Artificial Intelligence, Speech, and Language Processing Approaches to Monitoring Alzheimer's Disease: A Systematic Review

Importance of Task Selection for Connected Speech Analysis in Patients with Alzheimer’s Disease from an Ethnically Diverse Sample

Features of linguistic impairment in Alzheimer’s disease (AD) are primarily derived from English-speaking patients. Little is known regarding such deficits in linguistically diverse speakers with AD. We aimed to detail linguistic profiles (speech rate, dysfluencies, syntactic, lexical, morphological, semantics) from two connected speech tasks–Frog Story and picture description–in Bengali-speaking AD patients. The Frog Story detected group differences on all six linguistic levels, compared to only three with picture description. Critically, Frog Story captured the language-specific differences between the groups. Careful consideration should be given to the choice of connected speech tasks for dementia diagnosis in linguistically diverse populations.

Temporal speech parameters detect mild cognitive impairment in different languages: validation and comparison of the Speech-GAP Test® in English and Hungarian

BACKGROUND

The development of automatic speech recognition (ASR) technology allows the analysis of temporal (time-based) speech parameters characteristic of mild cognitive impairment (MCI). However, no information has been available on whether the analysis of spontaneous speech can be used with the same efficiency in different language environments.

OBJECTIVE

The main goal of this international pilot study is to address the question whether the Speech-Gap Test® (S-GAP Test®), previously tested in the Hungarian language, is appropriate for and applicable to the recognition of MCI in other languages such as English.

METHOD

After an initial screening of 88 individuals, English-speaking (n = 33) and Hungarian-speaking (n = 33) participants were classified as having MCI or as healthy controls (HC) based on Petersen's criteria. Speech of each participant was recorded via a spontaneous speech task. 15 temporal parameters were determined and calculated by means of ASR.

RESULTS

Seven temporal parameters in the English-speaking sample and 5 in the Hungarian-speaking sample showed significant differences between the MCI and the HC group. Receiver operating characteristics (ROC) analysis clearly distinguished the English-speaking MCI cases from the HC group based on speech tempo and articulation tempo with 100% sensitivity, and on three more temporal parameters with high sensitivity (85.7%). In the Hungarian-speaking sample, the ROC analysis showed similar sensitivity rates (92.3%).

CONCLUSION

The results of this study in different native-speaking populations suggest that changes in acoustic parameters detected by the S-GAP Test® might be present across different languages.

Automated text-level semantic markers of Alzheimer's disease

INTRODUCTION

Automated speech analysis has emerged as a scalable, cost-effective tool to identify persons with Alzheimer's disease dementia (ADD). Yet, most research is undermined by low interpretability and specificity.

METHODS

Combining statistical and machine learning analyses of natural speech data, we aimed to discriminate ADD patients from healthy controls (HCs) based on automated measures of domains typically affected in ADD: semantic granularity (coarseness of concepts) and ongoing semantic variability (conceptual closeness of successive words). To test for specificity, we replicated the analyses on Parkinson's disease (PD) patients.

RESULTS

Relative to controls, ADD (but not PD) patients exhibited significant differences in both measures. Also, these features robustly discriminated between ADD patients and HC, while yielding near-chance classification between PD patients and HCs.

DISCUSSION

Automated discourse-level semantic analyses can reveal objective, interpretable, and specific markers of ADD, bridging well-established neuropsychological targets with digital assessment tools.

International Classification of Functioning, Disability, and Health augmented by telemedicine and artificial intelligence for assessment of functional disability

The concept of functional disability is aligned with the biopsycho-social model of disability. However, there are reasons why the antiquated measurement of medical impairment continues to be in use. We propose solutions for a fairer process using the International Classification of Functioning, Disability, and Health (ICF) at the level of the medical boards augmented by telemedicine and artificial intelligence (AI). The proposed technologies (Level 1 and Level 2 AI) need to be tried in pilot projects. It will accomplish two goals, the first being the measurement of disability and not merely the impairment. Second, and perhaps more importantly, making the process more transparent in creating a "just" society.

