Cookie Theft Picture Description Research Articles

Can we track the progression of Alzheimer's Disease via lexical-semantic variables in connected speech?

BackgroundAlzheimer's disease (AD) is one of the most common neurodegenerative disorders worldwide and is characterized by problems with cognition and language, especially word-finding difficulties. The present study focuses on lexical-semantic features via five discourse variables reflecting word-finding difficulties, namely indefinite terms, lexical frequency, repetitions, semantic paraphasias, and use of pronouns. Our aim is twofold: testing whether these variables can discriminate healthy aging from AD, but also mild from moderate AD. Method105 participants were examined from the existing Pitt corpus (available on DementiaBank), which includes the Cookie Theft Picture Description task. 40 participants were healthy controls, 25 were mild AD participants, and 40 moderate AD participants. ResultsThe moderate AD group differed significantly from healthy controls in terms of indefinite terms, repetitions, semantic paraphasias, and pronouns. For the latter variable, mild AD patients also differed significantly from healthy controls. However, none of the variables could differentiate mild from moderate AD. ConclusionFour out of five discourse variables could discriminate healthy aging from moderate AD, while only one could discriminate mild AD patients. This is therefore questioning current literature on connected-speech measures in AD and calling for further research on the variables that could better distinguish mild to moderate AD.

Resting state correlates of picture description informativeness in left vs. right hemisphere chronic stroke.

Despite a growing emphasis on discourse processing in clinical neuroscience, relatively little is known about the neurobiology of discourse production impairments. Individuals with a history of left or right hemisphere stroke can exhibit difficulty with communicating meaningful discourse content, which implies both cerebral hemispheres play a role in this skill. However, the extent to which successful production of discourse content relies on network connections within domain-specific vs. domain-general networks in either hemisphere is unknown. In this study, 45 individuals with a history of either left or right hemisphere stroke completed resting state fMRI and the Cookie Theft picture description task. Participants did not differ in the total number of content units or the percentage of interpretative content units they produced. Stroke survivors with left hemisphere damage produced significantly fewer content units per second than individuals with right hemisphere stroke. Intrinsic connectivity of the left language network was significantly weaker in the left compared to the right hemisphere stroke group for specific connections. Greater efficiency of communication of picture scene content was associated with stronger left but weaker right frontotemporal connectivity of the language network in patients with a history of left hemisphere (but not right hemisphere) stroke. No significant relationships were found between picture description measures and connectivity of the dorsal attention, default mode, or salience networks or with connections between language and other network regions. These findings add to prior behavioral studies of picture description skills in stroke survivors and provide insight into the role of the language network vs. other intrinsic networks during discourse production.

Evaluation of Interpretable Speech Biomarkers for Monitoring Alzheimer’s Disease and Mild Cognitive Impairment Progression

AbstractBackgroundAlzheimer’s disease (AD) is a progressive neurodegenerative disease. Like for other dementias, biomarkers may help characterize distinct aspects of the underlying pathology, predict decline, and monitor disease progression. We extracted a composite array of 24 speech‐based biomarkers using machine learning and signal processing techniques. We then determined which of the given biomarkers was predictive of MCI and AD progression from baseline by conducting a correlation analysis.MethodSpoken responses to a Cookie Theft Picture description task were recorded from 2 subjects with AD, 4 subjects with Mild Cognitive Impairment (MCI) due to AD (biomarker confirmed AD), and 4 subjects with MCI. We automatically extracted acoustic, linguistic, and cognitive biomarkers using speech recognition, acoustic, and language modeling techniques. We later computed Kendall’s tau‐b (tb) correlation between the biomarker values and Montreal Cognitive Assessment (MoCA) scores. Recordings and MoCA scores were collected at baseline, at 6 months after baseline, and at 12 months for a subgroup of the subjects. To perform the correlation analysis, we used scipy.stats.kendalltau library in Python.ResultWith respect to the acoustic biomarkers, a strong negative correlation (tb = ‐0.50) was observed between the MoCA scores and features such as pause time, pause percentage, and pause speech ratio, while a moderate positive correlation (tb = 0.2) was found for rhythm standard deviation. With respect to the linguistic biomarkers a strong positive correlation(tb = 0.50) for the features of corrected type‐token ratio, average sentence length in words, and noun count was reported, while a moderate positive correlation (tb = 0.27) was found for the features of word type count, word token count, and moving average type‐token ratio. The other acoustic, linguistic, and cognitive biomarkers showed a very weak (tb = 0.00 ‐ 0.10) or weak correlation (tb = 0.10 ‐ 0.20).ConclusionAltogether, these results suggest that speech‐based interpretable biomarkers may help clinicians to diagnose AD at earlier stages and monitor disease progression. Our preliminary data suggest that AD patients encounter more problems delivering longer and linguistically elaborated narratives as the disease progresses.

