of approximately 10 minutes. Each trial consisted of passive listening and repetition phases (see Fig. 1). During the passive listening phase, the target audio was presented through the earplugs, and participants were instructed to pay attention to it while fixing their gaze on a black dot centered on the screen. The target audio lasted 3.75 seconds and comprised a repetition of the pa-ta-ka sequence, whispered or spoken aloud (see the Auditory Stimuli section). At the end of the audio playback, the dot on the screen turned green, signaling the beginning of the repetition phase. During this phase, participants were instructed to continuously echo the target audio, matching the presented rhythm and voice level (i.e., whispering or speaking aloud). After 5.5 seconds, the green dot turned red, prompting participants to stop vocalizing and to wait until the subsequent trial. The intertrial interval was set to two seconds. Each trial belonged to one of four possible conditions (see Fig. 1): Natural feedback: The target audio comprised loud speech, and no auditory stimulus was presented during the repetition phase. Reduced feedback: The target audio comprised whispered speech, and no auditory stimulus was presented during the repetition phase. Masked feedback: The target audio comprised whispered speech, and the noise audio was played during the repetition phase, masking the participant’s voice auditory feedback. Replaced feedback: The target audio comprised whispered speech, and the exogenous speech audio was played during the repetition phase, masking the participant’s voice auditory feedback. We presented seven trials per condition in a randomized order. Data preprocessing For each participant and trial condition, five parameters were extracted (see Fig. 2 ): ( 1 ) the number of whispered trials, ( 2 ) the number of speech errors, ( 3 ) the mean reaction time, ( 4 ) the mean syllabic rate, and ( 5 ) the rhythmic stability. The first author manually computed the number of whispered trials, speech errors, and reaction time. Praat software was used to listen to and visualize the acoustic signals. Each trial was categorized as whispered if voiced speech (i.e., with activation of the vocal folds) occurred for less than 0.55 seconds (10% of the trial length). An error was identified if the participant: ( 1 ) repeated a syllable (e.g., “pa-ta- ta -ka”), ( 2 ) exchanged a syllable (e.g., “pa- ka-ta ”), or if a syllable was wrongly articulated (e.g., “pa- tra -ka”). Reaction time was computed as the speech onset time and averaged across trials of the same condition. To estimate the rhythmic features of the spoken samples, we calculated the speech envelopes as the absolute value of the acoustic signal’s Hilbert transform 25 . Subsequently, we used the fast Fourier transform to extract the spectrum of each trial’s envelope. Each trial's syllabic rate was estimated as the frequency value with the maximal power (see Fig. 2 ). The estimated syllabic rates were averaged across tests of the same condition. Next, to explore the stability of the rhythmic structure across trials of the same condition, we computed the correlation matrix between their envelope spectra. The rhythmic stability was estimated as the mean value of the lower diagonal elements of the matrix (see Fig. 2 ). Data Analysis Statistical comparisons Non-parametric inferential statistics compared the patient group with the healthy controls. More specifically, we used the Mann-Whitney test for two independent samples. All reported p-values were corrected using a false discovery rate approach 26 for multiple comparisons. All procedures were carried out in Matlab. Machine learning A random forest classifier model was trained on the binary classes of patients and healthy controls using the data obtained from the five evaluated parameters. The metrics sensitivity, specificity, and accuracy were considered to assess the model’s performance. The leave-one-out cross-validation method was used to assess the model generalization performance. This method consists of several iterations of the training-testing data sets, where in each iteration, one participant is selected to test the model, and all others are used to train it. All plausible training-testing combinations were evaluated, and the predicted outcomes were used to compute the accuracy, sensitivity, and specificity. This process was repeated 100 times, and the results were averaged. This process was conducted using the sklearn library in Python 27 . RESULTS We assessed the goodness of a behavioral test based on speech samples to diagnose Parkinsonism. Thus, we evaluated patients and healthy controls in a task consisting of continuously and rhythmically (four syllables per second) repeating the syllables “pa-ta-ka” under different auditory feedback conditions: Normal (speaking aloud while hearing the produced sounds), Reduced (whispering while hearing the produced sounds), Masked (whispering while hearing white noise), Replaced (whispering while hearing an alien voice). While the normal condition matches the DDK task used in previous studies 9 , the other three integrate a modulation in the participants’ voice feedback. For the Reduced condition, feedback is not entirely removed but diminished, given the whisper-low volume. In the Masked and Replaced conditions, the feedback is completely blocked, but in the last one, participants get an external cue of the intended syllabic rate. It has been shown in healthy participants that whispering while listening to an external stable rhythm leads some individuals to synchronize the produced syllabic pace 28 . Additionally, PD patients show abnormal rhythmic speech entrainment with a model speaker 29 . From the obtained recordings, we computed five parameters for each feedback condition: ( 1 ) whispering ability, ( 2 ) the number of articulatory errors, ( 3 ) reaction time to initiate speech, ( 4 ) syllabic rhythm, and ( 5 ) rhythmic stability (for more details, see the Methods section). Two different analyses were conducted on these parameters. First, we compared the parameters for patients and controls across feedback conditions. This allowed us to identify the speech features being abnormal in patients. Secondly, we fed the parameters to a supervised learning algorithm to assess the predictive power of the parameters’ combination to differentiate patients from healthy participants. To start, we explored if patients could correctly follow the instructions or if they showed more difficulties than control participants. Specifically, we investigated whether they could whisper in the conditions with this requirement (i.e., Reduced, Masked, and Replaced feedback conditions). We compared the percentage of whispered trials between patients and controls. The results showed significant differences between the group of patients with Parkinsonism and healthy subjects in the Masked and Replaced feedback conditions (see Fig. 3 a; Reduced: Patients: M = 65%, SD = 38%; Control: M = 79%, SD = 37%; p = 0.073; Masked: Patients: M = 44%, SD = 35%; Control: M = 72%, SD = 38%; p = 0.008; Replaced: Patients: M = 37%, SD = 39%; Control: M = 70%, SD = 41%; p = 0.008). Patients had difficulties in whispering the syllables when presented with altered feedback. Given that the number of whispered trials in the Masked and Replaced feedback conditions significantly differed between patients and controls, differences in the rest of the parameters were only assessed between Normal and Reduced feedback conditions. This was done because groups differ in the number of spoken-aloud trials for the other feedback conditions, making it impossible to disentangle whether the differences (if observed) derive from the auditory feedback state situation or from speaking aloud. For all other explored parameters, we found that: (i) patients made more speech errors than controls, only for the Reduced feedback condition (see Fig. 3 b; Normal: Patients: M = 0.047 err/sec, SD = 0.072 err/sec; Control: M = 0.044 err/sec, SD = 0.126 err/sec; p = 0.072; Reduced: Patients: M = 0.1 err/sec, SD = 0.117 err/sec; Control: M = 0.022 err/sec, SD = 0.076 err/sec; p = 0.002); (ii) patients had slower reaction times than controls, only for the Reduced feedback condition (see Fig. 3 c; Normal: Patients: M = 0.60 sec, SD = 0.20 sec; Control: M = 0.49 sec, SD = 0.15 sec; p = 0.081; Reduced: Patients: M = 0.52 sec, SD = 0.21 sec; Control: M = 0.39 sec, SD = 0.14 sec; p = 0.022); (iii) there was no significant difference between groups in their mean syllabic rate (see Fig. 3 d; Normal: Patients: M = 3.7 Hz, SD = 0.82 Hz; Control: M = 4.05 Hz, SD = 0.71 Hz; p = 0.105; Reduced: Patients: M = 3.55 Hz, SD = 0.84 Hz; Control: M = 3.94 Hz, SD = 0.68 Hz; p = 0.105); and (iv) patients showed less rhythmic stability than controls for the Normal feedback condition (see Fig. 3 e; Normal: Patients: M = 0.71, SD = 0.16; Control: M = 0.82, SD = 0.13; p = 0.015; Reduced: Patients: M = 0.81, SD = 0.11; Control: M = 0.85, SD = 0.13; p = 0.128). All the results were recovered when the two patients with atypical Parkinsonism were excluded and the analyses were restricted to PD (see Supplementary Fig. 2). Once the differences between patients and healthy controls had been established by statistically comparing the parameters’ distributions, we fed the relevant parameters into a random forest classifier to differentiate between groups (see Methods). This procedure allowed us to estimate the predictive power of these speech features to identify patients from the general population and generalize the reported results. The five variables showing significant differences between groups are the percentage of whispering trials in Masked feedback, rate of whispering tests in Replaced feedback, speech error per second in Reduced feedback, mean reaction time in Reduced feedback, and rhythmic stability in Normal feedback. First, we trained and tested the classifier with all five variables and computed accuracy, specificity, and sensitivity (see Table 1 , Model 1). Given that the two altered feedback conditions only contributed to the percentage of whispered trials, we explored if both were increasing the predictive power of the instrument or if they carried redundant information. To do so, we evaluated the classifier with the percentage of whispered trials in only one or the other condition (see Table 1 , Models 2 and 3). Results show that performance does not increase by including both conditions, indicating that the Masked feedback condition can be removed from the test. Finally, we tested whether the participants’ age helped the classifier's performance, which was not the case (see Table 1 , Model 4). Table 1 Performance of the random forest classifier trained and tested on different sets of parameters. Parameters are specified as pairs of features and condition s. Features are as follows: whis = percentage of whispered trials; Sp. Err = speech errors per second; RT = mean reaction time; Rhy. Stab = rhythm stability. Conditions are as follows: NF = Normal feedback; Red. F = Reduced feedback; MF = Masked feedback; Rep.F = Replaced feedback. Accuracy, sensitivity, and specificity are reported as a percentage. Model Parameters fed into the classifier Accuracy Sensitivity Specificity 1 %whis,RepF + %whis,MF + Sp.Err, RedF + RT, RedF + Rhy.Stab, NF 83.0 81.8 84.1 2 %whis,RepF + Sp.Err, RedF + RT, RedF + Rhy.Stab, NF 86.1 87.8 84.3 3 %whis,MF + Sp.Err, RedF + RT, RedF + Rhy.Stab, NF 85.4 87.8 83.1 4 %whis,RepF + Sp.Err, RedF + RT, RedF + Rhy.Stab,NF + age 82.0 80.3 83.7 DISCUSSION The present research focused on identifying Parkinsonism speech impairments under different auditory feedback conditions. More precisely, five parameters were studied: 1) whispering ability, 2) the number of articulatory errors, 3) reaction time to initiate speech, 4) syllabic rhythm, and 5) rhythmic stability. This is proposed as a pilot phase for developing an objective diagnostic test. The participant's ability to whisper in Reduced, Masked, and Replaced feedback conditions was analyzed, and it was found that the patient group presented more difficulties than the control group in adapting their speech output (i.e., whispering) when their auditory feedback was modified. Given that whispering was not wholly impaired in the patient group and that the difference between groups appeared only when the auditory feedback was masked or replaced, it can be inferred that the patients have an abnormal enhancement of the Lombard effect (i.e., an increase in voice intensity in response to an increase in the ambient noise level). It has been suggested that (i) the Lombard effect occurs unconsciously, driven by a subcortical mechanism, and (ii) the activation of such a subcortical mechanism can be modulated by a cortical network, allowing voluntary control of the effect 30 . The pattern of results obtained in this study suggests that the cortical network is affected in the Parkinsonian cohort: when instructed to whisper, controls, but not patients, can suppress the Lombard effect and manage to maintain low speech intensity in noisy environments (i.e., listening to white noise in the Masked feedback condition and an alien voice in the Replaced feedback condition). In line with this hypothesis, it has been shown that PD patients have decreased activity in frontotemporal regions 31 , and studies based on deep brain stimulation in PD patients provide evidence that the connectivity between frontal and sub-thalamic brain regions explains individual differences in inhibition 32 . Articulatory errors are prevalent in the patient group when whispering the syllables. It has been suggested that PD patients rely more on auditory feedback due to impaired input motor control and somatosensory feedback 33,34 , which can explain the present findings. As there is no vocal fold vibration during whispering, it is more difficult for the acoustic signal to be perceived by the auditory system 35 , and speech-motor production relies on the forward model and somatosensory feedback, mechanisms proposed to be abnormal in PD patients. Accordingly, studies based on repetitive transcranial magnetic stimulation have shown that stimulating the auditory cortex improves articulation in PD patients during an overt-speech task 34 . It was identified that the patients recruited in our study have difficulty initiating the motor process, such as articulating syllables. Furthermore, the increase in reaction time when starting speech occurs regardless of the condition in which they perform the task (Normal and Reduced feedback). The findings are consistent with the literature; different referents mention that PD patients struggle to initiate movement 32,36,37 . Regarding rhythmic features, we found that while syllabic rhythm did not differ between groups in any feedback condition, the rhythmic stability (i.e., how similar the rhythmic structure was across trials) decreased in the PD patients. Studies on PD patients’ speech rhythm present contradictory results. While some authors report a slowdown 38 , others report an increment in the rate 39 or no significant difference between patients and controls 40 . Here, we showed that while absolute rhythm did not differentiate patients from controls, rhythm stability did, suggesting this last measurement is more accurate in diagnosing Parkinsonism. Contrary to the pattern of results obtained for the articulatory errors, rhythmic stability significantly differed between groups for the Normal feedback but not the Reduced feedback condition. This contrast suggests that the circuit responsible for monitoring and correcting articulatory errors does not overlap with the one supporting the stability of the speech rhythm. It has been proposed that speech rhythm emerges due to the biophysical properties of the brain areas in charge of generating speech 41 , and the interaction of motor and auditory areas 42 modulates it. Under this framework, the observation that the rhythm becomes unstable when increasing the auditory feedback hints towards an abnormal cortical interaction between the frontal and temporal regions. The main goal of the current study was to assess the goodness of a short and easy-to-implement behavioral screening test based on speech samples. Similar designs to diagnose PD 11,43,44 , and atypical parkinsonism 13 , have been previously reported in the literature with promising results. In those studies, researchers typically asked participants to complete different speech tasks, such as sustained phonation, running speech, and, as in the current work, the DDK task (i.e., continuously repeating the syllables /pa/ /ta/ /ka/). They also extracted different acoustic parameters from the speech samples and entered them into a machine-learning algorithm to distinguish patients from healthy controls. Such strategies obtained high accuracy values from 85–99%. Two main aspects determine the current work from previous studies. On the one hand, although it has been shown that modified auditory feedback impacts PD patients differently from controls 14–16 , this observation has not been previously included in the design of speech-based screening tests. We integrate this observation by asking participants to complete the classic DDK task under different auditory feedback conditions. On the other hand, we evaluated speech features on a different time scale than those typically computed (e.g., formant periodicity correlations, Mel-frequency cepstral coefficients, phonation jitter, phonation shimmer phonation noise, and voice fundamental frequency variations 44 ). Here, we focused on general speech features, such as whispering, throughout the trial or syllabic scale characteristics, such as rhythm stability or number of errors at the syllabic level. A machine learning algorithm trained on such parameters to differentiate Parkinsonian patients from healthy controls shows accuracy within the range of previously reported results 45 . These results present novel experimental conditions (e.g., whispering, speaking under different listening conditions, and trained syllabic rate) and parameters (e.g., whispered trials, reaction time, rhythm stability, and speech errors at the syllabic level) as valuable tools to be introduced into existing screening behavioral tests to identify Parkinsonism. Additionally, it is important to highlight the accessibility of the piloted diagnostic test, which only requires a standard PC, a set of headphones, and a microphone and lasts less than 4 minutes (i.e., only three of the four evaluated conditions are required, and each condition comprises seven 11.25-second trials). A limitation of the present study is its small sample size (30 patients and 30 controls). However, this sample size is sufficient for a pilot study and reaches statistical significance. Furthermore, despite the small sample size, the current work addresses the need for cross-linguistic studies of dysarthria in PD 46 (i.e., this study was conducted on Mexican participants, a Spanish-speaking population typically overlooked in the existing literature). Our major limitation is mainly in the number of subjects with atypical Parkinsonism because, in addition to being rare diseases, there were few spaces and clinical records available, which prevented us from recruiting more participants. Another potential limitation is the lack of control over the influence of levodopa or other medications since the patients were on medication during the application of the instrument. There needs to be more interest in addressing this issue, which makes it difficult to access reliable data and collaborate with other institutions. However, some studies suggest that speech fluency (or the lack of it) in Parkinson's patients is not modulated by levodopa 47–49 . The present protocol studies rhythmic and articulatory changes related to PD, MSA, and PSP in the speech production system. Although it does not address any therapeutic strategy or alternative treatment, it lays the foundation for developing noninvasive, low-cost, and easy-to-apply diagnostic tests. CONCLUSIONS Syllabic rhythm stability, reaction time, whispering ability, and syllable-level articulatory errors under different auditory feedback conditions differentiate Parkinsonian patients from healthy controls. An automatic detection algorithm trained on these parameters showed an accuracy of 85.4% in distinguishing patients from controls. The current work represents a pilot trial, showing the potential of the introduced behavioral paradigm as an objective and accessible (in cost and time) diagnostic test. DECLARATIONS ACKNOWLEDGEMENTS We thank Jessica González Norris for proofreading the manuscript. This work was supported by DGAPA-PAPIIT IA200223 and the IBRO Return Home Fellowship (MFA). The authors report no relevant conflicts of interest. AUTHORS CONTRIBUTION MFA conceived the project. MFA, OPR, and IRL supervised the project. APM collected the data. AT, APM, and MFA analyzed the data. MFA, IRL, and APM wrote the manuscript. All authors read and approved the manuscript. COMPETING INTERESTS The authors report no competing interests. DATA AVAILABILITY The data supporting the findings of this study are available as Supplementary Data. All other data and computer code used to generate results are available upon request from the corresponding author. REFERENCES Pilotto, A. et al. Plasma NfL, clinical subtypes and motor progression in Parkinson’s disease. Parkinsonism & Related Disorders 87 , 41–47 (2021). Titova, N., Padmakumar, C., Lewis, S. J. G. & Chaudhuri, K. R. Parkinson’s: a syndrome rather than a disease? J Neural Transm 124 , 907–914 (2017). Draoui, A., El Hiba, O., Aimrane, A., El Khiat, A. & Gamrani, H. Parkinson’s disease: From bench to bedside. Revue Neurologique 176 , 543–559 (2020). Jankovic, J. & Tan, E. K. Parkinson’s disease: etiopathogenesis and treatment. J Neurol Neurosurg Psychiatry 91 , 795–808 (2020). Armstrong, M. J. & Okun, M. S. Diagnosis and Treatment of Parkinson Disease: A Review. JAMA 323 , 548–560 (2020). Pinto, S. et al. Treatments for dysarthria in Parkinson’s disease. The Lancet Neurology 3 , 547–556 (2004). Critchley, E. M. Speech disorders of Parkinsonism: a review. J Neurol Neurosurg Psychiatry 44 , 751–758 (1981). Ngo, Q. C. et al. Computerized analysis of speech and voice for Parkinson’s disease: A systematic review. Computer Methods and Programs in Biomedicine 226 , 107133 (2022). Montaña, D., Campos-Roca, Y. & Pérez, C. J. A Diadochokinesis-based expert system considering articulatory features of plosive consonants for early detection of Parkinson’s disease. Computer Methods and Programs in Biomedicine 154 , 89–97 (2018). Vasquez-Correa, J. C., Arias-Vergara, T., Schuster, M., Orozco-Arroyave, J. R. & Nöth, E. Parallel