Search

INTRODUCTION: Identification of individuals with mild cognitive impairment (MCI) who are at risk of developing Alzheimer's disease (AD) is crucial for early intervention and selection of clinical trials. METHODS: We applied natural language processing techniques along with machine learning methods to develop a method for automated prediction of progression to AD within 6 years using speech. The study design was evaluated on the neuropsychological test interviews of n = 166 participants from the Framingham Heart Study, comprising 90 progressive MCI and 76 stable MCI cases. RESULTS: Our best models, which used features generated from speech data, as well as age, sex, and education level, achieved an accuracy of 78.5% and a sensitivity of 81.1% to predict MCI‐to‐AD progression within 6 years. DISCUSSION: The proposed method offers a fully automated procedure, providing an opportunity to develop an inexpensive, broadly accessible, and easy‐to‐administer screening tool for MCI‐to‐AD progression prediction, facilitating development of remote assessment. Highlights: Voice recordings from neuropsychological exams coupled with basic demographics can lead to strong predictive models of progression to dementia from mild cognitive impairment.The study leveraged AI methods for speech recognition and processed the resulting text using language models.The developed AI‐powered pipeline can lead to fully automated assessment that could enable remote and cost‐effective screening and prognosis for Alzehimer's disease. [ABSTRACT FROM AUTHOR]

INTRODUCTION: Early cognitive decline may manifest in subtle differences in speech. METHODS: We examined 238 cognitively unimpaired adults from the Framingham Heart Study (32–75 years) who completed amyloid and tau PET imaging. Speech patterns during delayed recall of a story memory task were quantified via five speech markers, and their associations with global amyloid status and regional tau signal were examined. RESULTS: Total utterance time, number of between‐utterance pauses, speech rate, and percentage of unique words significantly correlated with delayed recall score although the shared variance was low (2%–15%). Delayed recall score was not significantly different between β‐amyoid‐positive (Aβ+) and ‐negative (Aβ−) groups and was not associated with regional tau signal. However, longer and more between‐utterance pauses, and slower speech rate were associated with increased tau signal across medial temporal and early neocortical regions. DISCUSSION: Subtle speech changes during memory recall may reflect cognitive impairment associated with early Alzheimer's disease pathology. Highlights: Speech during delayed memory recall relates to tau PET signal across adulthood.Delayed memory recall score was not associated with tau PET signal.Speech shows greater sensitivity to detecting subtle cognitive changes associated with early tau accumulation.Our cohort spans adulthood, while most PET imaging studies focus on older adults. [ABSTRACT FROM AUTHOR]

Background: With new drug treatments emerging, the need for early Alzheimer's disease (AD) detection methods that can be done in a primary care setting has become urgent. Significant advances have been made in validating blood‐based AD biomarkers, but the absence of a validated, non‐invasive clinical indicator remains a barrier for determining those who might be optimal candidates for treatment. The digital clock drawing test (dCDT) has been previously shown to be related to AD PET biomarkers. However, it is unknown whether and the extent to which dCDT would predict AD as compared to blood‐based AD biomarkers. The current study used an interpretable machine learning approach to determine whether the dCDT, independently and in combination with blood‐based AD biomarkers (e.g., amyloid‐β 42 (Aβ42); total tau (T‐tau)), was predictive of incident AD. Method: This study included participants from the Framingham Heart Study Generation 2 cohort (Gen 2; n = 527) who had dCDT measures, plasma Aβ42, and T‐tau. The fast interpretable greedy‐tree sums method was used to identify clinical decision rules for AD prediction. Synthetic minority oversampling technique was employed to handle the class imbalance. The predictive ability of plasma Aβ42, T‐tau and dCDT measures was evaluated by both area under the curve (AUC) and standardized partial AUC. The permutation feature importance method was used to show the importance of each measure. Result: Interpretable clinical decision rules were generated as multivariate biomarker signatures for AD prediction (Figure 1). The dCDT (AUC 0.859) was better at incident AD prediction than AD blood‐based biomarkers (AUC 0.839). The combination of dCDT, Aβ42, and T‐tau best predicted incident AD (AUC 0.884) (Figure 2). The combination of dCDT and blood‐based biomarkers also has higher partial AUC than individual ones (Table 1). The total score and information processing scores were the most important contributing dCDT measures. Conclusion: These results highlight the efficiency of detecting early AD through the integration of digital measures as a clinical indicator combined with AD specific biomarkers. The interpretable decision rules have the potential to provide automated results that can be used in practical clinical settings, although external validation is warranted. [ABSTRACT FROM AUTHOR]