Evaluation of Interpretable Speech Biomarkers for Monitoring Alzheimer’s Disease and Mild Cognitive Impairment Progression

AbstractBackgroundAlzheimer’s disease (AD) is a progressive neurodegenerative disease. Like for other dementias, biomarkers may help characterize distinct aspects of the underlying pathology, predict decline, and monitor disease progression. We extracted a composite array of 24 speech‐based biomarkers using machine learning and signal processing techniques. We then determined which of the given biomarkers was predictive of MCI and AD progression from baseline by conducting a correlation analysis.MethodSpoken responses to a Cookie Theft Picture description task were recorded from 2 subjects with AD, 4 subjects with Mild Cognitive Impairment (MCI) due to AD (biomarker confirmed AD), and 4 subjects with MCI. We automatically extracted acoustic, linguistic, and cognitive biomarkers using speech recognition, acoustic, and language modeling techniques. We later computed Kendall’s tau‐b (τb ) correlation between the biomarker values and Montreal Cognitive Assessment (MoCA) scores. Recordings and MoCA scores were collected at baseline, at 6 months after baseline, and at 12 months for a subgroup of the subjects. To perform the correlation analysis, we used scipy.stats.kendalltau library in Python.ResultWith respect to the acoustic biomarkers, a strong negative correlation (τb= ‐0.50) was observed between the MoCA scores and features such as pause time, pause percentage, and pause speech ratio, while a moderate positive correlation (τb= 0.2) was found for rhythm standard deviation. With respect to the linguistic biomarkers a strong positive correlation(τb= 0.50) for the features of corrected type‐token ratio, average sentence length in words, and noun count was reported, while a moderate positive correlation (τb= 0.27) was found for the features of word type count, word token count, and moving average type‐token ratio. The other acoustic, linguistic, and cognitive biomarkers showed a very weak (τb= 0.00 ‐ 0.10) or weak correlation (τb= 0.10 ‐ 0.20).ConclusionAltogether, these results suggest that speech‐based interpretable biomarkers may help clinicians to diagnose AD at earlier stages and monitor disease progression. Our preliminary data suggest that AD patients encounter more problems delivering longer and linguistically elaborated narratives as the disease progresses.

18 Automatically calculated lexical and sentential context features of connected speech predict cognitive impairment

Objective:Early detection is critical to the effective management of Alzheimer’s Disease (AD) and other forms of dementia. Existing screening assessments are often costly, require substantial expertise to administer, and may be insensitive to mild changes in cognition. A promising alternative is automatically measuring features of connected speech (c.f., Ostrand & Gunstad, 2021, Journal Geriatric Psych & Neurol) to predict impairment. Here, we built on prior work examining how well speech features predicted cognitive impairment. Unique to the current work, we attempted to capture more holistic effects of cognitive impairment by examining the relevance of linguistic features that measure sentential or discourse context properties of speech, including the context in which filler words (e.g., um) occur, and the predictability of individual words within their sentence context, computed from a large computational language model (GPT-2).Participants and Methods:Participants completed the Cookie Theft picture description task, with data available in the DementiaBank corpus (Becker et al., 1994, Arch Neurol). Descriptions that contained at least 50 words (N = 214) were submitted to an automatic feature calculation pipeline written in Python to calculate various part-of-speech counts, lexical diversity metrics, and mean lexical frequency, as well as multiple metrics related to lexical surprisal (i.e., how surprising a word is given its context).Surprisal of individual words was computed using the pre-trained GPT-2 transformer language model (Radford et al., 2019, Comput. Sci.) by computing word probability given the previous 12 words. Multiple linear regression was performed using 17 linguistic features jointly as predictors, and Mini-Mental State Examination (MMSE) score as the outcome variable. Simple regressions were calculated between each feature and MMSE scores to examine the predictability of each linguistic feature on cognitive decline.Results:A multiple linear regression model containing all linguistic features plus demographic information (age, sex, education) significantly predicted MMSE scores (Adjusted R2 = 0.41, F20, 193 = 8.37, p < .001), and explained significantly more variance in MMSE scores than did demographic variables alone (F17193 = 6.85, p < .001). Individual predictors that were significantly correlated with MMSE score included: how unexpected an individual’s word choices were, given the preceding context (median surprisal: r = -0.33, p < 0.001; interquartile range: r = 0.18, p < 0.02), mean lexical frequency (r = -0.50, p < .001), and usage of definite articles (r = 0.31, p < 0.001), nouns (r = 0.26, p < .001), and empty words (r = -0.25, p < 0.001).Conclusions:Participants with lower MMSE scores, indicating greater impairment, used more frequent, yet more surprising, words, and produced more empty words and fewer definite articles and nouns. These results suggest that measures of semantic specificity and coherence of speech could be meaningful predictors of cognitive decline, and can be computed automatically from speech transcriptions. The results also provide novel evidence that computational approaches to estimating lexical predictability may have value in predicting the degree of decline, motivating future work in other speech elicitation tasks and differing clinical groups.