Representation Learning for the Classification of Pathological Speech: Studies on Parkinson’s Disease and Cleft Lip and Palate. Speech Communication 122 , 56–67 (2020). Rusz, J. et al. Acoustic voice and speech disorders assessment in Parkinson's disease through quick vocal test. Movement Disorders 26 , 1951–1952 (2011). Skodda, S. Aspects of speech rate and regularity in Parkinson’s disease. Journal of the Neurological Sciences 310 , 231–236 (2011). Rusz, J., Hlavnička, J., Čmejla, R. & Růžička, E. Automatic Evaluation of Speech Rhythm Instability and Acceleration in Dysarthrias Associated with Basal Ganglia Dysfunction. Frontiers in Bioengineering and Biotechnology 3 , (2015). Liu, H., Wang, E. Q., Metman, L. V. & Larson, C. R. Vocal Responses to Perturbations in Voice Auditory Feedback in Individuals with Parkinson’s Disease. PLOS ONE 7 , e33629 (2012). Senthinathan, A., Adams, S., Page, A. D. & Jog, M. Speech Intensity Response to Altered Intensity Feedback in Individuals With Parkinson’s Disease. Journal of Speech, Language, and Hearing Research 64 , 2261–2275 (2021). Blanchet, P. G. Factors influencing the efficacy of delayed auditory feedback in treating dysarthria associated with Parkinson’s disease. (2002). Lowit, A., Dobinson, C., Timmins, C., Howell, P. & Kröger, B. The effectiveness of traditional methods and altered auditory feedback in improving speech rate and intelligibility in speakers with Parkinson's disease —International Journal of Speech-Language Pathology 12 , 426–436 (2010). Kowalska-Taczanowska, R., Friedman, A. & Koziorowski, D. Parkinson's disease or atypical Parkinsonism? The importance of acoustic voice analysis in differential diagnosis of speech disorders. Brain and Behavior 10 , e01700 (2020). Postuma, R. B. et al. MDS clinical diagnostic criteria for Parkinson’s disease. Movement Disorders 30 , 1591–1601 (2015). Litvan, I. et al. Accuracy of clinical criteria for the diagnosis of progressive supranuclear palsy (Steele-Richardson-Olszewski syndrome). Neurology 46 , 922–930 (1996). Gilman, S. et al. Consensus statement on the diagnosis of multiple system atrophy. Clinical Autonomic Research 8 , 359–362 (1998). Boersma, P. & Weenink, D. PRAAT, a system for doing phonetics by computer. Glot international 5 , 341–345 (2001). Dutoit, T., Pagel, V., Pierret, N., Bataille, F. & van der Vrecken, O. The MBROLA project: towards a set of high quality speech synthesizers free of use for non commercial purposes. in Proceeding of Fourth International Conference on Spoken Language Processing. ICSLP ’96 vol. 3 1393–1396 vol.3 (1996). Matlab. (2020). Lizcano-Cortés, F. et al. Speech-to-Speech Synchronization protocol to classify human participants as high or low auditory-motor synchronizers. STAR Protocols 3 , 101248 (2022). Benjamini, Y. & Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society: Series B (Methodological) 57 , 289–300 (1995). Buitinck, L. et al. API design for machine learning software: experiences from the scikit-learn project. Preprint at https://doi.org/10.48550/arXiv.1309.0238 (2013). Mares, C., Echavarría Solana, R. & Assaneo, M. F. Auditory-motor synchronization varies among individuals and is critically shaped by acoustic features. Commun Biol 6 , 1–10 (2023). Späth, M. et al. Entraining with another person’s speech rhythm: Evidence from healthy speakers and individuals with Parkinson’s disease. Clinical Linguistics & Phonetics 30 , 68–85 (2016). Luo, J., Hage, S. R. & Moss, C. F. The Lombard Effect: From Acoustics to Neural Mechanisms. Trends in Neurosciences 41 , 938–949 (2018). Auclair-Ouellet, N. et al. Action fluency identifies different sex, age, global cognition, executive function, and brain activation profile in non-demented patients with Parkinson's disease. J Neurol 268 , 1036–1049 (2021). Mosley, P. E. et al. Subthalamic deep brain stimulation identifies frontal networks supporting initiation, inhibition, and strategy use in Parkinson's disease. NeuroImage 223 , 117352 (2020). Weerathunge, H. R., Tomassi, N. E. & Stepp, C. E. What Can Altered Auditory Feedback Paradigms Tell Us About Vocal Motor Control in Individuals With Voice Disorders? Perspectives of the ASHA Special Interest Groups 7 , 959–976 (2022). Brabenec, L. et al. Noninvasive stimulation of the auditory feedback area for improved articulation in Parkinson's disease. Parkinsonism & Related Disorders 61 , 187–192 (2019). Frühholz, S., Trost, W. & Grandjean, D. Whispering - The hidden side of auditory communication. NeuroImage 142 , 602–612 (2016). Mekyska, J. et al. Quantitative Analysis of Relationship Between Hypokinetic Dysarthria and the Freezing of Gait in Parkinson’s Disease. Cogn Comput 10 , 1006–1018 (2018). Rosin, R., Topka, H. & Dichgans, J. Gait initiation in Parkinson's disease. Movement Disorders 12 , 682–690 (1997). Ludlow, C. L., Connor, N. P. & Bassich, C. J. Speech timing in Parkinson’s and Huntington’s disease. Brain and Language 32 , 195–214 (1987). Ackermann, H., Konczak, J. & Hertrich, I. The Temporal Control of Repetitive Articulatory Movements in Parkinson’s Disease. Brain and Language 56 , 312–319 (1997). Duez, D. Syllable structure, syllable duration and final lengthening in Parkinsonian French speech. Journal of Multilingual Communication Disorders 4 , 45–57 (2006). Assaneo, M. F. & Poeppel, D. The coupling between auditory and motor cortices is rate-restricted: Evidence for an intrinsic speech-motor rhythm. Science Advances 4 , eaao3842 (2018). Poeppel, D. & Assaneo, M. F. Speech rhythms, and their neural foundations. Nature Reviews Neuroscience 21 , 322–334 (2020). Orozco-Arroyave, J. R. et al. Automatic detection of Parkinson’s disease in running speech spoken in three different languages. The Journal of the Acoustical Society of America 139 , 481–500 (2016). Moro-Velazquez, L., Gomez-Garcia, J. A., Arias-Londoño, J. D., Dehak, N. & Godino-Llorente, J. I. Advances in Parkinson’s Disease detection and assessment using voice and speech: A review of the articulatory and phonatory aspects. Biomedical Signal Processing and Control 66 , 102418 (2021). Ali, L., Zhu, C., Zhou, M. & Liu, Y. Early diagnosis of Parkinson’s disease from multiple voice recordings by simultaneous sample and feature selection. Expert Systems with Applications 137 , 22–28 (2019). Pinto, S., Chan, A., Guimarães, I., Rothe-Neves, R. & Sadat, J. A cross-linguistic perspective to the study of dysarthria in Parkinson’s disease. Journal of Phonetics 64 , 156–167 (2017). Cantiniaux, S. et al. Comparative analysis of gait and speech in Parkinson’s disease: hypokinetic or dysrhythmic disorders? Journal of Neurology, Neurosurgery & Psychiatry 81 , 177–184 (2010). Martínez-Sánchez, F. et al. Estudio controlado del ritmo del habla en la enfermedad de Parkinson. Neurología 31 , 466–472 (2016). Skodda, S., Flasskamp, A. & Schlegel, U. Instability of syllable repetition in Parkinson’s disease—Influence of levodopa and deep brain stimulation. Movement Disorders 26 , 728–730 (2011). Additional Declarations No competing interests reported. Supplementary Files Supplementary.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 05 Apr, 2024 Reviews received at journal 24 Mar, 2024 Reviewers agreed at journal 18 Mar, 2024 Reviewers agreed at journal 15 Mar, 2024 Reviewers invited by journal 15 Mar, 2024 Editor assigned by journal 14 Mar, 2024 Editor invited by journal 03 Mar, 2024 Submission checks completed at journal 03 Mar, 2024 First submitted to journal 07 Feb, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3937556","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":276294891,"identity":"86ed86e3-2b33-430c-bfd6-0d06f272fedc","order_by":0,"name":"Ángeles Piña Méndez","email":"","orcid":"","institution":"Autonomous University of San Luis Potosí","correspondingAuthor":false,"prefix":"","firstName":"Ángeles","middleName":"Piña","lastName":"Méndez","suffix":""},{"id":276294892,"identity":"692ba71f-fc9c-4a68-ac75-d9c08d0d589f","order_by":1,"name":"Alan Taitz","email":"","orcid":"","institution":"University of Buenos Aires","correspondingAuthor":false,"prefix":"","firstName":"Alan","middleName":"","lastName":"Taitz","suffix":""},{"id":276294893,"identity":"bdd23c8e-8908-433a-9222-afa67baf9314","order_by":2,"name":"Oscar Palacios Rodríguez","email":"","orcid":"","institution":"Autonomous University of San Luis Potosí","correspondingAuthor":false,"prefix":"","firstName":"Oscar","middleName":"Palacios","lastName":"Rodríguez","suffix":""},{"id":276294894,"identity":"e92f6424-0305-4c9a-84b4-2e5488d0dfa1","order_by":3,"name":"Ildefonso Rodríguez Leyva","email":"","orcid":"","institution":"Autonomous University of San Luis Potosí","correspondingAuthor":false,"prefix":"","firstName":"Ildefonso","middleName":"Rodríguez","lastName":"Leyva","suffix":""},{"id":276294895,"identity":"aafb5924-d94d-417b-8d55-96792feda88e","order_by":4,"name":"M. Florencia Assaneo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA10lEQVRIiWNgGAWjYBAC9gYog5+ZgY04LTwHoAzJZpK1GBwgWgv78ccffu6wkTM+znvsAWPbHXndBu40CbxaeHLMJHvPpBmbHeZLN2Bse2a47QDvZgN8WuwZctgYeNsOJ247zGMmwdh2mBGoZeMDvLbwP3/88W/b/8TNzRAt9kAtGw7g1SKRYCDN23YgcQMzREsiYVsk3phJy7YlG0sc5jE3SDh3OHnbYQJ+4eFPf/zxbZudHH//GbMHH8oO22473rsNb4ihggQQwUy8+lEwCkbBKBgFOAAAso9H7ZyyauEAAAAASUVORK5CYII=","orcid":"","institution":"National Autonomous University of Mexico","correspondingAuthor":true,"prefix":"","firstName":"M.","middleName":"Florencia","lastName":"Assaneo","suffix":""}],"badges":[],"createdAt":"2024-02-07 17:14:37","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3937556/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3937556/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":52102931,"identity":"7279d5d1-ed2b-4e6d-a3a7-fc501f39f98e","added_by":"auto","created_at":"2024-03-06 19:17:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":66892,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExperimental design of the test (trial structure). \u003c/strong\u003eDuring the passive listening phase, participants were presented with a target audio (whispered or spoken aloud). Next, during the repetition phase, participants continuously mimicked the target audio under different auditory feedback situations. Each trial was one of four conditions according to the participant’s voice level (determined by the target audio) and the auditory feedback situation. Condition 1: Speaking aloud while listening to their voice (Normal feedback). Condition 2: Whispering while listening to their whispers (Reduced feedback). Condition 3: Whispering while listening to noise (Masked feedback). Condition 4: Whispering while listening to an alien voice speaking at the same syllabic rate (Replaced feedback). The intertrial interval was set to two seconds.