A deep learning framework with an embedded-based feature selection approach for the early detection of the Alzheimer's disease

Ageing is associated with various ailments including Alzheimer 's disease (AD), which is a progressive form of dementia. AD symptoms develop over a period of years and, unfortunately, there is no cure. Existing AD treatments can only slow down the progression of symptoms and thus it is critical to diagnose the disease at an early stage. To help improve the early diagnosis of AD, a deep learning-based classification model with an embedded feature selection approach was used to classify AD patients. An AD DNA methylation data set (64 records with 34 cases and 34 controls) from the GEO omnibus database was used for the analysis. Before selecting the relevant features, the data were preprocessed by performing quality control, normalization and downstream analysis. As the number of associated CpG sites was huge, four embedded-based feature selection models were compared and the best method was used for the proposed classification model. An Enhanced Deep Recurrent Neural Network (EDRNN) was implemented and compared to other existing classification models, including a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), and a Deep Recurrent Neural Network (DRNN). The results showed a significant improvement in the classification accuracy of the proposed model as compared to the other methods.

Using Digital Speech Assessments to Detect Early Signs of Cognitive Impairment

Detecting early signs of cognitive decline is crucial for early detection and treatment of Alzheimer's Disease. Most of the current screening tools for Alzheimer's Disease represent a significant burden, requiring invasive procedures, or intensive and costly clinical testing. Recent findings have highlighted changes to speech and language patterns that occur in Alzheimer's Disease, and may be detectable prior to diagnosis. Automated tools to assess speech have been developed that can be used on a smartphone or tablet, from one's home, in under 10 min. In this study, we present the results of a study of older adults who completed a digital speech assessment task over a 6-month period. Participants were grouped according to those who scored above (N = 18) or below (N = 18) the recommended threshold for detecting cognitive impairment on the Montreal Cognitive Assessment (MoCA) and those with diagnoses of mild cognitive impairment (MCI) or early Alzheimer's Disease (AD) (N = 14). Older adults who scored above the MoCA threshold had better performance on speech composites reflecting language coherence, information richness, syntactic complexity, and word finding abilities. Those with MCI and AD showed more rapid decline in the coherence of language from baseline to 6-month follow-up, suggesting that this score may be useful both for detecting cognitive decline and monitoring change over time. This study demonstrates that automated speech assessments have potential as sensitive tools to detect early signs of cognitive impairment and monitor progression over time.

Digital medicine and the curse of dimensionality

Digital health data are multimodal and high-dimensional. A patient's health state can be characterized by a multitude of signals including medical imaging, clinical variables, genome sequencing, conversations between clinicians and patients, and continuous signals from wearables, among others. This high volume, personalized data stream aggregated over patients' lives has spurred interest in developing new artificial intelligence (AI) models for higher-precision diagnosis, prognosis, and tracking. While the promise of these algorithms is undeniable, their dissemination and adoption have been slow, owing partially to unpredictable AI model performance once deployed in the real world. We posit that one of the rate-limiting factors in developing algorithms that generalize to real-world scenarios is the very attribute that makes the data exciting-their high-dimensional nature. This paper considers how the large number of features in vast digital health data can challenge the development of robust AI models-a phenomenon known as "the curse of dimensionality" in statistical learning theory. We provide an overview of the curse of dimensionality in the context of digital health, demonstrate how it can negatively impact out-of-sample performance, and highlight important considerations for researchers and algorithm designers.

Connected speech markers of amyloid burden in primary progressive aphasia

INTRODUCTION

Positron emission tomography (PET) amyloid imaging has become an important part of the diagnostic workup for patients with primary progressive aphasia (PPA) and uncertain underlying pathology. Here, we employ a semi-automated analysis of connected speech (CS) with a twofold objective. First, to determine if quantitative CS features can help select primary progressive aphasia (PPA) patients with a higher probability of a positive PET amyloid imaging result. Second, to examine the relevant group differences from a clinical perspective.

METHODS

117 CS samples from a well-characterised cohort of PPA patients who underwent PET amyloid imaging were collected. Expert consensus established PET amyloid status for each patient, and 40% of the sample was amyloid positive.

RESULTS

Leave-one-out cross-validation yields 77% classification accuracy (sensitivity: 74%, specificity: 79%).

DISCUSSION

Our results confirm the potential of CS analysis as a screening tool. Discriminant CS features from lexical, syntactic, pragmatic, and semantic domains are discussed.