'Eye-Tracking' with Words for Alzheimer's Disease Detection: Time Alignment of Words Enunciation with Image Regions During Image Description Tasks.

Alzheimer's disease (AD) is a progressive neurodegenerative disease that results in cognitive decline, dementia, and eventually death. Diagnosing early signs of AD can help clinicians to improve the quality of life. We developed a non-invasive approach to help neurologists and clinicians to distinguish probable AD patients and healthy controls (HC). The patients' gaze points were followed based on the words they used to describe the Cookie Theft (CT) picture description task. We hypothesized that the timing of words enunciation aligns with the participant's eye movements. The moments that each word was spoken were then aligned with specific regions of the image. We then applied machine learning algorithms to classify probable AD and HC. We randomly selected 60 participants (30 AD and 30 HC) from the Dementia Bank (Pitt Corpus). Five main classifiers were applied to different features extracted from the recorded audio and participants' transcripts (AD and HC). Support vector machine and logistic regression had the highest accuracy (up to 80% and 78.33%, respectively) in three different experiments. In conclusion, point-of-gaze can be applied as a non-invasive and less expensive approach compared to other available methods (e.g., eye tracker devices) for early-stage AD diagnosis.

Individual Differences in Response to Transcranial Direct Current Stimulation With Language Therapy in Subacute Stroke

Background Transcranial direct current stimulation (tDCS) can be used to improve post-stroke aphasia. However, given the mixed evidence for its efficacy, individual differences may moderate the relative benefit of this strategy. In planned exploratory subgroup analyses, we examined whether age, education, sex, brain-derived neurotrophic factor status, and baseline performance individually impacted improvement in picture naming between baseline and 1 week after the end of the therapy, then whether the combination of factors that predicted recovery of naming and discourse differed for those who received concurrent tDCS. Objective Examine whether individual differences influenced the effect of tDCS on language recovery. Methods In this randomized, double-blind, sham-controlled, efficacy study of tDCS combined with language therapy for subacute post-stroke aphasia, patients completed an evaluation including the Philadelphia Naming Test and Cookie Theft picture description, which was analyzed for Content Units (CU) and Syllables/CU. Individual factors were examined using linear models including the interaction between treatment group and subgroup. Results Significant interactions were observed between tDCS group and both age and education. The predictors of a positive response to tDCS differed from the predictors of a positive response to language treatment alone. While baseline performance was an important predictor of future performance regardless of treatment group, responses to treatment without tDCS were influenced by age whereas responses to treatment with tDCS were not. Conclusions Age and education influence the efficacy of different treatment strategies. Refinement of treatment selection is important to the overall individualization and optimization of post-stroke patient care. Trial Registration: ClinicalTrials.gov NCT02674490.

Micro and macro‐linguistic analysis of narrative discourse in Moroccan patients with Alzheimer’s Disease