Introduction: Automated computational assessment of neuropsychological tests would enable widespread, cost‐effective screening for dementia. Methods: A novel natural language processing approach is developed and validated to identify different stages of dementia based on automated transcription of digital voice recordings of subjects' neuropsychological tests conducted by the Framingham Heart Study (n = 1084). Transcribed sentences from the test were encoded into quantitative data and several models were trained and tested using these data and the participants' demographic characteristics. Results: Average area under the curve (AUC) on the held‐out test data reached 92.6%, 88.0%, and 74.4% for differentiating Normal cognition from Dementia, Normal or Mild Cognitive Impairment (MCI) from Dementia, and Normal from MCI, respectively. Discussion: The proposed approach offers a fully automated identification of MCI and dementia based on a recorded neuropsychological test, providing an opportunity to develop a remote screening tool that could be adapted easily to any language. [ABSTRACT FROM AUTHOR]

Background: Digital is the new blood, where collection, use, and storage of raw digital data streams enabled by technology‐embedded sensors are replacing analog methods. Analyses of digital data extend beyond current applications, analogous to previously stored blood currently be used for omics/biomarker research. Today digital technologies are used to measure Alzheimer's disease (AD)‐related clinical symptoms (e.g., cognition, function, and behavior), but future possibilities remain to be determined. Importantly, digital data collection can go beyond current intermittent sampling by employing internet‐connected applications and devices that allow real‐time monitoring and accurate detection of changes, even within a widely heterogeneous spectrum. Method: Beginning in 2005, the Framingham Heart Study (FHS) began digitally recording participants' spoken responses to neuropsychological (NP) test questions, and in 2011 extended to a digital pen to record written responses. The emergence of wearable devices and smartphone applications has led to an expanded use of digital technologies beyond recordings during NP assessments. The Boston University Alzheimer's Disease Research Center (BU ADRC) leverages digital sensors in multiple internet‐connected devices that longitudinally collect digital data, in both in‐person and remote settings. Result: The first publication involving FHS digital data was in 2016; the remaining (n> 17) have been from 2020‐current. Digital data measuring cognition, function and behavior are being collected through the BU ADRC, where results to date indicate nearly 100% and 75% uptake of active and passive engagement applications, respectively, and 95% uptake for wearable devices (n = 60). Longitudinal assessments include > 3 months for ∼50 participants; > 1 year for n∼21 participants. Conclusion: Opportunities for improved trial outcome measurement are evident in the broad range of outcomes digital sensors capture beyond traditional research methodologies. Digital tools are being deployed for rapid screening for clinical trial inclusion, but more transformative opportunities are still to come. Replacing the sporadic and highly heterogeneous measures of cognition and related behaviors with more accurate measures that track treatment impact in real‐time at the individual level will shift the concept of clinically meaningful outcomes from group indicators to personalized ones. Digital is central to the realization of virtual clinical trial monitoring for precision treatment effects. [ABSTRACT FROM AUTHOR]