Novel tacrine-based acetylcholinesterase inhibitors as potential agents for the treatment of Alzheimer's disease: Quinolotacrine hybrids

A new series of quinolotacrine hybrids including cyclopenta- and cyclohexa-quinolotacrine derivatives were designed, synthesized, and assessed as anti-cholinesterase (ChE) agents. The designed derivatives indicated higher inhibitory effect on the acetylcholinesterase (AChE) with IC₅₀ values of 0.285-100 µM compared to butyrylcholinesterase (BChE) with IC₅₀ values of > 100 µM. Of these compounds, cyclohexa-quinolotacrine hybrids displayed a little better anti-AChE activity than cyclopenta-quinolotacrine hybrids. Compound 8-amino-7-(3-hydroxyphenyl)-5,7,9,10,11,12-hexahydro-6H-pyrano[2,3-b:5,6-c'] diquinolin-6-one (6m) including 3-hydroxyphenyl and cyclohexane ring moieties exhibited the best AChE inhibitory activity with IC₅₀ value of 0.285 µM. The kinetic and molecular docking studies indicated that compound 6m occupied both the catalytic anionic site (CAS) and peripheral anionic site (PAS) of AChE as a mixed inhibitor. Using neuroprotective assay against H₂O₂-induced cell death in PC12 cells, the compound 6h illustrated significant protection among the assessed compounds. In silico ADME studies estimated good drug-likeness for the designed compounds. As a result, these quinolotacrine hybrids can be very encouraging AChE inhibitors to treat Alzheimer's disease. A novel series of quinolotacrine hybrids were designed, synthesized, and evaluated against AChE and BChE enzymes as potential agents for the treatment of AD. The hybrids showed good to significant inhibitory activity against AChE (0.285-100 μM) compared to butyrylcholinesterase (BChE) with IC₅₀ values of > 100 μM. Among them, compound 8-amino-7-(3-hydroxyphenyl)-5,7,9,10,11,12-hexahydro-6H-pyrano[2,3-b:5,6-c'] diquinolin-6-one (6 m) bearing 3-hydroxyphenyl moiety and cyclohexane ring exhibited the highest anti-AChE activity with IC₅₀ value of 0.285 μM. The kinetic and molecular docking studies illustrated that compound 6 m is a mixed inhibitor and binds to both the catalytic anionic site (CAS) and peripheral anionic site (PAS) of AChE.

The Use of Mobile Applications as Communication Aids for People with Dementia: Opportunities and Limitations

BACKGROUND

Communication difficulties are one of the primary symptoms associated with dementia, and mobile applications have shown promise as tools for facilitating communication in patients with dementia (PwD). The literature regarding mobile health (mHealth) applications, especially communications-based mHealth applications, is limited.

OBJECTIVE

This review aims to compile the existing literature on communications-based mobile applications regarding dementia and assess their opportunities and limitations. A PICO framework was applied with a Population consisting of PwD, Interventions consisting of communication technology, focusing primarily on mobile applications, Comparisons between patient well-being with and without technological intervention, and Outcomes that vary but can include usability of technology, quality of communication, and user acceptance.

METHODS

Searches of PubMed, IEEE XPLORE, and ACM Digital Library databases were conducted to establish a comprehensive understanding of the current literature on dementia care as related to 1) mobile applications, 2) communication technology, and 3) communications-based mobile applications. Applying certain inclusion and exclusion criteria, yielded a set of articles (n = 11).

RESULTS

The literature suggests that mobile applications as tools for facilitating communication in PwD are promising. Mobile applications are not only feasible socially, logistically, and financially, but also produce meaningful communication improvements in PwD and their caregivers. However, the number of satisfactory communications-based mobile applications in the mHealth marketplace and their usability is still insufficient.

CONCLUSION

Despite favorable outcomes, more research involving PwD using these applications are imperative to shed further light on their communication needs and on the role of mHealth.