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-3937556/v1/a232a9f52a3894d875132189.png"},{"id":52102933,"identity":"88a54de3-fbb2-4638-a436-22fbf18a19ca","added_by":"auto","created_at":"2024-03-06 19:17:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":216576,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eData preprocessing: a sketch of the pipeline applied to one condition for one representative participant. \u003c/strong\u003eThe left column shows the acoustic signal of each trial in gray, with its corresponding envelope over-imposed in pink. Trials 1 and 2 are whispered (Whis=1), and trial seven is voiced (Whis=0). Arrows point to the reaction time. The middle column displays the spectrum of the envelope for each test. Arrows indicate the syllabic rate (fMAX). The right panel shows the correlation matrix of the envelope spectra (Sp\u003csub\u003e1-7\u003c/sub\u003e) with its lower diagonal element highlighted in gray. \u0026nbsp;\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-3937556/v1/f13be6881613647e8a45491f.png"},{"id":52102934,"identity":"4dff4f32-7c31-4fb6-9f3d-751ae9c24a26","added_by":"auto","created_at":"2024-03-06 19:17:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":176235,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStatistical comparisons between the extracted parameters of patients and healthy controls. a. \u003c/strong\u003epercentage of whispered trials across conditions where whispering was required. \u003cstrong\u003eb,c d \u0026amp; e.\u003c/strong\u003e Errors per second, mean reaction time, mean syllabic rate, and rhythmic stability, respectively, across Normal and Reduced feedback conditions. In all panels, colors identify the different conditions; dots represent individual subjects (opaque: patients, shaded: controls), bars represent the standard deviation, triangles represent the mean values, and *p\u0026lt;0.05.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-3937556/v1/8c069b18c9f8bb5f2ca97ec2.png"},{"id":52104793,"identity":"ae093948-560d-43dd-97d1-182400b2d799","added_by":"auto","created_at":"2024-03-06 19:25:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":865072,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3937556/v1/dcd79c76-ed32-4388-865d-4d900c78cf04.pdf"},{"id":52102932,"identity":"71d1e7a8-e810-4d1c-bbb7-b2d81bb7b4d0","added_by":"auto","created_at":"2024-03-06 19:17:44","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":304498,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-3937556/v1/bc8f199ed64cb8a1c85909ee.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Speech's syllabic rhythm and articulatory features produced under different auditory feedback conditions identify Parkinsonism","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eParkinson's disease (PD) is a complex and frequent neurodegenerative disorder characterized by an extensive collection of motor and non-motor symptoms, a variable response to treatment, and a generally progressive course\u003csup\u003e1,2\u003c/sup\u003e. PD is considered a public health problem, as it is the second most common neurodegenerative disorder after Alzheimer's disease\u003csup\u003e3\u003c/sup\u003e. Furthermore, the Global Burden of Disease study has estimated that PD cases will double from 7\u0026nbsp;million in 2015 to 13\u0026nbsp;million in 2040 \u003csup\u003e4\u003c/sup\u003e. Due to its complexity, variability, and subtypes, PD represents a challenge in its diagnosis because it is mainly based on history and physical examination\u003csup\u003e5\u003c/sup\u003e. Moreover, there are other tests to confirm the diagnosis of PD\u003csup\u003e5\u003c/sup\u003e, but these tend to be expensive and unavailable in most imaging centers. Therefore, attention has been focused on specific, noninvasive, and low-cost biomarkers.\u003c/p\u003e \u003cp\u003eDysarthria (i.e., abnormalities in different aspects of speech production) represents an early symptom of PD and atypical Parkinsonian syndromes\u003csup\u003e6,7\u003c/sup\u003e. In line with this observation, several studies explored the potential of different speech and voice features as PD biomarkers\u003csup\u003e8\u003c/sup\u003e. Several of these studies rely on the diadochokinetic task (DDK), which evaluates articulatory and rhythmic speech impairments by having a subject rhythmically repeat a consonant-vowel combination (typically, the PA-TA-KA sequence, which involves different places of articulation: bilabial, alveolar, and velar)\u003csup\u003e9\u003c/sup\u003e. Based on this task, low-cost vocal tests have been developed, showing a high success rate in identifying PD in its early stages\u003csup\u003e9\u0026ndash;11\u003c/sup\u003e. All these tests rely on the analysis of speech acoustic features in the millisecond scale (e.g., Mel-frequency cepstral coefficients or phonation jitter/shimmer). However, longer timescales features, such as speech articulatory rate and regularity, have also been identified as altered in PD\u003csup\u003e12\u003c/sup\u003e, but remain excluded from vocal tests. Furthermore, Rusz and colleagues\u003csup\u003e13\u003c/sup\u003e show abnormal speech rhythm stability in PD, as well as in atypical Parkinsonian syndromes.\u003c/p\u003e \u003cp\u003eIn addition to dysarthria, it has been shown that the effects produced on the ongoing speech by an unexpected modulation of the auditory feedback differ between PD patients and healthy controls\u003csup\u003e14\u0026ndash;16\u003c/sup\u003e. This result made researchers hypothesize that PD patients have abnormal speech auditory-motor integration. Furthermore, interventions based on altered auditory feedback have successfully restored speech fluency in patients with dysarthria\u003csup\u003e17\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBringing together the existing literature, we developed a modified diadochokinetic task in which subjects are instructed to repeat the syllables PA-TA-KA under different auditory feedback conditions. Using this test, we explore significant differences between patients with Parkinsonism and healthy controls in their rhythmic and articulatory speech features. Specifically, we assessed general speech features such as a subject\u0026rsquo;s ability to whisper, syllabic rhythm stability, or several syllable-level errors instead of the typically used Mel-frequency cepstral coefficients or phonation jitter/shimmer\u003csup\u003e9\u0026ndash;11\u003c/sup\u003e. Furthermore, using a supervised learning technique, we show this paradigm's high accuracy, sensibility, and specificity to identify Parkinsonism.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003eParticipants\u003c/h2\u003e\n\u003cp\u003eTwo cohorts of gender-matched participants (patients and healthy controls) completed this study. Both cohorts comprised Mexican subjects, all of whom were native Spanish speakers. Signed informed consent was obtained from all participants. The protocol was evaluated and approved by the Ethics Committee of the Faculty of Psychology of San Luis Potos\u0026iacute; (Registration number: 2131082021). All methods were performed in accordance with the relevant guidelines and regulations.\u003c/p\u003e\n\u003cp\u003eA total of 30 participants (17 female and 13 male) composed the group of patients, 28 of whom were diagnosed with PD (14 rigid-akinetic and 14 with tremor), two patients with atypical Parkinsonism: progressive supranuclear palsy (PSP) and multiple system atrophy (MSA). In line with previous studies\u003csup\u003e13,18\u003c/sup\u003e, we hypothesize that speech production abnormalities relate to basal ganglia dysfunction; accordingly, they will similarly impact PD, PSP, and MSA patients. Consequently, this first pilot of the test aims to identify Parkinsonism deriving from a basal ganglia dysfunction and not specifically PD.\u003c/p\u003e\n\u003cp\u003eThe patients were diagnosed by a neurologist, following the Movement Disorder Society (MDS) clinical diagnosis criteria for PD\u003csup\u003e19\u003c/sup\u003e, NINDS-PSP for PSP\u003csup\u003e20\u003c/sup\u003e, and the consensus diagnostic criteria for MSA\u003csup\u003e21\u003c/sup\u003e. The diagnosis was made before the start of this study. The ages of the patients ranged from 38 to 88 years (mean 68.3, SD 10.82). We did not modify the treatment during the task; the patients were in an ON state, except for two participants who did not take their medication on the day of the study. Depression symptoms were assessed with the short version of the 15-item Geriatric Depression Scale. In addition, cognitive impairment was measured using the Mini-Mental State Examination. The group of healthy controls comprised 30 participants (17 female and 13 male) who did not report having neurologic or psychiatric disorders (mean\u0026thinsp;=\u0026thinsp;64.2, SD\u0026thinsp;=\u0026thinsp;9.51, range\u0026thinsp;=\u0026thinsp;45\u0026ndash;93).\u003c/p\u003e\n\u003cp\u003eBoth cohorts of participants completed a short questionnaire about their musical experience and educational level. There were no significant differences between cohorts in the assessed demographic features (see Supplementary Fig.\u0026nbsp;1).\u003c/p\u003e\n\u003cdiv id=\"Sec4\" class=\"Section3\"\u003e\n\u003ch2\u003eAuditory stimulus\u003c/h2\u003e\n\u003cp\u003e\u003cem\u003eTarget audio\u003c/em\u003e: The syllables \u0026ldquo;pa,\u0026rdquo; \u0026ldquo;ta,\u0026rdquo; and \u0026ldquo;ka,\u0026rdquo; spoken aloud and whispered, were recorded by a female Spanish speaker. Praat software\u003csup\u003e22\u003c/sup\u003e was used to make each syllable last 250 ms. Next, five repetitions of the sequence pa-ta-ka, spoken or whispered, were concatenated, generating two rhythmic (4 syllables/sec), 3.75-second-long audio files: the whispered and the spoken targets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExogenous speech\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA 5.5-second-long audio file comprising a rhythmic train of syllables was synthesized at 16 Hz using the MBROLA text-to-speech synthesizer\u003csup\u003e23\u003c/sup\u003e with the Spanish Male Voice \u0026ldquo;es2.\u0026rdquo; A set of 13 Spanish syllables (\u0026ldquo;te,\u0026rdquo; \u0026ldquo;bi,\u0026rdquo; \u0026ldquo;ki,\u0026rdquo; \u0026ldquo;pu,\u0026rdquo; \u0026ldquo;bo,\u0026rdquo; \u0026ldquo;la,\u0026rdquo; \u0026ldquo;su,\u0026rdquo; \u0026ldquo;go,\u0026rdquo; \u0026ldquo;mu,\u0026rdquo; \u0026ldquo;rra,\u0026rdquo; \u0026ldquo;le,\u0026rdquo; \u0026ldquo;do,\u0026rdquo; \u0026ldquo;fe\") were repeatedly and randomly concatenated to achieve the desired length. All phonemes were equal in pitch (200 Hz), and the duration was set to 0.125 ms.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNoise\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA 5.5-second-long audio file comprising white noise was synthesized using Matlab\u003csup\u003e24\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n\u003ch2\u003eProcedure\u003c/h2\u003e\n\u003cp\u003eAll participants performed the experimental test in a room with low ambient noise. They sat in front of a computer wearing insert earphones (ETYMOTIC ER1), and we recorded their vocalizations with a microphone connected to the laptop (Marantz Pro M4U). All instructions appeared written on the computer screen and were verbally reinforced by the examiner.\u003c/p\u003e\n\u003cp\u003eThe experimental test consisted of 28 trials and had a total duration of approximately 10 minutes. Each trial consisted of passive listening and repetition phases (see Fig.\u0026nbsp;1). During the passive listening phase, the target audio was presented through the earplugs, and participants were instructed to pay attention to it while fixing their gaze on a black dot centered on the screen. The target audio lasted 3.75 seconds and comprised a repetition of the pa-ta-ka sequence, whispered or spoken aloud (see the Auditory Stimuli section). At the end of the audio playback, the dot on the screen turned green, signaling the beginning of the repetition phase. During this phase, participants were instructed to continuously echo the target audio, matching the presented rhythm and voice level (i.e., whispering or speaking aloud). After 5.5 seconds, the green dot turned red, prompting participants to stop vocalizing and to wait until the subsequent trial. The intertrial interval was set to two seconds.