Background: Early detection of cognitive decline is critical to identify effective treatments for Alzheimer's disease and related dementias. The inherent heterogeneity of behaviors associated with brain health may obscure the signals derived from a typical neuropsychological evaluation, partly because of the limited frequency of assessment with which it can be conducted. This may lead to inaccurate or poor detection sensitivity. Digital health technology using smartphone applications enables frequent assessment and, thereby, more precise monitoring of brain health. The current study sought to examine the feasibility of using smartphone application‐based tasks in community‐based participants from the Framingham Heart Study (FHS). Method: The smartphone application consists of three blocks of assessments, lasting approximately 15‐20 minutes (Figure 1). Block 1: Participants answer questions to ensure that they can see the stimuli and hear the instructions adequately. Block 2: Participants complete a series of cognitive tasks and respond to open‐ended questions which are voice recorded. Block 3: To ensure that performance is not influenced by external factors, participants respond to questions, for example, about their experiences with using the smartphone application. Result: To date, 318 FHS participants have been enrolled into the study. Of these, as of 12/09/2022, 168 participants' first assessment period has been completed within the two week period after being enrolled in the study. Of the 168 participants, a large majority (81%; n = 136) completed all the tasks, while the remaining 32 completed partial assessment. Table 1 reveals that the overall percentage of participants who have been enrolled in the study and the percentage of those who completed their first assessment are highly similar. Conclusion: Findings reveal that older adults from a large, community‐based study are able to successfully download and complete the first set of tasks using their smartphone application. As data collection is ongoing, the number of older adults recruited into the study is expected to increase exponentially. These results indicate that it is feasible to monitor brain health on an on‐going basis using self‐administered smartphone applications, potentially creating an effective, lower cost, and scalable approach to detect significant changes in cognitive functions earlier and more accurately in the preclinical neurodegenerative process. [ABSTRACT FROM AUTHOR]

Background: Early detection of individuals at high risk of future progression toward dementia is in great need, particularly in the era of emerging treatments for Alzheimer's disease. Given the easier administration of neuropsychological (NP) tests over invasive and expensive biomarker acquisition, we seek to predict the future cognitive status of an individual based on digital voice recordings of NP exams. Method: We utilized automatic speech recognition and natural language processing techniques to create an automated tool that predicts if an individual with mild cognitive impairment (MCI) will develop dementia in a fixed follow‐up window (5 years). We characterize the predictive accuracy of this approach using data from the Framingham Heart Study. First, the automated transcription of the digital voice recordings were classified into 8 main NP sub‐tests including memory assessment, naming and language skill, verbal fluency, general questions, etc. Using a large language model, only the participants' sentences were encoded into quantitative data. These data and the participants' characteristics, such as age, sex, ApoE4 (+/‐) genotype, and education were employed to train and test a binary classifier, predicting dementia withing 5 years after a positive MCI diagnosis. We also provide a comparison between our approach and alternative models using demographics and ApoE genotype or the Mini‐Mental State Examination (MMSE) score. Result: We evaluated the performance of the classification task on the data containing 111 stable MCI and 103 progressive MCI subjects. Our method achieved an average accuracy of 75.5% on the held‐out test data while a baseline model using age, sex, education, and ApoE has an average accuracy of 71.8%. The sensitivity and specificity of the proposed method versus the baseline model are 78.2% vs. 81.8% and 72.7% vs. 61.8%, respectively. Our approach also significantly outperforms the model based on the MMSE score by a margin of 14% in accuracy. Conclusion: We have analyzed digital voice recordings of NP exams for participants with MCI to predict their cognitive status in 5 years. The proposed approach offers a fully automated procedure with higher accuracy, providing an opportunity to develop an accessible and easy‐to‐administer screening tool that could be easily adapted to any language. [ABSTRACT FROM AUTHOR]