AbstractBackgroundThe production of narrative discourse or connected speech implies dynamic and complex functions at different levels of neuro‐cognitive, micro (parts of speech) and macro‐linguistic processing (cohesion and coherence). Such functions are classically impaired in Alzheimer’s disease (AD). Several studies have described these impairments in European languages; however, no study has characterized them in terms of neuro‐psycholinguistic patterns in Arabic.MethodWe studied the speech sample of six patients with moderate Alzheimer’s disease from the Alzheimer’s Center in Rabat. They described orally the picture of cookies theft subtest (BDAE). They meet the clinical criteria for AD (McKhann, 2011). The average age is 69 years (SD: 5.9; 61‐77) and of education is 13.6 years (SD: 2.3; 9‐15), duration of the disease is 5 years (SD 2.09; 3‐8). The neuropsychological assessment has been carried out with MoCA (Montréal Cognitive Assessment), MIS (Memory Impairment screen), Fluency tasks and Clock drawing test. Regarding discourse analysis and psycholinguistic variables, we used the method which has been described by Croisile (1996) and Boschi et al (2017).ResultThe neurolinguistic assessment showed an impairment of macro‐linguistic variables. The effectiveness of speech was clearly one of the most impaired variables as well as the syntactic complexity in the majority of patients. Cohesion has been lower than local and global coherence. These macro‐linguistic performances revealed a pragmatic dysfunction as part of the neuropsycho‐linguistic profile of our patients depending on their disease duration. However, at a micro‐linguistic level, functional words and verbs were the most produced in oral description compared to nouns, adjectives and adverbs. Such results might evoke the hypothesis of an impairment of explicit linguistic knowledge which involved different intra and inter‐linguistic systems in AD.ConclusionThe wide range of cognitive‐linguistic impairments which has been highlighted during the cookie theft picture description, showed the importance of assessing micro and macro‐linguistic impairments in narrative speech in Moroccan patients with AD. Such linguistic task is an added‐value to formal assessments and follow‐up of cognitive decline in AD.

Building the Research Path for Identifying and Validating Cognition‐related Digital Biomarkers

AbstractBackgroundCognitive dysfunction is a core feature of dementia spectrum disorders but simple screening tests are less sensitive to detect the deficits, especially in the very early stage of disease course using most standardized test instruments. This results in delayed discovery of disease onset and opportunity for early interventions that might be more effective in delaying or arresting progression. Digital phenotyping using a combination or hardware and software solutions provide potentially much more effective tools for helping clinicians identify subtle signs of cognition deficits and track them in a compressed timeline to confirm the accuracy of decreasing slope indicative of a neurodegenerative etiology.MethodsBeginning in 2017, 1,866 participants were recruited from the Dementia Center, Shuang Ho Hospital, New Taipei City, Taiwan and will be invited to use a smartphone application to collect digital phenotypes including information processing speed, response accuracy, error detection, and multitasking capacity and vocal patterns, such as speech pitch, tone, rhythm. Additionally, four groups of 100 new participants each will be recruited; cognitive unimpaired (CU), subjective cognitive decline (SCD), mild cognitive impairment (MCI) and mild dementia. In addition to the smartphone applications, these 400 participants will undergo additional testing that include the cognitive abilities screening instrument (CASI), Clinical Dementia Rating (CDR), Cookie Theft Picture Description, Eye Tracking Test, and gait/balance.ResultsDigital phenotyping profiles will be constructed for the cohort and validated against the smaller sample to determine those metrics that can serve as actual digital biomarkers. For the smaller sample, additional multi‐dimensional profiles will be constructed that add gait parameters (step length, stride length, step width, foot angle) and eye‐tracking (visual fixation changes, speed, duration) and determine to what extent they increase differential diagnostic accuracy.ConclusionsDigital assessment tools can be used for detecting changes in cognitively related behaviors associated with preclinical and early stage clinical dementia. Future research is needed to confirm the efficacy of these digital phenotypes to serve as digital biomarkers that have high prognostic and diagnostic accuracy.

Silent Pauses and Speech Indices as Biomarkers for Primary Progressive Aphasia.

Background and Objectives: Recent studies highlight the importance of investigating biomarkers for diagnosing and classifying patients with primary progressive aphasia (PPA). Even though there is ongoing research on pathophysiological indices in this field, the use of behavioral variables, and especially speech-derived factors, has drawn little attention in the relevant literature. The present study aims to investigate the possible utility of speech-derived indices, particularly silent pauses, as biomarkers for primary progressive aphasia (PPA). Materials and Methods: We recruited 22 PPA patients and 17 healthy controls, from whom we obtained speech samples based on two elicitation tasks, i.e., cookie theft picture description (CTP) and the patients’ personal narration of the disease onset and course. Results: Four main indices were derived from these speech samples: speech rate, articulation rate, pause frequency, and pause duration. In order to investigate whether these indices could be used to discriminate between the four groups of participants (healthy individuals and the three patient subgroups corresponding to the three variants of PPA), we conducted three sets of analyses: a series of ANOVAs, two principal component analyses (PCAs), and two hierarchical cluster analyses (HCAs). The ANOVAs revealed significant differences between the four subgroups for all four variables, with the CTP results being more robust. The subsequent PCAs and HCAs were in accordance with the initial statistical comparisons, revealing that the speech-derived indices for CTP provided a clearer classification and were especially useful for distinguishing the non-fluent variant from healthy participants as well as from the two other PPA taxonomic categories. Conclusions: In sum, we argue that speech-derived indices, and especially silent pauses, could be used as complementary biomarkers to efficiently discriminate between PPA and healthy speakers, as well as between the three variants of the disease.