Language in Behavioral Variant Frontotemporal Dementia: Another Stone to Be Turned in Latin America

Beyond canonical deficits in social cognition and interpersonal conduct, behavioral variant frontotemporal dementia (bvFTD) involves language difficulties in a substantial proportion of cases. However, since most evidence comes from high-income countries, the scope and relevance of language deficits in Latin American bvFTD samples remain poorly understood. As a first step toward reversing this scenario, we review studies reporting language measures in Latin American bvFTD cohorts relative to other groups. We identified 24 papers meeting systematic criteria, mainly targeting phonemic and semantic fluency, naming, semantic processing, and comprehension skills. The evidence shows widespread impairments in these domains, often related to overall cognitive disturbances. Some of these deficits may be as severe as in other diseases where they are more widely acknowledged, such as Alzheimer's disease. Considering the prevalence and informativeness of language deficits in bvFTD patients from other world regions, the need arises for more systematic research in Latin America, ideally spanning multiple domains, in diverse languages and dialects, with validated batteries. We outline key challenges and pathways of progress in this direction, laying the ground for a new regional research agenda on the disorder.

Temporal Integration of Text Transcripts and Acoustic Features for Alzheimer's Diagnosis Based on Spontaneous Speech

Background: Advances in machine learning (ML) technology have opened new avenues for detection and monitoring of cognitive decline. In this study, a multimodal approach to Alzheimer's dementia detection based on the patient's spontaneous speech is presented. This approach was tested on a standard, publicly available Alzheimer's speech dataset for comparability. The data comprise voice samples from 156 participants (1:1 ratio of Alzheimer's to control), matched by age and gender.

Materials and Methods: A recently developed Active Data Representation (ADR) technique for voice processing was employed as a framework for fusion of acoustic and textual features at sentence and word level. Temporal aspects of textual features were investigated in conjunction with acoustic features in order to shed light on the temporal interplay between paralinguistic (acoustic) and linguistic (textual) aspects of Alzheimer's speech. Combinations between several configurations of ADR features and more traditional bag-of-n-grams approaches were used in an ensemble of classifiers built and evaluated on a standardised dataset containing recorded speech of scene descriptions and textual transcripts.

Results: Employing only semantic bag-of-n-grams features, an accuracy of 89.58% was achieved in distinguishing between Alzheimer's patients and healthy controls. Adding temporal and structural information by combining bag-of-n-grams features with ADR audio/textual features, the accuracy could be improved to 91.67% on the test set. An accuracy of 93.75% was achieved through late fusion of the three best feature configurations, which corresponds to a 4.7% improvement over the best result reported in the literature for this dataset.

Conclusion: The proposed combination of ADR audio and textual features is capable of successfully modelling temporal aspects of the data. The machine learning approach toward dementia detection achieves best performance when ADR features are combined with strong semantic bag-of-n-grams features. This combination leads to state-of-the-art performance on the AD classification task.

Language Impairment in Alzheimer's Disease-Robust and Explainable Evidence for AD-Related Deterioration of Spontaneous Speech Through Multilingual Machine Learning

Alzheimer's disease (AD) is a pervasive neurodegenerative disease that affects millions worldwide and is most prominently associated with broad cognitive decline, including language impairment. Picture description tasks are routinely used to monitor language impairment in AD. Due to the high amount of manual resources needed for an in-depth analysis of thereby-produced spontaneous speech, advanced natural language processing (NLP) combined with machine learning (ML) represents a promising opportunity. In this applied research field though, NLP and ML methodology do not necessarily ensure robust clinically actionable insights into cognitive language impairment in AD and additional precautions must be taken to ensure clinical-validity and generalizability of results. In this study, we add generalizability through multilingual feature statistics to computational approaches for the detection of language impairment in AD. We include 154 participants (78 healthy subjects, 76 patients with AD) from two different languages (106 English speaking and 47 French speaking). Each participant completed a picture description task, in addition to a battery of neuropsychological tests. Each response was recorded and manually transcribed. From this, task-specific, semantic, syntactic and paralinguistic features are extracted using NLP resources. Using inferential statistics, we determined language features, excluding task specific features, that are significant in both languages and therefore represent "generalizable" signs for cognitive language impairment in AD. In a second step, we evaluated all features as well as the generalizable ones for English, French and both languages in a binary discrimination ML scenario (AD vs. healthy) using a variety of classifiers. The generalizable language feature set outperforms the all language feature set in English, French and the multilingual scenarios. Semantic features are the most generalizable while paralinguistic features show no overlap between languages. The multilingual model shows an equal distribution of error in both English and French. By leveraging multilingual statistics combined with a theory-driven approach, we identify AD-related language impairment that generalizes beyond a single corpus or language to model language impairment as a clinically-relevant cognitive symptom. We find a primary impairment in semantics in addition to mild syntactic impairment, possibly confounded by additional impaired cognitive functions.