Background: Diagnosis of dementia due to Alzheimer's disease (AD) is a resource‐intensive process, often requiring clinical assessments and neuroimaging which may fall beyond the reach of patients in remote settings. Given the challenges of traditional dementia work‐up, identification of disease‐relevant signatures from low‐cost modalities has attracted interest as a means to improve disease detection. Here, we describe a machine learning strategy to associate acoustic perturbations in voice with dementia status using digital recordings of neuropsychological exams administered in the Framingham Heart Study (FHS). Method: We used 118 voice recordings from 79 participants to extract jitter, shimmer, and haronics‐to‐nosie ratio (HNR) from the ComParE 2016 dataset. Each of these acoustic perturbation measures are commonly used in the clinical context to characterize pathological voice. For each voice transcript, these features were computed and and fed as input to a one‐dimensional convolutional neural network to predict dementia status. The data was randomly split in the ratio of 1:1 (at the participant‐level), where 50% of the cases were used for training and the remaining for testing. This process was repeated 10 times and the average model performance was reported. For each model run, the neural network was trained for 500 epochs without early stopping and the model obtained after the final epoch was used for testing. Result: The mean (±std) age of the participants included in the study was 82.6±8.3 years, and 21 of them were ApoE positive. The mean (±std) area under the receiver operating characteristic curve and the mean (±std) area under the precision‐recall curve of the model classifying participants with dementia from those with normal cognition was 0.673±0.043 and 0.791±0.052, respectively (Fig. 1). Also, the mean accuracy (±std) was 0.615±0.025, the mean precision (±std) was 0.714±0.121, and the mean sensitivity/recall (±std) was 0.721±0.179. Conclusion: Our study presents a proof‐of‐principle that deep learning on acoustic perturbation measures can predict dementia status. Further studies with additional participants and an expanded set of acoustic, speech and linguistic features are needed to evaluate the full potential of machine learning to derive disease‐relevant signatures from audio data, and potentially expand the reach of early detection and intervention efforts. [ABSTRACT FROM AUTHOR]

Background: To examine the predictive abilities of 1) traditional neuropsychological (NP) test summary scores, and 2) error and process scores in separating cognitively resilient individuals from those who exhibit cognitive decline. Method: A total of 69 participants from the Framingham Heart Study (FHS) Offspring and Omni 1 cohorts who had ≥1 NP assessment from 2011–2014 and at high risk for dementia were included in the sample. Participants also had to be ≥60 years of age and dementia‐free at the time of NP assessment. Being at high risk for dementia was determined using a cutoff score of ≥23 based on a published FHS dementia risk score algorithm. Dementia diagnosis was adjudicated based on DSM‐IV criteria by consensus. Participants completed 10 NP tests, which were scored using 11 traditional scores and 75 error and process scores. Supervised machine learning algorithms were used to classify individuals who were cognitively resilient, defined as not progressing to dementia within ≥5 years after the NP evaluation. Result: Among the 69 participants in our sample, 10 were cognitively resilient while 59 developed dementia. Average area under the curve when using traditional NP metrics to predict cognitive resilience ranged from 90.5% to 95.7% across each supervised machine learning algorithm (logistic regression, random forest, support vector machines, and light gradient boosting machine) (Figure 1). Average accuracy ranged from 85.3% to 92.6%, while average sensitivity was 100% across all models and average specificity ranged from 82.7% to 91.4% (Table 1). Average area under the curve when using the error and process scores to predict cognitive resilience ranged from 93.2% to 98.3% across each supervised machine learning algorithm (Figure 2). Average accuracy ranged from 89.9% to 97.0%, while average sensitivity was also 100% across all models and average specificity ranged from 88.2% to 96.5% (Table 2). All performance metrics were calculated based on five‐fold cross‐validation. Conclusion: Results from this pilot study suggest that use of NP tests may serve as an accurate method to identify individuals who are cognitive resilient, allowing for the study of genetic, lifestyle, and other factors that confer resilience among individuals with an elevated risk for dementia. [ABSTRACT FROM AUTHOR]

Background: Reliable cognitive impairment screening tools that are easy to administer and minimally time consuming are greatly needed. Given the high sensitivity of neuropsychological (NP) exams in detection of cognitive decline, we seek to develop an automated screening tool to detect dementia and mild cognitive impairment (MCI) based on digital voice recordings of NP assessments. This could enable wide‐spread screening for dementia and accelerate preventative efforts. Method: We used natural language processing methods to create a screening tool that identifies different stages of dementia based on automated transcription of digital voice recordings. The transcribed sentences were classified into 8 main sub‐tests including memory assessment, naming and language skill, verbal fluency, general questions, etc. Using the idea of transfer learning, we encoded the participants' sentences into quantitative data. This data and the participants' demographic variables such as age, sex, Apoe gene, and education were employed to train and test three binary classification tasks, (I) Normal cognition versus Dementia, (II) Normal/MCI versus Dementia, and (III) Normal versus MCI. Result: We evaluated the performance of the classification tasks using the digital voice recordings of NP assessments, collected from the Framingham Heart Study, containing 410 cognitively intact subjects, 387 MCI, and 287 subjects with dementia. The average Area Under the Curve (AUC) on the held‐out test data reached 92.6%, 88.0%, and 74.4% for differentiating Normal from Dementia, Normal or MCI from Dementia, and Normal from MCI, respectively. Looking at the importance of the sub‐tests in differentiating MCI from Normal, we note that general questions can be more useful for assessment of MCI, whereas verbal fluency would not be as useful in this task. Conclusion: The proposed approach offers a fully automated identification of MCI and dementia based on a recorded NP test, providing an opportunity to develop a remote screening tool that could be easily adapted to any language. [ABSTRACT FROM AUTHOR]