Parkinsonics: A Randomized, Blinded, Cross-Over Trial of Group Singing for Motor and Nonmotor Symptoms in Idiopathic Parkinson Disease.

Introduction Parkinson's disease (PD) frequently causes communication difficulties due to various voice impairments and there are few treatment options for vocal/communication complaints. We assessed the effects of weekly group singing on PD patients' objective vocal and motoric function, cognition, mood, self-efficacy, and quality of life. Methods Thirty-two participants were randomly assigned to either a singing group or a facilitated discussion group weekly over 12 weeks. After 12 weeks, participants crossed over for an additional 12 weeks. Evaluations were performed at baseline and every six weeks for 30 weeks. Objective voice measures included volume/loudness (decibels), held vowel duration, jitter, shimmer, and harmonic-to-noise ratio. Additional outcome measures included patient-centered quality of life, voice-related quality of life, MDS-UPDRS, Montreal Cognitive Assessment, and questionnaires assessing depression, self-efficacy, and overall well-being. Results Twenty-six participants (16 M/10F; Hoehn & Yahr stage 2.3 (range 2–3); and age 68.6 (55–89)) completed the study. Across participants in both groups (intention-to-treat analyses), there was significant improvement from baseline in average loudness on the Cookie Theft picture description at 24 weeks (end of interventions), corresponding with improved minimal reading volumes at 24 weeks and 30 weeks (end of study). Similarly, there were improvements in minimal loudness on Rainbow passage reading at 24 and 30 weeks. There were improvements observed in the Emotional Well-Being (mean delta −12.7 points, p = 0.037) and Body Discomfort (mean delta −18.6 points, p = 0.001) domains of the PDQ-39 from baseline to week 24 in the overall cohort and greater improvement in the Communication domain for Group S than Group D after 12 weeks of singing (delta −12.9 points, p = 0.016). Baseline differences between the participant groups (age, gender, Hoehn & Yahr stage, and several voice loudness measures) and observed improvements during the weekly discussion group period limited our ability to attribute all of the above results specifically to singing (per-protocol analyses). No significant changes in other assessed outcome measures were found. Conclusions Weekly group singing may improve some aspects of conversational voice volume and quality of life in PD. Some improvements were sustained at least six weeks after interventions ended. Further investigations of the mechanism of benefit and randomized controlled studies (without crossover) to assess the longitudinal effects of singing in PD are necessary.

An Analysis of Right Hemisphere Stroke Discourse in the Modern Cookie Theft Picture.

Adults with right hemisphere damage demonstrate differences in connected speech compared to controls, but systematic, quantitative methods to capture these differences are lacking. The current study aimed to (a) investigate if measures using the Modern Cookie Theft picture description would identify discourse differences in acute right hemisphere stroke, and (b) examine if discourse differences were associated with documented cognitive impairment. Eighty-four participants completed the Modern Cookie Theft picture description within 5 days of right hemisphere stroke. Descriptions were analyzed for multiple microlinguistic characteristics. Medical charts were retrospectively reviewed for documented presence of cognitive impairment. Individuals with acute right hemisphere stroke produced fewer content units, total syllables, and lower left-right content unit ratios compared to controls, indicating a paucity of informativeness. Presence of cognitive impairment was associated with fewer content units produced. Multiple measures of microlinguistic discourse characteristics differentiated adults with right hemisphere stroke from controls, highlighting variations in both the quantity and quality of connected speech. Findings continue to underscore the contribution and correlation between cognitive skills and discourse performance. Future work is needed to assess the relationship between particular cognitive domains and discourse production as well as to investigate longitudinal changes to discourse production during stroke recovery. https://doi.org/10.23641/asha.20778541.

Spatio-Semantic Graphs From Picture Description: Applications to Detection of Cognitive Impairment.