Analysis and Classification of Word Co-Occurrence Networks From Alzheimer’s Patients and Controls

In this paper we construct word co-occurrence networks from transcript data of controls and patients with potential Alzheimer’s disease using the ADReSS challenge dataset of spontaneous speech. We examine measures of the structure of these networks for significant differences, finding that networks from Alzheimer’s patients have a lower heterogeneity and centralization, but a higher edge density. We then use these measures, a network embedding method and some measures from the word frequency distribution to classify the transcripts into control or Alzheimer’s, and to estimate the cognitive test score of a participant based on the transcript. We find it is possible to distinguish between the AD and control networks on structure alone, achieving 66.7% accuracy on the test set, and to predict cognitive scores with a root mean squared error of 5.675. Using the network measures is more successful than using the network embedding method. However, if the networks are shuffled we find relatively few of the measures are different, indicating that word frequency drives many of the network properties. This observation is borne out by the classification experiments, where word frequency measures perform similarly to the network measures.

Comparing Pre-trained and Feature-Based Models for Prediction of Alzheimer's Disease Based on Speech

Introduction: Research related to the automatic detection of Alzheimer's disease (AD) is important, given the high prevalence of AD and the high cost of traditional diagnostic methods. Since AD significantly affects the content and acoustics of spontaneous speech, natural language processing, and machine learning provide promising techniques for reliably detecting AD. There has been a recent proliferation of classification models for AD, but these vary in the datasets used, model types and training and testing paradigms. In this study, we compare and contrast the performance of two common approaches for automatic AD detection from speech on the same, well-matched dataset, to determine the advantages of using domain knowledge vs. pre-trained transfer models. Methods: Audio recordings and corresponding manually-transcribed speech transcripts of a picture description task administered to 156 demographically matched older adults, 78 with Alzheimer's Disease (AD) and 78 cognitively intact (healthy) were classified using machine learning and natural language processing as "AD" or "non-AD." The audio was acoustically-enhanced, and post-processed to improve quality of the speech recording as well control for variation caused by recording conditions. Two approaches were used for classification of these speech samples: (1) using domain knowledge: extracting an extensive set of clinically relevant linguistic and acoustic features derived from speech and transcripts based on prior literature, and (2) using transfer-learning and leveraging large pre-trained machine learning models: using transcript-representations that are automatically derived from state-of-the-art pre-trained language models, by fine-tuning Bidirectional Encoder Representations from Transformer (BERT)-based sequence classification models. Results: We compared the utility of speech transcript representations obtained from recent natural language processing models (i.e., BERT) to more clinically-interpretable language feature-based methods. Both the feature-based approaches and fine-tuned BERT models significantly outperformed the baseline linguistic model using a small set of linguistic features, demonstrating the importance of extensive linguistic information for detecting cognitive impairments relating to AD. We observed that fine-tuned BERT models numerically outperformed feature-based approaches on the AD detection task, but the difference was not statistically significant. Our main contribution is the observation that when tested on the same, demographically balanced dataset and tested on independent, unseen data, both domain knowledge and pretrained linguistic models have good predictive performance for detecting AD based on speech. It is notable that linguistic information alone is capable of achieving comparable, and even numerically better, performance than models including both acoustic and linguistic features here. We also try to shed light on the inner workings of the more black-box natural language processing model by performing an interpretability analysis, and find that attention weights reveal interesting patterns such as higher attribution to more important information content units in the picture description task, as well as pauses and filler words. Conclusion: This approach supports the value of well-performing machine learning and linguistically-focussed processing techniques to detect AD from speech and highlights the need to compare model performance on carefully balanced datasets, using consistent same training parameters and independent test datasets in order to determine the best performing predictive model.