\u003c/p\u003e\n\u003cp\u003eEach trial belonged to one of four possible conditions (see Fig.\u0026nbsp;1):\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\n\u003cp\u003eNatural feedback: The target audio comprised loud speech, and no auditory stimulus was presented during the repetition phase.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eReduced feedback: The target audio comprised whispered speech, and no auditory stimulus was presented during the repetition phase.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eMasked feedback: The target audio comprised whispered speech, and the noise audio was played during the repetition phase, masking the participant\u0026rsquo;s voice auditory feedback.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eReplaced feedback: The target audio comprised whispered speech, and the exogenous speech audio was played during the repetition phase, masking the participant\u0026rsquo;s voice auditory feedback.\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eWe presented seven trials per condition in a randomized order.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n\u003ch2\u003eData preprocessing\u003c/h2\u003e\n\u003cp\u003eFor each participant and trial condition, five parameters were extracted (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e): (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e) the number of whispered trials, (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e) the number of speech errors, (\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e) the mean reaction time, (\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e) the mean syllabic rate, and (\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e) the rhythmic stability.\u003c/p\u003e\n\u003cp\u003eThe first author manually computed the number of whispered trials, speech errors, and reaction time. Praat software was used to listen to and visualize the acoustic signals. Each trial was categorized as whispered if voiced speech (i.e., with activation of the vocal folds) occurred for less than 0.55 seconds (10% of the trial length). An error was identified if the participant: (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e) repeated a syllable (e.g., \u0026ldquo;pa-ta-\u003cstrong\u003eta\u003c/strong\u003e-ka\u0026rdquo;), (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e) exchanged a syllable (e.g., \u0026ldquo;pa-\u003cstrong\u003eka-ta\u003c/strong\u003e\u0026rdquo;), or if a syllable was wrongly articulated (e.g., \u0026ldquo;pa-\u003cstrong\u003etra\u003c/strong\u003e-ka\u0026rdquo;). Reaction time was computed as the speech onset time and averaged across trials of the same condition.\u003c/p\u003e\n\u003cp\u003eTo estimate the rhythmic features of the spoken samples, we calculated the speech envelopes as the absolute value of the acoustic signal\u0026rsquo;s Hilbert transform\u003csup\u003e25\u003c/sup\u003e. Subsequently, we used the fast Fourier transform to extract the spectrum of each trial\u0026rsquo;s envelope. Each trial's syllabic rate was estimated as the frequency value with the maximal power (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). The estimated syllabic rates were averaged across tests of the same condition. Next, to explore the stability of the rhythmic structure across trials of the same condition, we computed the correlation matrix between their envelope spectra. The rhythmic stability was estimated as the mean value of the lower diagonal elements of the matrix (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n\u003ch2\u003eData Analysis\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical comparisons\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNon-parametric inferential statistics compared the patient group with the healthy controls. More specifically, we used the Mann-Whitney test for two independent samples. All reported p-values were corrected using a false discovery rate approach\u003csup\u003e26\u003c/sup\u003e for multiple comparisons. All procedures were carried out in Matlab.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine learning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA random forest classifier model was trained on the binary classes of patients and healthy controls using the data obtained from the five evaluated parameters. The metrics sensitivity, specificity, and accuracy were considered to assess the model\u0026rsquo;s performance. The leave-one-out cross-validation method was used to assess the model generalization performance. This method consists of several iterations of the training-testing data sets, where in each iteration, one participant is selected to test the model, and all others are used to train it. All plausible training-testing combinations were evaluated, and the predicted outcomes were used to compute the accuracy, sensitivity, and specificity. This process was repeated 100 times, and the results were averaged. This process was conducted using the \u003cem\u003esklearn\u003c/em\u003e library in Python\u003csup\u003e27\u003c/sup\u003e.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"RESULTS","content":"\u003cp\u003eWe assessed the goodness of a behavioral test based on speech samples to diagnose Parkinsonism. Thus, we evaluated patients and healthy controls in a task consisting of continuously and rhythmically (four syllables per second) repeating the syllables \u0026ldquo;pa-ta-ka\u0026rdquo; under different auditory feedback conditions: Normal (speaking aloud while hearing the produced sounds), Reduced (whispering while hearing the produced sounds), Masked (whispering while hearing white noise), Replaced (whispering while hearing an alien voice). While the normal condition matches the DDK task used in previous studies\u003csup\u003e9\u003c/sup\u003e, the other three integrate a modulation in the participants\u0026rsquo; voice feedback. For the Reduced condition, feedback is not entirely removed but diminished, given the whisper-low volume. In the Masked and Replaced conditions, the feedback is completely blocked, but in the last one, participants get an external cue of the intended syllabic rate. It has been shown in healthy participants that whispering while listening to an external stable rhythm leads some individuals to synchronize the produced syllabic pace\u003csup\u003e28\u003c/sup\u003e. Additionally, PD patients show abnormal rhythmic speech entrainment with a model speaker\u003csup\u003e29\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFrom the obtained recordings, we computed five parameters for each feedback condition: (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e) whispering ability, (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e) the number of articulatory errors, (\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e) reaction time to initiate speech, (\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e) syllabic rhythm, and (\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e) rhythmic stability (for more details, see the \u003cspan class=\"InternalRef\"\u003eMethods\u003c/span\u003e section). Two different analyses were conducted on these parameters. First, we compared the parameters for patients and controls across feedback conditions. This allowed us to identify the speech features being abnormal in patients. Secondly, we fed the parameters to a supervised learning algorithm to assess the predictive power of the parameters\u0026rsquo; combination to differentiate patients from healthy participants.\u003c/p\u003e\n\u003cp\u003eTo start, we explored if patients could correctly follow the instructions or if they showed more difficulties than control participants. Specifically, we investigated whether they could whisper in the conditions with this requirement (i.e., Reduced, Masked, and Replaced feedback conditions). We compared the percentage of whispered trials between patients and controls. The results showed significant differences between the group of patients with Parkinsonism and healthy subjects in the Masked and Replaced feedback conditions (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea; Reduced: Patients: M\u0026thinsp;=\u0026thinsp;65%, SD\u0026thinsp;=\u0026thinsp;38%; Control: M\u0026thinsp;=\u0026thinsp;79%, SD\u0026thinsp;=\u0026thinsp;37%; p\u0026thinsp;=\u0026thinsp;0.073; Masked: Patients: M\u0026thinsp;=\u0026thinsp;44%, SD\u0026thinsp;=\u0026thinsp;35%; Control: M\u0026thinsp;=\u0026thinsp;72%, SD\u0026thinsp;=\u0026thinsp;38%; p\u0026thinsp;=\u0026thinsp;0.008; Replaced: Patients: M\u0026thinsp;=\u0026thinsp;37%, SD\u0026thinsp;=\u0026thinsp;39%; Control: M\u0026thinsp;=\u0026thinsp;70%, SD\u0026thinsp;=\u0026thinsp;41%; p\u0026thinsp;=\u0026thinsp;0.008). Patients had difficulties in whispering the syllables when presented with altered feedback. Given that the number of whispered trials in the Masked and Replaced feedback conditions significantly differed between patients and controls, differences in the rest of the parameters were only assessed between Normal and Reduced feedback conditions. This was done because groups differ in the number of spoken-aloud trials for the other feedback conditions, making it impossible to disentangle whether the differences (if observed) derive from the auditory feedback state situation or from speaking aloud.\u003c/p\u003e\n\u003cp\u003eFor all other explored parameters, we found that: (i) patients made more speech errors than controls, only for the Reduced feedback condition (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb; Normal: Patients: M\u0026thinsp;=\u0026thinsp;0.047 err/sec, SD\u0026thinsp;=\u0026thinsp;0.072 err/sec; Control: M\u0026thinsp;=\u0026thinsp;0.044 err/sec, SD\u0026thinsp;=\u0026thinsp;0.126 err/sec; p\u0026thinsp;=\u0026thinsp;0.072; Reduced: Patients: M\u0026thinsp;=\u0026thinsp;0.1 err/sec, SD\u0026thinsp;=\u0026thinsp;0.117 err/sec; Control: M\u0026thinsp;=\u0026thinsp;0.022 err/sec, SD\u0026thinsp;=\u0026thinsp;0.076 err/sec; p\u0026thinsp;=\u0026thinsp;0.002); (ii) patients had slower reaction times than controls, only for the Reduced feedback condition (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec; Normal: Patients: M\u0026thinsp;=\u0026thinsp;0.60 sec, SD\u0026thinsp;=\u0026thinsp;0.20 sec; Control: M\u0026thinsp;=\u0026thinsp;0.49 sec, SD\u0026thinsp;=\u0026thinsp;0.15 sec; p\u0026thinsp;=\u0026thinsp;0.081; Reduced: Patients: M\u0026thinsp;=\u0026thinsp;0.52 sec, SD\u0026thinsp;=\u0026thinsp;0.21 sec; Control: M\u0026thinsp;=\u0026thinsp;0.39 sec, SD\u0026thinsp;=\u0026thinsp;0.14 sec; p\u0026thinsp;=\u0026thinsp;0.022); (iii) there was no significant difference between groups in their mean syllabic rate (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ed; Normal: Patients: M\u0026thinsp;=\u0026thinsp;3.7 Hz, SD\u0026thinsp;=\u0026thinsp;0.82 Hz; Control: M\u0026thinsp;=\u0026thinsp;4.05 Hz, SD\u0026thinsp;=\u0026thinsp;0.71 Hz; p\u0026thinsp;=\u0026thinsp;0.105; Reduced: Patients: M\u0026thinsp;=\u0026thinsp;3.55 Hz, SD\u0026thinsp;=\u0026thinsp;0.84 Hz; Control: M\u0026thinsp;=\u0026thinsp;3.94 Hz, SD\u0026thinsp;=\u0026thinsp;0.68 Hz; p\u0026thinsp;=\u0026thinsp;0.105); and (iv) patients showed less rhythmic stability than controls for the Normal feedback condition (see Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ee; Normal: Patients: M\u0026thinsp;=\u0026thinsp;0.71, SD\u0026thinsp;=\u0026thinsp;0.16; Control: M\u0026thinsp;=\u0026thinsp;0.82, SD\u0026thinsp;=\u0026thinsp;0.13; p\u0026thinsp;=\u0026thinsp;0.015; Reduced: Patients: M\u0026thinsp;=\u0026thinsp;0.81, SD\u0026thinsp;=\u0026thinsp;0.11; Control: M\u0026thinsp;=\u0026thinsp;0.85, SD\u0026thinsp;=\u0026thinsp;0.13; p\u0026thinsp;=\u0026thinsp;0.128). All the results were recovered when the two patients with atypical Parkinsonism were excluded and the analyses were restricted to PD (see Supplementary Fig.