Clinical assessments often use complex picture description tasks to elicit natural speech patterns and magnify changes occurring in brain regions implicated in Alzheimer's disease and dementia. As The Cookie Theft picture description task is used in the largest Alzheimer's disease and dementia cohort studies available, we aimed to create algorithms that could characterize the visual narrative path a participant takes in describing what is happening in this image. We proposed spatio-semantic graphs, models based on graph theory that transform the participants' narratives into graphs that retain semantic order and encode the visuospatial information between content units in the image. The resulting graphs differ between Cognitively Impaired and Unimpaired participants in several important ways. Cognitively Impaired participants consistently scored higher on features that are heavily associated with symptoms of cognitive decline, including repetition, evidence of short-term memory lapses, and generally disorganized narrative descriptions, while Cognitively Unimpaired participants produced more efficient narrative paths. These results provide evidence that spatio-semantic graph analysis of these tasks can generate important insights into a participant's cognitive performance that cannot be generated from semantic analysis alone.

Comparison of remote and in‐person digital speech‐based measures of cognition

AbstractBackgroundNeuropsychological testing requires an in‐person visit with a trained administrator using standard/fixed materials. Speech‐based cognitive testing on mobile devices enables more frequent and timely test administration, but head‐to‐head comparisons of in‐person and remote versions of tests are rare. We compare responses to a well‐validated task conducted under supervised (in‐person) and unsupervised (remote) conditions.MethodWe used two data sets containing Cookie Theft picture descriptions (BDAE), one collected in‐person, under supervision (Wisconsin Registry for Alzheimer’s Prevention (WRAP); N = 912; age = 62.2 (SD = 6.7), 70% F) and one collected remotely, without supervision (participants online; N = 93, age = 36.5 (SD = 11.5), 65% F). WRAP participants were judged to be cognitively normal, and online participants self‐reported as healthy. Each participant provided one picture description, yielding 93 remote and 912 in‐person transcribed descriptions. Language metrics previously used in dementia research were extracted: word count (number of words spoken), MATTR (ratio of unique words to total number of words), pronoun‐to‐noun ratio, and semantic relevance. Comparing MATTR, pronoun‐to‐noun ratio, and semantic relevance values elicited in‐person and remotely is important because these characteristics have been found to be impacted by declines in cognition. Differences in word counts may reflect different levels of motivation in the two settings.ResultUsing Cohen’s d effect sizes, differences in mean word count between in‐person and remote participant transcripts was negligible (112 words in‐person and 118 words remote; d = .10, p = .21), the difference for semantic relevance was negligible (.19 in‐person and .18 remote, scale = 0 to 1; d = .14, p = .19), the difference for MATTR was small (.75 in‐person and .74 remote, scale = 0 to 1; d = .37, p<.05), and the difference for pronoun‐to‐noun ratio was moderate (.48 in‐person and .35 remote, scale = 0 to 1; d = .56, p <.05).ConclusionResults show that response length and semantic relevance of responses to the Cookie Theft picture description task are comparable under supervised, in‐person and unsupervised, remote conditions. Small to moderate differences in vocabulary (MATTR) and pronoun‐to‐noun ratio may be due to differences in age and merit further investigation.

Machine learning of transcripts and audio recordings of spontaneous speech for diagnosis of Alzheimer’s disease

AbstractBackgroundLanguage deficits may occur in early stages of Alzheimer’s disease (AD). Analysis of individuals’ language changes has been shown to predict the diagnosis and severity of AD. This work explored machine learning methods to detect AD using voice recordings of the Cookie Theft picture description tasks.MethodAn open‐source dataset from Dementia Bank, which included manual transcripts and audio recordings of the Cookie Theft picture description task from 309 individuals with AD and 243 cognitively healthy older adults, was employed in the classification analysis. We analyzed transcripts generated using three different approaches (Figure 1), including a) automatically generated transcripts based on denoised voice recording; b) automatically generated transcripts without the denoising process; and c) manuscript transcripts. Denoising was implemented using noisereduce, a Python library to minimize background noise (CITATION). Automatic transcripts were produced using two toolkits: CMUSpinx pocketsphinx interpreter (PS) and Mozilla DeepSpeech (DS). The Small‐Bidirectional Encoder Representations from Transformers (BERT) model was implemented to extract high dimensional feature vectors to provide contextual representation of transcripts for classification analysis. The experiment was designed to compare the performance of transcript type (manual vs. automatic), recording quality (noise vs. denoise), as well as speech recognition methods (DS vs. PS). A neural network model was constructed to classify between those with AD and healthy controls. The classification analysis was conducted using 10‐fold cross validation. The classification was repeated 20 times and the averaged results were reported (see Table 1).ResultThe performance of each approach is shown as Table 1. Overall, the model that used the features extracted from DS transcripts with denoised speech recordings achieved the best performance which had 92.72% accuracy for AD diagnosis validation results. The comparison between raw and denoised speech recordings showed conflicting performances. Using features from PS transcripts with noisy speech recordings obtained better performance than the denoised one, which was different with features from DS transcripts.ConclusionThe analysis approach which combined transcription driven recognition and transfer learning indicated the potential of better performance in AD diagnosis. Further investigation is needed to understand the reason for its superior performance.