Longitudinal Speech Biomarkers for Automated Alzheimer's Detection

We introduce a novel audio processing architecture, the Open Voice Brain Model (OVBM), improving detection accuracy for Alzheimer's (AD) longitudinal discrimination from spontaneous speech. We also outline the OVBM design methodology leading us to such architecture, which in general can incorporate multimodal biomarkers and target simultaneously several diseases and other AI tasks. Key in our methodology is the use of multiple biomarkers complementing each other, and when two of them uniquely identify different subjects in a target disease we say they are orthogonal. We illustrate the OBVM design methodology by introducing sixteen biomarkers, three of which are orthogonal, demonstrating simultaneous above state-of-the-art discrimination for two apparently unrelated diseases such as AD and COVID-19. Depending on the context, throughout the paper we use OVBM indistinctly to refer to the specific architecture or to the broader design methodology. Inspired by research conducted at the MIT Center for Brain Minds and Machines (CBMM), OVBM combines biomarker implementations of the four modules of intelligence: The brain OS chunks and overlaps audio samples and aggregates biomarker features from the sensory stream and cognitive core creating a multi-modal graph neural network of symbolic compositional models for the target task. In this paper we apply the OVBM design methodology to the automated diagnostic of Alzheimer's Dementia (AD) patients, achieving above state-of-the-art accuracy of 93.8% using only raw audio, while extracting a personalized subject saliency map designed to longitudinally track relative disease progression using multiple biomarkers, 16 in the reported AD task. The ultimate aim is to help medical practice by detecting onset and treatment impact so that intervention options can be longitudinally tested. Using the OBVM design methodology, we introduce a novel lung and respiratory tract biomarker created using 200,000+ cough samples to pre-train a model discriminating cough cultural origin. Transfer Learning is subsequently used to incorporate features from this model into various other biomarker-based OVBM architectures. This biomarker yields consistent improvements in AD detection in all the starting OBVM biomarker architecture combinations we tried. This cough dataset sets a new benchmark as the largest audio health dataset with 30,000+ subjects participating in April 2020, demonstrating for the first time cough cultural bias.

[The Use of Artificial Intelligence in Alzheimer's Disease - Personalized Diagnostics and Therapy]

Using the example of dementia in Alzheimer's disease, it is shown which opportunities but also risks are posed by newer methodological approaches of artificial intelligence (AI) for the diagnosis and treatment of Alzheimer's dementia (AD). In addition, AI is examined in the context of an ethical-philosophical critique of technology.

Towards Computer-Based Automated Screening of Dementia Through Spontaneous Speech

Dementia, a prevalent disorder of the brain, has negative effects on individuals and society. This paper concerns using Spontaneous Speech (ADReSS) Challenge of Interspeech 2020 to classify Alzheimer's dementia. We used (1) VGGish, a deep, pretrained, Tensorflow model as an audio feature extractor, and Scikit-learn classifiers to detect signs of dementia in speech. Three classifiers (LinearSVM, Perceptron, 1NN) were 59.1% accurate, which was 3% above the best-performing baseline models trained on the acoustic features used in the challenge. We also proposed (2) DemCNN, a new PyTorch raw waveform-based convolutional neural network model that was 63.6% accurate, 7% more accurate then the best-performing baseline linear discriminant analysis model. We discovered that audio transfer learning with a pretrained VGGish feature extractor performs better than the baseline approach using automatically extracted acoustic features. Our DepCNN exhibits good generalization capabilities. Both methods presented in this paper offer progress toward new, innovative, and more effective computer-based screening of dementia through spontaneous speech.

Classifying Alzheimer's Disease Using Audio and Text-Based Representations of Speech

Alzheimer's Disease (AD) is a form of dementia that affects the memory, cognition, and motor skills of patients. Extensive research has been done to develop accessible, cost-effective, and non-invasive techniques for the automatic detection of AD. Previous research has shown that speech can be used to distinguish between healthy patients and afflicted patients. In this paper, the ADReSS dataset, a dataset balanced by gender and age, was used to automatically classify AD from spontaneous speech. The performance of five classifiers, as well as a convolutional neural network and long short-term memory network, was compared when trained on audio features (i-vectors and x-vectors) and text features (word vectors, BERT embeddings, LIWC features, and CLAN features). The same audio and text features were used to train five regression models to predict the Mini-Mental State Examination score for each patient, a score that has a maximum value of 30. The top-performing classification models were the support vector machine and random forest classifiers trained on BERT embeddings, which both achieved an accuracy of 85.4% on the test set. The best-performing regression model was the gradient boosting regression model trained on BERT embeddings and CLAN features, which had a root mean squared error of 4.56 on the test set. The performance on both tasks illustrates the feasibility of using speech to classify AD and predict neuropsychological scores.