\u0026nbsp;2).\u003c/p\u003e\n\u003cp\u003eOnce the differences between patients and healthy controls had been established by statistically comparing the parameters\u0026rsquo; distributions, we fed the relevant parameters into a random forest classifier to differentiate between groups (see Methods). This procedure allowed us to estimate the predictive power of these speech features to identify patients from the general population and generalize the reported results. The five variables showing significant differences between groups are the percentage of whispering trials in Masked feedback, rate of whispering tests in Replaced feedback, speech error per second in Reduced feedback, mean reaction time in Reduced feedback, and rhythmic stability in Normal feedback. First, we trained and tested the classifier with all five variables and computed accuracy, specificity, and sensitivity (see Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, Model 1). Given that the two altered feedback conditions only contributed to the percentage of whispered trials, we explored if both were increasing the predictive power of the instrument or if they carried redundant information. To do so, we evaluated the classifier with the percentage of whispered trials in only one or the other condition (see Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, Models 2 and 3). Results show that performance does not increase by including both conditions, indicating that the Masked feedback condition can be removed from the test. Finally, we tested whether the participants\u0026rsquo; age helped the classifier's performance, which was not the case (see Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, Model 4).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003e\u003cstrong\u003ePerformance of the random forest classifier trained and tested on different sets of parameters.\u003c/strong\u003e Parameters are specified as pairs of \u003cem\u003efeatures and condition\u003c/em\u003es. Features are as follows: whis\u0026thinsp;=\u0026thinsp;percentage of whispered trials; Sp. Err\u0026thinsp;=\u0026thinsp;speech errors per second; RT\u0026thinsp;=\u0026thinsp;mean reaction time; Rhy. Stab\u0026thinsp;=\u0026thinsp;rhythm stability. Conditions are as follows: NF\u0026thinsp;=\u0026thinsp;Normal feedback; Red. F\u0026thinsp;=\u0026thinsp;Reduced feedback; MF\u0026thinsp;=\u0026thinsp;Masked feedback; Rep.F\u0026thinsp;=\u0026thinsp;Replaced feedback. Accuracy, sensitivity, and specificity are reported as a percentage.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eParameters fed into the classifier\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eAccuracy\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSensitivity\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSpecificity\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e%whis,RepF + %whis,MF\u0026thinsp;+\u0026thinsp;Sp.Err, RedF\u0026thinsp;+\u0026thinsp;RT, RedF\u0026thinsp;+\u0026thinsp;Rhy.Stab, NF\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e83.0\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e81.8\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e84.1\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e%whis,RepF\u0026thinsp;+\u0026thinsp;Sp.Err, RedF\u0026thinsp;+\u0026thinsp;RT, RedF\u0026thinsp;+\u0026thinsp;Rhy.Stab, NF\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e86.1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e87.8\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e84.3\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e%whis,MF\u0026thinsp;+\u0026thinsp;Sp.Err, RedF\u0026thinsp;+\u0026thinsp;RT, RedF\u0026thinsp;+\u0026thinsp;Rhy.Stab, NF\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e85.4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e87.8\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e83.1\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e%whis,RepF\u0026thinsp;+\u0026thinsp;Sp.Err, RedF\u0026thinsp;+\u0026thinsp;RT, RedF\u0026thinsp;+\u0026thinsp;Rhy.Stab,NF\u0026thinsp;+\u0026thinsp;age\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e82.0\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e80.3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e83.7\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThe present research focused on identifying Parkinsonism speech impairments under different auditory feedback conditions. More precisely, five parameters were studied: 1) whispering ability, 2) the number of articulatory errors, 3) reaction time to initiate speech, 4) syllabic rhythm, and 5) rhythmic stability. This is proposed as a pilot phase for developing an objective diagnostic test.\u003c/p\u003e \u003cp\u003e The participant's ability to whisper in Reduced, Masked, and Replaced feedback conditions was analyzed, and it was found that the patient group presented more difficulties than the control group in adapting their speech output (i.e., whispering) when their auditory feedback was modified. Given that whispering was not wholly impaired in the patient group and that the difference between groups appeared only when the auditory feedback was masked or replaced, it can be inferred that the patients have an abnormal enhancement of the Lombard effect (i.e., an increase in voice intensity in response to an increase in the ambient noise level). It has been suggested that (i) the Lombard effect occurs unconsciously, driven by a subcortical mechanism, and (ii) the activation of such a subcortical mechanism can be modulated by a cortical network, allowing voluntary control of the effect\u003csup\u003e30\u003c/sup\u003e. The pattern of results obtained in this study suggests that the cortical network is affected in the Parkinsonian cohort: when instructed to whisper, controls, but not patients, can suppress the Lombard effect and manage to maintain low speech intensity in noisy environments (i.e., listening to white noise in the Masked feedback condition and an alien voice in the Replaced feedback condition). In line with this hypothesis, it has been shown that PD patients have decreased activity in frontotemporal regions\u003csup\u003e31\u003c/sup\u003e, and studies based on deep brain stimulation in PD patients provide evidence that the connectivity between frontal and sub-thalamic brain regions explains individual differences in inhibition\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eArticulatory errors are prevalent in the patient group when whispering the syllables. It has been suggested that PD patients rely more on auditory feedback due to impaired input motor control and somatosensory feedback\u003csup\u003e33,34\u003c/sup\u003e, which can explain the present findings. As there is no vocal fold vibration during whispering, it is more difficult for the acoustic signal to be perceived by the auditory system\u003csup\u003e35\u003c/sup\u003e, and speech-motor production relies on the forward model and somatosensory feedback, mechanisms proposed to be abnormal in PD patients. Accordingly, studies based on repetitive transcranial magnetic stimulation have shown that stimulating the auditory cortex improves articulation in PD patients during an overt-speech task\u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIt was identified that the patients recruited in our study have difficulty initiating the motor process, such as articulating syllables. Furthermore, the increase in reaction time when starting speech occurs regardless of the condition in which they perform the task (Normal and Reduced feedback). The findings are consistent with the literature; different referents mention that PD patients struggle to initiate movement\u003csup\u003e32,36,37\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eRegarding rhythmic features, we found that while syllabic rhythm did not differ between groups in any feedback condition, the rhythmic stability (i.e., how similar the rhythmic structure was across trials) decreased in the PD patients. Studies on PD patients\u0026rsquo; speech rhythm present contradictory results. While some authors report a slowdown \u003csup\u003e38\u003c/sup\u003e, others report an increment in the rate\u003csup\u003e39\u003c/sup\u003e or no significant difference between patients and controls\u003csup\u003e40\u003c/sup\u003e. Here, we showed that while absolute rhythm did not differentiate patients from controls, rhythm stability did, suggesting this last measurement is more accurate in diagnosing Parkinsonism. Contrary to the pattern of results obtained for the articulatory errors, rhythmic stability significantly differed between groups for the Normal feedback but not the Reduced feedback condition. This contrast suggests that the circuit responsible for monitoring and correcting articulatory errors does not overlap with the one supporting the stability of the speech rhythm. It has been proposed that speech rhythm emerges due to the biophysical properties of the brain areas in charge of generating speech\u003csup\u003e41\u003c/sup\u003e, and the interaction of motor and auditory areas\u003csup\u003e42\u003c/sup\u003e modulates it. Under this framework, the observation that the rhythm becomes unstable when increasing the auditory feedback hints towards an abnormal cortical interaction between the frontal and temporal regions.\u003c/p\u003e \u003cp\u003eThe main goal of the current study was to assess the goodness of a short and easy-to-implement behavioral screening test based on speech samples. Similar designs to diagnose PD\u003csup\u003e11,43,44\u003c/sup\u003e, and atypical parkinsonism\u003csup\u003e13\u003c/sup\u003e, have been previously reported in the literature with promising results. In those studies, researchers typically asked participants to complete different speech tasks, such as sustained phonation, running speech, and, as in the current work, the DDK task (i.e., continuously repeating the syllables /pa/ /ta/ /ka/). They also extracted different acoustic parameters from the speech samples and entered them into a machine-learning algorithm to distinguish patients from healthy controls. Such strategies obtained high accuracy values from 85\u0026ndash;99%. Two main aspects determine the current work from previous studies.