Written Discourse Task Helps to Identify Progression from Mild Cognitive Impairment to Dementia

Purpose: We aimed to investigate: (1) the clinical, diagnostic value of a written discourse task, and (2) the relationship between executive functions and written discourse within the spectrum of individuals with mild cognitive impairment (MCI). Method: To determine whether written discourse performance predicts clinical course among individuals with MCI, we retrospectively classified individuals with MCI as converters (N = 26) who were later diagnosed with dementia or as a stable MCI group (N = 45). We quantified core word measures from written discourse samples obtained from the Cookie Theft picture description task. Result: Written discourse measures differentiated converters from the stable MCI group. Converters produced a fewer number of core words than the stable MCI group. A measure of executive function significantly predicted performance on the production of core words in written discourse for the converters. In a multivariable regression, production of core words remained the only explanatory variable closely associated with the progression to dementia in MCI. Conclusion: Written discourse tasks can predict the likelihood of MCI progressing to dementia, independently of recall and an executive function measure. Correlational results suggest that written discourse performance was associated with executive function as measured by the Trail Making Test. Our findings emphasize the usefulness of including written discourse tasks in language assessment batteries targeting preclinical dementia populations.

Identification of the main components of spontaneous speech in primary progressive aphasia and their neural underpinnings using multimodal MRI and FDG-PET imaging

BackgroundPrimary progressive aphasia (PPA) is a clinical syndrome characterized by gradual loss of language skills. This study aimed to evaluate the diagnostic capacity of a connected speech task for the diagnosis of PPA and its variants, to determine the main components of spontaneous speech, and to examine their neural correlates. MethodsA total of 118 participants (31 patients with nfvPPA, 11 with svPPA, 45 with lvPPA, and 31 healthy controls) were evaluated with the Cookie Theft picture description task and a comprehensive language assessment protocol. Patients also underwent 18F-fluorodeoxyglucose positron emission tomography and magnetic resonance imaging studies. Principal component analysis and machine learning were used to evaluate the main components of connected speech and the accuracy of connected speech parameters for diagnosing PPA. Voxel-based analyses were conducted to evaluate the correlation between spontaneous speech components and brain metabolism, brain volumes, and white matter microstructure. ResultsDiscrimination between patients with PPA and controls was 91.67%, with 77.78% discrimination between PPA variants. Parameters related to speech rate and lexical variables were the most discriminative for classification. Three main components were identified: lexical features, fluency, and syntax. The lexical component was associated with ventrolateral frontal regions, while the fluency component was associated with the medial superior prefrontal cortex. Number of pauses was more related with the left parietotemporal region, while pauses duration with the bilateral frontal lobe. The lexical component was correlated with several tracts in the language network (left frontal aslant tract, left superior longitudinal fasciculus I, II, and III, left arcuate fasciculus, and left uncinate fasciculus), and fluency was linked to the frontal aslant tract. ConclusionSpontaneous speech assessment is a useful, brief approach for the diagnosis of PPA and its variants. Neuroimaging correlates suggested a subspecialization within the left frontal lobe, with ventrolateral regions being more associated with lexical production and the medial superior prefrontal cortex with speech rate.

A Linguistic Profile of Narrative Speech in Early and Late-Onset Alzheimer’s Disease

Objective. Alzheimer’s Dementia is defined as multiple impairments in cognitive functions and language impairment is one of the observed problems in this disease. In order to analyse these problems, narrative speech transcriptions of Turkish patients with early and late-onset Alzheimer’s Disease (AD) and an age/education-matched control group were examined.
 Materials & Methods. By using different language tests, the sentence production of Turkish patients with early (n: 23) and late-onset (n: 39) was examined and the results were compared within groups and within language tests. Our aim was to reveal the differences in sentence processing of Turkish Alzheimer patients. 
 Results. According to the results, Turkish Alzheimer patients use short and simple sentences composed of fewer words in their speech. It was determined that these patients performed better in Picnic Picture Description test, which is thought to be easy to describe. In Cookie Theft Picture Description test, which is a complex one, and in the Subject-based Narration test, it was found that the speech amounts of AD patients were less. It was revealed that early-onset AD patients used shorter sentences including fewer words compared to late-onset AD patients. 
 Conclusion. In the light of the data obtained, it can be said that AD patients have different sentence processing than normal individuals because of the problems in the language and memory areas of the disease. The fact that sentence processing is different means that these patients do not prefer the types of sentences that normal individuals prefer in their speeches, and use less words and simpler sentences.