Background: Detection of any form of cognitive impairment is challenging and the subjects have to undergo numerous evaluations and clinical tests. Hence, it would be of great importance to design a reliable and accessible procedure by which patients may get diagnosed for dementia remotely. The capability of the Clock Drawing Test (CDT)as an effective cognitive assessment tool has motivated us to develop an online diagnostic tool by leveraging artificial intelligence techniques. Method: Digital pen recordings of 3,263 normal subjects and 160 with dementia in the Framingham Heart Study (FHS) were collected, where all subjects have completed two analog clock drawings, one drawn on command and the other by copying. Using the idea of transfer learning, we first modified and trained a Convolutional Neural Network (CNN) pre‐trained on the ImageNet dataset to extract high level features of the CDT images, which generated a score associated with the likelihood of dementia for each patient. The proposed method integrates the scores of the CDT images and other demographic information. Therefore, the generated scores for both command and copy CDTs along with age were used to train a logistic regression model to classify individuals as demented or normal. Result: We have evaluated the performance of the developed models by applying 5‐fold cross validation on the FHS dataset. On the test dataset, the model (modified pre‐trained CNN) based on command CDT images yielded an AUC of 0.81±0.043. The logistic regression model using age and the generated scores of command and copy CDTs, yielded an average AUC and average F1 score of 0.92±0.008 and 0.94±0.008, respectively. Conclusion: Our method need not necessarily have access to digital biomarkers or clinical tests since the CDT can be completed using pen and paper, capturing the image using a smartphone. Hence, our method offers a cost‐effective and accurate screening tool to diagnose dementia and related diseases remotely. [ABSTRACT FROM AUTHOR]

Background: Age remains the strongest risk factor for Alzheimer's disease (AD). With the number of people over 65 now outnumbering those under the age of 5, the global burden of AD is rapidly increasing. Cognitive decline is the hallmark behavioral symptom of AD. However, accurate cognitive assessment requires consideration of a myriad of factors, including education, language, and culture. In fact, in a majority of low‐to‐middle income countries, as well as many emerging economies, there is no formalized doctoral training in neuropsychology. These circumstances necessitate embracing novel data collection methods in order to preserve the voice of the neuropsychologist. Technological advances can provide the tools to address the global realities of AD and other related disorders. Method: We have developed a brain health monitoring platform that leverages the relatively ubiquitous use of smartphones across virtually all countries and collects behavioral measures that are inextricably linked to cognitive skills. We are applying machine learning approaches to quantify digital indices into cognitive ones. Importantly, we have strategically selected mobile applications that measure behaviors that are 1) common regardless of age, education, language and culture and 2) can be flexibly customized to circumstances. Result: We have begun deploying versions of this high tech‐driven platform in the U.S. Further, we are starting to have an impact internationally, particularly in Asia. The health and age demographics in China make its population particularly vulnerable to AD. We present findings that speak to this risk and how our platform, and others like it, will help fill the neuropsychological expertise gap that such countries are currently experiencing. Conclusion: AD is a rising epidemic. Early and accurate detection is the first step to finding disease modifying interventions and providing the context necessary for maximized efficacy. The coupling of professional neuropsychology skills and digital technology will generate solutions that can sufficiently address the impending challenge of conducting cognitive assessments universally. This potentially includes resolving longstanding health disparities by extending neuropsychologists' reach into regions with significant economic and/or geographical constraints. [ABSTRACT FROM AUTHOR]