\u003c/p\u003e \u003cp\u003eOn the one hand, although it has been shown that modified auditory feedback impacts PD patients differently from controls\u003csup\u003e14\u0026ndash;16\u003c/sup\u003e, this observation has not been previously included in the design of speech-based screening tests. We integrate this observation by asking participants to complete the classic DDK task under different auditory feedback conditions. On the other hand, we evaluated speech features on a different time scale than those typically computed (e.g., formant periodicity correlations, Mel-frequency cepstral coefficients, phonation jitter, phonation shimmer phonation noise, and voice fundamental frequency variations\u003csup\u003e44\u003c/sup\u003e). Here, we focused on general speech features, such as whispering, throughout the trial or syllabic scale characteristics, such as rhythm stability or number of errors at the syllabic level. A machine learning algorithm trained on such parameters to differentiate Parkinsonian patients from healthy controls shows accuracy within the range of previously reported results\u003csup\u003e45\u003c/sup\u003e. These results present novel experimental conditions (e.g., whispering, speaking under different listening conditions, and trained syllabic rate) and parameters (e.g., whispered trials, reaction time, rhythm stability, and speech errors at the syllabic level) as valuable tools to be introduced into existing screening behavioral tests to identify Parkinsonism. Additionally, it is important to highlight the accessibility of the piloted diagnostic test, which only requires a standard PC, a set of headphones, and a microphone and lasts less than 4 minutes (i.e., only three of the four evaluated conditions are required, and each condition comprises seven 11.25-second trials).\u003c/p\u003e \u003cp\u003eA limitation of the present study is its small sample size (30 patients and 30 controls). However, this sample size is sufficient for a pilot study and reaches statistical significance. Furthermore, despite the small sample size, the current work addresses the need for cross-linguistic studies of dysarthria in PD\u003csup\u003e46\u003c/sup\u003e (i.e., this study was conducted on Mexican participants, a Spanish-speaking population typically overlooked in the existing literature). Our major limitation is mainly in the number of subjects with atypical Parkinsonism because, in addition to being rare diseases, there were few spaces and clinical records available, which prevented us from recruiting more participants.\u003c/p\u003e \u003cp\u003eAnother potential limitation is the lack of control over the influence of levodopa or other medications since the patients were on medication during the application of the instrument. There needs to be more interest in addressing this issue, which makes it difficult to access reliable data and collaborate with other institutions. However, some studies suggest that speech fluency (or the lack of it) in Parkinson's patients is not modulated by levodopa\u003csup\u003e47\u0026ndash;49\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe present protocol studies rhythmic and articulatory changes related to PD, MSA, and PSP in the speech production system. Although it does not address any therapeutic strategy or alternative treatment, it lays the foundation for developing noninvasive, low-cost, and easy-to-apply diagnostic tests.\u003c/p\u003e"},{"header":"CONCLUSIONS","content":"\u003cp\u003eSyllabic rhythm stability, reaction time, whispering ability, and syllable-level articulatory errors under different auditory feedback conditions differentiate Parkinsonian patients from healthy controls. An automatic detection algorithm trained on these parameters showed an accuracy of 85.4% in distinguishing patients from controls. The current work represents a pilot trial, showing the potential of the introduced behavioral paradigm as an objective and accessible (in cost and time) diagnostic test.\u003c/p\u003e"},{"header":"DECLARATIONS","content":"\u003cp\u003e\u003cstrong\u003eACKNOWLEDGEMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Jessica Gonz\u0026aacute;lez Norris for proofreading the manuscript. This work was supported by DGAPA-PAPIIT IA200223 and the IBRO Return Home Fellowship (MFA). The authors report no relevant conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHORS CONTRIBUTION\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMFA conceived the project. MFA, OPR, and IRL supervised the project. APM collected the data. AT, APM, and MFA analyzed the data. MFA, IRL, and APM wrote the manuscript. All authors read and approved the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCOMPETING INTERESTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors report no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDATA AVAILABILITY\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data supporting the findings of this study are available as Supplementary Data. All other data and computer code used to generate results are available upon request from the corresponding author.\u003c/p\u003e"},{"header":"REFERENCES","content":"\u003col\u003e\n\u003cli\u003ePilotto, A. \u003cem\u003eet al.\u003c/em\u003e Plasma NfL, clinical subtypes and motor progression in Parkinson\u0026rsquo;s disease. \u003cem\u003eParkinsonism \u0026amp; Related Disorders\u003c/em\u003e \u003cstrong\u003e87\u003c/strong\u003e, 41\u0026ndash;47 (2021).\u003c/li\u003e\n\u003cli\u003eTitova, N., Padmakumar, C., Lewis, S. J. G. \u0026amp; Chaudhuri, K. R. Parkinson\u0026rsquo;s: a syndrome rather than a disease? \u003cem\u003eJ Neural Transm\u003c/em\u003e \u003cstrong\u003e124\u003c/strong\u003e, 907\u0026ndash;914 (2017).\u003c/li\u003e\n\u003cli\u003eDraoui, A., El Hiba, O., Aimrane, A., El Khiat, A. \u0026amp; Gamrani, H. Parkinson\u0026rsquo;s disease: From bench to bedside. \u003cem\u003eRevue Neurologique\u003c/em\u003e \u003cstrong\u003e176\u003c/strong\u003e, 543\u0026ndash;559 (2020).\u003c/li\u003e\n\u003cli\u003eJankovic, J. \u0026amp; Tan, E. K. Parkinson\u0026rsquo;s disease: etiopathogenesis and treatment. \u003cem\u003eJ Neurol Neurosurg Psychiatry\u003c/em\u003e \u003cstrong\u003e91\u003c/strong\u003e, 795\u0026ndash;808 (2020).\u003c/li\u003e\n\u003cli\u003eArmstrong, M. J. \u0026amp; Okun, M. S. Diagnosis and Treatment of Parkinson Disease: A Review. \u003cem\u003eJAMA\u003c/em\u003e \u003cstrong\u003e323\u003c/strong\u003e, 548\u0026ndash;560 (2020).\u003c/li\u003e\n\u003cli\u003ePinto, S. \u003cem\u003eet al.\u003c/em\u003e Treatments for dysarthria in Parkinson\u0026rsquo;s disease. \u003cem\u003eThe Lancet Neurology\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 547\u0026ndash;556 (2004).\u003c/li\u003e\n\u003cli\u003eCritchley, E. M. Speech disorders of Parkinsonism: a review. \u003cem\u003eJ Neurol Neurosurg Psychiatry\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 751\u0026ndash;758 (1981).\u003c/li\u003e\n\u003cli\u003eNgo, Q. C. \u003cem\u003eet al.\u003c/em\u003e Computerized analysis of speech and voice for Parkinson\u0026rsquo;s disease: A systematic review. \u003cem\u003eComputer Methods and Programs in Biomedicine\u003c/em\u003e \u003cstrong\u003e226\u003c/strong\u003e, 107133 (2022).\u003c/li\u003e\n\u003cli\u003eMonta\u0026ntilde;a, D., Campos-Roca, Y. \u0026amp; P\u0026eacute;rez, C. J. A Diadochokinesis-based expert system considering articulatory features of plosive consonants for early detection of Parkinson\u0026rsquo;s disease. \u003cem\u003eComputer Methods and Programs in Biomedicine\u003c/em\u003e \u003cstrong\u003e154\u003c/strong\u003e, 89\u0026ndash;97 (2018).\u003c/li\u003e\n\u003cli\u003eVasquez-Correa, J. C., Arias-Vergara, T., Schuster, M., Orozco-Arroyave, J. R. \u0026amp; N\u0026ouml;th, E. Parallel Representation Learning for the Classification of Pathological Speech: Studies on Parkinson\u0026rsquo;s Disease and Cleft Lip and Palate. \u003cem\u003eSpeech Communication\u003c/em\u003e \u003cstrong\u003e122\u003c/strong\u003e, 56\u0026ndash;67 (2020).\u003c/li\u003e\n\u003cli\u003eRusz, J. \u003cem\u003eet al.\u003c/em\u003e Acoustic voice and speech disorders assessment in Parkinson\u0026apos;s disease through quick vocal test. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1951\u0026ndash;1952 (2011).\u003c/li\u003e\n\u003cli\u003eSkodda, S. Aspects of speech rate and regularity in Parkinson\u0026rsquo;s disease. \u003cem\u003eJournal of the Neurological Sciences\u003c/em\u003e \u003cstrong\u003e310\u003c/strong\u003e, 231\u0026ndash;236 (2011).\u003c/li\u003e\n\u003cli\u003eRusz, J., Hlavnička, J., Čmejla, R. \u0026amp; Růžička, E. Automatic Evaluation of Speech Rhythm Instability and Acceleration in Dysarthrias Associated with Basal Ganglia Dysfunction. \u003cem\u003eFrontiers in Bioengineering and Biotechnology\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eLiu, H., Wang, E. Q., Metman, L. V. \u0026amp; Larson, C. R. Vocal Responses to Perturbations in Voice Auditory Feedback in Individuals with Parkinson\u0026rsquo;s Disease. \u003cem\u003ePLOS ONE\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, e33629 (2012).\u003c/li\u003e\n\u003cli\u003eSenthinathan, A., Adams, S., Page, A. D. \u0026amp; Jog, M. Speech Intensity Response to Altered Intensity Feedback in Individuals With Parkinson\u0026rsquo;s Disease. \u003cem\u003eJournal of Speech, Language, and Hearing Research\u003c/em\u003e \u003cstrong\u003e64\u003c/strong\u003e, 2261\u0026ndash;2275 (2021).\u003c/li\u003e\n\u003cli\u003eBlanchet, P. G. Factors influencing the efficacy of delayed auditory feedback in treating dysarthria associated with Parkinson\u0026rsquo;s disease. (2002).\u003c/li\u003e\n\u003cli\u003eLowit, A., Dobinson, C., Timmins, C., Howell, P. \u0026amp; Kr\u0026ouml;ger, B. The effectiveness of traditional methods and altered auditory feedback in improving speech rate and intelligibility in speakers with Parkinson\u0026apos;s disease\u003cem\u003e\u0026mdash;International Journal of Speech-Language Pathology\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 426\u0026ndash;436 (2010).\u003c/li\u003e\n\u003cli\u003eKowalska-Taczanowska, R., Friedman, A. \u0026amp; Koziorowski, D. Parkinson\u0026apos;s disease or atypical Parkinsonism? The importance of acoustic voice analysis in differential diagnosis of speech disorders. \u003cem\u003eBrain and Behavior\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e01700 (2020).\u003c/li\u003e\n\u003cli\u003ePostuma, R. B. \u003cem\u003eet al.\u003c/em\u003e MDS clinical diagnostic criteria for Parkinson\u0026rsquo;s disease. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1591\u0026ndash;1601 (2015).\u003c/li\u003e\n\u003cli\u003eLitvan, I. \u003cem\u003eet al.\u003c/em\u003e Accuracy of clinical criteria for the diagnosis of progressive supranuclear palsy (Steele-Richardson-Olszewski syndrome). \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 922\u0026ndash;930 (1996).\u003c/li\u003e\n\u003cli\u003eGilman, S. \u003cem\u003eet al.\u003c/em\u003e Consensus statement on the diagnosis of multiple system atrophy. \u003cem\u003eClinical Autonomic Research\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 359\u0026ndash;362 (1998).\u003c/li\u003e\n\u003cli\u003eBoersma, P. \u0026amp; Weenink, D. PRAAT, a system for doing phonetics by computer. \u003cem\u003eGlot international\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 341\u0026ndash;345 (2001).\u003c/li\u003e\n\u003cli\u003eDutoit, T., Pagel, V., Pierret, N., Bataille, F. \u0026amp; van der Vrecken, O. The MBROLA project: towards a set of high quality speech synthesizers free of use for non commercial purposes. in \u003cem\u003eProceeding of Fourth International Conference on Spoken Language Processing. \u003c/em\u003e\u003cem\u003eICSLP \u0026rsquo;96\u003c/em\u003e vol. 3 1393\u0026ndash;1396 vol.3 (1996).