A linguistic profile of narrative speech in early and late Alzheimer's disease

Objective. Alzheimer’s Dementia is defined as multiple impairments in cognitive functions and language impairment is one of the observed problems in this disease. In order to analyse these problems, narrative speech transcriptions of Turkish patients with early and late-onset Alzheimer’s Disease (AD) and an age/education-matched control group were examined. Materials & Methods. By using different language tests, the sentence production of Turkish patients with early (n: 23) and late-onset (n: 39) was examined and the results were compared within groups and within language tests. Our aim was to reveal the differences in sentence processing of Turkish Alzheimer patients. Results. According to the results, Turkish Alzheimer patients use short and simple sentences composed of fewer words in their speech. It was determined that these patients performed better in Picnic Picture Description test, which is thought to be easy to describe. In Cookie Theft Picture Description test, which is a complex one, and in the Subject-based Narration test, it was found that the speech amounts of AD patients were less. It was revealed that early-onset AD patients used shorter sentences including fewer words compared to late-onset AD patients. Conclusion. In the light of the data obtained, it can be said that AD patients have different sentence processing than normal individuals because of the problems in the language and memory areas of the disease. The fact that sentence processing is different means that these patients do not prefer the types of sentences that normal individuals prefer in their speeches, and use less words and simpler sentences.

Description of connected speech across different elicitation tasks in the logopenic variant of primary progressive aphasia.

Despite its importance, in-depth analysis of connected speech is often neglected in the diagnosis of primary progressive aphasia (PPA) - especially for the logopenic variant (lvPPA) for which unreliable differential diagnosis has been documented. Only a few studies have been conducted on this topic in lvPPA. The aim of this study was to describe and compare lexico-semantic and morphosyntactic features of connected speech in participants with lvPPA, in comparison with healthy controls, using three different elicitation tasks (i.e., picture description, story narration and semi-structured interviews). In addition to a number of discourse features, we were particularly interested in the presence or absence of syntactic deficits in this PPA variant in line with recent findings. A prospective group study was conducted to compare lvPPA participants (n = 13) to age- and education-matched healthy controls (n = 13). For each individual, connected speech was obtained using three tasks: (1) The Cookie Theft picture description; (2) Cinderella Story; (3) Topic-directed interview. Production on each task was recorded, transcribed and analysed according to the Quantitative Production Analysis (QPA) protocol, a tool developed by Berndt etal. (2000) for the analysis of sentence production in aphasia. Differences between lvPPA and healthy controls and among elicitation tasks were analysed using repeated measures multilevel mixed-effects regression, separately for each outcome. Four measures were significantly different between lvPPA participants and healthy controls across all elicitation tasks. Specifically, lvPPA participants produced a reduced proportion of open-class words, a higher proportion of verbs, a higher proportion of pronouns and fewer well-formed sentences. For these measures, the difference between lvPPA and healthy controls was consistent among elicitation tasks, except for the proportion of well-formed sentences, where the difference between the two groups was significantly greater in the story narration task than in the other tasks. Across elicitation tasks that used the same analysis protocol (i.e., QPA), a similar pattern of deficits in connected speech emerged in lvPPA patients. Importantly, the findings replicate previous studies, which used different elicitation tasks and analysis protocols. Especially in relation to the documented syntactic deficits, these findings provide implications for differential diagnosis in PPA. What is already known on the subject Connected speech analysis can provide an important contribution to the language assessment for the logopenic variant of primary progressive aphasia (lvPPA). However, only a few studies have been conducted with this population. What this paper adds to existing knowledge This study highlights differences between patients with lvPPA and healthy controls regarding the proportion of open-class words, nouns, verbs and well-formed sentences. What are the potential or actual clinical implications of this work? Importantly, our results highlight syntactic deficits in the same group of individuals with lvPPA, using the same analysis protocol and across various elicitation tasks, which has implications for differential diagnosis.