\u003c/li\u003e\n\u003cli\u003eMatlab. (2020).\u003c/li\u003e\n\u003cli\u003eLizcano-Cort\u0026eacute;s, F. \u003cem\u003eet al.\u003c/em\u003e Speech-to-Speech Synchronization protocol to classify human participants as high or low auditory-motor synchronizers. \u003cem\u003eSTAR Protocols\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 101248 (2022).\u003c/li\u003e\n\u003cli\u003eBenjamini, Y. \u0026amp; Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. \u003cem\u003eJournal of the Royal Statistical Society: Series B (Methodological)\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 289\u0026ndash;300 (1995).\u003c/li\u003e\n\u003cli\u003eBuitinck, L. \u003cem\u003eet al.\u003c/em\u003e API design for machine learning software: experiences from the scikit-learn project. Preprint at https://doi.org/10.48550/arXiv.1309.0238 (2013).\u003c/li\u003e\n\u003cli\u003eMares, C., Echavarr\u0026iacute;a Solana, R. \u0026amp; Assaneo, M. F. Auditory-motor synchronization varies among individuals and is critically shaped by acoustic features. \u003cem\u003eCommun Biol\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 1\u0026ndash;10 (2023).\u003c/li\u003e\n\u003cli\u003eSp\u0026auml;th, M. \u003cem\u003eet al.\u003c/em\u003e Entraining with another person\u0026rsquo;s speech rhythm: Evidence from healthy speakers and individuals with Parkinson\u0026rsquo;s disease. \u003cem\u003eClinical Linguistics \u0026amp; Phonetics\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 68\u0026ndash;85 (2016).\u003c/li\u003e\n\u003cli\u003eLuo, J., Hage, S. R. \u0026amp; Moss, C. F. The Lombard Effect: From Acoustics to Neural Mechanisms. \u003cem\u003eTrends in Neurosciences\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 938\u0026ndash;949 (2018).\u003c/li\u003e\n\u003cli\u003eAuclair-Ouellet, N. \u003cem\u003eet al.\u003c/em\u003e Action fluency identifies different sex, age, global cognition, executive function, and brain activation profile in non-demented patients with Parkinson\u0026apos;s disease. \u003cem\u003eJ Neurol\u003c/em\u003e \u003cstrong\u003e268\u003c/strong\u003e, 1036\u0026ndash;1049 (2021).\u003c/li\u003e\n\u003cli\u003eMosley, P. E. \u003cem\u003eet al.\u003c/em\u003e Subthalamic deep brain stimulation identifies frontal networks supporting initiation, inhibition, and strategy use in Parkinson\u0026apos;s disease. \u003cem\u003eNeuroImage\u003c/em\u003e \u003cstrong\u003e223\u003c/strong\u003e, 117352 (2020).\u003c/li\u003e\n\u003cli\u003eWeerathunge, H. R., Tomassi, N. E. \u0026amp; Stepp, C. E. What Can Altered Auditory Feedback Paradigms Tell Us About Vocal Motor Control in Individuals With Voice Disorders? \u003cem\u003ePerspectives of the ASHA Special Interest Groups\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 959\u0026ndash;976 (2022).\u003c/li\u003e\n\u003cli\u003eBrabenec, L. \u003cem\u003eet al.\u003c/em\u003e Noninvasive stimulation of the auditory feedback area for improved articulation in Parkinson\u0026apos;s disease. \u003cem\u003eParkinsonism \u0026amp; Related Disorders\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 187\u0026ndash;192 (2019).\u003c/li\u003e\n\u003cli\u003eFr\u0026uuml;hholz, S., Trost, W. \u0026amp; Grandjean, D. Whispering - The hidden side of auditory communication. \u003cem\u003eNeuroImage\u003c/em\u003e \u003cstrong\u003e142\u003c/strong\u003e, 602\u0026ndash;612 (2016).\u003c/li\u003e\n\u003cli\u003eMekyska, J. \u003cem\u003eet al.\u003c/em\u003e Quantitative Analysis of Relationship Between Hypokinetic Dysarthria and the Freezing of Gait in Parkinson\u0026rsquo;s Disease. \u003cem\u003eCogn Comput\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1006\u0026ndash;1018 (2018).\u003c/li\u003e\n\u003cli\u003eRosin, R., Topka, H. \u0026amp; Dichgans, J. Gait initiation in Parkinson\u0026apos;s disease. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 682\u0026ndash;690 (1997).\u003c/li\u003e\n\u003cli\u003eLudlow, C. L., Connor, N. P. \u0026amp; Bassich, C. J. Speech timing in Parkinson\u0026rsquo;s and Huntington\u0026rsquo;s disease. \u003cem\u003eBrain and Language\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 195\u0026ndash;214 (1987).\u003c/li\u003e\n\u003cli\u003eAckermann, H., Konczak, J. \u0026amp; Hertrich, I. The Temporal Control of Repetitive Articulatory Movements in Parkinson\u0026rsquo;s Disease. \u003cem\u003eBrain and Language\u003c/em\u003e \u003cstrong\u003e56\u003c/strong\u003e, 312\u0026ndash;319 (1997).\u003c/li\u003e\n\u003cli\u003eDuez, D. Syllable structure, syllable duration and final lengthening in Parkinsonian French speech. \u003cem\u003eJournal of Multilingual Communication Disorders\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 45\u0026ndash;57 (2006).\u003c/li\u003e\n\u003cli\u003eAssaneo, M. F. \u0026amp; Poeppel, D. The coupling between auditory and motor cortices is rate-restricted: Evidence for an intrinsic speech-motor rhythm. \u003cem\u003eScience Advances\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, eaao3842 (2018).\u003c/li\u003e\n\u003cli\u003ePoeppel, D. \u0026amp; Assaneo, M. F. Speech rhythms, and their neural foundations. \u003cem\u003eNature Reviews Neuroscience\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 322\u0026ndash;334 (2020).\u003c/li\u003e\n\u003cli\u003eOrozco-Arroyave, J. R. \u003cem\u003eet al.\u003c/em\u003e Automatic detection of Parkinson\u0026rsquo;s disease in running speech spoken in three different languages. \u003cem\u003eThe Journal of the Acoustical Society of America\u003c/em\u003e \u003cstrong\u003e139\u003c/strong\u003e, 481\u0026ndash;500 (2016).\u003c/li\u003e\n\u003cli\u003eMoro-Velazquez, L., Gomez-Garcia, J. A., Arias-Londo\u0026ntilde;o, J. D., Dehak, N. \u0026amp; Godino-Llorente, J. I. Advances in Parkinson\u0026rsquo;s Disease detection and assessment using voice and speech: A review of the articulatory and phonatory aspects. \u003cem\u003eBiomedical Signal Processing and Control\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 102418 (2021).\u003c/li\u003e\n\u003cli\u003eAli, L., Zhu, C., Zhou, M. \u0026amp; Liu, Y. Early diagnosis of Parkinson\u0026rsquo;s disease from multiple voice recordings by simultaneous sample and feature selection. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e \u003cstrong\u003e137\u003c/strong\u003e, 22\u0026ndash;28 (2019).\u003c/li\u003e\n\u003cli\u003ePinto, S., Chan, A., Guimar\u0026atilde;es, I., Rothe-Neves, R. \u0026amp; Sadat, J. A cross-linguistic perspective to the study of dysarthria in Parkinson\u0026rsquo;s disease. \u003cem\u003eJournal of Phonetics\u003c/em\u003e \u003cstrong\u003e64\u003c/strong\u003e, 156\u0026ndash;167 (2017).\u003c/li\u003e\n\u003cli\u003eCantiniaux, S. \u003cem\u003eet al.\u003c/em\u003e Comparative analysis of gait and speech in Parkinson\u0026rsquo;s disease: hypokinetic or dysrhythmic disorders? \u003cem\u003eJournal of Neurology, Neurosurgery \u0026amp; Psychiatry\u003c/em\u003e \u003cstrong\u003e81\u003c/strong\u003e, 177\u0026ndash;184 (2010).\u003c/li\u003e\n\u003cli\u003eMart\u0026iacute;nez-S\u0026aacute;nchez, F. \u003cem\u003eet al.\u003c/em\u003e Estudio controlado del ritmo del habla en la enfermedad de Parkinson. \u003cem\u003eNeurolog\u0026iacute;a\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 466\u0026ndash;472 (2016).\u003c/li\u003e\n\u003cli\u003eSkodda, S., Flasskamp, A. \u0026amp; Schlegel, U. Instability of syllable repetition in Parkinson\u0026rsquo;s disease\u0026mdash;Influence of levodopa and deep brain stimulation. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 728\u0026ndash;730 (2011).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3937556/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3937556/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eParkinsonism diagnostic tests based on speech samples have been reported with promising results. However, although abnormal auditory feedback integration during speech production and impaired rhythmic organization of speech have been shown in Parkinsonism, these observations have not been integrated into diagnostic tests.\u003c/p\u003e\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eTo identify Parkinsonism and evaluate the power of a novel speech behavioral test (based on rhythmically repeating syllables under different auditory feedback conditions).\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThirty parkinsonism patients and thirty healthy subjects completed the study. Participants were instructed to repeat the PA-TA-KA syllable sequence rhythmically, whispering and speaking aloud under different listening conditions. The produced speech samples were preprocessed, and parameters were extracted. Classical, unpaired comparisons were conducted between patients and controls. Significant parameters were fed to a supervised machine-learning algorithm differentiating patients from controls, and the accuracy, specificity, and sensitivity were computed.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eDifficulties in whispering and articulating under altered auditory feedback conditions, delayed speech onset, and alterations in rhythmic stability were found in the group of patients compared to controls. A machine learning algorithm trained on these parameters to differentiate patients from controls reached an accuracy of 85.4%, a sensitivity of 87.8%, and a specificity of 83.1%.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe current work represents a pilot trial, showing the potential of the introduced behavioral paradigm as an objective and accessible (in cost and time) diagnostic test.\u003c/p\u003e","manuscriptTitle":"Speech's syllabic rhythm and articulatory features produced under different auditory feedback conditions identify Parkinsonism","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-06 19:17:39","doi":"10.21203/rs.3.rs-3937556/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-04-05T08:07:22+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-03-24T19:09:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"32cda65d-e092-42bc-82f8-4eb758324343","date":"2024-03-18T08:49:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"944608b8-b0fa-48ce-883c-aa9a09b050b8","date":"2024-03-15T16:49:47+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-03-15T16:15:52+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-03-14T09:10:57+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-03-03T11:10:36+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-03-03T11:08:16+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-02-07T17:01:39+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e1ae2a5a-d3bc-49f5-9c9c-f0fbac303eae","owner":[],"postedDate":"March 6th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":29124206,"name":"Health sciences/Neurology/Neurological disorders/Peripheral neuropathies"},{"id":29124207,"name":"Health sciences/Biomarkers/Diagnostic markers"}],"tags":[],"updatedAt":"2024-06-25T19:23:22+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-06 19:17:39","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3937556","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3937556","identity":"rs-3937556","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}